From 31eea8bbebb19428ba938bd37b8701758f40e269 Mon Sep 17 00:00:00 2001 From: mhjensen Date: Thu, 7 Nov 2019 14:11:06 +0100 Subject: [PATCH] interchanging --- .../html/._DecisionTrees-bs000.html | 14 +- .../html/._DecisionTrees-bs001.html | 14 +- .../html/._DecisionTrees-bs002.html | 14 +- .../html/._DecisionTrees-bs003.html | 14 +- .../html/._DecisionTrees-bs004.html | 14 +- .../html/._DecisionTrees-bs005.html | 14 +- .../html/._DecisionTrees-bs006.html | 14 +- .../html/._DecisionTrees-bs007.html | 14 +- .../html/._DecisionTrees-bs008.html | 14 +- .../html/._DecisionTrees-bs009.html | 14 +- .../html/._DecisionTrees-bs010.html | 14 +- .../html/._DecisionTrees-bs011.html | 14 +- .../html/._DecisionTrees-bs012.html | 14 +- .../html/._DecisionTrees-bs013.html | 14 +- .../html/._DecisionTrees-bs014.html | 14 +- .../html/._DecisionTrees-bs015.html | 14 +- .../html/._DecisionTrees-bs016.html | 14 +- .../html/._DecisionTrees-bs017.html | 14 +- .../html/._DecisionTrees-bs018.html | 14 +- .../html/._DecisionTrees-bs019.html | 14 +- .../html/._DecisionTrees-bs020.html | 14 +- .../html/._DecisionTrees-bs021.html | 14 +- .../html/._DecisionTrees-bs022.html | 14 +- .../html/._DecisionTrees-bs023.html | 14 +- .../html/._DecisionTrees-bs024.html | 14 +- .../html/._DecisionTrees-bs025.html | 14 +- .../html/._DecisionTrees-bs026.html | 14 +- .../html/._DecisionTrees-bs027.html | 14 +- .../html/._DecisionTrees-bs028.html | 14 +- .../html/._DecisionTrees-bs029.html | 14 +- .../html/._DecisionTrees-bs030.html | 14 +- .../html/._DecisionTrees-bs031.html | 14 +- .../html/._DecisionTrees-bs032.html | 14 +- .../html/._DecisionTrees-bs033.html | 14 +- .../html/._DecisionTrees-bs034.html | 14 +- .../html/._DecisionTrees-bs035.html | 14 +- .../html/._DecisionTrees-bs036.html | 14 +- .../html/._DecisionTrees-bs037.html | 14 +- .../html/._DecisionTrees-bs038.html | 14 +- .../html/._DecisionTrees-bs039.html | 14 +- .../html/._DecisionTrees-bs040.html | 14 +- .../html/._DecisionTrees-bs041.html | 14 +- .../html/._DecisionTrees-bs042.html | 14 +- .../html/._DecisionTrees-bs043.html | 14 +- .../html/._DecisionTrees-bs044.html | 107 +- .../html/._DecisionTrees-bs045.html | 91 +- .../html/._DecisionTrees-bs046.html | 80 +- .../html/._DecisionTrees-bs047.html | 14 +- .../html/._DecisionTrees-bs048.html | 14 +- .../html/._DecisionTrees-bs049.html | 14 +- .../html/._DecisionTrees-bs050.html | 14 +- .../html/._DecisionTrees-bs051.html | 14 +- .../html/._DecisionTrees-bs052.html | 14 +- .../html/._DecisionTrees-bs053.html | 14 +- .../html/._DecisionTrees-bs054.html | 14 +- .../html/._DecisionTrees-bs055.html | 14 +- .../html/._DecisionTrees-bs056.html | 14 +- .../html/._DecisionTrees-bs057.html | 14 +- .../DecisionTrees/html/DecisionTrees-bs.html | 14 +- .../html/DecisionTrees-reveal.html | 126 +- .../html/DecisionTrees-solarized.html | 132 +- doc/pub/DecisionTrees/html/DecisionTrees.html | 132 +- .../DecisionTrees/ipynb/DataFiles/cancer.dot | 16 +- .../DecisionTrees/ipynb/DataFiles/cancer.png | Bin 223889 -> 221592 bytes .../DecisionTrees/ipynb/DecisionTrees.ipynb | 144 +- .../ipynb/ipynb-DecisionTrees-src.tar.gz | Bin 294061 -> 294061 bytes .../pdf/DecisionTrees-minted.pdf | Bin 526683 -> 526688 bytes doc/src/DecisionTrees/DecisionTrees-bs.html | 311 ++ .../DecisionTrees/DecisionTrees-minted.pdf | Bin 0 -> 526688 bytes .../DecisionTrees-plain-minted.tex | 2395 +++++++++ .../DecisionTrees/DecisionTrees-reveal.html | 2657 ++++++++++ .../DecisionTrees-solarized.html | 2502 +++++++++ doc/src/DecisionTrees/DecisionTrees.do.txt | 125 +- doc/src/DecisionTrees/DecisionTrees.html | 2507 +++++++++ doc/src/DecisionTrees/DecisionTrees.ipynb | 2602 ++++++++++ doc/src/DecisionTrees/DecisionTrees.p.tex | 2425 +++++++++ doc/src/DecisionTrees/DecisionTrees.tex | 2395 +++++++++ .../_minted-DecisionTrees/default.pygstyle | 0 .../ipynb-DecisionTrees-src.tar.gz | Bin 0 -> 294061 bytes doc/src/DecisionTrees/reveal.js/.gitignore | 8 + doc/src/DecisionTrees/reveal.js/.travis.yml | 5 + .../DecisionTrees/reveal.js/CONTRIBUTING.md | 23 + doc/src/DecisionTrees/reveal.js/Gruntfile.js | 140 + doc/src/DecisionTrees/reveal.js/LICENSE | 19 + doc/src/DecisionTrees/reveal.js/README.md | 1052 ++++ doc/src/DecisionTrees/reveal.js/bower.json | 27 + .../reveal.js/css/images/cbc_footer.png | Bin 0 -> 10008 bytes .../reveal.js/css/images/cbc_symbol.png | Bin 0 -> 2946 bytes .../reveal.js/css/images/simula_footer.png | Bin 0 -> 2513 bytes .../reveal.js/css/images/simula_logo.png | Bin 0 -> 2138 bytes .../reveal.js/css/images/simula_symbol.png | Bin 0 -> 2138 bytes .../reveal.js/css/images/uio_footer.png | Bin 0 -> 18189 bytes .../reveal.js/css/images/uio_symbol.png | Bin 0 -> 11352 bytes .../reveal.js/css/print/paper.css | 202 + .../DecisionTrees/reveal.js/css/print/pdf.css | 157 + .../DecisionTrees/reveal.js/css/reveal.css | 1886 +++++++ .../DecisionTrees/reveal.js/css/reveal.scss | 1319 +++++ .../reveal.js/css/theme/README.md | 23 + .../reveal.js/css/theme/beige.css | 154 + .../reveal.js/css/theme/beigesmall.css | 155 + .../reveal.js/css/theme/black.css | 273 + .../reveal.js/css/theme/blood.css | 180 + .../DecisionTrees/reveal.js/css/theme/cbc.css | 144 + .../reveal.js/css/theme/darkgray.css | 153 + .../reveal.js/css/theme/default.css | 153 + .../reveal.js/css/theme/league.css | 279 + .../reveal.js/css/theme/moon.css | 153 + .../reveal.js/css/theme/night.css | 141 + .../reveal.js/css/theme/serif.css | 143 + .../reveal.js/css/theme/simple.css | 144 + .../reveal.js/css/theme/simula.css | 144 + .../DecisionTrees/reveal.js/css/theme/sky.css | 150 + .../reveal.js/css/theme/solarized.css | 153 + .../reveal.js/css/theme/source/beige.scss | 50 + .../css/theme/source/beigesmall.scss | 51 + .../reveal.js/css/theme/source/black.scss | 49 + .../reveal.js/css/theme/source/blood.scss | 91 + .../reveal.js/css/theme/source/cbc.scss | 39 + .../reveal.js/css/theme/source/darkgray.scss | 42 + .../reveal.js/css/theme/source/default.scss | 42 + .../reveal.js/css/theme/source/league.scss | 34 + .../reveal.js/css/theme/source/moon.scss | 68 + .../reveal.js/css/theme/source/night.scss | 35 + .../reveal.js/css/theme/source/serif.scss | 35 + .../reveal.js/css/theme/source/simple.scss | 38 + .../reveal.js/css/theme/source/simula.scss | 39 + .../reveal.js/css/theme/source/sky.scss | 46 + .../reveal.js/css/theme/source/solarized.scss | 74 + .../reveal.js/css/theme/source/white.scss | 49 + .../reveal.js/css/theme/white.css | 273 + doc/src/DecisionTrees/reveal.js/index.html | 411 ++ doc/src/DecisionTrees/reveal.js/js/reveal.js | 4508 +++++++++++++++++ .../reveal.js/lib/css/zenburn.css | 117 + .../reveal.js/lib/font/league-gothic/LICENSE | 2 + .../lib/font/league-gothic/league-gothic.css | 10 + .../lib/font/league-gothic/league-gothic.eot | Bin 0 -> 25696 bytes .../lib/font/league-gothic/league-gothic.ttf | Bin 0 -> 64256 bytes .../lib/font/league-gothic/league-gothic.woff | Bin 0 -> 30764 bytes .../lib/font/source-sans-pro/LICENSE | 45 + .../source-sans-pro-italic.eot | Bin 0 -> 75720 bytes .../source-sans-pro-italic.ttf | Bin 0 -> 238084 bytes .../source-sans-pro-italic.woff | Bin 0 -> 98556 bytes .../source-sans-pro-regular.eot | Bin 0 -> 88070 bytes .../source-sans-pro-regular.ttf | Bin 0 -> 288008 bytes .../source-sans-pro-regular.woff | Bin 0 -> 114324 bytes .../source-sans-pro-semibold.eot | Bin 0 -> 89897 bytes .../source-sans-pro-semibold.ttf | Bin 0 -> 284640 bytes .../source-sans-pro-semibold.woff | Bin 0 -> 115648 bytes .../source-sans-pro-semibolditalic.eot | Bin 0 -> 75706 bytes .../source-sans-pro-semibolditalic.ttf | Bin 0 -> 240944 bytes .../source-sans-pro-semibolditalic.woff | Bin 0 -> 98816 bytes .../font/source-sans-pro/source-sans-pro.css | 39 + .../reveal.js/lib/js/classList.js | 2 + .../reveal.js/lib/js/head.min.js | 8 + .../reveal.js/lib/js/html5shiv.js | 7 + doc/src/DecisionTrees/reveal.js/package.json | 45 + .../reveal.js/plugin/highlight/highlight.js | 30 + .../reveal.js/plugin/leap/leap.js | 159 + .../reveal.js/plugin/markdown/example.html | 129 + .../reveal.js/plugin/markdown/example.md | 31 + .../reveal.js/plugin/markdown/markdown.js | 393 ++ .../reveal.js/plugin/markdown/marked.js | 6 + .../reveal.js/plugin/math/math.js | 64 + .../reveal.js/plugin/multiplex/client.js | 13 + .../reveal.js/plugin/multiplex/index.js | 56 + .../reveal.js/plugin/multiplex/master.js | 51 + .../reveal.js/plugin/notes-server/client.js | 60 + .../reveal.js/plugin/notes-server/index.js | 66 + .../reveal.js/plugin/notes-server/notes.html | 396 ++ .../reveal.js/plugin/notes/notes.html | 406 ++ .../reveal.js/plugin/notes/notes.js | 122 + .../reveal.js/plugin/print-pdf/print-pdf.js | 48 + .../reveal.js/plugin/remotes/remotes.js | 39 + .../reveal.js/plugin/search/search.js | 196 + .../reveal.js/plugin/zoom-js/zoom.js | 278 + .../reveal.js/test/examples/assets/image1.png | Bin 0 -> 21991 bytes .../reveal.js/test/examples/assets/image2.png | Bin 0 -> 10237 bytes .../reveal.js/test/examples/barebones.html | 41 + .../test/examples/embedded-media.html | 49 + .../reveal.js/test/examples/math.html | 185 + .../test/examples/slide-backgrounds.html | 144 + .../test/examples/slide-transitions.html | 101 + .../reveal.js/test/qunit-1.12.0.css | 244 + .../reveal.js/test/qunit-1.12.0.js | 2212 ++++++++ .../test-markdown-element-attributes.html | 134 + .../test/test-markdown-element-attributes.js | 46 + .../test/test-markdown-slide-attributes.html | 128 + .../test/test-markdown-slide-attributes.js | 47 + .../reveal.js/test/test-markdown.html | 52 + .../reveal.js/test/test-markdown.js | 15 + .../reveal.js/test/test-pdf.html | 83 + .../DecisionTrees/reveal.js/test/test-pdf.js | 15 + .../DecisionTrees/reveal.js/test/test.html | 85 + doc/src/DecisionTrees/reveal.js/test/test.js | 589 +++ 194 files changed, 39006 insertions(+), 867 deletions(-) create mode 100644 doc/src/DecisionTrees/DecisionTrees-bs.html create mode 100644 doc/src/DecisionTrees/DecisionTrees-minted.pdf create mode 100644 doc/src/DecisionTrees/DecisionTrees-plain-minted.tex create mode 100644 doc/src/DecisionTrees/DecisionTrees-reveal.html create mode 100644 doc/src/DecisionTrees/DecisionTrees-solarized.html create mode 100644 doc/src/DecisionTrees/DecisionTrees.html create mode 100644 doc/src/DecisionTrees/DecisionTrees.ipynb create mode 100644 doc/src/DecisionTrees/DecisionTrees.p.tex create mode 100644 doc/src/DecisionTrees/DecisionTrees.tex create mode 100644 doc/src/DecisionTrees/_minted-DecisionTrees/default.pygstyle create mode 100644 doc/src/DecisionTrees/ipynb-DecisionTrees-src.tar.gz create mode 100644 doc/src/DecisionTrees/reveal.js/.gitignore create mode 100644 doc/src/DecisionTrees/reveal.js/.travis.yml create mode 100644 doc/src/DecisionTrees/reveal.js/CONTRIBUTING.md create mode 100644 doc/src/DecisionTrees/reveal.js/Gruntfile.js create mode 100644 doc/src/DecisionTrees/reveal.js/LICENSE create mode 100644 doc/src/DecisionTrees/reveal.js/README.md create mode 100644 doc/src/DecisionTrees/reveal.js/bower.json create mode 100644 doc/src/DecisionTrees/reveal.js/css/images/cbc_footer.png create mode 100644 doc/src/DecisionTrees/reveal.js/css/images/cbc_symbol.png create mode 100644 doc/src/DecisionTrees/reveal.js/css/images/simula_footer.png create mode 100644 doc/src/DecisionTrees/reveal.js/css/images/simula_logo.png create mode 100644 doc/src/DecisionTrees/reveal.js/css/images/simula_symbol.png create mode 100644 doc/src/DecisionTrees/reveal.js/css/images/uio_footer.png create mode 100644 doc/src/DecisionTrees/reveal.js/css/images/uio_symbol.png create mode 100644 doc/src/DecisionTrees/reveal.js/css/print/paper.css create mode 100644 doc/src/DecisionTrees/reveal.js/css/print/pdf.css create mode 100644 doc/src/DecisionTrees/reveal.js/css/reveal.css create mode 100644 doc/src/DecisionTrees/reveal.js/css/reveal.scss create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/README.md create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/beige.css create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/beigesmall.css create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/black.css create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/blood.css create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/cbc.css create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/darkgray.css create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/default.css create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/league.css create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/moon.css create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/night.css create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/serif.css create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/simple.css create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/simula.css create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/sky.css create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/solarized.css create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/source/beige.scss create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/source/beigesmall.scss create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/source/black.scss create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/source/blood.scss create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/source/cbc.scss create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/source/darkgray.scss create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/source/default.scss create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/source/league.scss create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/source/moon.scss create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/source/night.scss create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/source/serif.scss create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/source/simple.scss create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/source/simula.scss create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/source/sky.scss create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/source/solarized.scss create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/source/white.scss create mode 100644 doc/src/DecisionTrees/reveal.js/css/theme/white.css create mode 100644 doc/src/DecisionTrees/reveal.js/index.html create mode 100644 doc/src/DecisionTrees/reveal.js/js/reveal.js create mode 100644 doc/src/DecisionTrees/reveal.js/lib/css/zenburn.css create mode 100644 doc/src/DecisionTrees/reveal.js/lib/font/league-gothic/LICENSE create mode 100644 doc/src/DecisionTrees/reveal.js/lib/font/league-gothic/league-gothic.css create mode 100644 doc/src/DecisionTrees/reveal.js/lib/font/league-gothic/league-gothic.eot create mode 100644 doc/src/DecisionTrees/reveal.js/lib/font/league-gothic/league-gothic.ttf create mode 100644 doc/src/DecisionTrees/reveal.js/lib/font/league-gothic/league-gothic.woff create mode 100644 doc/src/DecisionTrees/reveal.js/lib/font/source-sans-pro/LICENSE create mode 100644 doc/src/DecisionTrees/reveal.js/lib/font/source-sans-pro/source-sans-pro-italic.eot create mode 100644 doc/src/DecisionTrees/reveal.js/lib/font/source-sans-pro/source-sans-pro-italic.ttf create mode 100644 doc/src/DecisionTrees/reveal.js/lib/font/source-sans-pro/source-sans-pro-italic.woff create mode 100644 doc/src/DecisionTrees/reveal.js/lib/font/source-sans-pro/source-sans-pro-regular.eot create mode 100644 doc/src/DecisionTrees/reveal.js/lib/font/source-sans-pro/source-sans-pro-regular.ttf create mode 100644 doc/src/DecisionTrees/reveal.js/lib/font/source-sans-pro/source-sans-pro-regular.woff create mode 100644 doc/src/DecisionTrees/reveal.js/lib/font/source-sans-pro/source-sans-pro-semibold.eot create mode 100644 doc/src/DecisionTrees/reveal.js/lib/font/source-sans-pro/source-sans-pro-semibold.ttf create mode 100644 doc/src/DecisionTrees/reveal.js/lib/font/source-sans-pro/source-sans-pro-semibold.woff create mode 100644 doc/src/DecisionTrees/reveal.js/lib/font/source-sans-pro/source-sans-pro-semibolditalic.eot create mode 100644 doc/src/DecisionTrees/reveal.js/lib/font/source-sans-pro/source-sans-pro-semibolditalic.ttf create mode 100644 doc/src/DecisionTrees/reveal.js/lib/font/source-sans-pro/source-sans-pro-semibolditalic.woff create mode 100644 doc/src/DecisionTrees/reveal.js/lib/font/source-sans-pro/source-sans-pro.css create mode 100644 doc/src/DecisionTrees/reveal.js/lib/js/classList.js create mode 100644 doc/src/DecisionTrees/reveal.js/lib/js/head.min.js create mode 100644 doc/src/DecisionTrees/reveal.js/lib/js/html5shiv.js create mode 100644 doc/src/DecisionTrees/reveal.js/package.json create mode 100644 doc/src/DecisionTrees/reveal.js/plugin/highlight/highlight.js create mode 100644 doc/src/DecisionTrees/reveal.js/plugin/leap/leap.js create mode 100644 doc/src/DecisionTrees/reveal.js/plugin/markdown/example.html create mode 100644 doc/src/DecisionTrees/reveal.js/plugin/markdown/example.md create mode 100644 doc/src/DecisionTrees/reveal.js/plugin/markdown/markdown.js create mode 100644 doc/src/DecisionTrees/reveal.js/plugin/markdown/marked.js create mode 100644 doc/src/DecisionTrees/reveal.js/plugin/math/math.js create mode 100644 doc/src/DecisionTrees/reveal.js/plugin/multiplex/client.js create mode 100644 doc/src/DecisionTrees/reveal.js/plugin/multiplex/index.js create mode 100644 doc/src/DecisionTrees/reveal.js/plugin/multiplex/master.js create mode 100644 doc/src/DecisionTrees/reveal.js/plugin/notes-server/client.js create mode 100644 doc/src/DecisionTrees/reveal.js/plugin/notes-server/index.js create mode 100644 doc/src/DecisionTrees/reveal.js/plugin/notes-server/notes.html create mode 100644 doc/src/DecisionTrees/reveal.js/plugin/notes/notes.html create mode 100644 doc/src/DecisionTrees/reveal.js/plugin/notes/notes.js create mode 100644 doc/src/DecisionTrees/reveal.js/plugin/print-pdf/print-pdf.js create mode 100644 doc/src/DecisionTrees/reveal.js/plugin/remotes/remotes.js create mode 100644 doc/src/DecisionTrees/reveal.js/plugin/search/search.js create mode 100644 doc/src/DecisionTrees/reveal.js/plugin/zoom-js/zoom.js create mode 100644 doc/src/DecisionTrees/reveal.js/test/examples/assets/image1.png create mode 100644 doc/src/DecisionTrees/reveal.js/test/examples/assets/image2.png create mode 100644 doc/src/DecisionTrees/reveal.js/test/examples/barebones.html create mode 100644 doc/src/DecisionTrees/reveal.js/test/examples/embedded-media.html create mode 100644 doc/src/DecisionTrees/reveal.js/test/examples/math.html create mode 100644 doc/src/DecisionTrees/reveal.js/test/examples/slide-backgrounds.html create mode 100644 doc/src/DecisionTrees/reveal.js/test/examples/slide-transitions.html create mode 100644 doc/src/DecisionTrees/reveal.js/test/qunit-1.12.0.css create mode 100644 doc/src/DecisionTrees/reveal.js/test/qunit-1.12.0.js create mode 100644 doc/src/DecisionTrees/reveal.js/test/test-markdown-element-attributes.html create mode 100644 doc/src/DecisionTrees/reveal.js/test/test-markdown-element-attributes.js create mode 100644 doc/src/DecisionTrees/reveal.js/test/test-markdown-slide-attributes.html create mode 100644 doc/src/DecisionTrees/reveal.js/test/test-markdown-slide-attributes.js create mode 100644 doc/src/DecisionTrees/reveal.js/test/test-markdown.html create mode 100644 doc/src/DecisionTrees/reveal.js/test/test-markdown.js create mode 100644 doc/src/DecisionTrees/reveal.js/test/test-pdf.html create mode 100644 doc/src/DecisionTrees/reveal.js/test/test-pdf.js create mode 100644 doc/src/DecisionTrees/reveal.js/test/test.html create mode 100644 doc/src/DecisionTrees/reveal.js/test/test.js diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs000.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs000.html index cfc8c9a63..263295b72 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs000.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs000.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs001.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs001.html index ce0cd4709..79fbfc13e 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs001.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs001.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs002.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs002.html index 826c588f6..54da0e0de 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs002.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs002.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs003.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs003.html index 7227d3372..12f9d3166 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs003.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs003.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs004.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs004.html index 95963bdbe..18a6fbe9b 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs004.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs004.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs005.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs005.html index 7c89669c1..1a1bd8838 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs005.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs005.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs006.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs006.html index d66b98492..500fbf4fa 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs006.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs006.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs007.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs007.html index 0127d1301..d511cd5de 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs007.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs007.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs008.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs008.html index 212fcf49e..11194998e 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs008.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs008.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs009.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs009.html index 697c502e2..eb5aa5671 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs009.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs009.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs010.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs010.html index d255f36e7..85b95dded 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs010.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs010.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs011.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs011.html index 88e444f30..4be724e7f 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs011.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs011.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs012.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs012.html index 2809ab887..72a1783d5 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs012.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs012.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs013.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs013.html index 4f42c7a2b..3b717abff 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs013.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs013.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs014.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs014.html index 02016f7be..c12a0d38c 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs014.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs014.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs015.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs015.html index 7fd9c8230..8b6cf4b08 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs015.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs015.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs016.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs016.html index 8c7b4e80d..60adf1836 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs016.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs016.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs017.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs017.html index 904e184ad..d35ad431d 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs017.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs017.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs018.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs018.html index 644fe7bbb..f1b07f07e 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs018.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs018.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs019.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs019.html index 8d2a18b42..528dedd74 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs019.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs019.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs020.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs020.html index a1a708d79..715646b0a 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs020.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs020.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs021.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs021.html index c0c27cf60..a050b8b34 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs021.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs021.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs022.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs022.html index cf1324afa..0177f1aae 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs022.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs022.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs023.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs023.html index 1e3e04676..eea8b3d83 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs023.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs023.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs024.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs024.html index 4be1ac486..9b371bc57 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs024.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs024.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs025.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs025.html index b6806a8e1..0120d48af 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs025.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs025.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs026.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs026.html index 7ab79cd45..b0014d62f 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs026.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs026.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs027.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs027.html index 785b455fa..e19d0024e 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs027.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs027.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs028.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs028.html index 7755c7d4d..62f8fa987 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs028.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs028.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs029.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs029.html index ed615163d..fad03ce5d 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs029.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs029.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs030.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs030.html index f7faa8f50..3431c71d4 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs030.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs030.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs031.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs031.html index 5c6217b3f..62a37dcb0 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs031.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs031.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs032.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs032.html index 711b9c8e6..42345f6bb 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs032.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs032.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs033.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs033.html index d1b16089b..89a1e8210 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs033.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs033.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs034.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs034.html index 04be98f42..b3329a307 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs034.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs034.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs035.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs035.html index 0750e65bb..09eed4d06 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs035.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs035.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs036.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs036.html index f234a1f54..21fd02d4b 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs036.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs036.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs037.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs037.html index da7b2e0c1..33e9ecf58 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs037.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs037.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs038.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs038.html index 3b1d454d3..cc87226ec 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs038.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs038.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs039.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs039.html index c5850d1bb..f08ea8d23 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs039.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs039.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs040.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs040.html index b63ab6261..c3d1f4f4d 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs040.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs040.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs041.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs041.html index dbd2fe7c9..d2471d168 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs041.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs041.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs042.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs042.html index 9483dcfa1..795ef5233 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs042.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs042.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs043.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs043.html index 45f547496..459bb57ff 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs043.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs043.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs044.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs044.html index 7e7fe255f..757cecffe 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs044.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs044.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({
  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • @@ -240,74 +240,61 @@ MathJax.Hub.Config({ -

    Random Forests Compared with other Methods on the Cancer Data

    +

    Bootstrap with Random Forests Instead of a Single Tree, own Bagging

    + 

    import matplotlib.pyplot as plt
 import numpy as np
-from sklearn.model_selection import  train_test_split 
-from sklearn.datasets import load_breast_cancer
-from sklearn.svm import SVC
-from sklearn.linear_model import LogisticRegression
-from sklearn.tree import DecisionTreeClassifier
+from sklearn.model_selection import train_test_split
+from sklearn.pipeline import make_pipeline
+from sklearn.utils import resample
+from sklearn.ensemble import RandomForestRegressor
 
-# Load the data
-cancer = load_breast_cancer()
+np.random.seed(2018)
+
+n = 100
+n_boostraps = 100
+maxdegree = 14
+
+# Make data set.
+x = np.linspace(-3, 3, n).reshape(-1, 1)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
+error = np.zeros(maxdegree)
+bias = np.zeros(maxdegree)
+variance = np.zeros(maxdegree)
+polydegree = np.zeros(maxdegree)
+X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
 
-X_train, X_test, y_train, y_test = train_test_split(cancer.data,cancer.target,random_state=0)
-print(X_train.shape)
-print(X_test.shape)
-# Logistic Regression
-logreg = LogisticRegression(solver='lbfgs')
-logreg.fit(X_train, y_train)
-print("Test set accuracy with Logistic Regression: {:.2f}".format(logreg.score(X_test,y_test)))
-# Support vector machine
-svm = SVC(gamma='auto', C=100)
-svm.fit(X_train, y_train)
-print("Test set accuracy with SVM: {:.2f}".format(svm.score(X_test,y_test)))
-# Decision Trees
-deep_tree_clf = DecisionTreeClassifier(max_depth=None)
-deep_tree_clf.fit(X_train, y_train)
-print("Test set accuracy with Decision Trees: {:.2f}".format(deep_tree_clf.score(X_test,y_test)))
-#now scale the data
 from sklearn.preprocessing import StandardScaler
 scaler = StandardScaler()
 scaler.fit(X_train)
 X_train_scaled = scaler.transform(X_train)
 X_test_scaled = scaler.transform(X_test)
-# Logistic Regression
-logreg.fit(X_train_scaled, y_train)
-print("Test set accuracy Logistic Regression with scaled data: {:.2f}".format(logreg.score(X_test_scaled,y_test)))
-# Support Vector Machine
-svm.fit(X_train_scaled, y_train)
-print("Test set accuracy SVM with scaled data: {:.2f}".format(logreg.score(X_test_scaled,y_test)))
-# Decision Trees
-deep_tree_clf.fit(X_train_scaled, y_train)
-print("Test set accuracy with Decision Trees and scaled data: {:.2f}".format(deep_tree_clf.score(X_test_scaled,y_test)))
 
+for degree in range(maxdegree):
+    model = RandomForestRegressor()
+    y_pred = np.empty((y_test.shape[0], n_boostraps))
+    for i in range(n_boostraps):
+        x_, y_ = resample(X_train_scaled, y_train)
+        model.fit(x_, y_.ravel())
+        y_pred[:, i] = model.predict(X_test_scaled).ravel()
 
-from sklearn.ensemble import RandomForestClassifier
-from sklearn.preprocessing import LabelEncoder
-from sklearn.model_selection import cross_validate
-# Data set not specificied
-#Instantiate the model with 500 trees and entropy as splitting criteria
-Random_Forest_model = RandomForestClassifier(n_estimators=500,criterion="entropy")
-Random_Forest_model.fit(X_train_scaled, y_train)
-#Cross validation
-accuracy = cross_validate(Random_Forest_model,X_test_scaled,y_test,cv=10)['test_score']
-print(accuracy)
-print("Test set accuracy with Random Forests and scaled data: {:.2f}".format(Random_Forest_model.score(X_test_scaled,y_test)))
+    polydegree[degree] = degree
+    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )
+    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )
+    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )
+    print('Polynomial degree:', degree)
+    print('Error:', error[degree])
+    print('Bias^2:', bias[degree])
+    print('Var:', variance[degree])
+    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
 
-
-import scikitplot as skplt
-y_pred = Random_Forest_model.predict(X_test_scaled)
-skplt.metrics.plot_confusion_matrix(y_test, y_pred, normalize=True)
-plt.show()
-y_probas = Random_Forest_model.predict_proba(X_test_scaled)
-skplt.metrics.plot_roc(y_test, y_probas)
-plt.show()
-skplt.metrics.plot_cumulative_gain(y_test, y_probas)
+plt.plot(polydegree, error, label='Error')
+plt.plot(polydegree, bias, label='bias')
+plt.plot(polydegree, variance, label='Variance')
+plt.legend()
 plt.show()

    diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs045.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs045.html index 4372127c9..17be7fe3c 100644 --- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs045.html +++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs045.html @@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source ('Changing the Level of the Decision Tree', 2, None, '___sec40'), ('Random forests', 2, None, '___sec41'), ('Random Forest Algorithm', 2, None, '___sec42'), - ('Random Forests Compared with other Methods on the Cancer Data', + ('Bootstrap with Random Forests Instead of a Single Tree, own ' + 'Bagging', 2, None, '___sec43'), - ('Compare Bagging on Trees with Random Forests', + ('Random Forests Compared with other Methods on the Cancer Data', 2, None, '___sec44'), - ('Bootstrap with Random Forests Instead of a Single Tree, own ' - 'Bagging', + ('Compare Bagging on Trees with Random Forests', 2, None, '___sec45'), @@ -211,9 +211,9 @@ MathJax.Hub.Config({

  • Changing the Level of the Decision Tree
  • Random forests
  • Random Forest Algorithm
    • -
  • Random Forests Compared with other Methods on the Cancer Data
    • -
  • Compare Bagging on Trees with Random Forests
    • -
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • +
  • Random Forests Compared with other Methods on the Cancer Data
    • +
  • Compare Bagging on Trees with Random Forests
  • Boosting, a Bird'e Eye
  • Adaptive boosting: AdaBoost, Basic Algorithm
  • Basic Steps of AdaBoost
    • @@ -240,24 +240,75 @@ MathJax.Hub.Config({ -

    Compare Bagging on Trees with Random Forests

    +

    Random Forests Compared with other Methods on the Cancer Data

    -

    bag_clf = BaggingClassifier(
-    DecisionTreeClassifier(splitter="random", max_leaf_nodes=16, random_state=42),
-    n_estimators=500, max_samples=1.0, bootstrap=True, n_jobs=-1, random_state=42)
-
    -

    +

    import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import  train_test_split 
+from sklearn.datasets import load_breast_cancer
+from sklearn.svm import SVC
+from sklearn.linear_model import LogisticRegression
+from sklearn.tree import DecisionTreeClassifier
+
+# Load the data
+cancer = load_breast_cancer()
+
+X_train, X_test, y_train, y_test = train_test_split(cancer.data,cancer.target,random_state=0)
+print(X_train.shape)
+print(X_test.shape)
+# Logistic Regression
+logreg = LogisticRegression(solver='lbfgs')
+logreg.fit(X_train, y_train)
+print("Test set accuracy with Logistic Regression: {:.2f}".format(logreg.score(X_test,y_test)))
+# Support vector machine
+svm = SVC(gamma='auto', C=100)
+svm.fit(X_train, y_train)
+print("Test set accuracy with SVM: {:.2f}".format(svm.score(X_test,y_test)))
+# Decision Trees
+deep_tree_clf = DecisionTreeClassifier(max_depth=None)
+deep_tree_clf.fit(X_train, y_train)
+print("Test set accuracy with Decision Trees: {:.2f}".format(deep_tree_clf.score(X_test,y_test)))
+#now scale the data
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+# Logistic Regression
+logreg.fit(X_train_scaled, y_train)
+print("Test set accuracy Logistic Regression with scaled data: {:.2f}".format(logreg.score(X_test_scaled,y_test)))
+# Support Vector Machine
+svm.fit(X_train_scaled, y_train)
+print("Test set accuracy SVM with scaled data: {:.2f}".format(logreg.score(X_test_scaled,y_test)))
+# Decision Trees
+deep_tree_clf.fit(X_train_scaled, y_train)
+print("Test set accuracy with Decision Trees and scaled data: {:.2f}".format(deep_tree_clf.score(X_test_scaled,y_test)))
+
 
-
-
    bag_clf.fit(X_train, y_train)
-y_pred = bag_clf.predict(X_test)
 from sklearn.ensemble import RandomForestClassifier
-rnd_clf = RandomForestClassifier(n_estimators=500, max_leaf_nodes=16, n_jobs=-1, random_state=42)
-rnd_clf.fit(X_train, y_train)
-y_pred_rf = rnd_clf.predict(X_test)
-np.sum(y_pred == y_pred_rf) / len(y_pred) 
+from sklearn.preprocessing import LabelEncoder
+from sklearn.model_selection import cross_validate
+# Data set not specificied
+#Instantiate the model with 500 trees and entropy as splitting criteria
+Random_Forest_model = RandomForestClassifier(n_estimators=500,criterion="entropy")
+Random_Forest_model.fit(X_train_scaled, y_train)
+#Cross validation
+accuracy = cross_validate(Random_Forest_model,X_test_scaled,y_test,cv=10)['test_score']
+print(accuracy)
+print("Test set accuracy with Random Forests and scaled data: {:.2f}".format(Random_Forest_model.score(X_test_scaled,y_test)))
+
+
+import scikitplot as skplt
+y_pred = Random_Forest_model.predict(X_test_scaled)
+skplt.metrics.plot_confusion_matrix(y_test, y_pred, normalize=True)
+plt.show()
+y_probas = Random_Forest_model.predict_proba(X_test_scaled)
+skplt.metrics.plot_roc(y_test, y_probas)
+plt.show()
+skplt.metrics.plot_cumulative_gain(y_test, y_probas)
+plt.show()
 
    
 
    
 
    
diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs046.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs046.html
index 09ecd3191..04cbb3018 100644
--- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs046.html
+++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs046.html
@@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source
               ('Changing the Level of the Decision Tree', 2, None, '___sec40'),
               ('Random forests', 2, None, '___sec41'),
               ('Random Forest Algorithm', 2, None, '___sec42'),
-              ('Random Forests Compared with other Methods on the Cancer Data',
+              ('Bootstrap with Random Forests Instead of a Single Tree, own '
+               'Bagging',
                2,
                None,
                '___sec43'),
-              ('Compare  Bagging on Trees with Random Forests',
+              ('Random Forests Compared with other Methods on the Cancer Data',
                2,
                None,
                '___sec44'),
-              ('Bootstrap with Random Forests Instead of a Single Tree, own '
-               'Bagging',
+              ('Compare  Bagging on Trees with Random Forests',
                2,
                None,
                '___sec45'),
@@ -211,9 +211,9 @@ MathJax.Hub.Config({
       
  • Changing the Level of the Decision Tree
    • 
       
  • Random forests
    • 
       
  • Random Forest Algorithm
    • 
-      
  • Random Forests Compared with other Methods on the Cancer Data
    • 
-      
  • Compare  Bagging on Trees with Random Forests
    • 
-      
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • 
+      
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • 
+      
  • Random Forests Compared with other Methods on the Cancer Data
    • 
+      
  • Compare  Bagging on Trees with Random Forests
    • 
       
  • Boosting, a Bird'e Eye
    • 
       
  • Adaptive boosting: AdaBoost, Basic Algorithm
    • 
       
  • Basic Steps of AdaBoost
    • 
@@ -240,62 +240,24 @@ MathJax.Hub.Config({
 
 
 
-
    Bootstrap with Random Forests Instead of a Single Tree, own Bagging 
    
-
+
    Compare  Bagging on Trees with Random Forests  
    
 
    
 
 
-
    import matplotlib.pyplot as plt
-import numpy as np
-from sklearn.model_selection import train_test_split
-from sklearn.pipeline import make_pipeline
-from sklearn.utils import resample
-from sklearn.ensemble import RandomForestRegressor
+
    bag_clf = BaggingClassifier(
+    DecisionTreeClassifier(splitter="random", max_leaf_nodes=16, random_state=42),
+    n_estimators=500, max_samples=1.0, bootstrap=True, n_jobs=-1, random_state=42)
+
    
+
    
 
-np.random.seed(2018)
-
-n = 100
-n_boostraps = 100
-maxdegree = 14
-
-# Make data set.
-x = np.linspace(-3, 3, n).reshape(-1, 1)
-y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
-error = np.zeros(maxdegree)
-bias = np.zeros(maxdegree)
-variance = np.zeros(maxdegree)
-polydegree = np.zeros(maxdegree)
-X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
-
-from sklearn.preprocessing import StandardScaler
-scaler = StandardScaler()
-scaler.fit(X_train)
-X_train_scaled = scaler.transform(X_train)
-X_test_scaled = scaler.transform(X_test)
-
-for degree in range(maxdegree):
-    model = RandomForestRegressor()
-    y_pred = np.empty((y_test.shape[0], n_boostraps))
-    for i in range(n_boostraps):
-        x_, y_ = resample(X_train_scaled, y_train)
-        model.fit(x_, y_.ravel())
-        y_pred[:, i] = model.predict(X_test_scaled).ravel()
-
-    polydegree[degree] = degree
-    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )
-    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )
-    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )
-    print('Polynomial degree:', degree)
-    print('Error:', error[degree])
-    print('Bias^2:', bias[degree])
-    print('Var:', variance[degree])
-    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
-
-plt.plot(polydegree, error, label='Error')
-plt.plot(polydegree, bias, label='bias')
-plt.plot(polydegree, variance, label='Variance')
-plt.legend()
-plt.show()
+
+
    bag_clf.fit(X_train, y_train)
+y_pred = bag_clf.predict(X_test)
+from sklearn.ensemble import RandomForestClassifier
+rnd_clf = RandomForestClassifier(n_estimators=500, max_leaf_nodes=16, n_jobs=-1, random_state=42)
+rnd_clf.fit(X_train, y_train)
+y_pred_rf = rnd_clf.predict(X_test)
+np.sum(y_pred == y_pred_rf) / len(y_pred) 
 
    
 
    
 
    
diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs047.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs047.html
index 2c10930ed..2eb3e3331 100644
--- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs047.html
+++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs047.html
@@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source
               ('Changing the Level of the Decision Tree', 2, None, '___sec40'),
               ('Random forests', 2, None, '___sec41'),
               ('Random Forest Algorithm', 2, None, '___sec42'),
-              ('Random Forests Compared with other Methods on the Cancer Data',
+              ('Bootstrap with Random Forests Instead of a Single Tree, own '
+               'Bagging',
                2,
                None,
                '___sec43'),
-              ('Compare  Bagging on Trees with Random Forests',
+              ('Random Forests Compared with other Methods on the Cancer Data',
                2,
                None,
                '___sec44'),
-              ('Bootstrap with Random Forests Instead of a Single Tree, own '
-               'Bagging',
+              ('Compare  Bagging on Trees with Random Forests',
                2,
                None,
                '___sec45'),
@@ -211,9 +211,9 @@ MathJax.Hub.Config({
       
  • Changing the Level of the Decision Tree
    • 
       
  • Random forests
    • 
       
  • Random Forest Algorithm
    • 
-      
  • Random Forests Compared with other Methods on the Cancer Data
    • 
-      
  • Compare  Bagging on Trees with Random Forests
    • 
-      
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • 
+      
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • 
+      
  • Random Forests Compared with other Methods on the Cancer Data
    • 
+      
  • Compare  Bagging on Trees with Random Forests
    • 
       
  • Boosting, a Bird'e Eye
    • 
       
  • Adaptive boosting: AdaBoost, Basic Algorithm
    • 
       
  • Basic Steps of AdaBoost
    • 
diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs048.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs048.html
index 633771aec..307d7d155 100644
--- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs048.html
+++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs048.html
@@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source
               ('Changing the Level of the Decision Tree', 2, None, '___sec40'),
               ('Random forests', 2, None, '___sec41'),
               ('Random Forest Algorithm', 2, None, '___sec42'),
-              ('Random Forests Compared with other Methods on the Cancer Data',
+              ('Bootstrap with Random Forests Instead of a Single Tree, own '
+               'Bagging',
                2,
                None,
                '___sec43'),
-              ('Compare  Bagging on Trees with Random Forests',
+              ('Random Forests Compared with other Methods on the Cancer Data',
                2,
                None,
                '___sec44'),
-              ('Bootstrap with Random Forests Instead of a Single Tree, own '
-               'Bagging',
+              ('Compare  Bagging on Trees with Random Forests',
                2,
                None,
                '___sec45'),
@@ -211,9 +211,9 @@ MathJax.Hub.Config({
       
  • Changing the Level of the Decision Tree
    • 
       
  • Random forests
    • 
       
  • Random Forest Algorithm
    • 
-      
  • Random Forests Compared with other Methods on the Cancer Data
    • 
-      
  • Compare  Bagging on Trees with Random Forests
    • 
-      
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • 
+      
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • 
+      
  • Random Forests Compared with other Methods on the Cancer Data
    • 
+      
  • Compare  Bagging on Trees with Random Forests
    • 
       
  • Boosting, a Bird'e Eye
    • 
       
  • Adaptive boosting: AdaBoost, Basic Algorithm
    • 
       
  • Basic Steps of AdaBoost
    • 
diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs049.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs049.html
index 0a56c63ac..720cd6745 100644
--- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs049.html
+++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs049.html
@@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source
               ('Changing the Level of the Decision Tree', 2, None, '___sec40'),
               ('Random forests', 2, None, '___sec41'),
               ('Random Forest Algorithm', 2, None, '___sec42'),
-              ('Random Forests Compared with other Methods on the Cancer Data',
+              ('Bootstrap with Random Forests Instead of a Single Tree, own '
+               'Bagging',
                2,
                None,
                '___sec43'),
-              ('Compare  Bagging on Trees with Random Forests',
+              ('Random Forests Compared with other Methods on the Cancer Data',
                2,
                None,
                '___sec44'),
-              ('Bootstrap with Random Forests Instead of a Single Tree, own '
-               'Bagging',
+              ('Compare  Bagging on Trees with Random Forests',
                2,
                None,
                '___sec45'),
@@ -211,9 +211,9 @@ MathJax.Hub.Config({
       
  • Changing the Level of the Decision Tree
    • 
       
  • Random forests
    • 
       
  • Random Forest Algorithm
    • 
-      
  • Random Forests Compared with other Methods on the Cancer Data
    • 
-      
  • Compare  Bagging on Trees with Random Forests
    • 
-      
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • 
+      
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • 
+      
  • Random Forests Compared with other Methods on the Cancer Data
    • 
+      
  • Compare  Bagging on Trees with Random Forests
    • 
       
  • Boosting, a Bird'e Eye
    • 
       
  • Adaptive boosting: AdaBoost, Basic Algorithm
    • 
       
  • Basic Steps of AdaBoost
    • 
diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs050.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs050.html
index c12d3cb33..b0f46a328 100644
--- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs050.html
+++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs050.html
@@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source
               ('Changing the Level of the Decision Tree', 2, None, '___sec40'),
               ('Random forests', 2, None, '___sec41'),
               ('Random Forest Algorithm', 2, None, '___sec42'),
-              ('Random Forests Compared with other Methods on the Cancer Data',
+              ('Bootstrap with Random Forests Instead of a Single Tree, own '
+               'Bagging',
                2,
                None,
                '___sec43'),
-              ('Compare  Bagging on Trees with Random Forests',
+              ('Random Forests Compared with other Methods on the Cancer Data',
                2,
                None,
                '___sec44'),
-              ('Bootstrap with Random Forests Instead of a Single Tree, own '
-               'Bagging',
+              ('Compare  Bagging on Trees with Random Forests',
                2,
                None,
                '___sec45'),
@@ -211,9 +211,9 @@ MathJax.Hub.Config({
       
  • Changing the Level of the Decision Tree
    • 
       
  • Random forests
    • 
       
  • Random Forest Algorithm
    • 
-      
  • Random Forests Compared with other Methods on the Cancer Data
    • 
-      
  • Compare  Bagging on Trees with Random Forests
    • 
-      
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • 
+      
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • 
+      
  • Random Forests Compared with other Methods on the Cancer Data
    • 
+      
  • Compare  Bagging on Trees with Random Forests
    • 
       
  • Boosting, a Bird'e Eye
    • 
       
  • Adaptive boosting: AdaBoost, Basic Algorithm
    • 
       
  • Basic Steps of AdaBoost
    • 
diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs051.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs051.html
index cbbe511d6..89e47b91f 100644
--- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs051.html
+++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs051.html
@@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source
               ('Changing the Level of the Decision Tree', 2, None, '___sec40'),
               ('Random forests', 2, None, '___sec41'),
               ('Random Forest Algorithm', 2, None, '___sec42'),
-              ('Random Forests Compared with other Methods on the Cancer Data',
+              ('Bootstrap with Random Forests Instead of a Single Tree, own '
+               'Bagging',
                2,
                None,
                '___sec43'),
-              ('Compare  Bagging on Trees with Random Forests',
+              ('Random Forests Compared with other Methods on the Cancer Data',
                2,
                None,
                '___sec44'),
-              ('Bootstrap with Random Forests Instead of a Single Tree, own '
-               'Bagging',
+              ('Compare  Bagging on Trees with Random Forests',
                2,
                None,
                '___sec45'),
@@ -211,9 +211,9 @@ MathJax.Hub.Config({
       
  • Changing the Level of the Decision Tree
    • 
       
  • Random forests
    • 
       
  • Random Forest Algorithm
    • 
-      
  • Random Forests Compared with other Methods on the Cancer Data
    • 
-      
  • Compare  Bagging on Trees with Random Forests
    • 
-      
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • 
+      
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • 
+      
  • Random Forests Compared with other Methods on the Cancer Data
    • 
+      
  • Compare  Bagging on Trees with Random Forests
    • 
       
  • Boosting, a Bird'e Eye
    • 
       
  • Adaptive boosting: AdaBoost, Basic Algorithm
    • 
       
  • Basic Steps of AdaBoost
    • 
diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs052.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs052.html
index 5b430bec9..61212693a 100644
--- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs052.html
+++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs052.html
@@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source
               ('Changing the Level of the Decision Tree', 2, None, '___sec40'),
               ('Random forests', 2, None, '___sec41'),
               ('Random Forest Algorithm', 2, None, '___sec42'),
-              ('Random Forests Compared with other Methods on the Cancer Data',
+              ('Bootstrap with Random Forests Instead of a Single Tree, own '
+               'Bagging',
                2,
                None,
                '___sec43'),
-              ('Compare  Bagging on Trees with Random Forests',
+              ('Random Forests Compared with other Methods on the Cancer Data',
                2,
                None,
                '___sec44'),
-              ('Bootstrap with Random Forests Instead of a Single Tree, own '
-               'Bagging',
+              ('Compare  Bagging on Trees with Random Forests',
                2,
                None,
                '___sec45'),
@@ -211,9 +211,9 @@ MathJax.Hub.Config({
       
  • Changing the Level of the Decision Tree
    • 
       
  • Random forests
    • 
       
  • Random Forest Algorithm
    • 
-      
  • Random Forests Compared with other Methods on the Cancer Data
    • 
-      
  • Compare  Bagging on Trees with Random Forests
    • 
-      
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • 
+      
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • 
+      
  • Random Forests Compared with other Methods on the Cancer Data
    • 
+      
  • Compare  Bagging on Trees with Random Forests
    • 
       
  • Boosting, a Bird'e Eye
    • 
       
  • Adaptive boosting: AdaBoost, Basic Algorithm
    • 
       
  • Basic Steps of AdaBoost
    • 
diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs053.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs053.html
index 6dabddbf2..255a7b826 100644
--- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs053.html
+++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs053.html
@@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source
               ('Changing the Level of the Decision Tree', 2, None, '___sec40'),
               ('Random forests', 2, None, '___sec41'),
               ('Random Forest Algorithm', 2, None, '___sec42'),
-              ('Random Forests Compared with other Methods on the Cancer Data',
+              ('Bootstrap with Random Forests Instead of a Single Tree, own '
+               'Bagging',
                2,
                None,
                '___sec43'),
-              ('Compare  Bagging on Trees with Random Forests',
+              ('Random Forests Compared with other Methods on the Cancer Data',
                2,
                None,
                '___sec44'),
-              ('Bootstrap with Random Forests Instead of a Single Tree, own '
-               'Bagging',
+              ('Compare  Bagging on Trees with Random Forests',
                2,
                None,
                '___sec45'),
@@ -211,9 +211,9 @@ MathJax.Hub.Config({
       
  • Changing the Level of the Decision Tree
    • 
       
  • Random forests
    • 
       
  • Random Forest Algorithm
    • 
-      
  • Random Forests Compared with other Methods on the Cancer Data
    • 
-      
  • Compare  Bagging on Trees with Random Forests
    • 
-      
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • 
+      
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • 
+      
  • Random Forests Compared with other Methods on the Cancer Data
    • 
+      
  • Compare  Bagging on Trees with Random Forests
    • 
       
  • Boosting, a Bird'e Eye
    • 
       
  • Adaptive boosting: AdaBoost, Basic Algorithm
    • 
       
  • Basic Steps of AdaBoost
    • 
diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs054.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs054.html
index 7f4292f42..0178c6e44 100644
--- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs054.html
+++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs054.html
@@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source
               ('Changing the Level of the Decision Tree', 2, None, '___sec40'),
               ('Random forests', 2, None, '___sec41'),
               ('Random Forest Algorithm', 2, None, '___sec42'),
-              ('Random Forests Compared with other Methods on the Cancer Data',
+              ('Bootstrap with Random Forests Instead of a Single Tree, own '
+               'Bagging',
                2,
                None,
                '___sec43'),
-              ('Compare  Bagging on Trees with Random Forests',
+              ('Random Forests Compared with other Methods on the Cancer Data',
                2,
                None,
                '___sec44'),
-              ('Bootstrap with Random Forests Instead of a Single Tree, own '
-               'Bagging',
+              ('Compare  Bagging on Trees with Random Forests',
                2,
                None,
                '___sec45'),
@@ -211,9 +211,9 @@ MathJax.Hub.Config({
       
  • Changing the Level of the Decision Tree
    • 
       
  • Random forests
    • 
       
  • Random Forest Algorithm
    • 
-      
  • Random Forests Compared with other Methods on the Cancer Data
    • 
-      
  • Compare  Bagging on Trees with Random Forests
    • 
-      
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • 
+      
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • 
+      
  • Random Forests Compared with other Methods on the Cancer Data
    • 
+      
  • Compare  Bagging on Trees with Random Forests
    • 
       
  • Boosting, a Bird'e Eye
    • 
       
  • Adaptive boosting: AdaBoost, Basic Algorithm
    • 
       
  • Basic Steps of AdaBoost
    • 
diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs055.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs055.html
index e5de741a6..00eaee4c4 100644
--- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs055.html
+++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs055.html
@@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source
               ('Changing the Level of the Decision Tree', 2, None, '___sec40'),
               ('Random forests', 2, None, '___sec41'),
               ('Random Forest Algorithm', 2, None, '___sec42'),
-              ('Random Forests Compared with other Methods on the Cancer Data',
+              ('Bootstrap with Random Forests Instead of a Single Tree, own '
+               'Bagging',
                2,
                None,
                '___sec43'),
-              ('Compare  Bagging on Trees with Random Forests',
+              ('Random Forests Compared with other Methods on the Cancer Data',
                2,
                None,
                '___sec44'),
-              ('Bootstrap with Random Forests Instead of a Single Tree, own '
-               'Bagging',
+              ('Compare  Bagging on Trees with Random Forests',
                2,
                None,
                '___sec45'),
@@ -211,9 +211,9 @@ MathJax.Hub.Config({
       
  • Changing the Level of the Decision Tree
    • 
       
  • Random forests
    • 
       
  • Random Forest Algorithm
    • 
-      
  • Random Forests Compared with other Methods on the Cancer Data
    • 
-      
  • Compare  Bagging on Trees with Random Forests
    • 
-      
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • 
+      
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • 
+      
  • Random Forests Compared with other Methods on the Cancer Data
    • 
+      
  • Compare  Bagging on Trees with Random Forests
    • 
       
  • Boosting, a Bird'e Eye
    • 
       
  • Adaptive boosting: AdaBoost, Basic Algorithm
    • 
       
  • Basic Steps of AdaBoost
    • 
diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs056.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs056.html
index 93ffabf9c..7d8c8d74c 100644
--- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs056.html
+++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs056.html
@@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source
               ('Changing the Level of the Decision Tree', 2, None, '___sec40'),
               ('Random forests', 2, None, '___sec41'),
               ('Random Forest Algorithm', 2, None, '___sec42'),
-              ('Random Forests Compared with other Methods on the Cancer Data',
+              ('Bootstrap with Random Forests Instead of a Single Tree, own '
+               'Bagging',
                2,
                None,
                '___sec43'),
-              ('Compare  Bagging on Trees with Random Forests',
+              ('Random Forests Compared with other Methods on the Cancer Data',
                2,
                None,
                '___sec44'),
-              ('Bootstrap with Random Forests Instead of a Single Tree, own '
-               'Bagging',
+              ('Compare  Bagging on Trees with Random Forests',
                2,
                None,
                '___sec45'),
@@ -211,9 +211,9 @@ MathJax.Hub.Config({
       
  • Changing the Level of the Decision Tree
    • 
       
  • Random forests
    • 
       
  • Random Forest Algorithm
    • 
-      
  • Random Forests Compared with other Methods on the Cancer Data
    • 
-      
  • Compare  Bagging on Trees with Random Forests
    • 
-      
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • 
+      
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • 
+      
  • Random Forests Compared with other Methods on the Cancer Data
    • 
+      
  • Compare  Bagging on Trees with Random Forests
    • 
       
  • Boosting, a Bird'e Eye
    • 
       
  • Adaptive boosting: AdaBoost, Basic Algorithm
    • 
       
  • Basic Steps of AdaBoost
    • 
diff --git a/doc/pub/DecisionTrees/html/._DecisionTrees-bs057.html b/doc/pub/DecisionTrees/html/._DecisionTrees-bs057.html
index 08bf5350e..6a1d2bd8b 100644
--- a/doc/pub/DecisionTrees/html/._DecisionTrees-bs057.html
+++ b/doc/pub/DecisionTrees/html/._DecisionTrees-bs057.html
@@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source
               ('Changing the Level of the Decision Tree', 2, None, '___sec40'),
               ('Random forests', 2, None, '___sec41'),
               ('Random Forest Algorithm', 2, None, '___sec42'),
-              ('Random Forests Compared with other Methods on the Cancer Data',
+              ('Bootstrap with Random Forests Instead of a Single Tree, own '
+               'Bagging',
                2,
                None,
                '___sec43'),
-              ('Compare  Bagging on Trees with Random Forests',
+              ('Random Forests Compared with other Methods on the Cancer Data',
                2,
                None,
                '___sec44'),
-              ('Bootstrap with Random Forests Instead of a Single Tree, own '
-               'Bagging',
+              ('Compare  Bagging on Trees with Random Forests',
                2,
                None,
                '___sec45'),
@@ -211,9 +211,9 @@ MathJax.Hub.Config({
       
  • Changing the Level of the Decision Tree
    • 
       
  • Random forests
    • 
       
  • Random Forest Algorithm
    • 
-      
  • Random Forests Compared with other Methods on the Cancer Data
    • 
-      
  • Compare  Bagging on Trees with Random Forests
    • 
-      
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • 
+      
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • 
+      
  • Random Forests Compared with other Methods on the Cancer Data
    • 
+      
  • Compare  Bagging on Trees with Random Forests
    • 
       
  • Boosting, a Bird'e Eye
    • 
       
  • Adaptive boosting: AdaBoost, Basic Algorithm
    • 
       
  • Basic Steps of AdaBoost
    • 
diff --git a/doc/pub/DecisionTrees/html/DecisionTrees-bs.html b/doc/pub/DecisionTrees/html/DecisionTrees-bs.html
index cfc8c9a63..263295b72 100644
--- a/doc/pub/DecisionTrees/html/DecisionTrees-bs.html
+++ b/doc/pub/DecisionTrees/html/DecisionTrees-bs.html
@@ -104,16 +104,16 @@ Automatically generated HTML file from DocOnce source
               ('Changing the Level of the Decision Tree', 2, None, '___sec40'),
               ('Random forests', 2, None, '___sec41'),
               ('Random Forest Algorithm', 2, None, '___sec42'),
-              ('Random Forests Compared with other Methods on the Cancer Data',
+              ('Bootstrap with Random Forests Instead of a Single Tree, own '
+               'Bagging',
                2,
                None,
                '___sec43'),
-              ('Compare  Bagging on Trees with Random Forests',
+              ('Random Forests Compared with other Methods on the Cancer Data',
                2,
                None,
                '___sec44'),
-              ('Bootstrap with Random Forests Instead of a Single Tree, own '
-               'Bagging',
+              ('Compare  Bagging on Trees with Random Forests',
                2,
                None,
                '___sec45'),
@@ -211,9 +211,9 @@ MathJax.Hub.Config({
       
  • Changing the Level of the Decision Tree
    • 
       
  • Random forests
    • 
       
  • Random Forest Algorithm
    • 
-      
  • Random Forests Compared with other Methods on the Cancer Data
    • 
-      
  • Compare  Bagging on Trees with Random Forests
    • 
-      
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • 
+      
  • Bootstrap with Random Forests Instead of a Single Tree, own Bagging
    • 
+      
  • Random Forests Compared with other Methods on the Cancer Data
    • 
+      
  • Compare  Bagging on Trees with Random Forests
    • 
       
  • Boosting, a Bird'e Eye
    • 
       
  • Adaptive boosting: AdaBoost, Basic Algorithm
    • 
       
  • Basic Steps of AdaBoost
    • 
diff --git a/doc/pub/DecisionTrees/html/DecisionTrees-reveal.html b/doc/pub/DecisionTrees/html/DecisionTrees-reveal.html
index debbea6c2..0d1365a1e 100644
--- a/doc/pub/DecisionTrees/html/DecisionTrees-reveal.html
+++ b/doc/pub/DecisionTrees/html/DecisionTrees-reveal.html
@@ -1914,7 +1914,68 @@ We will grow of forest of say \( M \) trees.
 
 
 
    
-
    Random Forests Compared with other Methods on the Cancer Data 
    
+
    Bootstrap with Random Forests Instead of a Single Tree, own Bagging 
    
+
+
    
+
+
+
    import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import train_test_split
+from sklearn.pipeline import make_pipeline
+from sklearn.utils import resample
+from sklearn.ensemble import RandomForestRegressor
+
+np.random.seed(2018)
+
+n = 100
+n_boostraps = 100
+maxdegree = 14
+
+# Make data set.
+x = np.linspace(-3, 3, n).reshape(-1, 1)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
+error = np.zeros(maxdegree)
+bias = np.zeros(maxdegree)
+variance = np.zeros(maxdegree)
+polydegree = np.zeros(maxdegree)
+X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
+
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+for degree in range(maxdegree):
+    model = RandomForestRegressor()
+    y_pred = np.empty((y_test.shape[0], n_boostraps))
+    for i in range(n_boostraps):
+        x_, y_ = resample(X_train_scaled, y_train)
+        model.fit(x_, y_.ravel())
+        y_pred[:, i] = model.predict(X_test_scaled).ravel()
+
+    polydegree[degree] = degree
+    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )
+    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )
+    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )
+    print('Polynomial degree:', degree)
+    print('Error:', error[degree])
+    print('Bias^2:', bias[degree])
+    print('Var:', variance[degree])
+    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
+
+plt.plot(polydegree, error, label='Error')
+plt.plot(polydegree, bias, label='bias')
+plt.plot(polydegree, variance, label='Variance')
+plt.legend()
+plt.show()
+
    
+
    
+
+
+
    
+
    Random Forests Compared with other Methods on the Cancer Data 
    
 
    
 
 
@@ -1988,7 +2049,7 @@ plt.show()
 
 
 
    
-
    Compare  Bagging on Trees with Random Forests  
    
+
    Compare  Bagging on Trees with Random Forests  
    
 
    
 
 
@@ -2010,67 +2071,6 @@ np.sum(y_pred == y_pred_rf) / len(y_pred)
 
    
 
 
-
    
-
    Bootstrap with Random Forests Instead of a Single Tree, own Bagging 
    
-
-
    
-
-
-
    import matplotlib.pyplot as plt
-import numpy as np
-from sklearn.model_selection import train_test_split
-from sklearn.pipeline import make_pipeline
-from sklearn.utils import resample
-from sklearn.ensemble import RandomForestRegressor
-
-np.random.seed(2018)
-
-n = 100
-n_boostraps = 100
-maxdegree = 14
-
-# Make data set.
-x = np.linspace(-3, 3, n).reshape(-1, 1)
-y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
-error = np.zeros(maxdegree)
-bias = np.zeros(maxdegree)
-variance = np.zeros(maxdegree)
-polydegree = np.zeros(maxdegree)
-X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
-
-from sklearn.preprocessing import StandardScaler
-scaler = StandardScaler()
-scaler.fit(X_train)
-X_train_scaled = scaler.transform(X_train)
-X_test_scaled = scaler.transform(X_test)
-
-for degree in range(maxdegree):
-    model = RandomForestRegressor()
-    y_pred = np.empty((y_test.shape[0], n_boostraps))
-    for i in range(n_boostraps):
-        x_, y_ = resample(X_train_scaled, y_train)
-        model.fit(x_, y_.ravel())
-        y_pred[:, i] = model.predict(X_test_scaled).ravel()
-
-    polydegree[degree] = degree
-    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )
-    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )
-    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )
-    print('Polynomial degree:', degree)
-    print('Error:', error[degree])
-    print('Bias^2:', bias[degree])
-    print('Var:', variance[degree])
-    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
-
-plt.plot(polydegree, error, label='Error')
-plt.plot(polydegree, bias, label='bias')
-plt.plot(polydegree, variance, label='Variance')
-plt.legend()
-plt.show()
-
    
-
    
-
-
 
    
 
    Boosting, a Bird'e Eye 
    
 
diff --git a/doc/pub/DecisionTrees/html/DecisionTrees-solarized.html b/doc/pub/DecisionTrees/html/DecisionTrees-solarized.html
index b220f97f4..1d6e730c8 100644
--- a/doc/pub/DecisionTrees/html/DecisionTrees-solarized.html
+++ b/doc/pub/DecisionTrees/html/DecisionTrees-solarized.html
@@ -124,16 +124,16 @@ div { text-align: justify; text-justify: inter-word; }
               ('Changing the Level of the Decision Tree', 2, None, '___sec40'),
               ('Random forests', 2, None, '___sec41'),
               ('Random Forest Algorithm', 2, None, '___sec42'),
-              ('Random Forests Compared with other Methods on the Cancer Data',
+              ('Bootstrap with Random Forests Instead of a Single Tree, own '
+               'Bagging',
                2,
                None,
                '___sec43'),
-              ('Compare  Bagging on Trees with Random Forests',
+              ('Random Forests Compared with other Methods on the Cancer Data',
                2,
                None,
                '___sec44'),
-              ('Bootstrap with Random Forests Instead of a Single Tree, own '
-               'Bagging',
+              ('Compare  Bagging on Trees with Random Forests',
                2,
                None,
                '___sec45'),
@@ -1904,7 +1904,67 @@ We will grow of forest of say \( M \) trees.
 
 









    
 
-
    Random Forests Compared with other Methods on the Cancer Data 
    
+
    Bootstrap with Random Forests Instead of a Single Tree, own Bagging 
    
+
+
    
+
+
+
    import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import train_test_split
+from sklearn.pipeline import make_pipeline
+from sklearn.utils import resample
+from sklearn.ensemble import RandomForestRegressor
+
+np.random.seed(2018)
+
+n = 100
+n_boostraps = 100
+maxdegree = 14
+
+# Make data set.
+x = np.linspace(-3, 3, n).reshape(-1, 1)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
+error = np.zeros(maxdegree)
+bias = np.zeros(maxdegree)
+variance = np.zeros(maxdegree)
+polydegree = np.zeros(maxdegree)
+X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
+
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+for degree in range(maxdegree):
+    model = RandomForestRegressor()
+    y_pred = np.empty((y_test.shape[0], n_boostraps))
+    for i in range(n_boostraps):
+        x_, y_ = resample(X_train_scaled, y_train)
+        model.fit(x_, y_.ravel())
+        y_pred[:, i] = model.predict(X_test_scaled).ravel()
+
+    polydegree[degree] = degree
+    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )
+    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )
+    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )
+    print('Polynomial degree:', degree)
+    print('Error:', error[degree])
+    print('Bias^2:', bias[degree])
+    print('Var:', variance[degree])
+    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
+
+plt.plot(polydegree, error, label='Error')
+plt.plot(polydegree, bias, label='bias')
+plt.plot(polydegree, variance, label='Variance')
+plt.legend()
+plt.show()
+
    
+
    
+









    
+
+
    Random Forests Compared with other Methods on the Cancer Data 
    
 
    
 
 
@@ -1977,7 +2037,7 @@ plt.show()
 
    
 









    
 
-
    Compare  Bagging on Trees with Random Forests  
    
+
    Compare  Bagging on Trees with Random Forests  
    
 
    
 
 
@@ -1999,66 +2059,6 @@ np.sum(y_pred == y_pred_rf) / len(y_pred)
 
    
 









    
 
-
    Bootstrap with Random Forests Instead of a Single Tree, own Bagging 
    
-
-
    
-
-
-
    import matplotlib.pyplot as plt
-import numpy as np
-from sklearn.model_selection import train_test_split
-from sklearn.pipeline import make_pipeline
-from sklearn.utils import resample
-from sklearn.ensemble import RandomForestRegressor
-
-np.random.seed(2018)
-
-n = 100
-n_boostraps = 100
-maxdegree = 14
-
-# Make data set.
-x = np.linspace(-3, 3, n).reshape(-1, 1)
-y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
-error = np.zeros(maxdegree)
-bias = np.zeros(maxdegree)
-variance = np.zeros(maxdegree)
-polydegree = np.zeros(maxdegree)
-X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
-
-from sklearn.preprocessing import StandardScaler
-scaler = StandardScaler()
-scaler.fit(X_train)
-X_train_scaled = scaler.transform(X_train)
-X_test_scaled = scaler.transform(X_test)
-
-for degree in range(maxdegree):
-    model = RandomForestRegressor()
-    y_pred = np.empty((y_test.shape[0], n_boostraps))
-    for i in range(n_boostraps):
-        x_, y_ = resample(X_train_scaled, y_train)
-        model.fit(x_, y_.ravel())
-        y_pred[:, i] = model.predict(X_test_scaled).ravel()
-
-    polydegree[degree] = degree
-    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )
-    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )
-    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )
-    print('Polynomial degree:', degree)
-    print('Error:', error[degree])
-    print('Bias^2:', bias[degree])
-    print('Var:', variance[degree])
-    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
-
-plt.plot(polydegree, error, label='Error')
-plt.plot(polydegree, bias, label='bias')
-plt.plot(polydegree, variance, label='Variance')
-plt.legend()
-plt.show()
-
    
-
    
-









    
-
 
    Boosting, a Bird'e Eye 
    
 
 
    
diff --git a/doc/pub/DecisionTrees/html/DecisionTrees.html b/doc/pub/DecisionTrees/html/DecisionTrees.html
index 98e5b068e..e66d392b2 100644
--- a/doc/pub/DecisionTrees/html/DecisionTrees.html
+++ b/doc/pub/DecisionTrees/html/DecisionTrees.html
@@ -129,16 +129,16 @@ div { text-align: justify; text-justify: inter-word; }
               ('Changing the Level of the Decision Tree', 2, None, '___sec40'),
               ('Random forests', 2, None, '___sec41'),
               ('Random Forest Algorithm', 2, None, '___sec42'),
-              ('Random Forests Compared with other Methods on the Cancer Data',
+              ('Bootstrap with Random Forests Instead of a Single Tree, own '
+               'Bagging',
                2,
                None,
                '___sec43'),
-              ('Compare  Bagging on Trees with Random Forests',
+              ('Random Forests Compared with other Methods on the Cancer Data',
                2,
                None,
                '___sec44'),
-              ('Bootstrap with Random Forests Instead of a Single Tree, own '
-               'Bagging',
+              ('Compare  Bagging on Trees with Random Forests',
                2,
                None,
                '___sec45'),
@@ -1909,7 +1909,67 @@ We will grow of forest of say \( M \) trees.
 
 









    
 
-
    Random Forests Compared with other Methods on the Cancer Data 
    
+
    Bootstrap with Random Forests Instead of a Single Tree, own Bagging 
    
+
+
    
+
+
+
    import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import train_test_split
+from sklearn.pipeline import make_pipeline
+from sklearn.utils import resample
+from sklearn.ensemble import RandomForestRegressor
+
+np.random.seed(2018)
+
+n = 100
+n_boostraps = 100
+maxdegree = 14
+
+# Make data set.
+x = np.linspace(-3, 3, n).reshape(-1, 1)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
+error = np.zeros(maxdegree)
+bias = np.zeros(maxdegree)
+variance = np.zeros(maxdegree)
+polydegree = np.zeros(maxdegree)
+X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
+
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+for degree in range(maxdegree):
+    model = RandomForestRegressor()
+    y_pred = np.empty((y_test.shape[0], n_boostraps))
+    for i in range(n_boostraps):
+        x_, y_ = resample(X_train_scaled, y_train)
+        model.fit(x_, y_.ravel())
+        y_pred[:, i] = model.predict(X_test_scaled).ravel()
+
+    polydegree[degree] = degree
+    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )
+    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )
+    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )
+    print('Polynomial degree:', degree)
+    print('Error:', error[degree])
+    print('Bias^2:', bias[degree])
+    print('Var:', variance[degree])
+    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
+
+plt.plot(polydegree, error, label='Error')
+plt.plot(polydegree, bias, label='bias')
+plt.plot(polydegree, variance, label='Variance')
+plt.legend()
+plt.show()
+
    
+
    
+









    
+
+
    Random Forests Compared with other Methods on the Cancer Data 
    
 
    
 
 
@@ -1982,7 +2042,7 @@ plt.show()
 
    
 









    
 
-
    Compare  Bagging on Trees with Random Forests  
    
+
    Compare  Bagging on Trees with Random Forests  
    
 
    
 
 
@@ -2004,66 +2064,6 @@ np.sum(y_pred =
 
    
 









    
 
-
    Bootstrap with Random Forests Instead of a Single Tree, own Bagging 
    
-
-
    
-
-
-
    import matplotlib.pyplot as plt
-import numpy as np
-from sklearn.model_selection import train_test_split
-from sklearn.pipeline import make_pipeline
-from sklearn.utils import resample
-from sklearn.ensemble import RandomForestRegressor
-
-np.random.seed(2018)
-
-n = 100
-n_boostraps = 100
-maxdegree = 14
-
-# Make data set.
-x = np.linspace(-3, 3, n).reshape(-1, 1)
-y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
-error = np.zeros(maxdegree)
-bias = np.zeros(maxdegree)
-variance = np.zeros(maxdegree)
-polydegree = np.zeros(maxdegree)
-X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
-
-from sklearn.preprocessing import StandardScaler
-scaler = StandardScaler()
-scaler.fit(X_train)
-X_train_scaled = scaler.transform(X_train)
-X_test_scaled = scaler.transform(X_test)
-
-for degree in range(maxdegree):
-    model = RandomForestRegressor()
-    y_pred = np.empty((y_test.shape[0], n_boostraps))
-    for i in range(n_boostraps):
-        x_, y_ = resample(X_train_scaled, y_train)
-        model.fit(x_, y_.ravel())
-        y_pred[:, i] = model.predict(X_test_scaled).ravel()
-
-    polydegree[degree] = degree
-    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )
-    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )
-    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )
-    print('Polynomial degree:', degree)
-    print('Error:', error[degree])
-    print('Bias^2:', bias[degree])
-    print('Var:', variance[degree])
-    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
-
-plt.plot(polydegree, error, label='Error')
-plt.plot(polydegree, bias, label='bias')
-plt.plot(polydegree, variance, label='Variance')
-plt.legend()
-plt.show()
-
    
-
    
-









    
-
 
    Boosting, a Bird'e Eye 
    
 
 
    
diff --git a/doc/pub/DecisionTrees/ipynb/DataFiles/cancer.dot b/doc/pub/DecisionTrees/ipynb/DataFiles/cancer.dot
index a8bce3417..4d258be18 100644
--- a/doc/pub/DecisionTrees/ipynb/DataFiles/cancer.dot
+++ b/doc/pub/DecisionTrees/ipynb/DataFiles/cancer.dot
@@ -6,15 +6,15 @@ edge [fontname=helvetica] ;
 0 -> 1 [labeldistance=2.5, labelangle=45, headlabel="True"] ;
 2 [label="worst concave points <= 0.135\ngini = 0.031\nsamples = 253\nvalue = [[249, 4]\n[4, 249]]", fillcolor="#e58139ee"] ;
 1 -> 2 ;
-3 [label="area error <= 48.975\ngini = 0.008\nsamples = 242\nvalue = [[241, 1]\n[1, 241]]", fillcolor="#e58139fb"] ;
+3 [label="radius error <= 0.643\ngini = 0.008\nsamples = 242\nvalue = [[241, 1]\n[1, 241]]", fillcolor="#e58139fb"] ;
 2 -> 3 ;
 4 [label="gini = 0.0\nsamples = 239\nvalue = [[239, 0]\n[0, 239]]", fillcolor="#e58139ff"] ;
 3 -> 4 ;
-5 [label="mean area <= 469.25\ngini = 0.444\nsamples = 3\nvalue = [[2, 1]\n[1, 2]]", fillcolor="#e5813913"] ;
+5 [label="texture error <= 1.938\ngini = 0.444\nsamples = 3\nvalue = [[2, 1]\n[1, 2]]", fillcolor="#e5813913"] ;
 3 -> 5 ;
-6 [label="gini = 0.0\nsamples = 1\nvalue = [[0, 1]\n[1, 0]]", fillcolor="#e58139ff"] ;
+6 [label="gini = 0.0\nsamples = 2\nvalue = [[2, 0]\n[0, 2]]", fillcolor="#e58139ff"] ;
 5 -> 6 ;
-7 [label="gini = 0.0\nsamples = 2\nvalue = [[2, 0]\n[0, 2]]", fillcolor="#e58139ff"] ;
+7 [label="gini = 0.0\nsamples = 1\nvalue = [[0, 1]\n[1, 0]]", fillcolor="#e58139ff"] ;
 5 -> 7 ;
 8 [label="mean texture <= 20.84\ngini = 0.397\nsamples = 11\nvalue = [[8, 3]\n[3, 8]]", fillcolor="#e581392c"] ;
 2 -> 8 ;
@@ -22,7 +22,7 @@ edge [fontname=helvetica] ;
 8 -> 9 ;
 10 [label="gini = 0.0\nsamples = 3\nvalue = [[0, 3]\n[3, 0]]", fillcolor="#e58139ff"] ;
 8 -> 10 ;
-11 [label="mean texture <= 16.22\ngini = 0.278\nsamples = 6\nvalue = [[1, 5]\n[5, 1]]", fillcolor="#e581396b"] ;
+11 [label="area error <= 13.475\ngini = 0.278\nsamples = 6\nvalue = [[1, 5]\n[5, 1]]", fillcolor="#e581396b"] ;
 1 -> 11 ;
 12 [label="gini = 0.0\nsamples = 1\nvalue = [[1, 0]\n[0, 1]]", fillcolor="#e58139ff"] ;
 11 -> 12 ;
@@ -30,11 +30,11 @@ edge [fontname=helvetica] ;
 11 -> 13 ;
 14 [label="worst texture <= 20.645\ngini = 0.202\nsamples = 167\nvalue = [[19, 148]\n[148, 19]]", fillcolor="#e5813994"] ;
 0 -> 14 [labeldistance=2.5, labelangle=-45, headlabel="False"] ;
-15 [label="worst area <= 964.4\ngini = 0.375\nsamples = 16\nvalue = [[12, 4]\n[4, 12]]", fillcolor="#e5813938"] ;
+15 [label="worst concavity <= 0.318\ngini = 0.375\nsamples = 16\nvalue = [[12, 4]\n[4, 12]]", fillcolor="#e5813938"] ;
 14 -> 15 ;
 16 [label="gini = 0.0\nsamples = 11\nvalue = [[11, 0]\n[0, 11]]", fillcolor="#e58139ff"] ;
 15 -> 16 ;
-17 [label="mean fractal dimension <= 0.054\ngini = 0.32\nsamples = 5\nvalue = [[1, 4]\n[4, 1]]", fillcolor="#e5813955"] ;
+17 [label="mean concavity <= 0.07\ngini = 0.32\nsamples = 5\nvalue = [[1, 4]\n[4, 1]]", fillcolor="#e5813955"] ;
 15 -> 17 ;
 18 [label="gini = 0.0\nsamples = 1\nvalue = [[1, 0]\n[0, 1]]", fillcolor="#e58139ff"] ;
 17 -> 18 ;
@@ -42,7 +42,7 @@ edge [fontname=helvetica] ;
 17 -> 19 ;
 20 [label="mean concave points <= 0.049\ngini = 0.088\nsamples = 151\nvalue = [[7, 144]\n[144, 7]]", fillcolor="#e58139d0"] ;
 14 -> 20 ;
-21 [label="concave points error <= 0.01\ngini = 0.48\nsamples = 15\nvalue = [[6, 9]\n[9, 6]]", fillcolor="#e5813900"] ;
+21 [label="compactness error <= 0.016\ngini = 0.48\nsamples = 15\nvalue = [[6, 9]\n[9, 6]]", fillcolor="#e5813900"] ;
 20 -> 21 ;
 22 [label="gini = 0.0\nsamples = 9\nvalue = [[0, 9]\n[9, 0]]", fillcolor="#e58139ff"] ;
 21 -> 22 ;
diff --git a/doc/pub/DecisionTrees/ipynb/DataFiles/cancer.png b/doc/pub/DecisionTrees/ipynb/DataFiles/cancer.png
index 22ae432a03c841dfc34de7df2ada031d576d0157..98014cc9c4f87884d2ee399136c4e95bedb8df7c 100644
GIT binary patch
literal 221592
zcmagGby!qw_dN~>Qi2Ld3W#(LpdcZwba#h
z_w#*#=e^$R`u&06VVF7R-22{p?X}iEpXFu6urP=*5D*ZsB*b4SA|RkFARr+5qTK~Y
zO4vu!!8b&EMKNK7qCbyU!GB&Et4q9IM3JDvpo-JSbRNL1=OPrm!s
zBRl!GO-@c4M}U7FG$xGSp#|}N*TIc!V;(LZNlB8Rxr;9SwyAD9tFW#PK5bv^R&1gB
zX|ESUolX!DL*5$+FAV|ZfBcB1;^*&ki_wV0r8;kXzPu-7g=H>StH2g3Lup=9Lq1n%8
zPUCm^HR2qznWs+&LoY7~htYp|=FLxYr#ncG4Q{IpdBmkol3&8-^aO&}Nru^Uhu2r6
zYQ;Pzj}!%Zoaqk^w!^-4IV}$ssEL2!VvD4a3DmB&`#w6Vfn9q0`kVE9Y`ZD8i?QzC
zI;vjMT<^5aYVzlCcOo~g%~aECpR<#QmoDt?$2*9-bL|}a+V|_ausw8QT}xE%x5KAm
z;c$Krvj(}L1AC9{IIBj&;ohUpKIl1gE?&F;n9O**=4_)0K1^^4H_)E03Ei#J71R55
zzPj$pH$4vb)R63j3c%~)s$V>P9hlXK>>7~yGnB$Sn$C3cE0$)pb;wF@W`X^&|Ls+4
zN4nQRVqu$nbL@#Un&oI23Mub0k5It#B|fv}M>R+)
zrQ^m3w1l?dISLX%00D$dmP{aU5@$N$5#lyY3^#5!C
z!4ZOyN(%xS*?8|~IkIDxtQWnRV$K(5C}nSd;~bLbhD-=`$FuF8by6Dj#GtaUu+Zr>
zHH1uX+eBSomwdB;&&~zujOWd1BVwrgHs{Et{%z9XWr`OV&gdJMXmG*xIKAK86$kOc
z=IBZ<&czD
zdW}h=?p~WztmwYFqj{~>sbkc9JIT@Zv{=`WRXBqT4|9A9iHQGsDA-8E`O)-hh%T6T
z&rd8z>jrHnoZF_R$fWR@1^o{k?Q>u!31sUPYCiuN8$u$ys5$Pm(R%g?QzUyc+*Dcj*EijAe4YBE
z<^_7C+1XjMzKw)uCJN`(n&RP4`0<(J{r6Xs6>7~L&OBzk^J%{yEir3RNZ^Ql)2qJQ
z8AU4+W6rU&Gb4fKqq?Lm38g$&j%Ee>o+hLkz0=wx{
zMc!-_rwlQ-WfdN?B60$|i6VVmx9ZI8I81s7;_xjQhfVbLdVQ4f87+UCZm}$T;n-w;
zFwOdkYE6fI8R^8AC>$2UD&sgKo0HY#U;20`bid6*#S>`kI`}`a?uMm
zP100e8~z3U9N8r7ILlEL5=ON`zhI=ENIq1>ud+d&ckbR@)ii7mZP6p)ock41K&RO7
zk>}_5I7#7H#p|{As}bB5oc9Qy6Q&Ay9ah`W5Wr04=YNKU^RuYebfgkrY
zdbfy$KGCc&^<0U+Q0qX
z8ZLBacka`)-iwWKs2<<+=(D$Jifw_WLlx+WoMyo%nD}#>w<;g=0D8;82i7!05IdB#y+VN%pTJ>2BamWfdCwkI3=bL?|M%{_)gt=6=U3gWOf
ziH@FuI>29RzYIHp?KbjwcAY26)!5Fui=Od!FP|rQd2DQKxG!kJx#0%{-ANWzkn_`n
z1Bqz5hIdXQtm
zEnTK;Es9=Av@)6N-5YYNA1YDy3p|_Sl^k5_!;@9UQ9LG>DUpe*qf5j@Ct`8Q5?94*X!z-kgL&MQuDsg2CwP7q)_O2KYcDRb%@%e`L&fydON3}}D
zuhO5D^Az(Q4pm!E`o_f(r0{v&C8{hY*!<+PCiNZZWRPXPX~khb8~?AE-#wy%x9?#VBb~;grp*lB3CoJjM$+_&v2(1@oj%w$L=?cvXw#KhUjR{$k2`&QwWvvSmMtx
z==sy{3a4brd=5C8^19j7kT9W|VgxB=YyBmlLIx7#o2DEq
zvdVg;TJ9dqMsk-oq0cuFI$l&~gft&DVlEw`9B8^4rd-M>G9p^@v2M=jhtY(P5c%#Z
z4nvzr1!lkc?!u&b94w0!*E)jk&OgP?@H3K#XH9oC)1O%CiX-dzLXV3>Bc0|rT%@9e
zGH;yeT4}zoPGBb0N80HG5e+6>E2*FsI(fg^Crn~c-1sF=pGyYm+_-{(!kSi&
z`0%9qDa0g82mf|qc%+Z*GVXEMV1Ip3e-|mTOAF?Js!H{w&VHEX3%9-MA_hdYB0zJx
z!MFe;_SXr$QkL@Xj@C5S+VsVoqXFL|GsT=CrvxVdFv2`hp)RBJqwNXB%=v2T@vkx)
zt1rSW?B={k)QWYUZNx^bv75hCY7e%xlgvIOU{HUc8AyDPz+qSyo5cT6kxI^*iR$C%Fte^bOgUOy$N
zSQgBBXY98Ek;{BFAWLJi=rn!vM5I{$1)*5rt~dIF>|#JkIeLI@zj(~Jnf>8RSg64R
z+__&Mt6}+LKYod3viv@ult=@@y&KZC$y*&k^?5ajYzxyZ^YaZ@Qyn3H3htQ`s+EV^l?<0r(t=Jj*Mk^Wo9-
zOTWq%8B5;v)5Ngk7%mqvJPM6<(k7`)e$W<(-!?}rRqFa%vv51Ex55Q8n1~lUGLVK!
zO1_BN`aU`T4+mxkxtL!u%K52N9;ZTm!yT9bcSDl&#}7Bh&`#8S9A%LHus{5ERyNaK
z$r!G1S-#x;NlDWr=3;r~yNK!3H+sK&#HNS1f03ImbQ%;|;0!Y?KCq-0HW_NEO=~B=
z#NAKGDom%OlaQS#zxKqVCwN!t_=hr>LZU__&3f}WJjwZm3)S+~h4|)+_*`?Bs|@Bv
z556qauD(v`4~!`oILg%uOR33{A=83d>ztQByV5*U&#k;sZ}ql68W5C*KAwnmklboI
z!XuV&lv&i%yfK<3FZN%$*cAxoF=dPEJFaGY%2g=f>@)7jI+T!%T$Caxi0E@&Bjj7I
zS6=DM!(AJ&@qXGSl?(5Tq*lsYSZn}cE||3QC?tuqU$L#V4{0<;g>+s(qx`MVX59fr
zo;ePcxVM=SO~F2oqRG}qLO#Ty@-SOGDT-EB@U4|{bpvbon+XDq;cDZ&g
zs1;L4;S(GvHrF6_>OnI5^crEl`)%6T&hv1AGP03Ms!u@+!EqAS+2Vgs=GvrwiFxYG
zoL)3lL;8x48$~iZ_yQ^$&?1%>>&3^2J5eFEGY4loP3tS;Zz#+x0Q->P*_*aK95*#J
zHCApfHdbNgdsg80N2i`5+1~a2%3mQHKdI{^Cx}RbJA>vz;x^NubA9kZEHYKVf1HcloUvS4BlFE%;h=%cLjex8);rJ1OGiR0X8^*llHQc=0{qI-B%I
z!Ev6y_!fRA{hVg$f)Fp@MnBDST2Om%@yNv`l5x2D;r_+oZlhjPR}<~SLmmI|TPSac
zx%n|{=-mAlRq#X_6PcPLD0?dH!-o%&nTC1ihEUHY?*_bHrHiw(g1x|7e;)g!`5`QJ
z2X6%R`Z@#=%aIIRUX@Cl!YNlj6?E*d}?MqbwCO*%5w=jd)(ON*E;1@>tIl?Qr{kYUH{ON1EdnEiwp91k(7BQE*Wux=;
z>^JKK*xdeTM?L3*=rz5{&L6ChupwpjSi^hie*EChT`bkj_oma>%YRaTf?kpB{mzp{
zqRQaM+-03rm6jv0qfyNiitP{A9n@Ja!X20ZL&x5AV~>C$EF&n2CpBed!IaAHAq+*l
zQ);${PbCwHER*!S9V9wd*8%ZI+NsZtF<=XSLL9Be(_|arblSNOKzA4dTJIlI{ABQh
z)eo8+UM*3ya>~3)v~t2r-3gxY=G`^BrsH2L>x!PgB^bvpY6~>%8Y(<46vZmvoNn^?
zPg|PMm;G+%FNH{QLiyX4{(fA5yk7%?W$>c)GcJj4Sl`ci7x~-X1>qYCLYv}WzC((I==ciVfuqtTf)?Sd
z(dj&7@l6e
zqYlb|R4ko=Exk$^mEQ&bM9HSxYsI0JRX3yPT0A0|)K<37+xdQ*u*F}(NNY=h$vTz-
z_0wL*oTF>YJ*hIGuY(B{y6#+FK56un7u}ZwOpX8jGpk5mT>%!#6Jw;|JEw;kC7f#)
z7w4x8;%pr42rxT@wiB`!hU)k8V$AwYnpsjljLM~)p`2-p
z!rH__h#y7jibCL2g$d72l3A-93lmocO7CDS^c3^wC4U*9^Bk3pW{vno09ke6cRx;h
z;SUJ$JcUyNh_3K%WOk{?PwW;pA|VI_!6;J8f@uiuQPpnJ=9Y0?t<$MN=1-^Ya=djX
z>1M(N+k=T_Ha3EHacL=pe2)Wgs4;k)Hkaf!x2KtXb}2sGO|)tTHL~Y|knK&H=z%%l
z)7XV9@}u+qW~$i?4g*djAI_{5I7{&QjI@!^;
z5-S<7WOA@cOKdJZ4_2oN89~p9SRF;%n=iw=Fe0DRh9`Gr{=$#oE4
zamDyoT=Aj=jQ={0y&qwS2i`wX*NS)!xytVS_{{R(%4Qj0gqE;iD@l9Ye%EZltM8_wy>&jGwdZ~-sG*-
z(a+PH{DFt2{p8DGxkPGnLFwJ=Dh05pXH#=`!n1c<8${h-_&1mJUg8zgMSYd5OT4$*
zU(j5EoCDyuFTk43UGtwLtWV@OM7?FE>l`ov9VD{U?Wr_~?3|JCL04C|#`Sbn|vtogGojva?Mtk$5)UsrjgJqL>G`g5~Z|)odiZ4Bggzw0Pw<$l2
zvs~}s@a=IgZ%LJB?kn;f=5_Yf*5=RacV86?xF(ZxPfK*6k}J0T`Gf7Ux3pVb3;2Q)
zx8U1|n@(Q_>C3=iVhYXnr|bPI|D+pDI)UpU%y1AVyDLq3)Gok~05$1Cll(2bC-k}<
z`cKXNPB4RZGLy7>?{>j*(Wu~hTE1^62eS
zacdlw9|hWj&wZvoo#nY5ZoQ6PVY}K=BJ)5W%T>-M#-Wz*L;f6<2FgZ_=Xu@i@yQ;6
ze6f(oKOp6WC(;zu%Z7_gRds>jAuRdrtw<=*@d+pQGpvuj)iz&g9#0UiIMoc*WY^>U
z)hG&9=5-Vdf-8xmx7rCJ-gGicDVDC#|CY;hfmOH3{j;P>0fnbKs&pz33J8%_Z=nwl
zIn0#-1Yz+m*nMz2#F*p&BxBxAC=t80iwv~&`1xu2ARZoP%H({!8W
zcOs>RB<(_X6aml+Sc4CnT=r;SkC;Bb4vI^YuEh{f;9&+i!4DL)OrzB@b7Ys3z?j{Q
z(Px8G$LP3Mfp+ilOXO8{BjK@qB8t;9#%FK;C=N@fFmRX;7y7#$Z4u@F3fOlf<#}or
zsGQGiwg=wGZ(gA8X=V#3YJu)9dx_5-0K9wjUJvSs#pARV0H}_1w_z4rd*09GCOrmA
z-$r-#_gkmkj?z?rPyelyGU-m`MLj$`q}jB+0a!pTG!XJeJSo|U9ud10LadZKYlDmf
zg~i!wy`t~Kjy(u7#KMYv_rMl9K0A44K9CST#@UCo72TeJ9a+?ErwaG;P=id4epFpnuRkcgssJ^UxxR@RLzyjb7B+0Ct>u5hfX1t==_NMEA84yu_8
z>oqKKZ<7?*?7&7ZIz3`@^R2S4h~-%x+8if*v2;I*!O^jl?cEud(Vs7!CDVX~uO#EP
zqv~*dsHSE)-gM`;W`*)mIQwgifkLHm`BcVR%Z`V}h$O9ntD1s$dW(^GK!fbKHC9Mt
zZZ%fJhVh6Y7wPBu-FvM7UR6HvaL;nI`>|sn62Yh@>mcBdP2P*NM$!v-(Ux;8;=kk{
z?qbUeZ4;Kxn|^l@vm7szAcZ>bM&e5g*yJ+eezHEcC+*9U$P|sAuJg7=9l&>~nI~b$PtHPcaOtf^)5(
z6*ArvMYE`202cLo^Tj!kBAz;6cEz!@1I~oaZA)8wnC2PZ`xRDg9n7Q&T^~S?Rw0j*
zK7K4;C2iKS`w<;C_eo_FCD0KV2qEK&Q1#vnbLB=)?bzyyw5s60UWaE1lJ7FfgC0;I
z6`AFkziA817M<=gJp}+E4JZ@2Us3Aqm%7u{IGDQP7=ysu5m$u(y{dlu>FV1xbKEk?
ze4<*&{B!PhhwLz`ai`WTKDWPVo_x|t>>8w=`zxiAtM>ZSx_B(wXh)R<*jF7*$t)#d
z_=~#>JU2-~F#Z8hm!)6QI7Z&e)!)#>0~1?N4-HddIYsS$0@hR413vdh#))H2TQpmW
zUMHXGmnq^YkTv18l_RCncgAs8kBYgt;+GUJcV_P_7;3-d(<9>4gL+!2+l-{~YqNW;
z_bhzI4SyzvZ8=r;2$xo-y+grEzt{J~bfCm1z3JOAS+BL--)_I)=C~(k7tl*Ti_$s2
zrh~2V`vCs_{P@oB-BPzkG`*fsNXm4D0Uv{EuD+>+uHY*GJ^`y~0)X0pRH$*68sVDl
zt)c@m!RxC`0LvPYyPXLrp5WF^)!M5qhQ&RGsTFAp9csI*YdYF^Q#s}>sxxqvM)GK7;;OX4}K
zY)!gdg!aAQk6XXJ$4j0AfW*A{xkTXBQ3Fjuq*4gktDj1GV6*w(LkGt3L>SivLXu|
z5efxMHY?*f)%n)$Qxl`l6f#w<1`02kQ9^=;;+}IPRL(Tup_knUQtsXoM1I_*Qs4ck
z$n(S0f52r^^$O+5!6LxY%1U=o@9*|6d_O;x2Vjx|eG8!565HN4Dmhn7e49@yZq2_+
z7e32_KVppdQ(}%`Hyg596yl@i*3NE5z7
z{im*xvhAA-5=PzOLNe0pdc4p#&@-HwwLyzS4fddQEqT7Hs{j7aYOfHHzK;x+luPAr
zgsK?9M!>S4x}B+!m6Vq&L0)(olY8!u!E`uZeBKVL>bgyk{0MEXPDuSB>oux$vyZ`N
z&Ill)r~M2f2_w4|2Y6|Mgbx9PWQC=T$lSc#r-b%NR85M2mk>$JR&YFg50)fjb!9)EtUkgoV>
zW2*e8t+l)wzt=@@FpGvv5#w^8-V_qbx%*U;N6_aen$ywZVV%ae%s?X#T!Rk!N@JWB
zNU2})Uz!W(rd|h?`UeED4y_iY1-=`972~6DK^yho6gHp3vYdLzR;CII(k(
zr#t{4>M-&G+c1LC%~3pUyat$Nuzt#Yk&o)vZG4Z2Me@R@`Uu*t;@Qaq?SXzwCdj_7
z;P}+d-VLw$upRxbuT{Re5BZncx8Qq3eTHWvY<@)&J10P|RIhbUkd%04&_}faTqWo@
zt^U9TK*;fi3&`Q-xe!JiQ|rD46#VYw55R4eZ`l8}r#ZdFztnXRqtXMuPqYuJQyhz)
zpMBl?mA+caeC48Vp`Q*@fltV6c>QBUiI+}Y57fP&UjA5_pYhuMN*d<%+p9GvE7k^G2N?J&yh`S->9|U
zUG1kWcbK%95Ul#9(VK6KDeyB0@7*W&eY@EoulGq$vuh8`#rPG%c6y^87e`#?UGuX?<_O>U%g0Pa=O!4jSln
zHrBDXrc&?*9-I*W%qb;x+~pjOH8CM89P3(J5OPnzZL49d!SkF4qYnSO2=m|RLMfBh
zAOg8;&CxQ40^LcYys#?H>gRccAF*ohuX>%=Ibc=VOdQ9C(%5B*hkMsVXGuf~>xw=H
z`qi2J8Xlm`?hs{ft)5`WCi?3x7=!Z8X8YO={b97Cj6JW*8DW*xlu;G@V0~)`b|`Fb
z^{OeWh3Nm-yR(6F_yMD4mDz^|7g(BSD=%W3*J)EUNVV{aA*~buV&vh*
zq`&`t^88JZ3}%6>O#>nK?I)iGLGa(~-x4I?h0VpKoyZ@J%VE8>?4K(NlpZe(9*)QE
zcvPXGIr?62g^?1VhmGvCEEnGJQ&I%sy}M0&A1*ie0b}|pGcyz86;Rz27R7o}_>w3#
zyJ6WumG0|OD!*dZflc)PATMftQe+i1aCk6G271B(3H45fgnPTbHH3sf3jZV(Lll0>
z_nO9$6n<(OoXYy}O14Ue9#Osy#B)Ad_-rl<>yk@kB~z56anPx=
z2JnGRL{Ft^tMdoFl(;uf4^~5_8eNHDGqeHZdODdJQ8a5PpR=;e0MNzg=#mPsGwzz3
zSoIpn`j(O?p}?GTYHNj8zc6+GbD!FJ*t-a~?U7o{BHjA8BOVJK3LHGTun8^mSu__Z
z5B3d$Yv6wD0;J^2mlnF8uUA(7Njy?4dxpAQfm}Qn`jsxRYQ8p~8KrCJpU-R-KZ+qUi??C*7P>h)G9l+>TPl;#h@*
zrN2Da2ua-Tua6r$Iybeu`B_MchPd`Yq){&5ALEc8NVt;9;a^RbNQtl&8&16
z9hXKqk$tuxI;qyEVZ9)gv9M%8;zU-Qw0un^*N-KY@uD?>X3~
ziYMC;-8$rxl{Ri29%VW`-Fhfhc+>08IImG!eukzZn3$M6Y7ns9Nt3Kp0$oCD*tU1k
zq}x;A;>y)2_-ek?m)*58?uAhIek->qy_z&SyWw6q0SYh)wxlGFtH2+x`Zvq~mBv8?
zcTVG*sRBpmH-nr{g&po_P&ffOzUYswo1&eB3tW0jWZX9K*P-x{6p@JHvcSG$UGrdl
z=sWNV{Vfsl@F2J&pKjexD|&hgCwLET{bL75BOZ)$J`YSU4$qT
zvisx*FAdqZR$KNy2I5T1c8geUn0A)tVE$jRWD=JALg4Ph)Tw{ZIdvei&YpPQpCiXW
z2k~*4?*jxu%nQ%Ovg}+XDB>@b+$`)I$ruc;B8@_%`8Q>5J-T=h70t!~@mMCJSaa?z
z!}j33HGZ$%$68g^uS3?R?V3rdZLBAc6QoEARGCCBeLjR!*yF3{Z0XspNHw|ALv$OO
z7rWFd2#7)ksYyJ2!5$#l+
zgA~eFAej}h904LL#!E-E?34xY+~E_uIDa|SQusacWddFA0Ha$!|Mb{1HAu~=Vzxw%
z?t}N+Ln_H@ZoG7x9v5dspobi^n-o)~Jf(aUpl_ErF?@b{Kb$#f`aKVTDnd!Bh4Ll(
zh*9>w<^LgQ*~L9oo}3(h&|Dasn!0>5ujh65AX+|;HqWW=tGy`jsS0O%a7tcBKb`)`
z70^EE3E;aSDmWRRPzF+OS7j3bq#eaVBOcytH0~_oyT62V;7TgtpPAFzd-Q1lRuU`w
z^ZR$ikjiK-MRyxiM^Xvg`8k*sj+VW{=H?~^`SfBd`iqG^-_I9Ci3-6I^QP1-;7Ds#NX;FJcbJtLxKZQf-D^PKUJhl^R~(eQ%_6W_^0@1(8#J
zFVWKz*pd$rs4m7z6maC<6c&Ww2=7gWjK9Tp
zF~lmmsY~N{Kx`j(-)gY9M)82Mu{3_Q3tDp9(?6sK(m)Tz;^nP>;Kk1v2YF`Je^k}^
z{lU~B+IcSV2r-28DN-jo|G|Hhf2x8n9O>f%w4dqvKfp-|W0DM$6=HIghs!X5@
zQ9ZMDvAAAIlx_RatSE&t{NM>eR{0Ui2OC&M-n{>34^K|ck2ofnKn|w$s`#3m>715Yc
zxwTgWwbZin4WNKfRPqPq;|?3GiUfwLdQ_Wwh7Kp3Z&-^bSy#*)y+qZGu15D@4ylTj
zXX<$n~qgGzXK
zT3u#wMAku7>E^h}syte6BLI9d1xT=GCcpi1a%fqS1%g07n9uVWm(~wC(PpaQbEa6B
zK{t8k-%=4b2PW5j(_D>u768YBBp)#nk5#li(mLlKpDSPDy9w4yaK=Y+ey6txM$cOB
zERK_%*%rw5&z>#8Z|bGCuwyodQQxB1XWF`bW6mya=`LNW&aee~KqiNhP1Ys;<`V$j
zUQGK^4@bdyg`u_B!{Hx5=ueyd8wb&|-5X*S+0oIx3k;iAx;F3?%DxICBgI-96aki#
zcd!M4wT6Mbz-u$+E~wjg(SeEc|L@*oM>++H2>x_W>ZH&S$rhLA*+)sO
z`Po&Uhm>Lzau>eNdrKh_Y7LEzyQcvW<`u4ha+3}{sivS@El|zNz`i`)hMJDlpkECT
z$zFyy$IYAp$C!MMu7~*Am(j5f7a|`haBqPQ@k7>cjY`*$U-Xb3>)E-s{KDZT_|o)2
zp$qgZ@Ck4u4!H+4ytv-*f?1X|(4JCFTD}
z)c^Izf_<|?$;)=b*Lqq*bPxbmOIW8J0AF58J{Jq6iWqR<<>LkoPZjtx*
zZ5E{d%GI}3zMOm$3_%1PDSW&aC(C$LRv2BEDDcD8khQ^lBFE>sdiNhbebo%$&b?DA
zj$aFT{qos%)SzeTDG{J!#HEoHt|aD&;E=-qB4GcEO3mZPNAz$S8Dc==igQqZKL+5@
zGzwZ*uzAH|Egu##cu$vl@Pc6t(aoNC3K4l0t4fE3VWdk_dJhm%29-x0B$?NpG0-9b
z8`>PAev5IMAwZ{+j(=cQIa8b^Zg4_l3qTqkM1y$T>CCH7MS`9Z!W>B>@kXBqj7CUu
z^yeDfX=zIP#-vmA7*LAspv@?Cnb$M-Rh!>a{3?E>Vr@?AFJ=Y1fXx<9IjU4tg6()oVPvD`x_o7s=!w3NA=6yO~PlR171olyzNx%*|Diu?fEn`brdb|zI@3SG?<
zZB@R=c=Z`+dc57C(BK@jSRE;h_peA!*kYGf3-d17Edke45LhZa5yEnA5gTWZefqMc
zGA*jD#lyMKL0a@`&Su!Ik20UV3p~Gof|*&2J({HF
z<>kFH-C45gu1P|vHbdS;;BjhITfNwTHCl1$Q!|1tPjoh;tv6i+_3nJ(LIRBQ`fQrR
zFj${lX8OLoVbbJha?svyPmuy?S`|^2$iY&Ex-u$?*sKZ6z*zklFiH2c$eECRA{M%l
z?nydLFapX?k#-}7gyGf1g;kkuzt!|X3#`iDAN9@8O0i@fhrghkE^%kv;Jw)df@m15
z#xiJc%II%m%;zrw;j_2*z(9dtw&tJr=o@8MJXh#r?m&15ae;*P>A|{**`%S@^|=Gn
zK|0{gp$}5yzTPB-V}2iIO)^-hCUH1$v!=NX7j?uzwvRYk%jq8YB3WD0vU$$E`ocBq
zmp3Mr`74IeUw_>)@K(GKgZy0Wm($^PW3c6<+W}N2us5KVzx~siRBaO>pvMo)_}Vgd
z=H!qRp@o95`#=0KbE*!&U}^v1335p^Zq^^rLPW3{2&H=t9j@wg(p=$Zc5`!_E`>th0sJ8!O
z-n7YRyyl~*p3}yN4>CHAY3+tRp`K*n>3@x%e%g0vOxd>bfHGyTOp|+@UrowJ(>=U3
zDLXh%eiH_W7{|b)*5JiHA$JYU>GaEy0J-{lZ|S!}zB0XZGS4G0=UJ%{^tWyo*ik98
zsw~yZjL;=P52hJsP=EVQl+ayKCTs6Xp4~bMRh&6k9|q~e@OfW3~E#p=7
z+Js(SkT|RlwXOE&M1Z-N0;t~9lbg6jK=~1MB
z*FpHNd6}2%kQdvM|0UlIv0TyY7av``Zk-Gw#u)$`E2HH(#E+o*-;rXz(lp~bv%W&1
z3In=Dx}XmdA*ZE!K-KT|TaizvZ-(@cxlm-|{>{mGFQPYqfz&t_9gGq~jWQ!4z&GFd
zW@v^QV&B%4dkj*uh9*H_lK^|fTHqr_WqubmwPC@%PN>JJ>bdXt$I>98!#ONe?fvha
z0$n=*%)n6sUO0g0*;fMrwcDS)^9-3w_`-y4kNEmBo&;lpK`wm2>fPUA*_vt$0=W`Z
z{ih-}yIMo`%g>ib^UXeY{M=!maj4;!h0^4NZi-8#c$bIQ!SbLA#c2N&SgP-lkH+?Px1UB*OX1U4
zN7l|U@3BePHdu{U{jy%B-=qkNX@U+eWi8|;A(jP3ASxAQFu*-K|{
zunRIlv@+Ce&CEpM>?i*7EPyoMG%%0LIRLEnSX7Gv3y5*-o@dU$KaMC7Me};NNax+w
z1oxwoI6+f!e43p8pU>?Oc_iO`x=5$~Ei6K?
zA^Q6V)B*w_YRPDM9~bm?r?r7jIX4()Se?WAQ_r(wV8%nfcCqa2?g}Na8<)KOO{`p?
zD&x@}N*WZ)tOYmo=e-?_50nvQ*NBzrhi(HyWkIAV&@U*Pi6Ip*MOk2z^FQV=A7~z{
zv`{$!v7di71*{km8v>d}%I<&^E_GN}Wl}Fe0277qUK$rqI!r3@a6uUugAWX3amsgd
zt#6A*igb7dPJu-!*cSyeUtxp&R}(;h`JrctBD#2u&bvPr9ur|fApVJ*R)NBwF!E-u
zM%H(ZcMZk0euk0%TpKJHga|COf
z_`~+dukFyMnV|=E>qE>jA1<(Q(Uk^pWla=rvM1lQkU^cNN=1NOE*xAm5?D_@dAav)#|(D$EB0F$^lc08JWE
zHM+}PRYjjWsKPWcgL*if;vVRKqyU?O&F=w6J{d&41FL&_dt-iosQ)W*#CKCyI
zBeA_{6$JU`yl@Y;DAmFyCnwhe2Mt-e7cr+L8UX=8OX*Az;c{Ty0Bs~B<@@mr!v!4f
zvls;yqaGtmidE-wrKi^O9L`|VAPq~T5V`x@dSYY)s}qw7z3OPCS0p457zg?O93&k}{OTZ)Tp?_{J*f!8)bjl;{lv^op-gCO`Dm
zU#S@l+c-w#e6AHd#(+D1D;I;VV?0prez{=3`0J4apUdY=F9BYstzaO!=Iv1voad_)
z44}D|W7*N(+%5zZq$`0}(A)Wj^OT=CMPYy29RR4IxgLSbRnO%fO^qEO$n8#sF&^L|!4bVXgfYV>=`dEfD5T`xEG(@TY
zMNT)RRCj+nyr~cZc-7BM9;f*VFkq0&)v2$-sFSKAGLsUJi;aBN1WtGDHQ0*6G=#AY
z2kf=BR}Px4rwJn*CV#$|r)CvvG`ZJ$(*N5oW$9x()~b1#s=+vtvxj_LRDh0#%(Qc|
zXL9vC1t56)WtHse7-jqh(OZD7Yy8S%W2DqWjjf(JbI$E>1D&Y0;Xg&Y-SYL;y*(rt
zX#@c?3I<+ti9w6+djZ&aIJ(ZFaR?BtFV5XK-ssZ2JJ89R0%T#t3me^6A;g0VB@Oa*
z&bv=R`;f83xpUsvnZ^F^oDcv&t`EX4>rKFbO5c%t{PuTZ-jhw>q>2}Ag^?))9EmIb
z6=!jhfGB~he_)e}bzRK>kh(I>Ke>Ds?Sgf4p*{~D&g33HFt}G9a&~?+&5+&wf2e1U
z)s1|E>(<9@V#L#&ROp~iao;-l39}>D8SG?{qGRx64Grh8(Z+E69Z0IXxSs#!EuLEb
zNd!vJAh5k%E3y>z--9*OR1>oUjlj
z@O~wH%qEkm!RqG15i#)r-;C+WwLyHnY9)?{^9sA&k8C#Oc=xyDdT71Y2I!e|8wf@G
zvD-j4+W}nfs~~*leU|Qc-pzcZdgE&Vww4u4+mG%mBzqXwSKj;<1@yiV
zEIK?5&bvrpOp{{?(@{>K8UmkNTO^vXNyegZ>LE7olTeZcJ8Tt*eZFyCkvD(1$Q3(t`eJINuG
z3w$qM>OT&Sld>7F!UDm1f!_FEPa3*%sC`&!)x5p}Z=~;d@z2e0u*qKL_W5_@0rvHK
zIE4tSuZ$+aBj00z|e>HC@^#RF!|=OS*v2icbDb+NjQ<|^{K=;j6J>HwspC9c!-hqaH5^;?N5DTT^O#T7Ifmm
zvViIfFE^69zL@4Elq*CCEVM+hcBz!(jtELhP=ZaztC(XUdPTkm!0T7#!ie-S6gc-z
z09l6h+-gi2WpNSZFTi3V%Wo`Wt~V#RhVOX#*eCkFrT@=NdZjz4Q*$?g>KAl+wA}(E
zCw5njR6>Q=NH>kZU=|Mkuf?ajk~$u#-?~ohtOeKG53*Bst9DPC!aA$ovGMmfLz~D(
z-c1kp^*qzPt4dok-!~L?NJ%%0_Dxpb18t>d+pv8&@&G)WN*_EMBTxou7gQS2#OG!+
z3k%3ZjZAGv+;FGX8Y}RaslH=(GqJ#n^Js@W$ZJ>ttwba;sMI1`6SWaQ>a9%m5|$P0
z!y)B)k2YMa_Y)w)!SdLlMIu50+)>&~w+sQP^OJQc59oIGRP1hTnVbN4cvMDefz7&9
zp4;*t(r?R?`K_ZCdNQmk__Haea3*D5&%^#S1U!CXIIjID
zeY-!-RjaAm7`iJUwNRtGa2ol{MD}316NDnA^+XMv;-KQB&-i)D-74h`!-0zW@L3Ej}XYkw^EjCcN$x
z*M_S-Bs5-ulE~K%&c?Q5Gx_iP$$Ww><#Mn@nei(0BkEM@PR}Mm02mk_bJ+03XVxO_
zi+3;RgnoY%{+<=LJ+vK+EkyKCV||Glng5jY;>+XBN)w76!PBs+XcO8t=hN}aqK;=#
z+uy6Rw@tp;bk)M^nxZ}6E@edoV0xZkW6eRpMO(I%9?Hy0s%sR1PJc#79@?k28ZD8+PJ$0ceex@cXxMf+}+(>Z_j*_dFTE;Pd}%7
z*REZ)YSr4+LFUNX8{Ivr5N#^5>(xmpr0n%9YX6^jwEg{GB$iLmu=+m@R>-k?cdS;!
z7YHpkR%`uXOZ=D*?ZaGH`@kXt)8(3`hV#|?uRFsEsG|$RBSB-~gAAq%)fN#&iS)mW
zOS1B{NO7`+LNEnNs<712?(X)#&Jb{go*yE0QdznVXVkr+k_VLwe^~kY=fJX&LkOI3
zZtIo*OZamc?f&PRjJj8F-%HcK1f-+R;lDXeg6NiT9lRotXuw%^kvx0zMeLfsl#omg
z6Gx>4RWi8CpLZY((URx={y>aZ1jEla0WGEJ1^2ZMx9b5hmC)`@`TqisIb@{RN=hdE
zY53nK+CocbX)k(jK983^EZ-lxqHX)E)&&h}uDX3_mOxhbf=ID!^3x`{bggOVuchYH
zRGf>Qkx;BD(h`Ue;5Pju
zOobEEd3|}#SFXZ#bh6v-zw`dp*DjN+t2}TBgxwn2dGPA~j5fIhcT)*6FGr<_utoL_@j9eNOp8H?uy$gV>
z^;N)K>z^TOpRl2{Titym1MT;;mpz%)PgBlL(I
zcjnUfM6pM~yPR@)g&X!+Wori8|HL4Xux9wOaeb=^k>|hy*OokFKr*;EmLWNU!QbC{
zG4Oibhh!~0-_T~lE7_8Mz1m;mmbuchTPQCnD?U|(9w!m?|GvVZHwmLL-7NpDoA16@
ze5}F#eM?Ga&Xwg@L{q)}@q1$7kgeT%y3@V!791@NhpsK8Y1$Co;gDs@wGDp(`hw@H
z!<$VEmF7)oj%mse=L-Eh0Dc;tdW2T@W}=pfYJrZmbX9(%`7K1KC>3OLy`nH@j{gP0
zpx0Yu`xBi6DI{_0bRKxjxnDtXHLnsn1)^cDWTBueE+)4GIT7&ac2D5V?K`U4o%aHT
zE3@rJvc%%&?0v1v*iU7vCrQypW3UxCW|^y%N?vjMnKmbitf=BfQpcSWbz2^X!+&OU
z!ayY3>_?j){&hIWMZwX213HNv?tpBdxbIJf*pnI`shd|lKU7{yL3Vy6e3{vOBSPY}m?-me3R1^&8?qbPn{
z(tDZ7E}Ya?KIuq`ZyNt)Zxv>#j95p|CNz&fdH#}p|GK=#q|g212oTKCW2fsnyKuL=
zpjn++-;2z*OLeHIp2_3Iiu-tb>~_R!heu|YE1J#hf|xla{@w3B5Vbn~Euw_(eoPLe
zXB~WQ^&hQ9(Wjq`LNq^YD4OB_K$$qc#e|VD@-l9Dil2r*Ko#hpT7JWDBIHUzF4wl(
zPwqbkM)z0UYWEF$%ue5UM6#Q-%^^O|WOTS&#$5`6>-rOW-$qxz*-JrnVzin4(tLl0
zRf3VhmtbH+SfKPWXA95$BfWE@i~W3K6}dl?oX6;>A@EDz>DoY%)7|5N!qiCOpv)?d
z+1s?=IHSK0L%gPHNZV7!$B`G>p)#;+tarqMS
zVitf~e}k6U<{2P4Q6Hsa8`~fn&FIMDm$PJEL*-R8o}*fQm^pi3uW}H$%C6OZ5*k_K
zNK#PE>}e&vIIc6J+wA)I`Rk3ALWcS%4_{-qqZkT%VTTw8
z%PJ>$m68|CPGED@YiBUF5LpJ(m1p|${K^$tk&?bxg0b*5EHX*XkB9i_3J-Z22(JU1
z7q;a7S*B)+5r+dqaPW>vP96uV?QBe{zq>HL>eXwjX=Ug4A(+v)jq_z%WRqvMm2-NfJifJ)xc+y#hcpOAeZU4jtv5U0PfoZa`6*u1nLG=9SH6X9M
zf7qMI?3!icoA>?C7~u7fucalofb#7=nauOq-*(>|o_Q*BrBPHIVmDEoy;FI03>elR|bOo3};_1eQs6|&NsZw2Uz5qU6uP3C>lt1iJ>YK76hrkna9Z1RsX
zbZhC{hFo-S@ZZ_gpi%EYRZQrt*5omrwj)jrq@u9ek}2;xjQd{7KbSK>lgiSNWals-
zS7X2^QM@8D(6vARaD%s2L#UKITT+z_mX}7@kkF}nU8p+HfS;(m6BMrKvDk(#ejYMIBdQf(hi(MUj9-utBf=pm=p(1?i)^gPHg!yV>Dz|0WK4Scu
zwc+pKTnwnD%RsWx6Z(U*tl!nc;wP{(}LdmEQUEME*q?5lU$
za=W4z&N_n`PWKST#AG%lq_3A%F`LiSyW)0=M-!rlAuRL>0dghygk$KWlI86euH9!x
z;wx~sM7l{@%EGGp?xM^#I;}6PA<+?F>
zzqD0ptl}wk_*k}d(V>`zu>t}7G}3vDR+$Pc1|HfuCz$D8Oa65D#ki_eL;9@8q(ktl
zC+1+iF%!r3j)|b_-ER@Y1%B?BqG_HJiChh(h^hVmM>f1h)%
z=%ThXtMSfa9Td9t-|axa9#<8Xq0qRx6qJ>fwO8V4yqKnW5$<9hmwYw=$G|^=H9F?s
zXkJ|$c;wj}Z8#he88hH;R?M589HglN4)ThZvWctP4i%Bv`|q`!&d~h1W|mPMwO)Y_
zX14tV!abcXiY|#Z2Icu-5Lci|DIx8is$rguc#H}YOW*g}hwds?
zN!N~mw9Z7}ifGB(woNv1@RgMse6q!4_Hw?Nb5ooKf3LRWe|CxSdA6PQ%{
zUU^mJ{1&q5)N%uy6faIx?UzQh;je7wifp_;%2e>xSpf{t{@$NHK?(4|!TA`7^4&7z
z8?akomH+eTEs;({J4Dnb+dl$tx*|lIQlt=&c1_d2@5y_FB`nD+EB!r!k5>q3Q{p}?
zZpO)J;$F-38y_H*9l5ZSIJ$u$BG8ax?`JOu#k9HY`{(%zv=p?V;6AWcJ`abIwdi7H
z!*o#g6~Hwn`@E&fsM{{Y_f(!rCQi4)&Yekk?3S?aChDOvUxG6v>@@&)ppJ?
zh!7zQ`X(g?I5T#l%HHW^rYPF8^r72QJQY&AMGQ6QYwIiZ%f6+X0Ci%AjNRDR`XK|h
z9-~MyvrE}O6G&GNMpPVZ))m5?rNC4cnx}ex88D2*N|z`yKkwupIGvqrE-MUp<9O89
zy->|ycj+*Ygmg5uV{*=Rg$~sQQ+|DfcYVXxcX6_XCRsw<$9|up&9Y{0eKFU7*7Jf)
zCv+yjPnTJIN3e|gvOJXy^HU#<4mkR;FFU=>^_1zuE~E5d|E-?g8eWM+bM4Zq3%0iF
z(A41gXd7~?wt8Es&(pj;(B#YpLCt-wC&dBsl{gG0LdNdM2)mA*75aU$ML$aR3Oy9J
zDWo5>?!3i}OZT_tZ8A8NsByX6;SC%9R|Z5zRv}9^KPy+iy)k>dqBpu%)L&6uz>#Y%i=`V67mv
z9_nI^)T2SreJ;uF^qs0zZ-z>Bsg}k-leLi7b(tqgGUl>Cp{e?e-mA9T=4z=fu~tIZ
zKD!H+)e3S4gEk8Z?4!pt41=plF9P6iVwqkd!a@4&a#seLjeXSX-{W^aui
zezC3f1WWVMgZAzx)&kxrG<~0Slct@;5I9kr5i}O!xRgQvmYv6(8?7vthL`-D^aFwS}=x>V>8r87hfh~7dlgiTksl8;6uGO)+%mmOBd_S%UG5(T%8Y}#kVc5K6Vk3y
zw)-ulKEHlGh(dL@MUhLG*Bz2VS}#H+Mpd%4f30^g{`~1mML5z9k8Y=0HCQ7_RBpqK
zY9k}ZX|7>86Zh4ER=LY=hcqjt!XT1&gp6a8how(m8k@wcG%9vJWJT>~Fl(Etj@kIg
zY_%4*2)fPh?waD$CuA`hFbfvlI)ySCNo)SPB>!}u_cZXZE
zZQ{S0)@p>Z||%ZLEmZ+)D#Zx
zS4S;*AXhmHGQ}t9>)znR*p-N%19!>m#-bnL5D*lQWI1C}f=BOgMe_-`oRq{P%e9~RS;6fS5HTx~+*hbcBfoSd8Ub8;u&$Bz+%Uc{wu
zoAc$k2;SF~mKw-bKiDi+8iZN8N|GgK*SJVv?D_86iPdtbPrqFAu)z$ax(%Gw4s{e8Y7mwhkb2DoxSN8p_#}!}}Z=yCgW|+w$
z@Az_q-@f*w9qSssy{kOW%(FSa;`vTkDSf-)V>pH_V*1rW17AqKvHpHW#WD`nE(hn`
zeYo%MW3$4oa(B;fE#o?)A7Yk}+K$J9Vvcij*m1w9T6IY4Q;Hvv7Za=wX`m_@SsE4`
z1o`_DD&)?+a-vv@+6utIz5V031ekAa`n`16g#t
za!2&@c=Ee}2TW{EX28P>08py0Z#frno1I!_%B|&n-{d45L)=)3a!8+5hj>Tl6Tlrt
z;&o34<0NI
z&U?gfZ)sOUS+$iqRYnK#zh>=sYZ@yo1@U?1kTesF$BRLUG#@c=R3rkzSGOv}Q197=iU^k3
zlCHL-My&B|8fs@=_5uzEu`)C5Pn+1OBU5@>t{RWj(hsT`KWjgF+|#>%wF?hl*Zq##0zW@?100c4({hKwi2h#xtQ`10EApV(eR=luUb>R;Uz?RG;w^sQ*
zW2NyxzWr6hCc-RKyuD01zn9KDvgS0<4?~T`j>zNQ>?*K}iTCpJs~Y#pMg1K+75Tr>
z5zZ%0i16g~;JbKLLc<^T6Z;K4ktKa?;>lpRxG&4tR=fk6=K8$B|kdYl$+L=g^@uG^!P5UJXxjxL_cOy9`OzM`{My$=ol5{l3p`9_@SaHo2kp;{3pYhNt5vIjD^+^kBkKTnG(NEX6Xt<)c6
z;|HKE?l0yO=wZ9Dj5Uvr*zE?~Z2?B!h9tHTRpt>_z0MYO_8j
z@|;N1N!pCVa~wUMB_6ec2LxSYL2H>UY@^+P2f}T$Qgr|>CTaadZM2o2h9GUF%+T3)
za_gbNNpebgjW4hO&cth59_nafTjg!;)K@_}hVw@hD
zU3iXjytzV^DVo1*dJHNw7t4?9@D3@LALb8{eqG<%P2oTtJy~fNk=g?GU+~$o*xyGe
z*ecu2lnE<+G%*|wWWArqdiB_K-CJx9k^6dUfwZM4gNb&;>kR-MbxA?
zM*8T(t9$0WbLL1g$oC=n3LN1nouM9r&MWT`^8rr;3~PSjC8|Fjc6|ppH@7e{$9n`a
z)$(jL|BH+Xp?O#!N&R)~N`>c0p~s-6XUEJlrKaGz@q-k!!?Jqk
z=k{F^amUh4HcxHos3q1P#X|=O
z1IrT6C=ROPUy);#=z0Y{j!14nQ?MFhp(b!($ymCadp|nl=DW@0x8Yqt%Ve~5{```s
zILSECHJjFd6lIJ=BcAFmzqs*G+K`sL-xs@tDA{B|ynIj4KW>TLQv&4U9E#9TPC05Y
zA<{*gHWZIf0gaD-?qaGya(U=4PsO|ibfzD}k!3oNi?UK|Yf26{^|3G%?+n|G@X0+r
z<@y(#`}W<*gcjnyeU;x>y{~ScxT@2X@M|O4e!m84oL>FbCAFB~Ok!-HGK#>I#+cMl
z9{7wK@&g$u(i`0r+!k>9)1_|EMa~BltH+#h3*bkfd6$6ORmnh2#CTPpml^0;*Zi~F
z7AI6Jay8JCleVNQ6u`-_)Wh3C(=%1VfzbesYCavAC7|&VVXep|2~PVBDyCKHjBt;r
z)sjS^;$AiBvg>9)%miyRX#+~TyltwR4=}EL~MQB`f16Ra#rky&+nCa
zu|4sux$*O+Dvs~UQw+1xfaQg=GkVb?XcWdfyIei$#r>5xLI0z8c6EUqE)G&g*Ux$L
z_{)?W2iMmNcJT$3Xrb?%-_@%ycIdrnYT`0%%A(W=czP8{?|kD)k_zgJCHAXrBvL@6mh+#%7ZkWe9?>X9=s9%i
zZzm#Nf2-=c%OEKvJnLMJCcNdXqtCq=0E!?P))ckQt@Gqtc|M7WYh#?W6%K~TRl9-B
zT&6v1o<@_}zy$HV7QG|lBg8lp_(L1te{YrJ<(&%dCO>va&s@Y6$7kQVxO=((kcdyj
zZ{V-lJeNL}mDo>Bwd=34i|Xr}X=m)~zhzSeFphG|Neal%3(9@{8FD18q2IW9
z!od=5qk1W+I9Era1x}TVAyuWHH?wRqNoD)j^G>u-v72R6@7zDXV3D1Hn%`J>m^&9#
zT%;oJizjBHj`Z&ADI;#vtU!OWq2kO-Jmm?Ph;xrG5%|qDNK`gZWUY9AAC&qFT!y$>Mb&4+1nscu{PL{#)SUa
z>lNb&RLWlo`F|1up=2%I&6fikiPxve&sAf(!|vrQhzyj&Ir$z5rxTuefp)?B3nG4O
zfHi|^T=7wHwb`Rv7}1f}s!O5ftK$Yq>FOaAmSs;h1A7?hl|iKYOgDCzwvw@f$N2*;
z(V{ZkNlk|Ms!e}_W>fxv@bt5)y%C>LeT-l8XIO>9@o9ILK@PtH7?+`bQHSlAYk+&J
z*RXuFs^47rqASmFOtkFJ{L)~d1F8H6D^sPu85ovW;va17Vdnfl0DYGPH=MzZ!@2Y9
z4^QWgV!1#&-5KF3+GHQ+&(gATkG2ehuV4?5>~@Z!Z!$$=yX^^v9x?5)^LhfKBxKtV
z9@{9;-C%&ohCe^yN{yzRCIE>$vT=e;0Ye?-EFUFDFtKl)BDQ2z0!1gW+tAS>j^g0r
zFy85XozdE5s>)7X6f$;sUjp(k*ODj=U>$9t!d|X^y{n8kQ5!xW63(hJ;%0DUoYjBZ
zF-SCuKnK(Zi}Uif4%Zx>HLH{ck(|v_Wy)?ocEQ^(<66c!HtTvZn4Yv&%f>{m)QgEq
z-`CHrd4Ydf9D44}l&EVI=IwDH<8$%xAUkQim^){U#HN>H0HF)(q%XGZ<&jqAICjO(
zQ~8M~r9V%rZ8gp8$lWRA+-8h2o3*}gMVH*Rww2dn(Sghi{P?ovZj^2p@eomT?x51O
zv?6`WzDZ_!sGVa|@g$*vI}YDayN;O|^`6K4kRpJ|*fGI;^fsxqaCZ=&Y?&Q;OV`P#&A&=CN8fB`sk=!&GF`h*1;cy38B5#1W5@$Ul43)Ml-?x
zQ}#GZLLE7Kr1zU2A7L`^M_<{nI+`hys^Zug9T>2ebVcV{k=wGV;YlwPnD(0+p3hez
zwdMu@=OXZCa;41_@>+B~COsmmcV#rf)4ORC^+z{`7BmPndW%2yb!&v~vX#x_X%gga
z0}sP?J)N~u?4wQFfOfc(&$|2AbWpsd9!0x>=po(qp?bOE%BX{6mMK%Ni1+n3J<%Ia
zOvX$AyL&ZU7%LNZ3pFEWo<@(!!*bLW61U+AiCmAC;~K&z}eXwmls;j
z%i7)n`_QOvDfck1&?onhKdXeSosjJFPi+-JiIj$6;XnB018IAN@{A7p;~CuCq6%g5
zFJzzrGqI=D9f?S#*b_S`5-<>shUknQ?hGiGG2+CpR{Dp};Dc+HghC;Dr8ga_`9mp^
zGWmdhM#*dtwauO|M4?6E316j3%A0>pwv{=3a|ww8K!0kH@Dw=H}$7ZexCV}K=oO;GC5YLpCVPA?A;fr
zzu3#ac5j0pl620!$&~)&9QQqFMGylS7FsBd&*)d=Um=O8BJ2z(`tPLo)DH73RhG*>
zNXTW)`eSo)do8=aJ(ON1F1g*6kltw`4_HgK+jiTERs+u
z6lTtB`QkgEPOmc?z?b8CyqJdbAJ{sITnb@(mbSmw9uTREv`gWF*%Qq)-@
z@C*(o_o0cHOm1|wcuhYS`6Jgv)J~t(e+b8h1dM1N_%TdA<16eTU-;!QyhHofb}=E|
zVl$*)RQHwt@r8Cb!uAM)mrRucn?!}1N1&lVEF5j1&
zkt`*eHu}(oKl6!8;5=0dl2Nl-b`B_Vj?6`F#>K^BAT|CN^rKpQVsyV3S7J@VeK^iL
z_IqvV8|+vKz&+obP4CfuER(;`KwzOt=-N{tmqAba1!_wBru;|5z>P
zyLUFcuprL`Qj7C7cDC}PUhJtQndD?)98Cg2Q^b1shfYrO_um)SgR&Zq#)H%5AV@*T
zG{lqF;hD%$j?|Uh*66lqKR`s-ZcS_NYuInk-fBuH2D3!jM~_2cdl02#D?JVcx8*le
z*xIxPzxTqGI1X*Ir})y#&iUxj63ybnt~*9>=y>%Zx`-Z(PLaF<8A3YXay3@~Uf#za
zq#aOSY=HG^XEDxGsr7g@9+i4Jc);vK_hHzB25~YQ>L;AB3bBmO14qO704MGl0*)dB
z(}fvIA1}v@_S^J>0t1kO@(Di^7gd}EtXtC+=9ww$7jgVM
z#<+)h?{jXPuEp!)*d#OXRN0W@FUgk5gGLP_<8&m4_s>4zLEzsu_X29HHQVRVK|CEH
z^D@i0~o<{;#X?S6q;XdwcJNPqig7$nqyNF}{r;
zcTbn^QeWLbM^<(Hi#ilg0vdml7jn)F2&#Bq(j-IeTE@%HzZ|;lR0JPw;Xi=kH4t6H
zBDR-5-|b;RXiNXIU53y2=O~X1tD<|8`JYRcR6;jc>Xg!P++lcJLnx-SO`X!8pfd;4
zdqk#n(p?qw8*oEe3S`63kyz}Us&ja>)*t+=$s`pLvR3LcTKrlOBE5Ihx>m5ILU*SM
z59=m_d+rw_r#gLGWgYFo5H-AjFI;{b(SV|$C!uhVjAjk&bm+F9KDqi;Qn^3w6vii9
z^wN{#XoOkCY5>9A!)ouZ0vhXW$0*x`RhLg8rJEQ;PxE&Eyy5Ly+Fo2AymA+kJ(Jx=
zYnUe+o%{p~z;JPNYq`AJ(x2}viWPunj{7BXT;c5E<0ic*92V%N3Y*LW^KsX~N1;Ov
zyeEg-Map}V#ms6yg5$C_?kiM5;u$t7Swv-4neP8fF;roE7f*?*ADZBq@&0g%!NP=J
zgzr6xu`fPQh}JOhV0ypQv|Nv`Sw*njC0Ea7SGm(Ws@xkIS}Phc%?_KqJksASBLH4soseV!hm13ftG+&nSukIS;GnM^5I
zVLi__;bNLxKCa>r%?Tj6WFBGJMBtv-QJQ&5&sOHEi(%*v9e0{A=|T-`*%xg2SR_xF
zYYrs9ovD-IV$mnKlWd)Gf9&cU5oN5Ga&TYH&25Z#sWj}EYhx32Q&;u!Sgy~wvd9rQ
zcL$8`c}-fMyApDrnr2&my&RutQ4sl~_u^~S%&};WfJ+oZ-q$&Rc6Zku9kyft%^a6ay6$K
z(mSPsnsJ;Xj%JIk2F++cVSbz;=aCWgJ14i%;)V^KIwte5_&^;ihe|#JwY?Ibn{yQr
z?*c)5W2a|m_O4Qdr>oQ;^KWCW)Ba^lhUI2a@PN4@J#I2;mPB+j6UP=nmj=SEL#Uh;
zKHY|-y_gc)mXtWIc)Ww1;ZY7Vt@qNEgEtdfh+z;jU;*}}9uev<-)KL-FgT{2b>%6A
zQXRXLUZ!oC-vQWI9J`I#tnV&aU5GYrx0}{>Hh#rS?zn&D?Q$H@)h6DG?b3{4`tVmY
z_&WrglMy1_-=V}l{J{=9I)6oS7dsXgEb}6uU+zpvMUMyC+5~65=u6skeanK!JRAm;@^_gz=6772d-GV}
z-tp6Eu`J{HX!8f63P&!<+bQ}aDbvkEnf?f32>6WV_1`KaB6tv@VCah8O~De$F*gic
zyg~z%&Qk!<8>%<=YI8pyjZaSw;!)mHFV-}5P@Zm<-e`|QL_>E}%^D|6>|rh-sYHV+Q{F$GS?m80T+MDiYZ#__
zXkJ^#wZ*u~lHWKENJy<9hEJ=i6S&%j70b*BC$sxL!f7s|WFK*mXgz5vQ)tdhwXNq=dro{D0k{RK=Hl)(4eGQz*P=F9gMrOU~rTm)&N7GWspm!UbY6(%pii|>rfHtZx6hh_|qw7)r?
zbJR|L`>-`?HpilRBI)m5*=)&d^2c)o2+<+5UnERa8u~w{j1W48;;l@GnsMHB+%d*F
zpWId&=5?we1*!)#*%Gt9Z2FVYV@+K|So<`!b)4IC=R6~@jF&npO?q^A=FX~ziuDJ(
z?OK&b%^qc?0_f43V4`A9(&)L8OUXZJ&)B
zzP6hPTHSO)cH5t7Byx{DJY3G
zCC`aAoZ*$jm%a+Ay*C7RJ)lw_9#91kNFQO0%v*0=2Xxumnerm+^h3XN_x09Nh7@h7
z0C_Dl&2|K^uV1f9^FhY-f5CHS!|8!pP~P4`!17S`NTwTPD50kBAT~(-)q#7N_zmZL
z&nY#)wMq}}wf+L@o
zny5;bkOYcOjf;#_zZ$m>GA%3lb7GWBH`f)>tG)CG92Sn|-{VZroSApYw8+)jdw#S~
z!b5#o2^;`@29MwBH?a$hHv??$+~4dEwRP6&V@&szTSkg($L0uX-MXIrr
z#k8P#e$UF2`6*pA8Z)c?@!xTjBvy0uTJR%$TQx1FWqRZ`^kFVRBMLM!1!qpn=)+l%
zSZgkXk6M3K(6=3H^U~b*{?+Ggb!Lh-l3B_(MX{~ka8_8WjAC6DO1$GY5KH5&u;R<
zRbFtT&KYS%e)Q}58svkQRB>M)#N0V!YN&Qn0Q8#TctlYfaMu1$w{|l$1t~y|so(4Y
z-bAFTUE@c&97m_-7p3&Ux9&#e(m@;dwk>4^{SJSM&A&@f6h1;Fs09gyKh}oqhx-^b
zr09h5#44ZP(lj~mv&!jd{nbQ&zYfc7jgkgm&i|wgBE|B
zYUNq%-GAN#d4E(KG^qd&)xN~)pM73odONcyWk?zK6jAp2TeDP*-mUmRS$LomB=rpc
zZIG*CO@>heRiJ0vCM#F1gaiD_aZ+B3|fL!Tbov!C~KR=j1L!R6@)*yD4|Q
z#hyc-a&cg=1rio&lleaQqmN<%B~8c^`^ZNus{a``<|_nhQ0z8Q=>ORU9Rcs`4sA(S
zOjK&I34>hj9ee@tx?#-4N|U(5)*;_r2CelVrPeu{b$Tma_mCx3i?KQ*9v12pLB!2l
zI*jt0DY|T;FBHM4aw7;ub#S?-&Hlo$r5lhM=uAMQj@m^IkQpoY9OwEgBXzb+;oxe6
zl>xGXbUo^H!-?&nJ&qTKxC_S~a7#*WsJXn7(eYDd3%QQTVv31;MfnSJGbIy*$G6~c
zM{>gV&w`Nody84jwoTuDZU0n-bG=g0p0zf5mdy4+Osj;n6VioGd*h=h+|Zw@oe^ALD-~P#Wg&4R
zUgL*(dnn^yp#xWyxgZTkt-<2|bJCbVNaT$TEV`opKQIJwCJ0p%-dcA)`$F^0efzC#
zbeD^XA5{Lvc5$ie%@4SjrIR3T&)YRV={;tbN*xtJ0+k%8aPP;+ak~{Fd&<*@?Y`13
zUdE3#2u3}|+e=Iq^Y9w=mvB4)=Jh&9vZ$V!4+^!K-6LxhYoG#1_WETaC9BSURk&m7
z7-;9H-K&je`m%OzhXzi^BEiNGaCA#g3x924+GNy55xnPO}d(!ne_|++r7Aqt$`nDC$J9j-@rD5oNb@S(r%TiKFH<
zZ)yfT_H`NF(Q#33Vrsidrf|=~YB5jMuu37L!I1#nf}@ytRMNR&BZ&Swx=St*awBr&
zF{#qF&NeKss+vzBL}yp*W-6F4-2qZm56@e_fn1RcVJ?+qKV64p7sgYs%-?eNz6Z^2
zY!7aHA4#;Dh59&h!WX`i8gnhioN{y#Lxe|yB;RZa|J_(PA5ua>y=3t>N{#;^dlK5S
zDL(Hp?fwe6GZN+EtXq^>3>wR0IJ=lOx6g2S(+A}nKiRCcev~8?u?_QBn$OQoWSY1+
z>w7+(m2Ev5g#=S++zn}*%iZ^lNH{e*>@UqpMm*V=g9{%L2Xw&^``4tm6A)JH!%qLt
zQ{{}@9`7PK@`2GuBD8`GR()4{ih;zcI{e?G{+4jrHWi2}?jQT!;p;F0)`wgSI)J1k
zm#5L@au4fJ5o+Nl`U~E>!wAqL5?fQ5jo6@AkeK9X+t0)@8$a|fNE)p&7B|_4NAFZz
zDfzS$T`sZ$jeeNQcP%ip(ns^OCVg^Vg%-f)75+;C8r6Zl>X`3$wY6k*{cHiLfF;0>
zG}Q#G$fEQb3euPFA^J4Gv@JV!T*WTWizAo_&E^Q6Hu?ZbBhI^e!=CDCxA8foNXW7jmV{^!drzqppSLRtO(bY&(OTS%ZGbju+zVlYaXyubA&g(8s-L5b>x+
zL->LAbKn>SSaG?v?K!)YyIPY@k#a9WJHe`N0gAYLq+z!7VnBon#8o9^8hG}fF9YljwtgVR
zkfh)pWGX;67`;1i)+h9sTi@o;fk|Ul(X|18IA9SSu#0OkN>i}bs21k5ubmHoDex~*lI6?ccdw07h!0bW;3mn~
z-`2HYZ$n_({O-Gx&cv)E)=ko^(AHzXV>KW
zRL(D!^V6KjQw@-;Ps0lhP
zbFjo;=_o{j<4>^&15Jt><7O+Pn6isF5sJNRHTqOf9(&k3r!7zk
z!2q!c>x2@fY7|n6uHI30H**9<4-%&?Ls})&PiUh}gO$b-WohT`?bk!m3aQx#h{xE%
zAUl6g*}1Pzky%@;`hb)nmiIVC)ub4kg;~eQn(zV+K#VE6IK%+kgDtx1hvt%uRpIcC
z^6!hr`)PY&s7)`;>$tfw8psq5nwLsnw#w_?b8bCO^#?hUDsM_`mr=^;9a6cq2WvE>
z8QNh>Ed*z+mMVl|DT~ce;0-*M-!%Vyj?EG;$YC4tu#uiol|pxf!#D03>m3}v$zd%H
zcUV@K9LkZc+a_mq%@b)y7?^)7K^qt-Kfg?!AH|*y6_t<%7*!IP;-sUd9&Yu=$N@HM
z9=?3o?ywIjWUuya>OlMb`tgtStkVyxT*GTAHRVGYYB&=XszYn
zbmGrnJJh77*pD*BmN=SJ*Hq0sOMRsd+a_dGxiykp#3o~`btB=TNKWyLP1GHz)yG^Y
z&6PMB)PxYoy=JbcYqy?s`^UegDz2#W{=bZj5(>%4sG*>%KSDc5JLM}B?tRNnQ6XsB
z69$(K%*;fhoDSg<4v7!}?L!eu`MvVYu{mA*^m%o4?DP$jJ2=$i
zohbH777eB+7X#lObTZoxnj+~BdWu!Vq%N`Wso*6dIv3vW3-0)_;zg~7RLa>emmG^e
zyyr}|t*n<+qMYVCRTcWrI1^*DVHoggal-+cXkWkx=*=n>`X-3QJf5gT
zXj1AC5ft*(URrr~;rxf%wE~1p@SwWN_TEf4n)#&TzNfhF!*x|N9Zn~V*Mab&IaL)^
zt!i>Ie@*=#+@9yRGGW^f%QWq>yCDtNHz{r!m9`W_aCs8)!hE3>Xj?5*W7=|Zkn4k#
zKO(vld8^iYX%jKK-31DdbpnS6J=fPReP8opA27d=ueJNwU(KsT1Yr=WNVX}}Yzz{C
zQCJ%^ai{JJK)}q1QC50wF}$I~|Axx;CL##g;nbyCO8jreBqe^Epw+-Hx0+TQ$Lk4l
z(6U@Yk8h7lzeW~E^*CHRpa#%rJ1#af!8dXoR)jiuWbz?<+pSp_Lje-hwn)-&qzD0U`36kWBcreOfvpO+QnSh)8>W*s-5zap)Q_RsEV7
zemh@@DYU_`>Jb8R_&^8_}{q3r-!1`cIQMRP4sS{&U#`-
z9XXRSh?#RhJ4!z}oXj|FV7_^;T?v;wI7t94d2~Q#wO1u%XK_1ktRw{X1>lZoET)}7
zX8SRG+jel*K~WyCqWZNk|4tmuu=U~&k%cZpVPKyY)LogzSEEhG0LtxGN~ojSDYQGn
zx>4$+`}%%*0H^|5HFm*M&@aIDRPZkIT9?j$FmagkIpu47;2d4;c}+-fXH=t|lR!YT
z6Mxs1RKzVlwU=e8>X(Uu%DkwecXttM)*K$X&(cuSYCr?M0_&rARSeNpA~LJwy`!*f
zap4te`5hCPw)L68A^K4H6q<;J%u=&_pAjhoL3V;M?d;kzEXU+UbU@aPt@LM`_|7&9
znYkgNtDt^xatNtmRT9YdL%mUela=vPvjI-RDNisu$mUdNch`~&T3T#qKv6*<0~WHV
zXof^xcC$869Zg_VTo|CLjx}P2UAeEi7~OQCPmvvn)2Gx{(B<+sytbsyNNPcG;);xtw;&s;T1`quc<-zE5_@euf>7U
zYGjTj;tPjml;Sj!<@V_weBs`x37*18XWH!AgcL8P)l6dDX5WM^&JV+%ycoHRr0yX#
z2c!NhnF>f8(($s5q76$d1Us;%jx}O(fVsR+&`L&o#~-p&F=ixZ%6~jYnuBFYUzHp(
z@HV)Mwv`?OwQ0?}GV2#iB}N~7(Q)SXM+pQUn`2-_#xbQ8x*ucw-l{lmHFAu|UA5IQ
z6x^`hr{uOBcT2jW65{)gl&tH%=PvL`1;4dL=JKY#k
zGzE?Sw;2DV^}Bg^UUM-{d;F$1BY|Mq|n`Zcq&+VQFCgkt^QKqdbT
zWPC@^Jok~bH)itxD_O;mp#K_(ig)SR>5m@mx;~)01mx^}6BE<$d`1-#qpvcH`<)AjGJM+|u(MmjBmtUmDtD+AvLQ1-DMX5nCGZc^GO
zgz7dCpuCqX)Hlw;1J3m>y(G5L_sn}Y&2dy9k7_rC79(LrkCQ*_#$6u(qK~rA@+4qa
zu3U9xGS-t;5w}Rvou$7&(uc2ps)xu!B<2n>ruolZ5@%(?%v~HX{WkN!o%8c5%#=(I
z2|RxqgAiXoYlRb=n|vf@nEfLyh=PW8PJg}@t)3aRrAgTep^~1ojj)yZX@gdhkj9$p
zsbaiB66jx?##Nh@r|?~qbRK}O{yq9Ot$BvU$1(i{DQ9H45L`YI)D;KY_Ao7Dw>t2`j%xK(?J_{_t+5W?*LPjrm2i535oMyb
zkSOvDi!{0gGhT1qdABb9T^HX41tei_c_ACV|MTn$TG3V@+aC0wKyW;o^-h_|W0x%L
z@%$LiG8m@p_DRqg<@`f~xS6qiXy+Z|3QB+SJa&mhgB~9qP=o0B9+;R3#n~=i@HW8C
zs1teQL)Wt0ZZS~&T2MT7W$a=0MqmVr!+7h)&WXwG)-Wv#y7XS-Xe?VoOEFuh&W33w
zau-zRM*m)Zqfe2V>4Y88W6i`kVE_No_0>^bWna94D5U~QcSyH%x0JNfDJ3P{B}g|&
zr_$Zs-QC^Y-SG}H!_4o!wVwary6b!Ixp$x4XYc*tpwTK=;DQ=%Za?#t&=P&q5w5dc
z6)N=7THKl{Y+z2U@(LzBclPxI@j}A|pY4MIQgf0}W8mp@Maav#ts-V#$;|1kju(ds
z9Iw2TFTDq#7jM{h%!5jI)Djp{_@*E?9}JlCl;X>Cu{6pjpkAeU6iZOVYJ8ftZ$_a0
z*xQ`Egd>yULBn?DxnoOj+Cq>!ZyOsnJ^h$D*pw}JrD)-Owa>N}TvK|9RB1@X^>DEN
zz~5bhkugIP@vjuP>3m$AyECGQaSL5~iXugVpcIXw9!SrZ-d-kGlRA7k`7CLRyXHqr
znbTTAt6ul9&UZ%3B<7{Y_^CUNL7&vD)f*#=TwJJUr=L)<{-{$tYUNLNJP4V?Za
zkA-t7PBc)YTakH6)K`4L*{dcu$VrR~CcEJObRpS6n2u>dLjnoTxA>CuBF?_N3TFG7
zn6S?#52r9$B5d*BUcWt$gRli5xQzyrOO7-t>Z`&m0-~Jc4F+P?6{PFiF@k
zqq>QqwS5!f!7IIeh;=8q3T3s6t%d2#mJ_-xk>Bex5YZ8fQP@M4`t=cSg~ES1Z^h1i&_jWP$txn)Te1a
z9dj&7!d|lUE?4LvrCn&XEty#26#C<+V=Fk&jv@_uq^xPO8Y1^N-LzdMnta71OxpJ4
zx_!B;G{je}iWM@Xt`?d{BOF0&_&QbM-H8f0X8{}5n?+)_=@bT=@3uGgNPZhVo_WHw
zn>jc{qOm;{`ztk)XVOW@$5WG44d;y*YatT~;s4evJZc#?K>~bHwRbO4te&ET#TNV!
ziZ2vzb4Bg1=ld~+weGoH3xCG$!3w$d&+HVes;kTD*l)x@n+2RsvK>EW(!|Pjo9ZcL
zzIR*uJ-Z;@DF0LR{C>QJG`lnDz4o?T_;o@^#-jhr6mN0bxg`Tr-7Qn_uXVFpJd@iND)@0s%9Un{ai^1agNTYs~rik*E|s+^qB>c9TE
z8ZzER&qXj5HMN6)A63Go=$>|$Y0ql06zj)nD%}swFuvb1`x-iBHvceV;IS9mgJZO{w??jPF`F$#Mlj%~-4KbA7r%aE1>W%7FEd{mzA99Y^O
z(_8*}5y_*k>~B3#(T6k1toKsu@|bW(;=jwZIv9`t&JEtac}#%iv2P{ez*N2qkWt&g
z8Y5p#lqk|Pi2Ye=SEMr|v&(J`3-SO=*%#bzE-T0e)^G4>C;lhJ83wEwDY4qWcCwL!
zq;dE;xlD1o(owu7wOyp~(ziVm{e$e)Owb^~PbOMb$*!mkvfimx`TkT~c30fvN!6YM
zlsn0FD-KvBp(PwlY(JP^XXmGCPhQD8-^%y?A+waoP}|vIW;^nCs*&UfD4x4(P(FN*
zb5W_RuFr8OTyjjh^^k!>&Io*#Vk9{PSI@eE#!vF3Qv`g2Of3#xArnnz7J*)DdB!)QY`k*}m`
z({Fk&ls|2FFrzgkgc~b=Cii<0hoB8@5H~>{ffpB?*dZqtbZsz}4D;I6YnCR4{I!wY7IdO59|)cQhE{j7caH0}n?_3ArNLoky`YvaE0
zU>q(Jv((hY1HGMc__}Aorr{ZJnr=)CilgLEt$)`0?A(ohy~@VJ^1G7H{lV}oT78p=
zh2_K4lIe~lp`+6*A|jF;|^XosQy<$EkX#QK`Fw*
z;U~A&?Sc({vCmivpyjlVCs+LsI^lMaaXO$JLiQywFPI?iOx`C`^T%!y;w6V0X=f
zA~oUr$7Hr?mB#8lWizYJUufM^eG`MZl}37n@c~)p$786(mNBTLo6+w|T*Q1oR~@{V
z>%kCOv-GIf3g@Pk{`}`Eu@W^mKL#8WvEX`OsqwJhvb`cU3y_S(cKXr-FJv*LO}lkC
zGk7Lmsrzwc!3INUFOE@^-usc}lz-Fwg#T-am=Ksu0JSubsC}|@f_@%D6Kc@8l0kNU
z#mdARDV-%h`m0^P_7IuvI@7UfHDEkIS4+{KEB?JXBjwf_T+OKr+5oFxJki)f#8D;l
z-R%q1k>=+{OB-G5zmr$>Wlrn!VvgGMRdX(GC`fM;NA5EPkPZhlHt*{L@u~0BK=n@l
zG+coXIW2Fc{z#X45i&YygzkpZITgl=WtCtaCDQ+stmQ@osH!EsvIe;)J@a&lJ70@UBU00_bk3b2
zqe%Z$Q%F3nmV!O8Z$gGhiw1(Dg1)&78ZM>Z33xT8H1$`07#`Av{HABswYhMaWMCwK
zw&U~jNRkrOmv~oU;r=CZ9RoC<)ZbZKG?r*5_QjQMu+i^QDhFy#;mqUKzx9wrIo{k6
zu4p;ID&K{|MCo%gCBvY>5{D`>|1cN+y*krOG_UpI%2}}U#9FC*9aLb^8P0z+8+EX>
z7%VADyb`d;K8U@I@Qd9jrhadr-7h;5M`WB^laE1aI@UYJ#<2g=TXuLts*JDN
zRoamv3q;~}pej#J3dkJu+6L4#pV6VmvSw<
zh5p^vCLV!_Bc;<^`OEZzdH=nRyFv>q7HeH~#Y>ldhf`x!hzPh2T9%KX;-v|%&rZgj
z=wN1y%1xQLu1WdKHGkUQO)e_Mw1fi2`S=wmNtXG*fAv<2kGE!j#+aD@-Xv2L>&fK?
zJsyuCBf{OAuFI$JrWIfpR~$dcW;1C^SX-o5{La$@N7r8*j@0z=q=n|EKKxiyZ~+Yg
z3wf-8R5p?LcnKdOowSr9HOTCDR;kg*+_K6F6P(Ix{a%5j5y(&=cPZd1~1P9GI?OT);^e#uJXG_g5
zb)65t_-Aj|E}`|bo_?3@DkK{^FKZa`;$gGt1f}&TB@=j!;Sn{V^TY0M^W-Nu7JlM#
zwYRryfXO)fuOla4YEM=esP)G*rkh|zQu=@}ON)RS=JP&>Xu=B2m6Arq3sNWNBog4i
zrP5CPPf49$BPsHW6Af8wn!Bd}6cO?uvN2bwq|kX@>N0JuV@ZT5wWmv(&DQ{w4D3RR
zH!6nGlvKeSfV44woe-vuAIZ|zWB$V7fxcK*tl(haP}#Ib!>N_siUgXSRFBJ^3%$%`
z+Dm(3!heFPIkEo7!yTQnNFh^>37*&n?F%nL9--k9IrWkbWaov&D4a=p#_LRt{GQNb
zjNyziao|3(T+O%Zk6bKWx26{|ZYppj#aH}{yh{F*Qx8TQ3tJ?9Ob*~8BZEn_pEcJ#
zs_-4w_-g&uYy(0u%nN72_(Gsr(;##__nt%8u4EKWd^EZQ#F>k(93rAc-|k{^&=!i_
zafDaE!So;~AQ&4xWu>{Rb2@{|An(dnpg~J5VRYe0o00M#ton`--k-!lqgM9?M?<-A
z@xifEmM!vS|K)4%{$$WHjaX@zD|4ly_8vEob0)YQcA><~+}u}~GV8zc4;VLuxW=2;
z_LqfIotsZm@}2_{TOw14i=?KB)T}Z8O=#%;ED9-GVE^P0?yn;3AQInqa&;AK9Tw~{
zCR9sX$-kB#i(l3_v1Pi|NOO=>l4={JSu>nV$5${9AaE!u<4}d!#@lu6
zzP4cZ;W|$5d2=RZFcT<(6Qp
zx_~s#7R%3tT0-(t4Gq$KyS>H38ryqZU>ikx`q(@ZGp#$(|6^XAELz1!JN
z@s#SC?0YY1e$ArGcVvmQXvDvr{mO2vi{;Wgjt=OHg^+@=Cr?k%4utsiCc4BAYs{rY
zq3d=wG#FQjT%Q+{{_B&aVuRA#B#ZsDUcrBC)-&#Pm@AKYm&nYjb9^vrQ&f&^n6t=4
zIb-;vT0Wu4<>jaK+
z>xzAE{M92lYtm^|&KTDQF*P^wpnIB-A_K@1bzu#_O6n%?v{u
zJFd}ONV}mR&*M3JaJ&i2rXkH53sMR1P3vBz+=sHu;wM~P>8E5HkFQz6PtTwJsVaFM
z&7D~mO4SQSY9XDHVT1l3W{$E1(%ZqWtyEN)^GHl$c3JFb`3r5eue}Q$77W4~iUd-Q
zUks(GW0-xIyo|0gD!IR7VVL+qG@{g;9v_TVzw=qDYj>AxFE|O+v~5l0V$7~0EW8(&
zq6P~a3|pF8VUWtA{FhoF?TRSM%kM_0)S38CcF0XfiZ|X5w#Jitis9?S;dHlsKkO&l
zUkvrO)>5zX{`+3Nk@rt
z6?#pC-FdtV$!L#W>NvDrRceL5EqBU@wPW|s9KcV~+_0-iLUSuyWP7VtqV2-+*@Im;
zZ-Q2Z7sZrk-@7dQ)=DSpTDk071-;tMb^Lfc*dw#
z5`~Pi%RwU!vN@jNXA)>%fgfn%>h(@BtO;p9)SE#cjDND0C#6{
z!=@|OG&Ux`$tJXPg@@VONwHEAPXj@~=3m_mjvOIjd~_t>Dt$k^c+$?85Lj4PwO*|^
zhMe;@ao$$*N^kYnigq;h_l(}UX*eBbZ_{uR+H{zJ?2A>-E`&{eNf9@faD_tW4vf_B
zPyM=!#|%u$fk~=F1RR)RA<&mpzTCKWlYKwt7#uqFvEtcpxXm7F|JS&jF8Mg!%t)}e
z!UF+iW&K+*K64~J`BYR?AtZsVyN{#E@X)@?vhR({w+|!>O*5j;R=+$?s&b$HdCwaR
zD?Ym5qaH{}a+$c%wJ+l8D7&H&HnJq*5ZQ5Mbb>TydRY>DaJ8XR9vRtw`|gzz`~O`~
zivKX}GQh%1^E6HYlv7kzR^rorLYqs7zH>2lo|5%&e3+g`uQ?AGcF>28IbVk$bq>sV
z0maq6*btn<9UIl@Sw!uZnGWhNmt9}O{Z?R-S}T1_DnCv%$gjxOK;cZOz*Y|->K_~q
zS(#7NF>ZaE^o)n$cGm}@yu4hqD;89$X`&54rICATz-)Ed{+C;*G(Y$=G%9(uASCeh9Z&4rou?7!M2KvQN*rGSe*-d
z;SZ(X01-t7i#^m6#>D1ko6(CU*~Q@hA4M$Z;4Ar?GR+lf=nKd@nfR
zR&Dqi*1t$Mb>O7cs>-wyd2hG;T-B*nO=x02$dYw-4`ALpbLN=Yxu=2h#-vT%qHHza
z*g8SGJf28f^?1o@u=_)POe96ULOSY#$g;D5+w9HjIL1ICwyggh1GW0s)Q}DJE|I6|
zs8(Oz0q#3GlY>Cku?Vqs*whk{JnZGiz;;ESwh&|mZ%gl~*u!n1d!`Ee=5~Sy()U{6
z{Q{WE?xOaI0;azlS~aIL*d87X?OX?^KYzFx_s`zlzU#QOxe|>iEW_xUE-a&3^NRy@
zCK=Ap67Z^y8Qoimq{n8aPGv3Pc_D9ePcM^AAst1P`et0uG358S0-=N7uB%GsfA*8L
zy0%T&+2Z-Bte_xgpilxeJwfGL{9+b(fQ?x_=EuVlyWG|oAUJgcxj<)R$BQI1?b~GP
zqmLo|AE^5^;9Gx?_aJ>N=G%D_-=sKPe4tvyjoN6MGCXwCgx-e}6GdjIn9*E7gCKH{igU
zbu~qZi1f*5yqAImpdRgbBYMda{v?y1x&#vDJAc0)n5u1%Q8=<7QM1)YAx|Q3R<4T=
zU?z+#W?RLB3QA`w3zdNX$wNPtoQQ&We){6}z`KdspL%y*aNfwoYjlFxxegiK?{?|~
zMMkZSWH~HpV=@2T16v?yvL`T6o>M`9Jz!cpJ?sM`5jshEKHwQd+VetXP#F}Hq77dZeY_6gl0X9ba)1V5)
zod;i{XZ0t=EBZLq{DiChb4ibXZvMT0dx^>*dV^H!cneV&w;`$Rk?~=&c59^a~Fx
z_{fj>k#4eIS+EP6w
zdplX>d5u#32udY?$lrrA=srOguVTaZPeAz;yl`e0Hgo3QoY)u-<*
zt{E$RNjyA~3GAMzM0(`ZC$l?4(Z%b}
zn2o31n>~rv>TULCeTjsF7EcCvpc~2v{#Of!^YY#^CA{5F?&W*(uU|p2!r7fS6Qh@I
ze_-!t^J+>r=jPq8?-;V3dhsc)L^}THsP+ngIyV0y>M&ciff757_i0lJ
zUIwyqv@4Vi51|c-bc&^1yhkQ>mgrFj5o~^VC_|*rIYFabO6KJ39Liz){d9j0D~cxV
z)zy3Lw-!(P4BDuSLGRPOcFoEMjlcQ-@*6Ct>pk`(Lh!*a68u#yIkpVMTlw-uQs4TGO
zs&bSbyQYf%{YSim8_e_x7Eb0_mV}(8s>@v?>#{tp2G#cTyO-9&)u#P%AajGtBA~H6
zIfg$usVi8(eK#19Y&?=CuCMszd_`1qrrad@a)fnvt^zL&Zq-e{ytfh!8#2;E>&D}I
z@6N=-)`N>#j7_jzoS)8@%i8j>{k$#ti70awnvvx@#iK>P;I~y)
z82aVeprR$)Y@H*TT>7V`(JVQcVja=boe2bh%{gdI@oyIyP@p1ukD2h^@6oTG27rL$
zh{KDE4v@$)-$cmfCy+l|9yS_6Pj~ipCU?*iDYv
z=TI=HK7>tk?lph?GcvQSSvhQ(oBK)XhN$|J<3VGr$D^wltEpHtyWZ5+VfyB*cmI?1
zSx)0HfUnFs>bmg#FNc5;>`6p}mx)89&+rmJiiBcev!5_riMhT&Ph!!BCM{cUT%y*A`$B`
zki!<)w
z%>a}(f{e#SGvDhCjvv6P`w>j&{sSMe94X}H?qc8oKu||~y$4xI0Eg5oPQqm_%d&h5Ivn(Qs?yUdE_NXLpwklqIf0ztD+366s+DGZ
ztBjxBPuqy)J@MX$WxY
zbiWGK2yWDHz4L@YS{>S%AfjY~gN*e4F?$e6PpVX+*9L$Cn;uM$Ryu?M`k73*vC?4>
z9n#Y~J@Oep9E3gPEArJewWq82P>7cc5BC{%)&L7R0Mu%xxIWyo+U{d#D-^tu$M?PE
zhC;sq{b{%LFH`lvO@-)Ct1cDMS<*P7@vKGH12xV!%-QmJ+1TM3;!$rb*L$15M^fjM
z3}Zd5z>5fT0^`Ha-78h}O8V)ue^?k-I9*wQkoOT351q*nSu%+`5J02)5~|Z6{(Y;P
znH%qDrc6Wopokm>Y3h7I5USA5bD0WRQkP=ouLl#~?U^t`g%=x+l!
zztXjK=5kZSl24xu69sel@Sav409p%NeaTWPAvs>_$=ICZ2Ih^1K9uj@2f8;N2m$VH
zK-0)ld%F2D1?CZi94G@2xji4xSMx`qlqcJ;egThqn{pS@{=~O#gC0L3B2cQWw%ma~
zKqRV+?7Q&gunh$KMoCK|!4TmBo@Zr@1TY?~kQNQzm&mtf(3iH=VD}~)TzSB#
zIOz2_{Wd3H9f3_L!^>BI`){Lu;QuDTH}%APYk~Z?)&opQ00}Tdn}OPjkgEoo=+Uf-
zVI)zjluNV57S63b+#R@H0%XTY+|+h_YY;7(}2@VYu{(e-O_R
zMdDfMyE7yq-nE*Ox>ArHFJVnR>*)p#JV2u%0pSWc8|>xuXSfb}dgTjFQs;CQJ2nD!{I
z|L8J0U*4r=&mygsmt}PKfR<%+>J{JB;#tj`bzFS^-HFF_1OV&(@9t0`x&h^#llO?E
z__E=!DM6!yLfO);nYE%~S|aml`zDp6>7OM>CtmPGjQD7lj`EOQ=Y%f7lul0Rf>b
zgj=^8d2(h(8vyVfo}HDI?u)rTNdN>NKGdTHf^nyn`1kg9m_kpro=EDZekCxmAP7L)
ze+3}dBC(8F=JIT4PmBl!0WW|LkmECD)8Nhz7J`9&14M&2R8Rph^Jc(3g5Sxr5`H2n
zU|h&YWdOB3oc^@!njX(oMvGEd=l2rrU
zu<3K!tnW>2t*wO~A=v@w=HQ_7R?2@ryd?qjdWicGVffqUGHy2qG)gc4y7B=YlPr09~rgM)9(PmO9>E}
z3_k-baYeRo`VDk|TKs;<F%K80`)fSWO+2rFf5oiu6BIzG
z^#NeZKdP4P#o0N&2fiit)3uRzIO{=Dtg;|J-lWdWNuuy*hcdc;`DC9?@3BX8y05sQ!cGNUOalyrchI_a|m9J<>
z_?%&2B?@5H5x$I|P_Xwf+U7=_iPeo2I7NKA)W4CcQki;`%)A6`>QE@S`TL51)|H*{#SXeUpj~X?*YRZm@)Vs=!?yUmw?N
z`=$R!Z6KD35|vgxpeKs9WqKNO7bbD*0K}~Zf4CyQ!#??_?!3Tt@N|-v4&!{<=Vt_s
zR#3(e_CQ1r4mEemu-Pnri$~Ec?ycP(1Icq13_w4sseOy&g7^rMUac1(+{?&RYs4=I
z=+pe>k+2C>rKO?oCF2!`7LM|3z$PaH)7=fb8_#BWxIf47Yp%|6Bj^L-<>gUIm(~Ah
zNw<(kB}w32S4#Ca=OFqaMzdRIFT3i$xB&p=p9f;T>-5Dkb%vAwDz*VmIm_(aoJ@{9
z^)InWX-3tpZCd_DV0v~72NVHLF$8cMR=z!IAvfs&;LH87JLVUiu(w6IN!DulK`xh0
zMuDZGq9PP%d$F(PnnI>l&DfvF<)19%zYG-a{qXAJ2mK?A{HTu_iRym1`tP0LCF9jx
z?~MVb6?(qj8NP1{Q>7HVyH%L|0s{WTWzB48unlBPM3WT{lpNXE1N;RYs73~68kz-D_E
z6cn@ti?WkF36%#LRBM)zU5o)n#@{d1h(JK+cy7EV7#MLBONYm)!c~$|2C!BQhk#G;
ziSY1nNucn>3V_Z2%x>8Y81XrXW9VZv2!RU3KPAM!aNJC&8Js`a7)a6vZf>^)H+TcX
zj`=8BjnTr%Vs886RaU1P%{z03o&N3Ki1I{djIH*@smF?&af{$(cJmbkxclv
zX~IJ0T;To_@VP^ECns)22>y>=Tt!6ejtyy1S4RDNg93^tNMxv#oLn
zlWG*{!3#_i5~VyVhue=Ob8wPjdBNG7vlY!2jd25&NOP(6{1@xq@cV|)EmWPXc
zYvJ>%+`j+WOSc7Y-qZz!veCcfppgZNlamu!=x0R*iybAkY0MfLL39iZhup>64J26SN7r5QND5>}9rDDpBXAZLO3t=DprSp9Qmtm2swrRMq114}L{zaKwtk
zZ@RE3cN7=;N_}1f@iI=9yBktfUg%W;`5IBZSqGw-NTCsQjHp9P~F-mEExC}!&cp;N%TP9
zM;8UbWZIU$WW6tf>2AV!>DugkS1AC4>1`ad>BsfbevyM{Y}3v#T7%|GtA=k4P%nMz
z5f>Ng>}1|{SgKR26+n$HGlc{WuyJRaUq^A~)h~umMA5&A+x}^eyEprpr#Aw@P+e{<
zuqT#dV!7%~{drl`*L_-VHg(S}e*sP@?h8C3$z4UWi_@kU@d=r9D}&R`F)XEG9U1$z
zo?|3UL;oV}cFn$Nrq0bURE(|ryM>4PMs{`8YN^-oADm3SB~y<^vl~u?*o;!Dl@>;}
zpBb1>%NPl8ykt?=$dLE88Lh}^Gw8sRV3=t5vCW|1b@P>GPZ<2?qme*TzGoSuvqmZ7
zUgp|Re;H9n$);-s+wJ9+oa|fSW>3lq^YOd46kl>ZaJd`_)vB#18k}C>LCn{&m~4OU
zO{7Oj;;9vQ)8Qe4P5G7$^&!(GexddupaPG(X=BR>Swz5pS>r(C&XuY{OuXP!Me3nL
zBlw#2&efM&4KSUh(p?odc08jq9;>m~>D?91ma4v{i5cz+y+ypbyR5LLJ)}6o->ip~
zu{FaiIg$iToI94bHr9_pOC1zUHEk`%gSS1Ev@-Af{wH*7#7zPBQtAegpwHXaeahv0
zD#y=JfPdLfNcfkHY9qjhv-gjCyA^_u$c#8yZB02{`pu7VJamvE__f7!8$8zYUotQt
zuIuAr}le)Eh@GA@8w?p(009tAini)M})!dmj`WASRs1LfBBe%5r|94<&mk
zYNTrQ@Gb7uT%glAzpzQH^vIjBZER|4vS26Wl7T5S4QTd+BM_i)48R+e+p|FY2Ia=%
z+1R5{Fr>Yy7;n*#Zu5nnI9_BC=GE>3n*T;*0m*_=xRDtUGs5d8jzT
zc`qgComMmHxRo&;!xRtZ#}Ip3@6$CU3
z+Cd=W*?E;cYWTza7s-PgDs>9%IhvNfzIXZ;vBo?F_@2%1)8=Hv=(_{v)7Vm&3eyGo
zdvvO7%cV+&i-QH>u=k?URiDgr<53u0p_kBUFs@AnHBU#`(3y1KT*e|#`~Gm-A4#G7
z7?kjAz?TV$!y)_!gLpPRVi2O;**+1)r|F-s*iN*s7IK=0I(0LFh1T4H@sU#yPa@b#
z!hduVm)E^8zH6J?YihLyBC#Q#sa8|KfE?STM)UZ%QmDfq^7|4Aw|=#nyZUyjR6Q^{
zg|2~;=(bG`(%^e)Cb+guSo&O#|Z-2SX$L{wt^E@N88q^<~F>&==9B}imq#m
z4#CmSE_Nykb!V8@CeIbdai({aYiy|5D-68UxcrHNz-ygvo$fq1FR<9%9^(wpv|srD
zY8)UqJl=yw#wvN2LLwm?h{@!V|ACzPQSuvUc^~s^&jhph5kVrN5Ryg!Q~%UF(EqCe
zwRed{Pn&n;!z~Wxh6K`l2HE6FgVO=A-zRo=>pTnmDCfpE_NLtSKXp4yQ>-C4T1kx#
zLn7xC@U)wjR9Ia%UdYb0^7q`5%P>kbPOO2Vw^6$q6iY$RPri3vjaCeIEUHaMarHg`
zP}VP1^oR1)FGA;GR*HerI+nwvo97L;*qw=lG+y!|E3}dvOZBg3Ptc?C9q=ty3#hg_|BVwP#|LuTfcO$PG_JbP_IXs|
zV?CJ~wB`0UlTG`U%NG3$ze$hv#k~!jRa)j+uuQ!B4s#UnI@gyyV7X$KdUd6&Fu8=2
zQkzxu-Z1Y7|Nb0D@NF;|CzgdpvDHOEkx37d?0#sd5d*0YLI)5tTb#z`C5~t#@!!hy85V%z&kV328S;pyI
za75XtO(K+#r7WdccTWbRiQN_#a4PclFAsXDIUVnCr*PhWH5)S)!M?b8(X-zZ^mVT*
z?l4+WT&r1E|DZnE%;EBux_DA0MK`pRY&C3@N-yI(C;x~isd?W?I%@#4f7ts+HcaBj
zZ%#54nf}}HZ}dZ<)Wq@4eM2RJG4So6{usSSY8JMp@u59*irB|3(8MBana$>R-mJql
zs@%nVtG{g$WZz8;CDv9J7dFEY&U
zmW*m|PTMr<^~C#^HUsrDqZrs6qAZ$`EQXcs*I{d2Owo4;vZKfiwpT^=FUZ5YYl&6}
zv)lG
zE{ZdRu0EpMa_L&)QabK6CF*W*l3bUtU(u+Ot>Rrz1{zay(g==P@UEp-e35w#$*73>bII)BtFrso|ebVC&p|g
zj9*@7o>3JDa0*!n<=ZT%Ni`SJ<=^42C9VRFemeg7=sAqeu7x>$j+|ATuf@+1u
zh$**D-Nvx6ai5LOVM22`l4|oM0AmoRV<28Qa2=;_{sDy<^sByaYqmNDJP59P9RtE3K6NzRP?lsa9vqx)um#W{#9}R&1Oq)zG9o
zWd9=D4TE~Ov%y!3E;FRrt{z{1q^L2f5uJduVN=*1ur+eq*5mtE>`mbIV;BAn>}Z+<
zAASG6fqdib~vLWY6
zwip@Txr9$Uk))Y_zIKSPBAE{Nx!f34+pWEY$sLUsi|l%rlR@gg+Zk;+h&4mQm3)gN
z6Pg=WgJRp-d&g3=w*0)-;X#R9dLk^XqR8pDclDA=zV?dejgKI$W7{wZZ;ku82qp_A
zcUaqLrU~w(-8}?(EDvE$9?HuosptB#hs8p6aVVgarI1TcZYH-=B90ZcT|pdr#%|~f
z1J=MoN2{88r7-IC74}Ms3yli$ZFZ)b^E&Q5mopmLe(mSV&&k{q-Jt>f8oxKDdp2c9
zn@r7D`n;tw*)EWkbV&u&WZ_E2E(-f5%px{*u-~-{!cT;gMf5YCpNV$$D&i1$9wQa#
ziy-l^zl;(O+$DdFoeQDWcS=E2uHNC}XMh-J#>6t8QNWXZwaj-?8(g`CdO#cSPlugVdfKN?ziJyh3;vzp{`T9Z7S0^&Y}P1SA`
z+TH*+!1GAe;*=Rk$*`jOFl5V7+yI>~^37NYX0AzM-BsrLX88e!uJggw0hQ;L$L29(9zh2GvEpbC@Xe5{HSlt?zdm-i0B4yr{d>p
z)ca73#8;j(q?LPbmB)US`F^P-LQLqBvy-6nEQ(>B0rQ)qvG|flc=wI>o}7~1CT-(AGDX?s8s48
zPDlcI4TT*G*KBiepBHH!ld;ZZw0fD;&^rykvYS2pC^nE59t!ZQXo@1eAn;|y0J?1&NF@2?@V>sEWpT(K-V+Bog**9YDK|;9)8k72+
zh3$v(P(pju{g9ObetU~QN_E>VD=?XO#((M_Hepx0A~8xLZv#Q*>yJ|xAv#ZH6edx{
zgzYL+?!76;-ZwvwVIlb6Dskx|iSOw_{?A+MU555ey}+UKJ*0itbe*ODN5=fpOzDP*
z^zQ4aFg@SJzB_@EtP}mqFXH7qgv)L0E
zD<-~Ws@YhicS_$#4HVbAmN_^)GY`#OQ<|@P_yJpjnoakhht*R&h9y(Vi_4g$U3l;O
zu=E_$b~mtRf~)VmMtWY=M{Ixy8Ua`E{?=OrME{yWzJJ|?PJ
z`}8{PTI%luwC!BDTz6o-a@iyd2A0BEgNj*c
z5~juxrd@;WwNRXCE;$9RSC6F0y%S{9Exgt(&xtB53FAq7HtV7e$c^?S(}q*N+Hf=+
zUq8DjnDEnPbv}WOH(CB|9Eq&6wwH{0>YP~1Ul7q@KJ1@(;EUFPopzQLE;K|&L(sOn
zw1W`xGdBx>+`UXEB$rshlqZ6otMN*(&uvU7yyD#Fc6>fD>e~eOz0YEL7Crr-(nCTg
z>rSd8*$0EEcze4a-j*&1xxFN|icr^mvimrWdLpUTAtSd6I+V%w_;TTZkz?Zy>bXJ^
z|D2Pg;-|Ya!rhY$(SXoYt01%yGbt#>c4ttkI+M6ij1P%<@FFP#j&$Gnm(tDWJ6KVS
z{_gF(4QFJVh~lX~!dC%4A_wezp$Ml9_X}GpV&v-H*Lc`5H5-1INh;4yES-CWINhES*Zow(e-Rg1*sL(1NP>8l}m
zz&_(g6KCynoaDQeLB^xTv5#Pla+81v5@1}$fAv9KrRO*ar0SEQhZzX>n
zmvfLWov^pJAj^)HG;_d{d&((y;2ZgNl1Yg<%MI(_=@vp(=#}ea#uCGhOIOi6U`}sW
z-22I%P7nwgH4JAp!YP&qxZIx;&AVI%FmMkHMN<{UyiK#J7GxEP99H$xT__{{Ru)DzIbM^WDNQ}lo(cR?
z(^+#JzG3@7;{SX=xIjkh+B|Lbd(47lU>+=alkFg(Cpp_kMoQ;ODV?(Xu-bMJfKe;|{Y&zW=f?7hBg(bze$(X@o4G0JT5C{}|HKr7B3w3N=<
z0u4OkQs*`t@_t{K1sMR&1pTX#Sz~oZh~f=@*zZ|YGawi$oWDrLJ8?Q+m`EtTxAc;%VBc
znEId&$_gwV?ws#f{9pT1iGmfg?U>@3<1ykAt$r+{eOkg)`wThdMWJTW5gByHQmnSK
z8PlrADwE->VXMnAdhRLCFZdS;i<=d&6w!sOklWjl2=(OGA~P?y>H!_&f5@rjY?A*aD0U3ZyCH7htq)Bw_V)oKda4O3ycvj
z5LX-GVsdJwzVTS^`bbp^b$BRunxme3Sfb_M{yQ8o^cfn%(mzV;$`kgp+5k!HOp;gG
z_kL2L1$lq$=dwU26mXWBinBY{cy!-7fj{B*)Nyj&_akpQa)U*ojWxRO&MjIigC$g6
zGQ7I*me}6$4I*-<%jY-iuzRFyK_w2zf3FbF_H2$9ZuMS1llU=mKnma^`^1Q^YB4QG
zyG5VC6hxy)78lFK>3|U!R(-PKgX^T_A~v!kIn}5Pa$UXpSzkb8_rAIP<9X9@pIEJ8
zRW$4tuSh#zK(Tf~|H*&%!DH2QPG-=%mnC9?V7BPbQo^VK{I2z)wXYO`@caflcKHu9
z$jZT51Kvb4@Du%{X6(2AQE-0g3&ZGon^*Xv^qU`mQ3g8V{ly6LFij>B$$+=<16Ok}
zTpIB_Ht(RZ+fyrEpARQdJw6Es*~;SM_wSn??vvR@MXNRJgpOFsb?hWU-fYk77DwEF
zupTSd+X6uqww>#8ceEy38Y9@uf37{a=}ZJ5E*l4uC;Sfn7el!<1Cdx5;#{xg
zf_T&TnAT5x8hD)+gL-07wf7gSGU=UPL(w&QRrI`gJAQGO0V|^r8x5G5ncH07WN7op
z{VozSO6Ar)|4v9p(ji9QG+cHe&iISGi|k0Cb&bpi7&i-8_VTI*Qqx9^Mv~%_ZZX^H
zGB5(ne*MhV^hj`sXJ=UzUk(iZ7Kq*Z9OJMOHr>hy2wRp%hLk;@njN%X!L>
zv7ow_J|Y}GUT?eT?wARxwXL~6ps$XaiiW{evm|4~B*xXGJ&&tR>&0_rIjH5!nDu+8
zTZRWNYkR!}DaET#nnY2E+Oj3lD9wYLvZwF$TEb21c%lWo8Gh%s*v1Msl4{Bceb{=HdbMx^6{_0=O
zU%eBIe#LO1Cms|b*SBA%JahXl62=QDqv2u@y92rM;SRJK6Z62@akP1BPZ?dMb690F^7?Qpt{79sNj^F-vrLdbt4>OUH
zp>;O!8X%&owFr6N>e<_i=O+0^0pALH;3ZN<)kemy+rt<(&H+AA<)@D1$_}syhd^U#
zA@^2{}frriqbzOuu+p9-^=s-!XAr|IMfa<8+`0l+m0FH3o6gnMIZb`FPpZ*
z-llT&JczwPFTbXGddFZ<^%g?J7^_6FPPzNcmeXP(=5-`Gf5cFBGRb`xeMGL^R$E2s
zagQEUV-;2VocgnSBpjv;E<;@igA9YItcdBC)E-9OcQYw#wT;RA)emEpx?EmlR^O4%
zh}4O*?K3bfFZIV4JxbOUKNm(An`DeoQL#k}cWeuth5RsSs2deS!i|}QbqF06ua1uI
z>z|=g@nU-$u_8dIJu+3Xwr_17AncOD*VfT0`ur8N-69jg9jeNaKdUFIwr52-5&pd`
zq>FdpSRNVD2%!wg$|&;0Ebhjshx7Q<9#B+2x33z|n{!eQVn=g!cKT
zrrU`-FI%1Kcsx>(P}#$=wTP9)Gx+uF42rqmIh@hn`H{}?H=0|;)YLrzHk3ekPg!tI
zKXwN5eJe9e-AR-L#LTo6j>qW;Q!0_6_~Xk~Psf!73@WIsZRXVNuFDJ`C%p70t
zhOJquRIQZ~3C;#Z%$N9}0v*nw@C!1Q5)+hW+=gw{qUxMcR=crjdE<7u6^re)m^
z+9sn5f2h_?qoIFRuUsc}t~mNd*^!>1!HZ@=)S=1XEzL?)6G=euvdMKS#if{7{x_S}
zvnhyMRB=5r`!l&5{yEO_CONBSg2ygQKNGoP&-9)v48
z|HNQ#IM`D}Tbq#oM!giumIdNZn6QWmjSO}9Ea3h0H~!wF+K!T!PS7ryVGb#d^5x^j
zgCzzEaA-f03jTVB0{%B$PSTSvSH$tgEU!z`xaQY_>SDDH%!X4y$bH>+jantV@{msQ
znJxysdSb_k)xCXb-(!5{buCxnw}YL1EflZJ~{7BdYB78EmArPE}y%G
zkj+?&j<{jNeE6{fq-$Kw)yasaoykdP^S)OmhMQM}bOd6xB57f92n3r+1m*N{)0*%L
zRi+WZo>S-_Uoo25(uRAgrJn!p3}~R#IcR>Vt!ZMGH_CwIWF6UtejSeS*l_Rn(MzLI
zmyw*jO8M7e?4HnHr7$dny`ve)3<-+VNOaesCf<|S@`CduLUYFzOiB9dw
zgzq;)!OLeew~0|hBKp)sthTPM(<>CXM6}}x8oG=()tZvkFWz{xU5=fZ&W0j#bf6zU&>G<|3@LQVTKlRQ;Q|x20dLquYX=?bwLd^
zgnv=)sUu}L!HpAo92S0{)f^h3}HGJ{euZ}vu
z_aF6=`Kk7-XBelz6dM_P*z|f{`;4u{=LayAcFxqR!yIw&d-rUhO$*+1PJ20w&
zeqtskjQON2H`fRaVW~ViUN0!;$>dau|LcPmE9GG_m`*H8ig+Y(ldvWl>C-M4OtO!{M?xmiDBSfe=ED$nm##h(N%
zW)Ev$Kn_U*@fw_Ap=5>(RMwgt`t#iRtL+2RLaHn4xw0HSf%5)d{ct
zNAU?f=WN%w{3mLq50^2;a(k$Ue@xQu=ti(nm261(95turi~N59&4xE==vNSA-wPhR
z?%x0v9_yE*#kRokGDWsd=&vb
zl2&;vt7`jmhmZpX<5%TT^PU^#vcF~qpkoTIFN!s47?{k_QQ&~-+nH)Cbn2_hu9=ae
zD4a_7ER~oSZ@HSqlIKI!UbzJ4_8_>WsE1huwSihYA?4_N0R3F0`-+0p-b-`sGbbcp
z)#xJF4W5OK&AUig^!@feKLJCtb))k1$T3v;!?1PyliDN9IvOCYx
zq9GBlyGh{7Q#pUkNcRLqpmZVB=_1uM;;DQC{%>HMxZfzK=eM3ab%|D+vy7DMai9;Z
zrt4sg9@&&|fSzQf{5A`78ADD(RJx!%5-vzF-(Fknr!jc)p-n69E(C)fn{V1*+)(Ld
zQl||a5F9B|?~p&n7qiRv!1pEmJ$>K-ZWnyM*jFdbN)y{V#j@@hQ9Cn9b{PAJB`xvW
z-|FXtL*zrJrp1_)g6!HjSz|0`roeAoN8yZGUVCl3|xgr<(Vs2zrj{FPR
ziqVEa-JH}Ht1=~oO*~E(@HPAj5IR7`bxR&_+E@n$4sO`5|4VW;e&|Nj#3ojEI8Jtz
z3KC+l(jA}QT5@cH)ihtXD=nXX7+t}2t(JE^3O3Wrj+bYr%`D0>!C?6aLn)UOLsW0uQ*%$Z49n;fOTwn&M}KZ8wrjXL43a<4T_vZRjqx
zEt9cShvF%EOT)lQPD53G8;#`Os<2NRxtWonNM9U>Yj>NdMCkJr
ze={u)!=yGJtU9kNObCFY4Y|K>-f5P}YACYEA|cIG9t7f`+@J>#KXE)R0h~taxaRwe
zY^0La+j$k!^2nIn3)C!Rkv_D~d*?cC!iVhi8BD4{Ub3Dd^&;tKY(8LDBmtT9W=3PE
z7DG@4Q5X%A+*>bKHi-bq|<7xfbfJa^XsZj1Jcx#}y
zZ(paqF%q$C$yE2Wc;v;(h+Wn15qKw}dSkM+wBH|v&iEv2On$
zWD`+xAU}uVUy--d`D?-y7JcuC3CAj1FsVp`tx!6g0dJZv
zFB
z)4Ex6;v1h1W+FYlh1jm?BiUDh)b7B<(X&MgK?mB2bShEXZ3
zz6vf5`X!&9>$>5F6Uf{AdmKYuppB(T96@W=qM*iP68up!Os%gZLVK#Z`HP3Z|uV+;z3s!c;x&F>+P$Gd+1@QVeK*SgfwBD7
z1AX6pYS!n%X6d0C5(8|%p+4Agxds@FhS+td+n5>_5=C0Z1p_L6Qxhh~h}<;BCgfQz
zl4xJdYX-B>()4~N@2_HyIdUD|kKefYmD8&@9RgMwM#A#fUf9E)0W5(6{5&SSvbScW
zg}Dr4K*9Tej4~w|a_Q_uk)!JLR(tjLlw$3=FxNUGCEvU#sl7b_GLw|_^8O)pA_*i)bXZxZ
zKBq~=K}kb*<|z&3)*M);l=@y^V36v(JcHTPo
zm((Ip!za~If%jIv-EPCwURF_Q4$o1}-M&U*Mbyi)58j6*g3lBaYg{_8-te8DT-O#$
z^q83>4gFlutLl-rN5$2i*ps~M0{*8KZ;mExx@-1lNaJi;^lL>_^5F
zzz;!5bK1XUYbYq`GJXFL?bhrXXv8hB8eM6L__A5|`-FMCE?ixPt@#tBIOFn;TH*2M
z1Q}Rp*D|oh)~Y8ust*^h+=*0GM}|`pxNIL@tBVG+1%bQqD}qmhCyP7_I+87emchJP
z<-jYpnicQsi5t1EoiZ&y!YIDj@04|G0mB8|hG9_}LhENfL?PfrasHbtTdjDSRAG#u
z#}Y{}bi>LYT@99xF5ly(L)adm$k|Ena5D6)dS8f6
z?NCS?D?B~YXYp@(WYDsSz6roon|pJLB{k1WM+W{gD6~Ug;@FQI}*VcHh`dd6CVGdd`3gyV-
zBn(`-8!RPD*5Vj!B-4tjUlUvP6Jr{Py$uc5viQW)t_mR2RxI3-t@POXDt
z=s{E*lnBLhBS|W4XD(-e9|dGsV_a@_+lZWITA$W5zeQjlP5@g<#ENxMJ!6^b;M!Nll@hdz9I2eaO
z^hXC)&I7x`BUUnt{F6vdj6EnBQpvZcV0MI!trlJ5;0}Cu2wMu`?P_J@-%H}=@xwF*
zOh%ffGz?3P2b
z=V}r7g33f|dyRBNyo~5F{C4URD0kIMk9F$4O(i{CYbFnvp7`vrkNiCcyXkCVsW-1#a8zK!(dhpM2dZ&Qk`P0!l?+#4Iu#ta70t?D?{`m3vz=8^9M4hL;
zUTJKAVieC=VFXp7jHUe-;&uYO%DD8dT)7)~yE*6ZU3bY+REL!uXj4U#
zEyiFDj6u$1uusOu^@e!;9NI<@jS3(Jpj+yl754ANsg%WAIV9rS`jM_oqhVSVNiIk%
z;((}AWZHg&cQe<=U3f&r6=y{EK}nhH?w~N5VS(
zE{%ne45Cl94w*O>t|*yPHlI)Nw?PIM1t9ayRDXtbST&R@|BpZdBNta+oETU2e&ZHF
z%>{>d!3YrQ>D58&8{$*EgBs!W3Ng@UET5zbE~g%)sj9!=sIfflm|~{9itK)(x6oDg
zWvhJ^ySR)Q*r|%0G5|#g$NapN6SLN5^Kpgc^fgl^_O0Bw%e=hIZy5DNec~vxIW=9*
zm1em#r}^l@-3x5~Nh9Yd8-%|Yh^F+a-K7m_wF}Kb+U-@n8b0(wC5Nh!CMks-5*2?n
z#8o(PDpvdA*BD_L3Mtt#J)rZrcx6rJ13uev=iP;d!9Vc&mS7Yo`Yicte673l$S)FK
z@CgmX8{?9!`yK*8in;Ohk#Vpqw7(~ikcvaJf{0e{vfa$JTL}?&hkUb7j~+|kh89Qb
z7NYxrFCB4bmW4n4tk%7An-ruj+pe-en1j_COCOomln1@^jo?=0Kk4v+o$&
z4i0bEpW)y=8sN;ry=I=g(~I;vI0h!cm0RI4*Atv+&NO1@
zH&X#GR|l|G@T0#e?YH}W#O@3AsdN!?(>lE(H!1AlbB>~5$I;l;7$r^e*!w%w}5jW*ntM~nzNdL#l8xE$tmp$h~MvVf7
zbKZ&g3xVG@ixTz=wIUE+>9=2LdI3{T|7d1QRMhfs&#$bfrkYUN*}yXz!@4O|ne8{W
zn9k#3DCTjcyppHpl{`3KcWHBfwaRYH7MuRFSx&K1;VwAW!q0z
ztiNh{;HOLBVdq;7xg8NHP1mU25?pO_TmS=hucX~kxb;?$zaG=^o`f8)>O!j(RZVEvyGYajjriBiRHM&F5(YxE43jf^D)Ujwy{N9yJEqf}B
z^MWNmKyd?(>ysLp`mjatWg1hFZZTGEzKU!W%t8XV@Xc!$<1luRJp8;VSpxigEax6D
zXebv{4iE|AV@;@?34E>~xpanVjy>X;C07mXMa%txYo1%;)A|AipUp~$D6h;SryXTH!z
z(;q=CpfZrcR_kNG{Yo!TDd?IXx#IeiXFr5zYWbBkkXE~+wyOWSiSAP%>S64w=EWtf
zd81(@ZvA)w=CzB8w^~rcgl%C@8HcX0+Ib)~@jy9Rzbg2yAJO-)o-J!?tJb__2kZbh
zI!i%LSD^Zuro;}L%E%kuW`tTowz6&!pJ^c>L%So%EukHZ+#_%!P;to*mFYLy*Z0p
zZNZwLJgSXO=3b3ql+b8Dum`tYI%`8;J#sA}b%$GCkyh*l%M2W8xqI58No@0E91^}M
zn80OXW~~obpZLEKOF=M!AcnSoRRIU4BQsl3v{O+@l?b23iMzJ)yyncrpx
zNx2X@`A4x1p7aF;5Yr9EFia&y9u@SIRk`Fl+DE3aw4y*Jd#Lr0m6m#_q|^e&})RqD7bsDPOxbGB?f8;+EE
zRK`7G1;v*WzBbdl5450%Meg_L2s$ChE%~>WXWw4Gh$+<09|NOuDvMH`t34tbvnA0=
zIoDtYz(Mws$$m?Ds2RjEAWd~=H;pAx&*%NAa+OsgBAFq8!j%K7{W`KI1lk09_
zpvty9hK=C4p$cqGo^{n!oPz1i^a7b)-WF#$mO}6LLk|^BN%N`TbjVS$^IJ4(WR!x|8hJ-Q>g>
zn<;D==ev<`EWcs$obpRD699cn9ERW?#a!3wfzIs5h>(i@fPwD$gMnKF9FU_mtD4nM
z8(E$!Y4jS#WaR(z0@xm@11L)RjUzD%k+p@BH6?>iFPat>8KX3738$z1J&Y?9Z|#4DE0I*<7IxBeVfdKrr6AY7a^twEI6Q-{YPWf0iQZ9`1E~x_d?7$(P8%eD_Q6y4%8HQ$5F|u@=+6
zm2kg6lSm)I1MM|GYKkCq2AxZ@c#Wld+{Y|UeZ1Vp!xrnDyk#61T}Eg;UlX5Tz{Bzy
zX~fi$p&WJWpw7XNZ>Dv-*(u)Ag%p|IxcNitEzLLP>+h6KWI5^vr8eztp0}xlEl=O_
z!)ujj*bDg`C#hw@5}AzuSfV1%-5T7@Qabc`pZ+HQ2zv#r*W;O+#Q%H^`j+qu^Xc^&
zZPU$Fx;5#q+o9~76BISsO7ARpM&|i8E^k-7;ELvl^KqJo{~=;sgr4Ob!FjaokRlryTe!F4%bpbw_W!?pf?vHEJ|RdU%4a^tLUJzaUg4PdCQEp+6HyyG{6uhoOm#?*gv
zdwcm?MfNV9@VK{RVwHnSOF6cSY~#5FciELGMCmtPY%etX{)JH@tw`K%$3K
z$1PCBMAZ|ajZGUmGqkOl7}^yjRIa(_ny3C>aT$!Zk)*okfL$#Bw>20dqegqhZtKS0
zD@7cTGl>3t$L)-T8!9Z*1VDK&)34TskaL4_@Yg8t{NG!2p*b2t^>N&DVoyXSlc_+joh->?KiIR%v
zbI2{o?%z|rn@@1yOXcZiW%22>Dr#RS2GVdG+{vWxp&1i8H$b_#s&_^R%7g6nAVtmi
z!n7$zeWKDW1J78*ufX7ivFRasA
zi>FWi)!)P$X+Y=@92VaN`r(9r%dI)Wrmd1}9j2iU$^B-69nBZs;wgA^J%8xvg>#T<
zUElsvX1)w6{vG-+zbGIRqkOovz3T17OThJ#*o|l_mxDK{9K85GzaZadrfEElw(Hs`
zT;Q8PQo!%Awrx
zuE%X-S9CuC1}|`;*7sL&xiBvejjz)*_GzI^&ZTjuT5
zx(V;$Or`}U9FoIDXR)msM7G^wR7Yx6Xa3az{)EsV6u`TFo?AH4{l8#gWsN_{YGtct
zSelo3q|w43y#Kz`EtGk4D0n-pB_r*(W8Z_Nke4^6W(uiVYpr)Zb*|g|?!1D%lRCCD
z;q`-?vo(80o$%v%+4Rxk!Li~ZS**Oqj3>f!=AgZQIWolUA>L*ono$|)S;;vkt=
z9q+3RE!#p4i6a8%|Az+3pvcG$rx#YT+RV6}bUL7_iM}eO5*|A<#!a2$d0sDSYk0kU
zLecf|U76VV!&L|vG2>JKvAVp(20*Ic^0?20d;-{ZjO)F<_er))x#t?I&~elwi878DEK0n13ECOH-t6f*rLMpAg}07!IuqHG)s%lsLq2
zaia|spx?9}b~|~{%_w4dg}ZYZe@?HZbS_x>eW}?-9r$FY#p8PQME+#m1Os%4A3^$Z
zjk?;yqcpER9t%eqPTkWWjXGu858qD?WC}pG5%LK4f@JE$(w;2Rp7EE+_X)vxj{fO?
zrs8-bhK6b2sZ|?2Wq!vni`?;bhOBIkWS52R=!-{W+%pfv3VRzD21)$Xxgp=L5U$>P
zwL3bOgn~q!=b=%Vg4yOXg2u%J$hrioeyU@*xPW&c2MGtf2hLy7W*FX!R0em>O*h;|
zedJhbxgU`%TIW^Wr7r{xxZE3v5fn
z^=Pl(`XBRx3J*mih#KZu6KdL57OZ)AihAxRH|j(hrpDSIeoRiZ*|}>%pbIGNwLBmw)qTUgI-IK;R|9Yrl=@oSs63+
zf`Ru&8eS-<=ht-GIB>!QFp1aL*h|rIZ~s~K3Hvhs{v-0X+XJ+d)%(J8)Q54TQKeqD+?Kq(v
zgbOs4Nhlx}O?4T?&o}MPWxX!TF=k!5wA-{WMV381hOuD(h&Ez&W83Z>UAEOM-}XKF
zwbV!rD(b9yc_6x>-Te#2s&f@e;F2`}?InWLyc@c1S4ag
zgnI=sqPeBDz|ALttAmrI9jaEpso`DcwjM7wLYY6w0we8L)85cWuR^8gbKT}*OLRL>
zw)(>SprdG`(Qf4;9ge6@>44|!KV=8pGN(~8#Y8Gg+ByFY1x!&UfbH33tGDnAR~ftO
zU)mGUZTk1TXkzKf@m1*xw$^=KLiS(_H^Ly+P!_Lni(ymxHSYpvW7!
z?((a6>t^SvM>InyGX8wo-vki1foLX^L?#y&)uirU!W8x_)!wC|qOBJw@$7UOWCuRrc|NqTFSt*LPfgEG92YSRBd81`CK+F{ct9hEtEGLZ
z0+n(f!LK7NFKzeAC+(i{yzr+gbO1Wj#@=2WTU7wLBxy$C$ZMI*syA8$Vu9MJgKqWk
zQF~95HB_F*MQ6;e(Eze0Y;z~sw*jak!SD{nW&|c&jH0?_0e2ZRoW`VSFP~$20O-69
z<7rEuq*2}Vks?yCuE2WG=HaX(=Zj7I`wgBo4q=aK&W5Bd$Q>bL-r1E8G1${J)f%t#
zyMFWMKLBgrT|-6y23!vXj8%58fG61}IQPodviP&W&_pUIJ#%mo)vBZ~hWBV7T#*NvGn!ljy_DSY@i8}93#
z4}Vus*7fKJhhQ>y=`LppkY;v8LmzNzmb4xzd8tA|8;$F?uWuNCsYnG}s_LrXUt87G
zWKwpD1wJ
zsA>kY%YGdVMS|F^K35=Sb~PM@VTz2B$UIq)`uAhzq9SY9aLju2>SN@?2w
zu;5ILaf0Xkb#PK8{jmo3j9kv8PBy28RldEts>b_w(tzER;+QLXAd0kcjV}Z#%A^nM
z^RLlFR_G~>teI`fzfaQYyF1v6dX?2^*OVg6Ht=TR4eR5RyD-_9YheE~8Aa7sn>6SG
z?+{rxNM^I8mBMrglqFMYb2JBDcdrb
zgyZx)wi6_Qb|dzg^|Hkp--;Hp
zI6pU8mu(rXO7zTHR7S8JlblQXJ^ny(&l@_{2q=Dp}|{b6!^1>bH24$dQUex
zOc1qsxe-d{P@$9?{_F!a|9VtJ)|TVX21ca
z&l)H&Z|hqOg}`Zy|KzBYpZ{Q$wg5kJYuDH2yP9ojookILsviCbJGm4+Hbxp0Y9^G11wO~BrS52Q
zFS>XTKd!pX$}g1~0ZwX}k25Y;hH=Bir?H!}LN;&H$*38WKihb`FWWS}h0|Eb*moC_
zir2hM8NV$NCD*zW{5MLfn+3LiO=*T{5LKOdDCHC=bh6gK^RZox;sXpmnGnr2$Miq4
zemhq0Jbc{#82Ad@dv}5QPuP&T+z};}Ps}25{@#_t4`wyI{0=7WlqcB=7~RyE0tO
za3@cbud7d?^YmCA@J;JjBrORmFnuTn;5b+hNACd5S+^F!r(5d|)m<~4
z`1tTg^t*fo;g`fmuXo0mmwT~v2=C{U;Y@8y-D_lJnV=NBQX?+SR9wZX1@H~Zn-)ED
zrrn$HMGhizEIsETuqg2gJT)ee&ye3f3uG*!d(bnCi@
z$vPW|8Sg43u+2)H>^s^&(I0#}oKw+*=pRFWgrTuus_BLyROEKOQnNn%=*9-|r4?7q
zJFrU6k^=j)qmS~J3#!N#@M{mtWzn5K&63}>2in`H5~LzfQHWtcG%zww0k^U<4(Ld0
zV>}&kmr&H)kWkCbDE->>arV>!3REyr-fX$(pki$kE_G?vCJV@Y@cWaj^2K
z=*iNe*N68vzRb`gq*EEGT1x+N4}c4}m#-@Z%J%0#N@(KbD4sdvd<#wC@7_&LRr$v}
z&GwxD=X7-6k0E8LMIsrLAeBjei`H6zmt5opnI%_gNzFpQfK$7HN%mVhq|HX@yZ1^G
z#1`{p<(`5l+k%(YJD?y=?;4*3C+j>?^}1k_q54VPdMS*#<1m;$kGGPevrFG9|9f?~
znQUGH+tK{Pf-~mLfqCO>o!CQ{k1!K-r_~JX)s8q_k6c=
zCoVU>$>&={ZFyQjF^gNFygR)5rLmrdMg^|UnTr2ZkKz2w*-CC#!@K<#?pwzCBwh*W
zD%sQ0Bc2N^7`Sr}^VIdW=ZZ+S%Cg4r=yy#m!3gXwdi`$e)D)82f0uT^LayYv$MQm5
zSGg`zxse1d3<5GW@y-7Vy=kG3u3wSl@3JQvqa;3c_V=2RhT0iDww>CU)kLF&&j-vf
zmjLdkW%Wh&P+!p>sA$n&-EL(=-OHnsSSNB#=MNlQOQk>j!e;>QSYEcGmu58Xq`H+|
zXY8786w&m*<5m%(u-|Os^FQ-O(@XE~jFvy7Y)JZpcj6n_*?SD8qxT)ekaorPNBS>S
z^9_qwk13pWPy>Ds4T>q>^Rh3yvUALE2YFIdi!A;Pr7hvO7H%QhA31)!DxvO;eU=$8
zsiHrYGPhOKm_m5pU~x@iW~`|uzl0y&G#{BZPwaGu;4nVRlLdkFUT#MvI4;#jfi%~Q
z8lAAeFuyf$@-8JfFLQr$dG%+luoXxAysSl`{(bxR%oXLV@~x8T-dL5X{a`vCrbAhmLuu+V(~
z8G61{Q(x)cbbXqkf1WU}OWku@=t5Iw8J5rU8m#K_by5Y@o+ojg#hOBCB-;A@EZ_N!
z7*gzBihnDrXZ>{TRw${b+zg`2a5Ur>+f{;L&T=w%N{St#pHDx-asR7MNyTFLwcVwj
z=`-}&dlWdmdnCkxmKADS$wX&pbzAXxPxcYJCK>1$8TpP^`-*>ed1cWr_<7jhTyzk9
z`G+UR&U92UO^~0Yy*C#e&Pdy&F}vFbV{lieVPPdvGs^`e*InfwaNcz;Fy!20Xx{(v
zHyI2##5Sry_naZE&l16&^DV%dM)Z&EB0t!6<{I-{8G2vK){)S<5-N32{;HTsD
z!?HNpdai0#tSyoWIp3?-!tzGkZfU~on}+mgc3YSc3+|^+_*UoBJTNSei%%8d(u+7Q
z-Q$6a&e}NqSgb}FIh-2%{$SgXk-RILOF8gN4uvu^x_xG;jo=nx_c9u){f)GINbe&0
zL0z!JPWisJ7or66v093_M1mYLoC%}z4ywsQTK0@Vvaz_$_OBV4bgzB8%Uc>rB7!HH
zaH%(sc*Hg+3p^kS#2s~cmLj^WfN=!5Y?D>Ge_$1TYN0eY(t3eN_`|_kqhngu_uFCU
zmU4X-MEa7r+|RZmpV3JH)5MmF!vvIfRPX^-qZpZ)k1Eg-*h9m{i-`#03k@gTFWrXi
zy-(+LRi7?var*2nV#h*31dXRX1*ff-7cZo9lt_;KeLdn@m$`MFZ>tgWvIPI~{GXnE
z^DRF`rH7Gvb^hGuTO$K)X|Tad?M=x&Dn$
z(3g?Y=2|6do{f)aY)-QO#B`ne+Q4DY;ytK4rEX#eAx*Q*;bC11o^=rph{dKckH;`_
zS?NCE%HI?pdy>T=$reZESES_MZqz~f>BWt-kcw~cJU7_1LE@LeX_X*5(k}Xg`5Vol
zKCF6c4%9e>f8k4dFiaLoX1l=nvs%8@m%nF=dS>mW@K_i+sW{gd%k`@1bQ&!tng9IE
zEWlHa>Y*Kc%QRrqb=fDP1~CSf*t+v5VB1I2XGn#H8l2f&S*c!vL(k#_-@QRz%6e8w
z@SX`Pnqstn&xH0f^YA9-A9mSKW16+HFXSUvfN2VtApZ}wP_?3n=fsw(K7w<`N;9kR
z$}9(oo+>ly3rAk&g!{OmdZ)SGSSE93bd+YBDb#$r=qp0n7A~uO;z4P6z?Mt>eN}Z8
zqZ+0V$`LvEL?8$cDKST+)Rf2NkZy09SDvSdZ!x?ig_h)LhZbM)LQA%`mNw{Gy<)p{
zQ^@E(DBqRv4i3Xx^8+-U(I0&O-(L#*@x`#bR>YIkqK%y%AU{F)U_bN7L1IHW4I&{f
zRU~P=>Uo@_KZ%DV@2;)PV3$TdW)-AyT8&2p{y(zbF+9%qixzIv*ftv5Nn#_dn;ud%a)hnh(!Bv+sLjt-V$d;-f0wU+3wQxoXylSdxrn
z@xIyt<+zP~-C+f!hx72t5o7ufl`)atiHH*#e(ZX(;(50e_dll}e_OL-jLb2%<{^mvbg6HIyxd(TuDy$w06{
zRhqxRphCZ@ogqRKXFM=kpgk@yrMs4>?J;Tj2Hr(kP^v@PA(
z?TkYB(W;UAc=X2AHksci@l+)F>12eu257ssUX+OgEx3&M5@o}4(p+)
z;WtTc4rk^r>kw46Vy_|9*8lD`_+CzT8ShXvT=g78C)iJH8`G~ByAUXIaOm-HMeK%(
z8Q>x$hd~*UAo1SS6)lQ98xJ=1e889Cqo-<_+*GQ5;W4n*S!;458yAopw)9oe0qPV5
zsbQPVRporcRH2PygmE@iGev%>#GgBroiE?2Oy;UnygR>5TyZloc@6Os;LBcm6T}b*
zi4qyw5DkLnV;=?;2v-qeCNtKUYr4zddJ5laxC`-*B|+HH*0?7QZTaheI_~ym28*y6
z1?!%smgGDJ4k`4kPpInbZq+id-_F(Z810Igduo?wap@Lf`JUDH&SpkSCxbTOFtOqilF7;h*MS8j!cLR#A&UcIrjx(_L
zV9mYn+iPYyGP*W0hWI`9>l8o>CEZFSr3p%li*hz6Nd>V*uH`|Ydm5K-uZHpIoF`l;
zK*h_g2oO@thKE?WtN64z)lB(wB{J{33v2`R1?uz4wcsCd9((@%v4|a4uhj?oUdyR)
z237VGM-Fg)=)fnxGqUn9oRkCrc?&*{(P*7YCRa+8<@=EP-%O=c
z!n<@Km=i?>cVygsH3~s%E~XRyoJ`Ug8ktE?1fx|8eCz2~WnBi9J_h$Pv4Cl-l*A!;
zNC_KXKVWv~A4Py5G*mbx6fb|BM+ht-5arA6*NEcE4GFl{ongDA(8hLd=lyT^_r#HL
zePtL)J|n}Oj=77eRhzl>DldFv-}?Y_6y3)Dp~7KHz-c=fgK14l@=Jo-{Q~`@;XLAM%{(|zuny+%lq3x7
zca&eBDVv}R2j>mZ)%`Us9Xlldmwk_eVr`s6Na$w~H}OqICW`cYygOBL*%ZD#R?d@U
z44;qBnw6K9>)T){f!@n*tGBn{hXOgRrbN|~U01b~GJDb@e!hm=IX*6c@5h+n)8z5j
ztc7#|Y_qnmKTmQphn`07XqQ_Q+Qz!vBv9t6V?Sue+VfnW$37In&x?$G?K>q(t`{!-
zNj;%;+n{>39xbi*oNm0VX((3cqr_VzM;e@o(ESm`zJgI)JZ_DZr2jINlpan8akjuf7D%af>AwbN<-%Ne2B|(yi+x`pU)ZMTX%hg2klb{rjV2d*B)^#P`$D#89W>GH!Z~17}@xNdde6MJKox_eE1G
zPn_NS82P`yn<2QnU~0X4EM}Fl#g92vCrElDOJhO4HnnzMJ}H}%2GR`aJtb!ZG=0`~
zZE^h$iAfiS;``1vn>$DYLyi^)ON094+)G1xI9(#@%5gl85iBQHZTf77PW8j5MhI3#WdVckhD8$ZYrra@@2k8
z?j*8UUc}~HMR~Y)j|`iKU1xUH>w7-bZYK?tVR8y5+Lq)&oyPfeIS<5o$H$p3;5eJg
zJov}ta+}4UINM=yL@l!k^MVfVM`W&x=e~tb+$ibqqFP|Ti%pC#UQFo3H6J?Lm*k?;
zbN4qifq|LoaGGg4nEmORx5vK-4U+Mm1}X|A96gK~yU6EZ>8qy;9V#ai*JJq4Cu+Pc
z$Vm|rBR9I{_KEl@0&pLBl@z9+oY%Quu_n0F#dPGvJ+YM
zBWjfYV(hzLpORNn9?eePFELr^xv@@dUm~n?1phA$WeV=ydN60JGq~I>xX*gc0B_b^yh@&Aim0>W1&T_bHafu+ncU
zv3W^}0#bn9qWOybPlwum4)}Q>%AnS51Jy_5_QY6TVQL3Z_V#h3J7#RP2D_b*k)_33
zT$NZ5_jL)1KiVf6YIA;ChU*UFV{;+S=N9qZXXIKM5)sS)?zMsKXa8RFe>X5V#=o+>
z0M?crcp!<|Bl3#M@V@E7N&xPcy2~JKXr@m`V5SUGgt+jNe^~1{_-Qo$>o-ojztmsX
zOXNeREPK);^3QZf6&;weDX3^WNqoe^nqWXePiaJ~4>51ODokt}sUyy%ya#t;;Dlj8
zu0TTY$c6j$kad5y0dK8@*8iHw`Cf^VNb8I53I*W#NDBz;o8E`chJiz-M1s{IX&tydIcB*f%{X-
zaqXYg^g%W9y2|a<-|++2v>bW?gd2+862z=l9y~Q7JU;uG`C^kngx%@Zu7-*|Dns|jVK1s!tVETYw|#HUe80hTaScqSF#McCmd;J}VcE~5
z>Ne}K@5NQ!ZQm{J$7gZA@cG6(UJ4mpWSoaZxraYQw~PB+Xs?l~%=n!MxIceW&}oHh
z`_rF8FprcBMtoN82QnWG78(O#eTivrPt@mG
zs5V8=?3<`Sw#Ddv8avr5>TUwxx|(qQzK=wM5%<2k^TD#es~hr!E8l0T8w2*;-yd|x
zHQV7s6|;haTWH(qJq>4vMFLnmZPv}eT2&C7WorGi&c5MOA8{L-TkEf`X&n}nfByWL
z$x~&QD|oZ1XuHR_Cl>PjJT#@_*CEe3PRK>5N_P~4aytaP*QhZ>=miK$?!)i1{2EtU
z##Sl+kSU)6M7$6FYWw{$>a69B1>p34>=AuyIVLIB=OEDM_(hZ9U|;>^@$g*R!MT7;;m0J3BI&~6@k;eOV3gCq5a_uLQ-CL)oq(vU~o8^&!#Veh+@97
z-`ydo
    5DaiO7zzSCMs=I~Q0NlRn{^(8_I5Bs#Qt;%O`yXx=_rKCa}#?E-EL72t5
zOBKo!(M~yKqen=0@&*V}Dd#;1eB{k>y*LW2mO`}s0i6xx8#;X?aM)m8c2y}L*jMPn
z>bS;Ff%~s&6DG?I?k)!5Tc}bi{_oN-BtyMEA$qB7lZd+1!Ihpg^YK4#Jz*oG%}$N<
zX?uZZdg(${xGlO+)3A^Du;(VA=ytn*_`(DuVK97${{qTTqarB`g
zT%;b37xKRR5)bmxKul42*5wkNsQ@C4#?|@sDv*?uOS
zPI=W%sN0qfW=8?!!x+w4PpI0oK1r!gB&pdx6-Ms>2+ZyoxKF4n-T5+sH^X1TyNkbG
zk4Vx?dVD^Q*5y9VPRsaK79tZYPc97C}M&MCC`6eAH%|;
z{v5svbe?g-9y00nZ3(VV1EswgD5!b(Hdi15MX#oEzN9_B{@^Ip#$e^>S#<7(MI>ZT
zCw#Xfq_*Qmw}RmehnR?KP(6A_Z98Um}hEnaGib$NwT5qa*|Ycw}kJ|C((961UM
zR8&A>yuVbmU468ZlZ}i3Gyt5ZO2ox`&j)O`RIxqB*#xX5uIi>(!1wpikDWB>=yUNFrXybZ#99HpK|fpUlCAC_u2<9TeKhBX{GQ^6nX*fw?eRkiqp>55+g4PV$J>=YIbn!1Nq!e09Q3^Mi)u?=N@^`Zwf}_A_B74;b^%`|!3{
z)?crx@t}=eScbxzwo1rc&kFQjP0;NoPR&Dqhuc;4t-CAJ?4EyjF)Oz>
zGiH>0?7&dsjgIf&R|7xIGZU{-QoFBQ1
zvEOqFYefy~sfwQG1s>7x@71`E&!vUD#mLRwFqvs2JW@37+H6DqjcgW%!4sWcl=^Gg
zrror_UY@Q-Pj~z5^lh$i(<%<+mzEoAMeeji(J7D(+S?AHS{AZcvCKfSX3(7Hq==J
z@Pl6N@Gqk-H=~Sio_mCyr_*u^Z!u82TL(%yNp^~pZ+hqW#^1Xp!y?>+bt&E*vI80x
zOMj~E9d6b5D&Ua?1T~bX!d+Cn8gjDEYA;j(fwAdF>?wBGh|XrfI?mEM1B%i-MvWW6+ShuN75kjI7e9mB`Ek;N){Rmj>EY-oPV9|C4)yVc&%@x63wF
ziUMwKvU-jbJIvw-fv~PwW4ymD{fuR|^B%gLqT9uMQ4SeLj
zPUE>?*kzlm8q%-FmqME3?}-Zqpp7C=-@Efwv_uIc2~7FbbMD`eOT+1UO(8~{;i;i~
zi%f6g{F?wSAWYLqbdA-VTGUf#F;LJfLovH44EG2rEc~NP%1)~b`>*x4KA|=gzg-QA
z`NIGzsGs*&pZOjn6*3Bwi}>vwUHTXL){=`yR$pIWBD~~zA5fpf<|z7G89_V4rH&dr
zc3QP3ap@Eg>A~HG31;r%d(klqS;_+|==wkFhxUA#$T@bB<4kLslvoY~&c?-Xrv9P~
zOw#)wX{r-hTUylB;;y!ryTLROR1=HC;~R>CJjg)jTR6Sxm4qZ+fmJ8k1TiU8N2*Im
zJrpQ)XGgkTs(s?=!QhN!@K9-Op}c`T=COc&j5LNRnQXrnK~(fP2CtO9Iw)Qt()JvC
z80BJm|112^{hcJi2;Ll#(F(ZUB{P79t6=7rCM=#Wo93r-Z31swqHN10Bh4CZ#>dv=AuNu}mU?G(va)IP^*C+`G?dx8Ny-
z=6%Ki=nRW29FBze9b3q@t)jKxf#XpcILNK2(9>z&_|DUXQzGeY`(qV;KO(c_h7RQ<
z{to$3BsG0c{;4?NOWVhHtt;O+Q9xj;?Ws1}H;PmYDeu?LMuE;h(w}n5pM*D_6~>qo
zyc9ka@|FDFbuq7J3eJDbF@oVaYpKjb!bJX|>m3hnd-uqu=hqO##}V>ZTz8eLWOz8r
z55nU&8-uY5YZw76u-2}`J#0n;n<#D+duAGq5|&WWw{YhCfVwY$3K3Ps8s@moFx*G?
zVb&@Qtln_W(Ny#)I+8k*Xn|jk#+t`jgAXj+QMfP6*L&`eA(L7~QUzT+76D__U(;Te
zm}Uga;HO2%jV`n^XsJGN&OLKi-sa2^57pkc)t5x#iGml49f%(Qt+1!AL<2{9=l4|A
z^pC0;Vqa+?GdN3-tmzRrPNs>Ftkfj)c{y1;3`j6*ehv>;WC#(JfmTB(twFs4`a)UU
zo@g7Y?LQZ4zhq@!+njZqtBkH==8iC%{go7#jCA@?sgn_3xSdGgL~b0&Gs&iBe};yH
z5Aj)^Gsoq3iC|05xg@+I9Z8!Be4$u1SN+Uup|eDwh7EuC$BdUG6O$z-GimR6cbb|E
zMbY4m&4*%R+GO9Ae8~8GbM1OA?y~l$!4d@wJv%NT1kxCGC>GE38n|C?kOkAhj*W0<
z>&vBfdzj8HqNB};Vc~F6!g(Bp~C$-2B<8K=uy5f1HQ#S4?;pxq{#FMc%*ipCmbqWzZ)pecUpt#?M
zlPl^nglxxEResNtpa^)#@T|Qti
zP({DCgc2vHb?~jK(*Vbr9xBxaw-_lO5>h_9YCA)jTizH76G?9SnNN6qLq2&i#!y#v
zenXL2yAiiP+hBi+GZSP#kyY{(x7^8Nw>Ee*JD#vhQc!_VMS%BUwPHD?*ZgeIMc(tBjlD)!=)~Q&5!i5$8GB
zX6SpSF8jH7!Dr8l
znw?{JzNwhVO|QUhXB1=WsB#YHV%FE?UKUrlRcpf_ozG#GOMhPWjdAPZKt|ys=#{X^
zCx<~v0hi|VIjJMKrMG3XLezTTR&g^nd(DT5@a9o!a&TnhK?Rs98?%-BkX;bp^D2HxKUjxL+1TKD>Ds&deD2rXJT
zkam`)jRiGR_%AZX6E80FRryV#Te7P+{4J#NV7%(zZOcaT`PZe=!K38EKLiVmH&8+l
zPA6+_5|$@%xzKT?dE#43I|OInPw98|b$(V2+TcU+3rz8e5TIz}hcSC)P}4h@4>a%f
zuh1+e;^|n?I0%02A5?hN>}O-vO$iv@iWb494v4*Hp_|
z+Fxq~QO+Nr%+*&&GF^<{dYyre6JKz&Gv2f~b=ku;Ug2E~x_7B*cCDBzp=Mb#3rmwC
z(1nMEG@(BuOPr8kM!}IoN1Q)&z?954GjbbPe|Sl_J&-||yMmxl9uth@JW)tM1vU0c
zKMQN#WR-@ElN=WGCI-sg8&9ytA_cEm`7Wp@^**ibERy+IefM5W>fCnds?rFpVL!Mr
z=6}U_@X?ApWW~uFd`=q^TL>$^0|q=duC<$IDXZbaJ@@l!8Ti&!K-vi*=C^|hKBt#+RBWQw=w@VhvrWYn7x8(t$Jn}L
zxQA#fd@i0D$I8p@djp1(?dX6SNRkVbOwQvY*~I@QJLsR^g~18_ephf<@86`yBc|KD
zpN2>g|Fqqy9*-v>P``a}4F!}ZhGvZAm>rm*g)=VNqaGM6CdUImINGDAV$2}UZabuV
zN0$B2#U#*+G@DBDLorj?Sj2&c!L+aE#|r6%gm=2OK9dE;q6^U``Ww03K0VDZQ5aPG!z&6mMkh*&)x#
z1dueX_gG{wa7$re(o25W@Gr>iJxk!O<>YULh}%_+U#+_sEHzk$uJ{uVv)s+i|(Y{wEPm?vnx
z*f*p)n$mrgbG}Oz=~x#4c0I^iZuWq}EcB~P#I_U5(P`;Nf8Z%=>RG6y`I~+2v`CIY
z3Gou~Nl`5}jT<3Y%A-g%5t>Yip08^~dNNVR24&b@uRe@Y&1N}3<&ZhwYh{*f%i3zN
z-{W+GVBVjcXWcjJrOd4}bE&FAH90w$$4b2{Xd$yHPfl7riAwFQ{yFk$=|SLV8i<#J
z@!u1wicza}@Jq=Q>Hjscx5YN+jG2_PN6%cW2sRQgXW35OjH7Pac?P@>!6$t^6#Q3V
z2m{czHgD}SnP>UjY=5)NeVuhOhk)lBNiLII3D+x#LtD&gkF7(=L1ka(xWC-G+jyHh
zKsb`ezw^z$ABfI1H$d``uj61v|M86h>?ZGKm{idD*80l((Xz$NalsfK+?fH<)9LmM
z9hjN)xe600)&q$dW?-+`OgC9d!JPYe?XF)kMqSE`)vQ9W2
zK9LAAYX_Ma5{kJ*8$A`u(|WbN(7lwa)KA#FOH-)qV$*A~*LPwRop&%7xHPly^}l+o
zMS5_6k_tz@fEh4Lf#laQR20j{lbJ>LCAJ7Dqb(Bd-*nu9x<;-i)iS7%4t$T9V%pT}
z29)U@_h%z_vOpZPzrweX8t_7akvGC*ao-^|*gJhxQSR_TnagQdn>i3|-%6E5xbroc
zJI6}&#VXDSNj5r=%qsrMAYZG}2)|s9F3`epjxDVmSJt!r99qgZM=0(pv=kX9eyI5E
zBn11@>jMNc@a1DI)_caXU(V_ET3wk#$WnFcZJS^_%QfGi91iSs5aV#UeejRNa~7n{
zW69F`qpbp30ewK@&}aS5a9G--?ti1lLFe|FxB`n#qPP5I4djC{HTjvSfD)glE0LoC
z*AsRz#8SBtbX?9{t|CyQ7}hD3YVc}M05-onT~AfSqoF&qK%*{I0R0lwx)
zN)HAeD&GD{dl$c{b;mEeG)gDx3(j5YR)QpNdKK63$g;DJ=4v14vpQ4Rx9s)h2C@Fp
zSeI29c;>E^0T+=mf0lj>z!RFCDt^q$Qfl@oxR`ZT@!0F%MHB6K6ZHF5jwO^M*#V3)
zb}f%hW34|=Re#)lu@NlRSl?A>W;jxNSILM2*CYSKc%#7>uVwiF4zI$lgo*RN&$h{`
zy`CKiwZ1{sX`Ps<6RZyNTv`3@w@zxJg)P);b%^qc#$y?fP6%e?e?hv5%P$iq<4qo|
z&3GmGw%Py7DC_b}9nMu6e+WGLOFoM<6WnKIF7p=lH{of_le(&hf~_dBy*9TD=~Xc@dlQHbUZ-a6Mrr^(&*urEojR_gZALYcY*
zS9aw&R!x2P@xq)m1my=^l+!Qn_)$$kDjE(#EZH{Lb*G7PFGgyppAEEs49bHAB8BxF
zbTZ>a1u50)Czo(VJ{x;B{DLF_8El6g3dg|Vooa68`&FSSvwnvWwDk2kn
zncuud_|D|e%d>`w_L(|S`&-AYKrt4?{wh-MocYvT8?B}Vd#QDRr2W9xPL
zQU%BwR$FH_5di90z+D4oi9}KJ1YVJ2zvUk?E9Vfd*>Cndh1D9t?TJ_pL_bh5
z97)d`zbR1P;4y5l`Su7Z0(-J5!FxtE)(wVQ_W^b`iCKjA%6=`<=N&oqt#^R4EP&Rt
zX?j6VX11S?yXJKwoONn9u1LaQ^}(8@Y=byoGFFXxl>8q9k}R$VR!T+;`U_KLA}lUG
z1b=ZcyVP~K*P;vZ5nZANew}WT+++U@tIE;<+cwTGW1}-SK41J>
zs+tFkwn1;%F!!lM90og4dgr?A>z{6azXpbU>II{H-!~7QiKLpcxpc9)-c_j4qlQUK
z<_neX6MyCl6w8TyI@hx>^g0qW02pOljFs1r!3iq>?0ZuuA~3iY-3$?V
zHc)2Vk{m>#zEh+fS1?|T+Z9M3SVyG>esRnREO0$(0V9h}Dsu2tYKk5@7?)ds;daGX
zpJUPNw=Qe7w~0BJGZT`3Abh>3a0fQ~`AHJzf1v#5@NVj_-N<*n&ko+-k!nl+y~lXi
zG%UJL`R%sd!T$eG3xL*m+tdVPqWvHL0*!@>QEuw7SiZWr(^>P!Khl{L;@uoaSC{Mm`+^H>ka$N=Vo4K?E{C&!;$Ra^yjfbd_Y+~&rZ)xBI4BR%
zvR}WTyVPa?{xF!mX6}BW5>kBnMqhkuB7>=@V85*NqO__b<$V+g`_~;e2%_L#XPnL%
zarIOir(k3{-$-x!FA|Azsyv#bMrNi}cH1iDI1n_C3<=)G{qao)h5d-BiyVGC)I+rxF`i>df5s`Nyf4QAFd~mE}YTnzz0{a%UjWZ-Z2;P{xIMY!#>#5
zISa#Z!B-~JQu!eQq|zSdP7VaB&T9FP6hM{KBL5e%xnDjX$1Q!1kV^n(?1b1)d@_Y~
zKMAVt4+CrVf`n|ND)PNaF(x}V-H56-eEbAP&g86)Ly*UqdP6ua_}s7tP~o3W2#5pi
zK^p_fxl%PwbNOx}S3#d-M=bD9=eRO_p%)0{uVvR}chH&dbsWe(om#XjTA=qPI^BCN
zR*N#i`(*y$bGfADt!fU!()Y|Ih8op%_)9`f!Y@g)aLUypXn9SLBARoEgLvIC0^Eez
zeB*#o9cwiN^QCCtkBTfmJPU+2Ye8bm%GC=@6Q_P(=B#pJ&t?=BHcEKt!P6d!lo*m+
z8QGI9u*EOi4;^Zqt*quE64JtXRV
zG+LX*_t*+jUlmes9G2g?VO#X&cB1nH$}`jGsny$>qf$lm^leBa!k|p&VfpwLQ_O|t
zB738_FuKd|RDnUG2D{1YA~5}xkzlFs$?!|J121h&;E-az=~Cx%yN%#|AHi)(#o&y}
zAu9R{Wc(v@_}6OMQs(>{{F_6LLR`9PDNEYEsjiQCZw6jBf+q8Zsk@#BZF>l2TCf_J
z^ts%z65RPAQBXxXE3cX(1dQLhM~|X836GzR3W|32k`raj`zb_@MAu+|oz+om6u8};
zK5i0ys$-Dce6=UB^`>C-UBrAqlY2&Kv#;)oF+(_3g@tZJ~x7U9~
zksXjZJdYQ>mGf@ih@edVx;!y}A+4TX`L>x{pkxGoyU!GgKjxN;TorS}=suOMrAzm%
znN~;OVnY;vJOB_SrrpLE-tmq2*u4={l}t9xP&d1hbm--E^OrKK#SIEUy^>rFq*AdN
z?k%a1d+LW)qSC>JJ(U5AO!#^`{!F9Q40?K*W=o8bhkBeeLezkBTyJyo`mJFDBEpMZmwG9?C
z+$tiV5!$E0pA?9$0d0_4rlDv(Kbs63f7&isY@dFgqStWzCY9{O$c=uUsw3PYEqOKY
z=XTkVUCx~9Qe#f3m2
zz=*QegNT;ZHX6h!?EYMk0)SUamOt?W<(XjnLTS^+#$&48c=W>Rbeu-6ZbtRyVBn~x
z8|igOblCa!kNP5q)g!gV=I#51|F)PX0$?C`-*<)nbmv6Dw%Z6~4pmr^`#!YIUTt*h
zql$l6x?zed!*Lm1B{?OrwMz}OK6bx)ncy4Ots?@H<@}X%lC1x3pdgB9s(U)wtqWW4
z3Ri(^=a-1?WP&E;d~wM$UbmkM18J$u%4MqN4Ejx)1r?jMUnck-CG2A}0-QUocGXn%
zCIW@jDkRtJ8nP$PJ(2rsmp!_R5M|LV;KZEzZJ|!k&EC81HtH_78#D8a8k-k$
zYVrL(OM*3MGw)*q)A?cZq3jk`q@Xz(Ea7aa{Jx_
z4Hlbj
z7`JFwCs$A&w=-v~9t6)i(MwuWa$pv8DyG}?q!$j()gTY0XU!JXla(i#j+2Y*j(lUF
zUTt?6Ms{%b7vBjcr#!o`sFizqbPkmgS}W_~drNLZGWrxnCWoe_cTMkWKJ@zLooGBOiU0qd~P<+4(`DqKMYT!B$yoB`j=;%iH4BgRjF3z%a@$>Oxt(C~@?IvrwXhDii
z0Unjy2V!Kd8H-hI-v{^O2jSsp-jjRtEw_x+>bUi#!kBe~{u8;>;NT<&0BEz31t2AP
zSNYA%Ad#E`wlkG{+fp2Vt(nkC9JNwfLULYtHwTSnT?YP8XuLv#|ID|Z-1xXhnbqNg
zlS4^lpX0RR`N0oJ(LL;a^3l)x?f-TpWE!s^*$T>M(dl?q(u`i#Z`q|!+xO~4z9wZ)
z`IAP?eSQgFbE(c1T?whv2lmc>dxw)bDl0b?5TQ
zXxc>X$yk6)YDr5!3V`q-AD?m!dvt5(evCfpwk3IEvFD_|nj<;{mDmoRHE_E=<9Fh5
zz`Hvf_7=nkDwX8IbrSFpw-O$Ih{Sdy`^dL>y<(
z2R|MXq2XawXo@*!rA%?D^V;3uW^J^f1$wxuHstYNRsKMm8$?vi+35>Kxw=OYuGAUP
zzv91rGtsffom31rjQ{?8Ma?sjES8#D(ARmAhH^F@
zJ;3MoC@yL5pSJfN=KhZre&(j{t%K$(t_gGb1t78O(ak
z-4D+@;6T*MxSK`aft9BWCSSuDG|4x)5ckYR4Z+wm3&e{!*uXS95>L9I)GKqY
z6YD8((L%q!FK;;>q%WR`%$X%8F2D~r<9W{MrCZ1iOmOkip|{liGz15T3tOsYO2dJH
z&A&QoqA^Ko2c#fy_F
za~r#EVB;W%#NBs*sJYVY4D`MXw-Mh+dp#H{wk}4&!U$tLRqNS=Q846UJps3NY)wdu25aU8xMVV!sz5#j1Z}
zHUeFj&rH=How=8WExo#Qi`~txwvVd@7rhBhoJki=T_oFqJTKd`~dHd51V^n^*?H1ly}H(*!^ZNkn$P&823ONyH16`J?k*
zuJl$1d1e67-zdutTE3dMBV7&#ZJn}8$tmM4R;x(0jDwNu8b+l*7AI<+7hjLl|7}XK
z{|Prs1~`q|yr%P28jWp1-(>>m^>n3mW7WI6(bF(U!4Z2_G`bCr*sl;Hw;3iq{v%BP
zHe~Pu{qky)feBO|XhtKKRMdQYdzvI2ax)Ngp^pK`rq#zg0*
zWNKNVZVtMW1Wxwr#$j6($DfP
zAg60~A*iQFI>c*rbCsFE(|9Gto7>|9+A4{ea?e`AXHAXc&OLjyV8k{iJo4L?p%M+#
zdMESjGvG35xo?ZbD5mgD|a<96IZNVJwwRhMC5i{kc(u4aD)O=ur521Kt4kLqt*I;{;Ifw_%Wv7l3H)F)0Az{T>rdXA-o)$EFHt38itu4rz=N1InGM_dkUPU#J
zJJ;~bA-;dft%r9B7V}fxZ*c%HyDHQ(R`FRRvMI_uG63f6SHu%CM#6>K+<_X;%JuMm
z>advfNjO9)bWSaof-xdnb?O!kSA>h++Qg=j>08sez-1psOlAwh1wsVlNzzR>t=ra@
z`cny-fUs-dn;u8yO@f5aMw!iPFhiF@eQpV}tR}{izh$y}G(X3V8x6C6Qypu8D}L>g
zoZ1FY)s3tb6+FI~jkyx@AlM=gD6iLHLc7?~63*-xv3>AIhG9RRiksBax!%bJpJ7t}
zW|^N=Z~o70WS()f01W+dKv)8O^>FDP`}AZ4QfIA6BXnAFRIGa
zW>mP>n=S>wD-QaR6{2NDf7-OcM$O<|8Kv0)G)_8C!`}eAY2Ukk6O{eqTbzhPzVGKj
zkul1$O5O`aP=D8V#(>FA_J6)*C#)-InPk<
zjc8m({Rw(-pIbtYochO_N_a0w>~-3HDx-}j)ksweP@6a05P`kzxxnk=LRPySKsRwo
z1WcBQDViw&HO?xfih_iyCL6NGmzI>?_gD%1z^#h%DqR@@J^BNftb4
zeN^qY*xiq`#>dX-8k=)vu}_aUth<0j74JRuEV
z)d*pseb?(kt$I#`_ow^ypfoW(4#Vz8?-_d`j}Z{c`zd$WDTBJB05?1lEw{~&K?kmk
z`=Rz{KYIW8-;Kjng_eMy0-E={i4KU$ZESTMd(V#DfDte`xF%zHqdf6Kp5b{%4Zd+D
z;GS;e?&O9|H*azJ0SmhZ&Ai`;|-DtO_}$
z1U0-Q1~rHwQ
z+p!HV#h>NKs^a8%mrT$ZANPK|E>u@O>$rym?{OFi>q<^o^|Qhlnx>}(jWhUqNhp4I
z=XuAhnM;SeJhut)p06xgAX@~b6|F`I6XyrI3syL#td@!;u$v9Gh+nBkLbS{zs9{&v
z5WkaZX;#b|1NALN+sykXg$LY~Fm26UuEycoo1ii{7o?iPtBeQIh2vgRHcqw01XghgnGRG7r^;8?(i5
zlP)vtn~2N&q-r_#{Nl4XK#c+I$7&}*quYiTr~nAsJ!x<*d*WhbTumA
zKg(=e%gse!{^32BPKp)I_dq1E{hf+NW_P{Xxflpd9s%k$ZgGAT9F)rp{19l(Rk$Ao
zo_cNVD@)jnisjU6+#NzYolk33pC=!`bf$uDU%mp3JfI*~J-JI`8&
zg*XGu{f>V?3Po?Knb}al>HE-wk(B=ih8Sc~G2Wn9PgY7d$LF9EISVYCYrhpp3xSv0
z1-RVW1^}q@qlAb3a~p*1w2T}(lxl?cBE=_@7D^JV*J<<4rkbMb;8pLu
z4NV@VaN))8;8wkE@jamfEY3g#F|M{Vp>&RbVl78)?!I<&lDa)u$QZ2W#SHH@sy4cc
z8#wGQcv;OnczNd!#Z#P|#_^jyG&DLq&f)odSaf^Ym7=(DmRFrFCNuUA%WH1cGIs|p
z&&BJKi;|i0d3w*r+M1s8M=1=>7cwy{{?6tHML-0L<1i7I=4sZxW6Zi|NMnDkFX&o`
z)UFTzkuWRed~(;~tnFP)XRPquH$+Xq>w8&ZFy576w=Ck5;U?|ht+smgWAKvVr1wmiq%RSTsB3?Wdbg4FVKQyTm;)drSQUQ!*b
z8?WbrP(RgUP+i5#Gw6J%^N7QUs`dS;jHa(g$^7Fg`uAiRUAVy2Z%0TwJk#wxi#_26
zLZ@$)c!%u0z>&c4nl9q~H6+Ybggvg_gzNW#6aVIC_FeuuA)$0uGjB_FdISB{32~{R
zzmK;tFt)0zxq7-YrBkTQYQJc5?2tQGOI(nmD9zhk_u+Ti)LO{G1M5B9n*nRaw
z2A!?C2YM7Vv9Kp`W0^gXzMIc#n~y0@O`Kil5A$698O`UvVo-r^FASyiB}wwM-90O`
zHe$yFowuZhGfIUSQzFfEAIwTiYzBJTjdKmU0NpnZ_XS`kly@)^-Rz`0gD)%=q}HBFndZ4
zXqw#g{BkczYG3n9?fj
zw<7a034Bd5L;)VcZg50%ZZ^gBWd!CO~JW&?X}9_!X&gC`2$yC+!Z-uG~;~qg=0-F)_BoUFnTzCSB(0iwYdZ>5~|gbwa4rCo`^ZQY};ua5VK?ki`_yHoD@|D4|A8TH3jAMYbd4
zl0MB+$Dx=L`O>{x*XB;~8v<;yw8LOiARg-1^X{KOiHB5bvXFz4Kf{v`U
zl9X|!6<$Lv^X{&@)Le=FAUz~At@@3}^3;Vm*fn@|I8&R|znH)3K$&f3=g{tHi?hwMJW#pSUywuLdVOove9)3i
zlz{Td)M@LWa|@5PV*lxA>S{eFMR?!jZM!>WSLW#LNit>)(GTs*ma(PoL{z=~dZM(E
zQoGj;(PhskH-}m?E8=IOyb%5Cao|+lTD&)*WP^BlK{OL#!~pNPv+gd3ixfSjJ^EF#bNrKKCBK{}Q0ZVufoAYDoe
zNOyyjG$`FlcQ@QQ-`{&b_q~4}csR^6vuEwS)}Gm$L!3BbW0*LL=rt!pE%(3LJH)=U
zwEI6o@*5M)Wlg>V1Ww~g#74J?N_KDsTAs}u+7X^k7V^MKO7
z!h#N<5v+|tj6AhYzjZa;-5-C2GItG$*#xb4u6=S1$Ev&+bq-bg?*lf-va(KO>_gWw
zo~LOMH$Pg0k@XVyO{XZ=7!#ZF`7yu?(-x}#=~F9*R|YZWf)Za(P$ItoMX8~xh+8`W5b-I3ZhMcqY;>^e9VJG#>BS74_b0hCUg9Huh0=EK7ymva>CunDD=B8*eF%#z8LK-ktJL?zP
zjuxYeciPUW931F2TW$BRbQ^Q#Y&lTRc{XB*nb9a*DI5Z2&{&3!P`$XZO9jr*q9tE&
zOw9ihNJH_;{QW1{v9wR9t)A7MVnKFauo@?2`lHWV9rpubHd&-;f>EL;h
z!r%tULd%h^C&E3Srod%-eo}DBmIe^am;TY4FThFR=oil}w6~|Fy`Gozl*~fA74qbse6wKLT7C
z{?dap4^x_0O;I`)@=EZ=%eTgYk)Zt6RBafUb#umg@-gSg*SuN`OW45Rh8n9J7e_5C
zCaHiNlI9xP!aOtYC-T2r>zie~1(SyG;VFvUn~NCimUoGsj|+D
z*zyAym{Q-2I{GKlhfj8fIDelx%k1B2aG1=m7U(_8!hbxRE6{W6qp=Ioqo7j-MtMyws?>)CkNGL-|3>$7cjiQKYC1t8=`5MJDC*=^y<{_CEw3V~|l
zKHa6WW(=3<9%}Y|Xqx%9c+7y-^+0PR`?d9NL1@(|vI^69+S4&6e}O*GgBht`LKhUT
zfIN_cWA<=*E+!98if)NvstT^vy{$7p{W;pa`0FleW_4~S7pZnG7O5WQ8yk>tNqDEL
z>Mg1+gD8%k7pRWtvbXgK5Iz5xW!p5ZT|YLSP;z0Tl;qmhdKhSBH%=pYmX)_}hxUWZdhM8Y}p`RMsRKBA3c78tgm3bDe({_vt(%kF{
zc!$pZwX3Ex5vTpRpX79pqGc2RA*Vx}&gb&OuMM218?SJTJI_UAyW9UOhkW3;J!*g9
z$1&Hmof{_V&Qj68l7N|qIUdGEnr%Y#;PwXf;Gwz6{V|6ER-%`qXf{>X%Y}hP<#nxL
zBbjjY`7ET$aoV1%WbfgifY(mucFg
zS4-i3*DNoukBZ!I#B)RP)mJG_j+T9@6l9%aiLNbJ4ZC~3>|^3TeXzlHU;T`;^UfH4
z??YuwUoHKs?RzN2vBY|Vy7vd~z?E(7A5+GL$>{mN28W2-f7>7Kf4mp0JC)tlAtF&f
zAexuNMfuiMDPu3x(-8gCbj^ni&Z|u57-;`{bD`}Vvoa~Nmwd3Kw5j#Eb~OTagnl+7
zxwd<_@mOR2wqtA_0#^=sHp7!~r7HG4J;qB4ISANR^!o9d7DFv2z}HU=HJurH+`WJU
zl-J4a2W?4!^F@Oh8k1WT7-I`xb!bDM@e^_76`cQKM5Fmon|fUJRR$rM$(;IEyZ1F#
z{!?=-Nj!CToIj8GrV)BfKy)FG%~NGG)d}T%meZ(wgfF+jNLS1Gurh*ihutr@pONqX
zjt!za#zsC1okUG9wpt`kkY@%-Ryn#(oMGpikq^)KBPSNcYI6yK1T4
zUi}BIS&EyD*UZT)T_n%_pF7UZ8w0r|f{q8c;8;@?NV95u0eV_r;vbSw3Bxap7;B_o`iQAH)vlc{2cw78J|4#z1`mvH>{jLwxuv?
zkX3o8#h}jhul}7$7sDCR+qKL$;q+L8)JTLm0e1K>hVJUr
z#*9d9kb8WG)qV5Ngue(mok=^%PJ>0NX9sv(6jf^xBCtfhN+CnjnO{s|y&LqaVHS=A
zqknRHb8DM4#gyr_SD5r4d#pVc(}fN4FP7&oBP{
zqH9)rt>Q|0;vqRAe@oUq!MM_y)cHvQG3%`yT%dVQ!<JT17s`LbJc-d=if4@VGSEZE-{_ivJSYtqM;yvZ?sb&U@2rB~^la)Rehf0H-EbxX&aHGk776oE$5(!alA=JT`
ze8>0joF4jp4h4U6L>zv2Rw&(Wzoj;8?^8BPj`v`pLsoAzz*6IO(m7`;mDo}w4#cQ)
z9Qr9)?ajBl^fnxvoukB(rGTE^fjEAte-B@S20Jq|gA~mpSmYqR+C@OiJMFQ|eL;iiT|WK@J5^AcWeT^z^?SMHO?D>bhjVK0MM?kD`kNsEnodzq+`sADw{|j)^T@}lAgPR^_3O;b)
zf_!W|0UXJL9yv@$DtIEG;q+lyOaV#ci
zOfH6WwYXPkZ4S==$(4Moa9d50(V5B?;VOCEV1&|#l2}=~rA_F9e%n8tq5K`=%zFyS
zml87d#&5kuh3YAByFh;5Bks)5kk6wR@yjs%qt~C~&$N{u60MN0U#caNGHyQ3yuG!7Y45h>#O6E{pl0hMb$y|H&aJ52BokOf+?}MTL>v?gj={8Lha_zG0#)vRaYK3}wCD-_U5#C201Tx{S_*n9lfkQS
zNX~cCu9xiPs1&gB{TF3smDaplqlo&*F(I{5ek=x_8ex2hV1rMXp8_YP`=V@+$RSJS
z@iOPlz;oT+);E}OqwbKDCGUAWSVV4J*QNNU?wkLT1{pd(w!9s2-+ALjPbUgKet?hr
z%%y^?MUa-6>2Iy!t@Q*>>#?s7eB4nqsM~c&comvZL;s
zH&pj2#oTIHzoEq8zB-!lYWk-whbDx?Gy=x
zdy8M+Gbg-=Pvm=7BWA^z54&#UCPct
z>(7{K!T)Q7xVN7>FM*>#l6l7g-P&Ac1+IhsuthmM2;G@e;xwzqbXHxp5^a9W)7jkX
z7|X~5Q>>dCEM`JdBi`mcJ_4B^S6S@(Kd@+V##nuSobnnwMj#zIw<=Vgcft4~%Q`T^
z*NF6F4Q?jsrpQB#UvFyvGb9P@v*?|$DEat~jD*`dM7Z61L$Lban-&}Ytvc1gI}7(2
z{3ge9;f~xhME3;K(Z46#Q2n|3pL3o@`}a=)Lj9w`wXxjNSC*-lh*@JaP0j-Z7{R?0
zO-_GZ?d3EpQ3Ahflx;j87$BtK$`CLIwf3%ux3(g9c%a#f;gG_H{uj?At&cYdPYdoZ>(A*W
zsxM9ztM1(s&s?`fBPWwY-2P(SvG9j%NvV}07#$AdMw*(;v2!EYa3#b)IV
zO#@f$=2K3vjK&IsH-{EENi-Q}3+CBN)Zr^|Cdyj6n;jCX>r_XJoY<e^Kk^e|0g<~dSh-g=`uJFF!Jky+~
zW=D!?y$0v^UTP7Pl%SK8QE&j}nNe`gKRkGNnAjI0OV4S12=$->kZz$AxqzCb*w!1Y
z2qF$Hd-T^wUz3gA(+%WAWBmD4mWxC8JMB7dtc}UhYnaqr%OO7;BUC50X07|2x9Ri-
zZ>lf$AC)0`2RY2IS8n?5=+oP*uYMPpw0mE&n@vfgrnO;qa-l8dtQVeHz&#ta_>q9d
z$6_@hPMB|h7HgG>ct1qg-N$)(o!VwhtqyYq@^=rWzc~^1B;>N4V`ul6UcIhRD}0|O
zXocN!MG;5U{_ghv>dp+#WFRJRR_pmvW%Gs#QsuTyj`8sdeSzx<=E0#J<(@|e!)z4=
z-W!7~Sew;Jwcc8TP3>jHj^ldI`n5=Z{${*lRkIC;tJ*8ctR)#V9M}UfmO;XcJj@B$4{TdLNP>{tmn_A{y%DPe)V?<-
z9a^80nZXPP1@L0sT>RM?MNL!yUD)7Ze2{Qoe?VIHxfCnP_3IfjLZN!kcj{aho@6&?
zAT_1548bd$?h|L0>hUqpZ`7B3Tw2j1{8j5L7#&>{;q}zGP^zbVI^HMaStht`)i5wq
zSK~kt$YA|-DJm&0PrWgRQMaz-Q8JJ=8O~KcwNi8>jRTAO6JK4q4t>JKNCio2e3oPk
zBbI8hE!DMObAR+!GyGS7E=tX1T
z-cP64_1s92IIgWxg;b%M!$*xFR4GmDyBxK7$tE*mQOlsllZ&Y+`^|O)M_{qM_AAls
z9|-*+l~x$@^H5>ana+{
z+J*7JI+}2F>iO|U$K7k)Zz4x<#>c{t(gF*meAl9o2(tNm-DBYrafU*zt4*Kr}?OUWVqHk1tKFxpD1AQ>HE~65T
z)dbV;22(Z~;;_6Qn9v9ph`zu9siIC7$(`I_Y)AIL
z++xAEp9^IFDnV^m6T;`eStLA`#Dw=rD*96*!j0Yby8p)mP>;ZI9rjaS6yDlB4&ggY
zCN*09c5ggZR@NIOv>$E`i-PQ~!+Max&vZ2QD;>dRHoV@^39Gf>J{l@@cDpChd1|HD
z3aJj}8)?@$a-TBJKw8yh&Fm`Ck4}$wZjzDMo=1=|VWH$G0skZB*GxT5ANP;3R0TKN
zL*_KR$_Ao|$>pyM#N7+Kk26DjizFwjYt)f3YBUUaPE3MOJaYDUwiJj_IB3EG534&xw0lR>ii}Fdvyx4EbamyShF%ze{qR5p+x?`zHNHSQiUkiJ_#&<=;-dtItFcJT
z_bk-g-_-e^f<#rMjAug;En9b(7u*k`#td3s@)NQew1iC2$y56G|J#{u;pqweEY?El
zfi<_wZt>!RSOAgaI=d^*pfxn6F?g~Oy~XRC{#P<+6jsdl>=LxcSyy=fco_+zDzfA_@@4*L&3cT+{u>C9%@bV^K5EI!{`VO@pA=!{F^
z{O7J`gq67W_lX}h8)9^(&fPbI$_Fn^y3
z4`U=@{Jbe+`SzO2&(H}Er)ca|qN!lFo10soh|l%9T@&G;c0JjX%F0AM-Xt|X7nbMU
z16}{9Q84PBt|!*(*H>hGjHYcO?YJ91BZJ8VJgD`5r^%i}h)~I{iyQITFN>Eyn
zFc%c7qunwh#b+RkHD0#dHtjQxwa0D`#omhKixqZ0zQp4fMvLf`3u~I;+mO>1wA``F
zZO8?og(5#g!!Cxf*XGZ+BXBs0JMJSYc9A)F
zbh{~peJAcwU3ymzy-fuF#^{W5_S(inu+8-v^52nuv6@<3T(C)PcSWt8MVGm4&zv#z
zTWpJqQ{)?7^LMeI(>&sgG2gq%8=H?6z`*1X_ayCfL<1GRq;}N`Cz@?|O~JC>al~b!)eoa!eH~g~nK^m8@#
z4!0wwK4&^mfA{7S{R%OtFiLWyla*hybk*tlz$8X&@e(t&Ao}0Q|(n-H$Tk&{J
zxHZ$LyggNcMjzYm$osH#gtPQ~vMxj}=pD1xEwMH?RLkf40RLLEq~v9FE2Z5+o&C^e
zWjvPWkDLmV=NmZCDW?CPw=zLduBJ&6De(Gum8vaKyjIxRD_7IQ)gava*B->)%9n?8
z#kS((h)HJL>5SukN9{(r-CBoGZNsc#n_a8zwC9)H-S(t@w!PEuWQO(KJS(*(7C&`{
ztB#DCuSt~+2+b#t=rMtqBN
zg6Vm=hv&`|zv=%BQnE!dvBZBKv(9}eij=R^JJ3R5r9@!Cr1Kv~CJ
z4qS>A_oC-`oZa7YH?S=6?X@C4SGFhF#j>BN@BEA_ZBryTXG<^1{%sDYt`Rfpw=&A7
zFu%>S5E`n1c20f5hoSoScv$*53>icMoLFn3qJR5iPj6056mo?Pk|f5MS4C3(G&&fy
zsn%jO;o*RIEgs4BZ~<>1j%>Q1ZGiCEEt$haI)+R@Gim;*%xsEStjnKT^lb--A~WW8
zr;FpCZT}H8SWgwmr3rZaaA&^hp{C7*`UM!+GxQ^u1OgV_5W{xwT``&stu{}mTmou-
z;6y;rCsCv(G|rcKFXhz{pr22eU}`(&q3V8xIF!3=)tk0DCq
zvCHT@cHjKj`=Wba%>Tip-qXkab4qoMObiK&f+DK1=Q7P!bJcK6>P4s-#-bd*`g=@x*f+4Gz6CUWpb+`Rf%KRovu1f(3ME=OB*hP`is%Fs
z?ebiH6c$5gCBX1B7!hN}JJ{E0vBeEYSc=3Y_Xir=e23~UY1+i)HiBQl-d`O~Un8+b
z5Ka`7NjA`z=oe0nvx%*3jb<`x{UB{xv)UR-?<|OTks8kF3wTs0z$nC~2Kiuwz`THf
zwqSv4=JWQ7kG)}hK3B}vze~QF*}YF;M($XQ{GXAiK#e3Xhu^9Z%uo}}$#Lg1wbl=P
z?M^=-bW*;&%a?HZ?&Ut=?adNi=;$9JkYUBh0vVKZcH()l-I1|rksKXO$8*RI;4EH1
zzp#q|-mt?r62Rv-u!bccN~ek=D64plyIwf1>yF)FLHFQuC>YlL*lk`2g@r`6h`-_L
zxz{Qi(EQNJ1$PAS-k1KjSU1$Lzy_(C)cy;^SYfliFIO6gIIhp&v*;RwZ4Ow<=VQ&+
zg;4VfkjMi#$*5iwW2jEu#cjx)1{b4IrrEdA7Y+C6yl^xqppr_Z(+}d<_1$LnM~7y(
zKMS9Fk%F$e7l78M`p7BP|CY`dArP!)(6KuKtNsh2O@8^P{V+0q=bu`8N#g(8$7sNS
zuF*aeDuCxt+~CnEgkx;yYd+5YFjx~Nxc~3%#JcbUfjO_;oc_)8^A+Xs4rI}9VyE;F
zTVO@v!wUp(v3_^CPsnYZtUin;8#>AK@n-~9lmD}ysC7e~#($);m=g2#2cUgtke|Cw
zQ1g$wdGPt@{*P}y#{r7}^oS~V4>oQW94ts}z=-rj6R^cy9WI(&?)?GGBqyB987c<^
zy};u0IxD!!%I0+-?+HQ1$rBsO6%qNN-y8(IE8>oy>|X>J3rcp2F;ak+Wb&!3Qacl0
zvsFz4m}#6gewL^g$K|jPQVLMU;)7Ka>q164A*(cKbzd+M2LF~(tso-h^*^7yG+;LX
zi}9dXyr2Hr9cl#@NRI0{%sMq|JCpf${y-=YmZyRFZ&@(|L4p6>_-e7>msDs1$aYU
ziYl3S@&%f)zP>*GP{a+j5&=bQWd%&6s2+Qi!vD7x0Fvv@1dP>m5!wf5&8)C?kV(k%
zd{;Og6d)Z?E6K;sZPJhWc#p3gpk4iG7v>fA`G3eHf|=lP;^0Hx6N(b9l>N>s{`O*b
z8j3GB>){lRM@!8Pi|nOPmIM!j0^9RjyDPx}YDNA?2P^F)M4aY70kAn9%yIlK*O`9i
zp2}_5>3n_ydJ)Bg)qa6aP{1*jC=tpp9mOoZ$?Mm*FX>!f@m!ubEVo>uS*N~62y}?<
zs{s&Omp25&JhmnAk1{BQcwmt*lcA8nm@str{H~7-4i7a-)#xfYa?cj1vQ@
ziOSarD7Kx5|8J9vb#cIg2@zZEfH++p=t#yAE>U)Mp@qut@#yp2q!`JbtZi6))GA)^2O2~@E4OH{jG
z--dLBoSmTaVnhk%#VC*24f;U>K1a~&+hq<3uYEazan@V2SA5wj|9hQdYRBUhq2E6Y
zNT!PvQLt?kfq6>#p!g6_$T1+nOTw32>H+y~Bu9t@xGszcgOhU}7k1}~Jh8Q&#+s7W
z++H?3ASR2xkfM~1AtV&^WLL@+j$P{xrUCe2x!NHFwib@45A7(}=YatMU|L|-ndL|j
z0!5>`fKc=q_*BPN@7$CDNRrF?yk3Ud)~aM40OGahZ&J~B=G;$32f$%{+MlV&8%a&0
z#3aP>e7NCY9G8>}1e+PLpY={RH-cn{?$^`*d?qE5Cq*FG`yauS0EDG&VpaYAp|lQ_
zkWA#~f0pgY1ALbNi7Jyam_s04t@nnP7yIS~kHDhwPFMb)fMp!r1wKOK
zXvL-ZeuSC7Ao!&Mu%X*eQ+1_!hv$k#iQJ<@Nt==hf-~2^o2VQN5S!?$|GnIwHByZ-
zg^Imml$O+M8yi8Y`O@oi)h3=#5BzEcvVmZa?_Q7QMMhlcf`RzL0f>gcX)vjS52Vzh
z%YW;vrh}q)Nrk*|5m4}5(=@67w=YoxK~VCkD*y%>_2yXp)miAm*8=;W*|3eT%1!=`
z25G1JZsMy%a6!#nDpdlF47%x1A_AyX^Hc*%3yn3Qz6bc;aFM4QIGH$77_wi9bn+le
z=e;(WFB25^-xEf$P_Y{AD=zf$6|pW$L63tuEEIg^2}4AEL7D$dCy-4nz?lV9BN_lN
zwZ6A^M<)fpBZ&7#^JfOP%}+QU(OfC3HWpZu{umen%PmS0H{ZSvQW$hiHRYT$kx&2m
z8=#8Ei3Y(Vv->@Gk^<7$=&&jc;`vKvTGnqY^w5!ZrBYLdJfu=H+=937T0}|+PzD=m
zV<5h=_-S)Ab9T{-?(2UezQP0>9U%&rDaiWI_xWh_i;uq}iVzQgH;0256$XpJ!59ud
z-d*;BwC|IVlIRtCIw1Fsw`XRwa*2{6ji!pZ
zl>SyoFM_%V0M4ve|HCoPGfvy4e4QgO0WKDKxo^
zh~K%OmrIOF+|(whM+co4BIveD4q{yd
zSWt@K_jGFkMg7jFyEo#YuQsFWV3I)3xR|6iUsK2pTwZd#nI`!`9ArLmeP1RlV-U#<#DsCGUkaM{6DfKhtn(N`p?PR
zFLuBVn=DmlsI&g_YP~1)Tv;)>0SZwf(15_-Q>{QQNK{lh0ABTcYxI(Q<4t>zp~sae8Y6A$@eogE9(fg1pT7J_RP(%(|@vq7zGOhR;}J3HJ=
zlx6p4B#y3L&Y0?QiyPb~o|(I+s8HsVy;#>Z7ACLU>VUR2;1Xo8Kwy$=u$@17di247
z{QWKDZS(f1D-ce+-$!)70sZdkFdA$`e@!9_#*2gb^IAt_@K@Nd`nVa>A^hjkBWN(6
zD?Y!WqNM3sb-vh{?6ycvbHVaF-9V5<*Qn6NBONJVo&dhvjSiIn>Snj8DUa92`Zy_5
zQUGj@xWTi`RF)29buJDxyJ7Cnfo(}Cg%5fo2|jF}8Ek!+C;Ry!#~9xsxf
z1_3`!J0chQ@jjU`ki(?k7)-PGg4bXI6`bBIoTUHWOa+P$`Mcy&ET(`LPOl0-qhkUr
zMMz=M3j_EsAdAHnEH}ueQv?6@p3XAoIbdm@N+N6xh<<7OE=9}RnxaRU{+0rTIKa7v
zSLiiv0BwLg^%uwO08hFAR&enve|iF|*pa@U5GNskVg>H;Ta5mT4R)6sQNM>!u&6_PvJQL&Ky_)Gt-!T4#AaDv-_~;~
zZhtb^2jlfmKeN)41U!x^-!La^;m>D?dVhkmYK$WniUR7x$K9#I2nrED%EQgZWBf>p
zG3yv&?mdfWVHbQFI$Nmo>uN?C6liieMP)8D8%mU_T@e;uuM#-ow&z`J=qXgu4@
zlq0hyTf}Uag1Bo-r2%k0@c@*RM8Q2^1_9QMMm#g*RL}p0a5`;+e$yd&q698BI^+Xx
zMkwIUTw^hQR3VtSGmIsh{jNR*!P&`a@>*4Ws!XdgB!Sm>P!rRU8Y<{Wvcr8;^*x$Y
z_nRr#A$;e`m?Pw~wIZK*c9C4Xlm_KF@K6GXLAND!TN=~@E^Z`r$ttXB-RaniFQHEAmnp(k*_Js{W6e{ljCeN7KGyH3mS-|JW~ztAxpt8d+D2jm09qP%%@pb6+j-bIzG>d~M<8j<%HYZ443TOQD2<0cavqRMQxh
znqgsnnU=rRVw{{*z#TP})nF>x>nyM*)rU9|=IC66g7v}Sib=q@r^#VTcUpVVW=gwP
zLB|N&yizd5q*B4=w&7k_?wGNn-O=Z;mSAhTjlP{Jdf-9%L6NYL%f+vW`k^}*(M@p3
zfE6S1yw4RJk(U%Cb3rZUfL8!x(yS7l-dKay;o%&C%%AvJ1+wRdE=
znh*D9K=xBYjS|lpYoeR#%qMnM>^&+e3^fTq*ol`opZ12)SCAJ!SEu$
z;QDA;$Z|N-yWEcBqxpV`eZN4I3+|X;+@5ptfX;f6p$(|OEa0SIvLc}%Y;R>g40+G3
z`Z|KMolk-aEO#zT&9Lk0FTVV!f>l7HoX-Bqsca0+*jC!?S6gTk>1*f64M1#~+*ULC
zTO2aO2Q+IHZDMOv9MHBqES!u;j{gWWIm{d~3zlzcFYYHf`dBII(2uQDA2zo*Z${Bq
z-?EN7Ofaf`o1?!qAGmPcc8EYRb6HDAv8r|0dNOVux99;{<#z(t&Et5K)-!=nYzGW-
z!HIFD#EUkMlbp#SMG}Le?jhITYXc@WS#G&bGp-F*cHoAnxnJ3QQDc8PWbrD^aetPAE8AX95@uA%R0E@PEv>}pO`*QUuXQz5IK8Kefvy%Y!=w-(I3iQ~Nv=@m0@OVtWt
z$-;JiGHr3gZcrU_ODFaObpfx;@y{O)h)cFj{V@@))#5ZLspYj?Wmfsf6g8^`*yms?0s?KP4T|qN>Y}l-}^|`HZTR6I>ZO^Z}>N%_*H%wDL
z&$uXJCdbkI^sxt|n$FKJE(HWVpYTt=g=pTu_fF4JoUm%klD0m+A~+VFUUA57ahCxp
z*5$W;!JWy!h`4tA0lVdYTO=QRQL4Vt?j0G4#~J2hGZG$*ycV5s8E672NvjK6t106A
zSVVem{B~ez0fo~{d)jGLwWV@HH2B#;{dfVVUc>l-UgJ_@
zjy(CfNbEZE;#_v?HKw2>5~+X_ad;e=Oz7F>M5znGM6~vZKcy&k04)g11}%S=d`|gP
z9h+pbBZh+B^7*2U*n;(Nsd<{1_C6>em$ECq2E>qiV%#>NS-0ZKw<6CE)7c{p{r3Zq
z#N*y@v_JctO+RDe%6HW$z5hkr2{E~rB(w}|HTS~pb4HF+j<2wdoR8b;pifULq|NV%
z4pUf!U?t^u{wMpGwRRqa
zjzasAAIaXP0a8a!5`!u{VskZ`@{vB9dxdA||7ZCP-Na*%@eN;r72bV!eDWFR^27Qt
zfd$^(ic|U#Tg9w)lSLOP^dL!I#Mm4Aq&Hfubq=fK;Dki2;-1BTez=Lsv#@rpJK0=4+GS=>uJ*~-%kDSZ!`=dW|DD(q^v>W}Uzyz5&yx$h
zw?S*PONy*bF}qUC;*LlDJEOm-p#0a(N7+gGYjS?iu8)J$%^4(_(Gb`6B-VDc)nlHHkIUryyjecX3?Rsqtle`lFxt1?5w^Y++o3bRp3<)cSn`)%$<4ej4q+aRRVBFK&>ZGTLACB`>KB
zr?5x
zmo0@>@-9*?-8Ro^!`Tq>xTobXbfRzgpaFFYUX;oH$Y;k`w&?u(YZrXX1&`Z1mYB?T
z)ZM8fG2n^5xQJRvEj4{6_M7zcCT_HxgatVzxLFtNI!o06BFb|V-vadK_!otZAWu7`F~DyhY0rG1{m4qBmOaw>jz*e1AqzO8SZlgm=uZ
z+*@b-N7ku0ocphvpF4cs0#$}>5JnRD48
z)T}fN24ekv9g^HJEqoXnNs&-HztAk9(BINrxzcD@gjZ?C^9y1D`;tK56!+oH*a2CV
z(tYoG{^SbEH47vawE)xHKRYl(f%Qd-nby0+zt(QG@^RiTFVs73_7Eu$O%O$H41Jq=
zy3|Mb=WfvML=xMACaphPkS1(Wlsm}pWPl}1vd~~xM(_odT+mZAOl5)mC)S!mEKrSt
zj}v^g3TO)*NGudLhrf^^;Dm@aQlDb-6g-6alDII5tWlqw9r`>tOuN~3oZ;K>+KT};
z0&8#KTe_LMb$+jgpk~`89T4p_IegS}3YVr!6Pa)dbMJchejl9ii)8wP-Fw;VR-lmc
zJ2)Au+spo;9&VoJA>z8V!r%ou
z{q1k=ggYVr2fGhc(cYuw)=P1tr)Hzy4wb6y#Li~2p#%fGRuHS_B||H{E}4LkjRUOq
zWYa~s4?jMr?oZxJSMsT7HrVO_iX*i@doN9_M1LosNlrFY&!ljP?II&tIEMFp$h$H
z#TXa#V_~1*N3EQl9(y6L*uT8F{Kdvho0(MX<>J^sdxcO*1SfXw^0rlgbgNpgfc)@a
zVp5#ShVar3{#}E0*cD6Wp3O*O`|lv0sod+Fn%98C!?#-kI$l!9`YV;~*-ISx6gv<~
zVlk#Ltzk?62Mm$N<|Bf6=L}9cO0Z%7nkNP)S-OGvr&{b0Cp~
z|7rHCf)BA>u>#kQ7r(v4dV{)ZO-N22y6rC1GLFW0SZfsC=0r1{6TP8Ff|rJGVnv>1
zWYYSPRcwO|ky~8_#~$9B$>)DeVJ!JF`lL%uIqZzLsFeE1NW
zKY)C#L};Q^b%mL32*#nnf@SUXYEPvN>zeJjAYt3-+bz)YQo_LbC4YHmL9g#l{E8o1
znex&=FM-2nZZXM=*Xe|;6j~n}S(5GyBzO$E`=>D+_lHP6j9}uhoDT%saK)1+D;u^;
zlsbif7H~2}tw_@+I2X10d94qD@nI7Y+rjKsVf6a&dk~vscVlkW&i{5xsDr|XG+OhA
z1W5haW}fv8Kvj`|4oH>7W8#bL`UcxeZ~aa;5*CujlfFk@qI)!w*oS79ynwSQq=>7#
zOB>gH0mA)8+ktwL#sp?3YFDdkn(rEn`K>OOcW71dnGMI1xSLD{9UnL@b|(7Q`578&
z30Qg?41DT}Ob%b;sWjUTjpxnR*)jb3V-?dk-JfszIViV}Ge8n_^wXI5>Z0}gyT-6R
z@hXxpkgeSh<^qAxw*lI8PwK8+WaOVVj`giHnGr5p#t}Iz^PZw!k9Anw5?R>HwAlVF
zznm|sCD)ryhp_fbjngEaq`x&gn_x(qE#Mwlnu-_(+FpWyhkb#jI#nsf`(>|Z;etoK6j0P-r3r4LfmrCwE1f+
zdOP}X*0DJ=DrfM1qdXs8i*VRY66?=3h)QrcE}{M#PJy|FGU5sSRxd=j4?aA{!g(uV
z)Oq^rZ97+gG44EUE1mYASE7+!Qr0IqgybD}OnyT?P>FRs(v?0DWg
zMD!JLu6*hzH{_@LZti|mYx8q!c(AZz)P*5#!>~9%TUC$EWNK{x<~mg{F3}UI@^)MG
zXC&z>n(qbTPgp*x>|R_KEc^bh9f)>KCH&ud4P1OI8%g=3*Y5JQ2&*^WHstz541Scz
zGua`8Wf{fbEY{K#*V$)1!WT?j(USk=t|F@R+RIJ-Q8@q;
z-WH2drw%hpx-D)&C;MKsdd7tAUuD^(Hg@N((Cw-2`OC{kEWIC$;!oeH*OMsyk|*|x
zd=T2G?Cmn~K8kz$c2j8{66AJ!OZD>(<6y{{g!BzoXkn?M)iK*AcEcsJY-RUE`5%AP
zEz%~`p$m;G6TD^oA?O=QQ8O^1~pl_P6KZj=das;POBel{&8l%dOmxgbK@;jD)M>l8mJf5>bCaT=IV|G^eYmdnJ
z{jld>rApmXWab%n=xbNoK
zXWvz=rRdY;&TXY#>Rl@e_O{m^>-hE0XVkMU^4UNdcFUcvuJ~mMg)cnza&W~7Y(@dy
z5q242iEYa2ILAKyCeu}Vf9!a328YY0iEH5V5LK>IyT|4CC^J^J==OeT)a!2tzsC#&
z$6@2)+-uV%qF}t_>vQiq%|L>H4k~Tj1KHTJchGOvbn!Y6(4gB
z?`g+)SmNzQbvw138mts4(#C7~idsFNzYrf_{H4lne<&jDM^JxW>39ynX*8s4mZ#PC
ze1K^kUzQ5H*q!)cqV?g2zRvniSf$-!w%bz4N4THyH)JjKCHv*ng`|i`2^SsrCs=Qb
zoqWb)GFxH&?6u*tGyb?I^g0CH9;p`C3tXa~;kPex_q2z{Pdcp;{;vG+N}|
z3qN$fPU>4qbRT-n{>Uh&9Kr+(`^m(eECJKf3
z9!jLE2C1KG{DhP7L^_lvB_+T&F7r?uzEsio*%D`!h_*c{uv%Q
zZ>8EiG~JZxiub9{rK$>gKQg_K6N>PP{oVCxs>Cp5cN)I*{^)1?El!?YcZi|+BeD3C
zkwFlArW{3N1yI*!wEI$
z=)N=cENd6ag*!=&zA8nTpVET<9~GYh-j&Hv&Ttb2`*-YcArHnhuplc&O0
zErX=EckH&BQpR=_OBoGm9eA5NDrfADYwE@q2Muz%O`D4pCO9fhKBhh3%chIrA=^3i
zVn6TD9Zam53T4u|#U+j&7VFgw_l?4mqnWgss=tS>Jw>vZpS*q@MX3
zJh6LqY6nU^(2gEIg^}Z@==fMMqHC}8{=almFPot;dB$99-@OO9vuD-iKY98L<<*Rm
z;r)7kM&b?`GXYDL7x&X=&rrBRO>AB^SC>ETubKF3)_hGm=gO8DgMRIXgzAN_W3PVL
zd*BG>OdgGRaa?fu>UGrn@n@7RSp-}D-i^mko?^VpOP&7e70aS=+fJA?auCv_beqxe
z$CT;V@y|Z3u|GqR-K*AbRvzI6Fl6)uOd2%^c~qMn)vX)mjXyVU-@%`%?PW_8LB_PH
zy^-Tnl69N6qg}Ix>Rfcvfq8Od#qu>9FlFXkiG{xH(j$T|%|!aDh6cdQ&tJ#G!Xw8~
zu6Q97E>;qw$4$_SK>I~k@d?erj`(A;*
zhepsS`Q7_?*m&k4`gdxHLH+yS&*`&}E?qib>HIVGd|t9-368Jv=$UD)?9TR!JF4ld+L
zJ9hd!exJ4wTmLzTnCkt69^!p|Zh|nOL#gst;qU!NkSkMKOq#nKPoBR}p4xdZq-(3s
z-Z5C#8$}@Z95D%t*KO0!!cqltp?%|;i0=|p5nk`mTY022?y-7T$(|uK26t(Rlu6u(
zogX^%Mc%CGap=SuY~FPM(V|2`=VoaU8q)40)6^rFZy9U+t
z0EweMoVsus?S31HuFA8VZJ|AX@e*a5c1FEQCDnj)aID?_4`z?-jq=U9sCKFEhq=pe
z``!cWJ9-j7tG2Tq8gVV(tScH;Esgq>OX)DL{;-S%tG8fq*Op!-^$@>x&mkOCp6**_
z{f?Xpd)m=c=P-5g8hviUh4n~MjhnRu(G_0f`?SbcgFVRIgs$Q!<|KlU=)Z1!tmC
zKCVoeGJ4V1`0?Xw!juFk*SL}s&;L+!7F-`lfW)j
zs+1deRMlvuMJY`@huZ{Cdk`}bql
zu3g%bo^hNB#k%Qb|Ni}Z(7T>yPiBv6l9+Xd%mrd3p>FSDx-$&pyDT
zm+z1>QDih%e1jqRFDag+`--zTbd}%Y;ZRzAv$s(3b*3G|lh^MsZO?T*aNR+be?k$O
z6BHM)LlLAgB7}0|T#6@g=UhDN;nJgbFlEU>#@gyikq|;gueLO2k)u5V@W4JFa4tatfqGO*X~X3@D#mb35dSHlDtZ
z$+g|?$=4LwOPmfUej;qB;F{a$QYa{L}W_z_ZuyQ6hD8^eK}}Jv)P$
zW_s8T0Lgj`o
z%g!pE=#}D$!dt6y-%srlLF2=F1|2Sfx_uYa0g(QGj=-u_tJKXahVu4ue0iKysJUhA;g>PAm5Z4$&JqmWX0IlvJ+Vkm`!M{?BU;g;ULCDTO5iWMuY!C$<;V)=4ZQrhvJ8Hw8-{fFSxne(Vp
zt`xTH*o|yO%4+G8dNnHJ0TwJ;FUln1tr-yX$yr1y*m=
z5_=9DQo7x4HMa$85=9no8^ulrF
ztzD;TMU6qIP$7{)k>Dh(pQ$Oxdk-E_+ZQ@5|6EZ7KCz%t5@yM&b!hf;HzZJGe+>1K
z>RH`7lY>9RY>OJasL{9;5@*hb|BV`}Kh*49vnI^8DB?;yYbjY7W9Kej*1P-&JbU4i
z9x2b6HIs=(@}!CJLVd_zymHk&%1|ufB`$w>q!v70H|ed2^#myr)>%AgOx4iEbg|eYf91xpWSAluWHY
z3>D_JTwm$|d8_B7GqItRLXJ%75H+%654@{4Z)4fU9T+xw4oZBV6VKFhsbQA^__^Qj
znoQ4^iFQjpn)dXNs2da+*l_g
zu|9I-NPSGNWD@Emy-Ci$c<};z_Uu8ECQUS1O#=P*-+xDj3>oyh9KE1P$oBlGi%)qu
z*2BRs%8rn%@7S@UCTCe0%c@teo|gY2SsPWoH-@d(;kt6wJ?qifMvWTUp($IoY+4eH
zZKRAB2L(9WfrB8+Kfa@0yLL@W$M75^OqkHqffx3h=YtqAExp|7{cZ2JVZtok(Yqw_
zIenYp(T{)v_%OTIsx)OuNzxqeqL`R#d$sP
zoN`dYD_Kci^9Cc^GJSfyV#y+uPUc3g+RGa?Tu65*nS4p3D5t;BxS>ep)yMCm!;;g8
z7dZ@~DM!Vjn=WZtkbFs^qRsaSHMh}INqx}BlhD(cq-}d82|}mEzKYx@xlK+blzJYf
zQO{&fN3Pu8AraM
znUXx=ph4O3hcxFx>0CPfHB#+PsYu$As?Fp>dVK6Z$>QI)UeTm(Lp2ycp27@rP?O)R
zImmM8{Y{*HQ7Qp-)$Dx|$DKb8h_KCzkKDm9HSj@#m{PufDr}ahe%dq7+WhKco14!)
zp*+4{BBS!qw|8FI2M#jS&g!DubWQP9mDRa?tjKv9y}^60K92>5Zh_Z_XU6M*g9^pR
z96%X`-B#t9PBXtpdt2M`E+}kxy<~|W$vrAbjBt3VI4*Kqwst|C?_91P+m$4GIPE~o
zw58L=#M+iwF|&Sp+pKy3Xmi_aDD%-7l-CunMK_+v@o$C&l}R2JOBc=mtjwSo;N}V%
z6nidxc8I`%J-d)8c1VZuCW0QFdpcxC6g9Q!ff@}kq&wDvl6&}r6w@TNp7r9QaFS5Ku!Q)~Qn42jf(aeR%G+9Rk+#+dp+
zFmA_nJ?hHz)-rf(Ya9umfABVDssRc<=;(RhN3GCaS(QcUp3>v(uLw_X^=&pD_k0M^
zI19m8Of@ae%x%pqE>m;rzdFxzb+cb|iae$oEHg9p%|d3`VQrc0d?+g8rkZM&zY
zEY4oI46`PVLY55aG_l*Tbw`j)KX&4@BHIsR_Qc=awackR*k{Nn4DIuaCUui261`8i
z_KMJRdqJevIX&ImG;QGCj&%!WAbpyY;DVrW(qz}fdhMze+~c`ZAa5>}CY745`39{T
z*VCTu14fL|I_LAI{Grp9QQo|1au?O)d3`12@=N>H=sjoz*xt39w&KsxLlD;`Hk#I}
zh3@?ZgI?vu6zNP7+AE`;@yzifQKM2hZ`EyE-orUT5B?U7ZXU<>L9`Tfqp;kM+
z{zJJq9=SiVeXYCaKFj6ogSN^$-WVlk=;0qeTsW}b!^cii-g%KdcmPsko_6nn+qGx%
za!U+yXUn2U_wo8a!}8|HhWxp*a~He!9n|M^!{V8UpuFC_Yn-w0+pw|TP44Y<{P;+c
zB#D-e`n8ywccQm5OF|zZz`?{`Mb-4lgLjcOO&VNQAN2I9@(q*o>KmxqFVip2D_gK;
z^(vGuQx@|Zq(NS#-zxTP5U7zU22#ZgqZWs)jQI-}plmsh&)KhS)gk4@nmwHxspM5z
ztELP=s-%gKCrdh1YuyvgYga(upPS;T8W>-tBukE+I*+_r(&K*H1i<;ct{Klp9GB#Ko^|8)-dpCfIr47?vx-zg68&-eRmPU5X_5xls)y%vTvl
z6{%mdTu~2Aq*wX$C2P^Pel=YlBO^zQfI+JM>$mRWuSKhrM|*X2X)PeFDqht``#lcS0t~eeW+Zb
z0OqaQq`mj)NI(z$)C#vusgr9AOeY#wFRQ)#b7xM6V)?S8YRQ5aKYOVr!Re9CL6D>g
z7s~O19_YP)YJ_5Wv+F^TZfXF*o^gi2p%s65h&rzb`DkiT
z#;R@2x@q^7e2uL!eb%%dHKmg!il?M&+z4coo?iac$^WrVCB%;t%LF^z=FE^1`xQay
z6_M_+080ct%{^PsiG#_1{q>h7hABltqsKD|X_CDpr|H4|%P+s^xPk=>YJGc>$b||O
z(qX-N^#cEsEaKwq)DkCwN+U^oKIz7`VQN_ie
zNt%ks^Aw)15nav_uNM^o_?ny|d{dKYaLb
z-F8Z(kjN*QOvx2P?y_Hw;!M3}OTs&N@Zi|HcduUChWJoQZ|vBydXZhGBO!h1(j`6k
zV4Md?z_V>^chRCn^?M_aiAmkMb+wZL2N1aK4vk*sbLPy^vjTL0%^!N0bCv?V?m3Hr
z11FS=A+gU{1=j1`Cad=BPov(UjF~nP17lq4gZ-dvjjb^~J=U#t@Z8hx-MeR-{Ig@u
z!r&|eIty~G77k(%Kg!(jT5)8Y$-(Q6n18L@*@1m~QMEuMM^%B};FU7OQYr!0P%X0y
zQYkNL&O%rIb{JbDwOY5$?Ke*%o!;}!^Ci$_anQNrk~8|-eUM_+*qyM^#$gaGd>Ag(O#Z@`tbt#(IJ*^w(at}@N#MgNp{TciBRn^{*`1|J^
z+IfyLBe{p|b(CC(&sQ=$$rXuCzNNv;<7k*GF1nXUqP)-FVMD+BIHA0+>FrOc63)b*
z^wE=@oJ+nD`_#b1J?-Gg!l+~3si>7N#LMG{hV>#xE)u;?I?8uTU)3k
zac?S_%e5`DxJ%fmVlZgqC8cPP#63CNOfPJDx2GSlOAiiIfgA5*5|JnJtj*a~e`%%E
z;K78NY1B0w43q=1p4WTN-e}Kr&Kzi``bNv6-WAzz67c4HkSE&PDW~GhLh3ZOf4G<~
z2No6{zO9J!_b8q!hL)S*`R1TOALTS@&p4A}yN$;^d7>zaC$r87FpitNc5IDY90wb$
zmRL!ro;~@(8f16hjVE&ZzOJiga%@%uKD7Sjl6m>NQ_B^^P*?T1NI
zDML*68yUUoDJx?h=-hGVB=&s(fD$YqN)O*!7(hE*T5W_E#h>b92Ku3K*{7ZW0Q2-K`x4%5aCcORf&x)7?(Q^yRm
zX7RAkK?Hlok0#D87(Q-_UetB#ig`$sAfBGZ@J@YLSdDelC^=+nEMxcngJ@j0hSsBB
zyJ?#mJXokDc{pHDw?#Yrq-G#crssNXS5voa$`46n&}ril^F&7-V%KDYORh~dK_
zoEq4tyv#+Vi%J=mJNF-=!+_E5MFv+dpl2>!#Sc+R;;+HoF>C31Oqst@d0Jo9ggj+y
zj75taNt5wB<~v40C3SOO4UpT~nTuByA@{+$TNX>D$2VnG&Z>OI
zC|P3&vqbUSp4$%|J;8~ym(+lH7|{FQ*p^wJ`r}z2=&3(>{t8eAhKseXRi5tN(dVyT
zrm&U(TEAl-N~)O%>;B&3LHfnCgF5sLIuF3EgU660VO%A%Q(Mmvh#e!ENuoXOW5xLV
z-GK?{iEe8z)r`B(-a$gRy@yAS`uTv|E9GcG$s((TQ0Z;cPF=VfZsoz;>Gpi)Md+-Md{L|wHyr7c^1cbt5Gf|Qlv;)D#wJ=
z3!c~+qdX920$7c5E2iBfpxFjuMdOwdI+PGH7EU>}IFfy6)25A{fk1~CdX>+dITQ4}
zH^y*E1UWL>g<4;!K`$Dw3#+jlD@kwS*k5~T8SkAn4fx*1GPZBuu4Qp3tIGir>ehQk
z$a8>&|EVX>nG;`QH`V*QOQcjjE+VXQ459bBv0qCi)v_jB0Q0iqD0VAp6&kN`&QvE)
z!kPKTmN(CAc0Fu0=b}jZ1qW|wDG_oS^m4a#s`HTP;{oMIilvIKISB?OQnUX((~GAh
zF3EUW^zd$61!Zn%5kiLm`}kCO&a*$7b5V|gJ(Qe`F*nZTg?qp9B>8k7O`xP~3_;vf
zsVme{4v^Ug|D)SiMDZ@aDAM27$jNXGymS_+-|Qb10}QtlRdG`!<7xc8P%?+I#kS^E
zn++jPiDE0cZ9T7C?}Lgz#@Up0&>KCPl7KPoVHo);dt^L4`diOGrJY#0_D)+R=~F{_
z+FN-cPkADRg{@KT!B$?ztV_mg$JWRzy|u$Bae%}p#`O(|+47@z@$0JdS~keYG1)7J
zUi7!_+_e`K9A)AJW{%aq2jwe_&rM()iTP*X*u7IJfM?*pDcCOR%+#fym-m4qwV}&uNf!SJ&k2h4-&j(YO
zkioRv4`(eLR|6%)kCHyMrn!;Sr?ZPTkhLE7xz*k}wTwmDAq+rhO(&ctt=^vd1U(T#*%(8#Grs_xtt2w4K|u
z&^queBX-OfxTh9!Wg3=EyRSTZ_@bXeN!Qrg;+1Q)$Nu-HO0*TPT)^y7e8T)&)1WoAgQAFGt&d)~I36NL&X0?vux-qu%oU
zz}6y0@c1sk_(y837h{yRF-G~Bbjg!wxtNTl8|dI7dDVm=MOEm~j
zvb#o^BBmOLTYAtZR|}%9*tA1S**t&oN-vsfo8gw489*|g&+F#3DxgmJ;#&XSnvOFW
z_}{x0D|%GDkSpbH!iRH{W#qsFN%rJwAj8&}o-+ol#`I~Ee|lfxJLZ^~i&3_4-p|@+
z&1c;Yt=`!lI^i9J&JUa=>6wwRq}d}!=~+#)C1q{PB#`^v<=O(cIpIn{*b6PX1%UWzevCvvY72^*s!7Y^rn=~`t|EEckW!QUAtBj
z+9bs(|5L13F^6%cF0*Allw+jCi!qYv*0llXJ#K5(FiHc(ixP_x
zE!#{&n{qEC=WW;FU|xFNd)K%Y0F46|W)O!S`fR6d*-Xa)8G6oBDu{K_I4D9~C|%=~
zF+aZnIJiPfkRXB9%jdv?SH|nq3?O)w*0cBrkDsY^%#=5u4=rrS;O^3}%t{UQl}g5j
z%&L#iwy*i|T;#q*AH?#=9I_axm%RR
zKBam{Vo$G^-Xzys)z1G{8A%e<>R
z>^ZxEBtLcgt0|cs+xj>lVg^O5?QpbDd^pfS?#p07U3WTeCQ+X6w#M|fQZ~%bnI2x*
zO{KHX!6ZuYRaan+|NK`Ln0qUI^A-a1
z@cAy-Coo0*``PLf^C3I*^+qIah-skkBVQO(!j+T1Y;;RSd
zUghm?wZlpW%}oEXhTGd;+cJ~8Tn))@de_%hi_?}@dRx3vQfupf8!y|-ilpksxw=Ir
z4{=G;^Ns`qr`tj5WX6=XvCZk~1tq_y`E5{yEDMCrYS`YSXL%GO2}NX@`y-
z$NVKL;Swt*rp{T22a4F?c9mR4Glc
z-7Q~;BiJEc>?Za`;FKwo5#sqRG|!dT8>yX>XWpdm1E+#
z;PZ&B*YNiRwJ4qPt|^f!8h$Pi-@^;ENz3gPrI9#w?=3Zx>bV+J45uW0;-Iklz;LFu
z%%gPW=U&3Zi4&A2Y-^_!E$20Go>yKR9e-+rdRd|=4Nk|zl2a}s2Zi|LH|dkK4yyAI
z>>q!>)&^!z7!i%LN7v_Q(Xy4+s4w29ECqGUon@}^+B||vNk;Eji
z<^X}QJx7j%-u%D!`pE=&x#a+Wy#+44%xEWpXWsElo;b1FdkMA6sl`s4G-;C7g69IE
zGAlN>i*?mB@F{xCm@QbUkaQYLL$H>{~uPpx!%g%eNCRHTMH&9inIk9Y4j
zgVKRCO6E|$g+^%~67ViAF6h^MaZSB{mE;T;Sp6EKr#+3*HC>Cl
z-BUE@viXv?I2TICkj&-cx4f3d{e2Dn)?HBCj>q4M!HHVs#$%tH49%?#Zi9$Az>7a$
z+h!I|FMrxz#lxie=#@{&7!vi~;fXRSo%fAxvSPZDs~NKK5(cO~mrX)h`<}XEeb!#o<(Qj)2i4CJYT$v!XY9)Avar&FiyXgwfscBhURHgfh^Sa!^O4rOY^ZGXIS#FPU-WmD#
zn8Wj<@+FIEuk3Dp25I@4)f={;e#`b)vuUgHkdL7C+F8PdOIL8`?me}D={W4%yI(JI
zX6u>ac};Gy^m4y`>$cWlkF6GGE0{00mcyC9bR|xlK8wcfx?sqdiAwi1B2J&Xp!MTx
zRxXbPe~#52;9S`C{znbq@!3w2&W_Bhx^44CS28hN
zg!RPkjaau}y2jUIbB;H(tk{{j|>iGNsu+ef}a9URXDc8uk@`XsQ>NWdJ2$HgDgBQ4{{e
zx&O0w72r`^UG!9d;_iVYkl+M&cXy{~ad&r$7T4nLQi{8~yAue(-M#Jq-}5%JOqShE
zHoHJ%Z@$gUJewyYZ_c~-+#BL7Y0+|tJ1F!ud5}#Urw-}U)%fcvy!@1gwX3P0IsfK&-FV3r(2{3tB@gvnYOv^%@^N7{gB^i=3dP0X0sV~Hcy<(}OxvB~8v4eS
z*91TQckS4YR=L0JJ*&SSZMi3HNvl7Hx&7dL|LW`8JZC%{*tbUkclvO9_BAU^d&{CZBc?6FK>>2^W+hq7|aWgCvKsq&>Nhd>}
zO9466wA4OoQqz&CkB<+?Sx@Jy8ubV@@i}KAcRioO)as_tf`y>HrFs(BB}hzgnT}Kg
zk_r_nsPx^sbu;plQ%juu@YL=n?|Fv~9k6%rUZwED`jA$J^|KaddR|h`u;2Ri>!&oz
z$=l8a26J5Nrg^nvI(R*B;DF-Yr}7N}eS+K6<`0aJu7y4TNRvZ~9%|dO4r~iK@~J`I
zzkh#=KT{eSg52cupE6~N;`Tp({=BkeAZV{GNYKHo=8vFUxpJ!S*q>+5o>lE&8yFrK
zVO=@?w3J}G$BY@H9MiI20wV!k|A8rkY3zMJP|=Or;g1DKrnWKx^a;BS-|4h)uatpS
z>`+c%Nh?D>dOB34k_v?y`j;NORv3$BwXb0=bcRaB!MC6O{er4QW!C*8{mOB{G}fGb
zsI;Pcthwrs(Lk5frd@@(B#R*!$$imse<+yC8+2HC8I$(hQc8XS6-FNOejBdgkdHWv
z_J{5G2W-owGtA3I3Z6vv1!`x{KXMoOlgs@niw(Ajy>;ff2xgPVeU1R?1mdY#K4RMq
z;lcmG=sO^wIuXbpRv`sWb(4)jeSXe(sg-|6lr{&;Vt|jdfzE`?A>UH?wkpp&%R}`5?j>bwW;rU*@85`uo@lc8qNKm9LdefHu#9Ni_?h9D
zwnIDCw#?K@G{`UBe~U&^jwbcQ0cs6-fKl{mwRm$|wlANU^zg?O@Ws*(3&o);9iTK7
z?lfCuNZI7_(0q>u3ooc_d>9h!=gKB!kN0yUEe{>vPOhF3EtXzZ+U|^}#lrMjse}XE
zWFQ%-6V{o)JQqa@j6b1b=v1$NDR)I-qf7eTceU$BtH_T({Af`#B1z&zm_1=6ns(@c
z9K{TpOdjrT(44FiuqR)9&Bg{0*Gn|Fdv<80fcdWdhspUj&^vb6$OHZP~d;fcZWC;@dQ7fYF2c;O_;~QNDI#
zWG+-%#gjw7f42^5pOic6x9k9^X((f*&p6rcKYXm}ABGjmms_0icH*u-;q^cA`an5G(xte$jZK|A3H7qXQI5szz2>atciz9&ddGTPc$wEUkTIG%c_JoFnPxq)gJRYD
zRNjTwgzuLh{nAVx&5I{0
z#1v5sh-56s+W8nq4N!8qS8vl7wc7Pp^eP2&WWtzUZIzbs%watwydTn+uBVD(J!b#W
z%c${gI(O2Jxh2p)u1`C39y&>Jz>^lI#jh37uxe?=O>eCYBmCM*r?l5@-d5V{T3cAQ
zb{>XRD_sZ^=l^AB%kZ9R5y2}}a-6?<1GR+aCm@LuIU?vBmcL_XFDb40@pG0d?)gk<
zQUsLU7*8j%D<=&`r4~Jry;2huUaCM&^ljZx?LQNmo(#NgPH?+$;lh|WaiY3@$S+O+
zRs&n9anJPBB)0@C>&Mn~teJ21>eb4rC^ep`sokST57it3^*nZSb5rMT(4c{$C+XR<
zr=qa2R8M9uRjQP7T1wEK${+eVFkU<6mb51Xg$W$f$ts2Iqcd3wdE#rd&ojKkGnc$L326m?8hBn&eK57E-6m(@(2oYV!9IO9Em$svas7NGA$|
z1Ii@F@8aZ?PEaXq{|GrZB`|eX3H|S}HV>cIWY<5n{V1
z3CF+22%)u+>8%N`kvigJm%mfDUt->odziZKmS99KMhCXm@@sMIv$^%sp{{7I7nyKe
z1ruBAQp8>Dlk$jkGUd`4lx0h#j)Qh`{fw9EXUe|YiiKfd84pl<-<;^5Pu%A?by;;q
zsH=~vW
z#C|3`)Rre899358D0HgfBWVcwm;=!D=cCzlB)O?gxwM#um+c8LQDWJ8#~%9p+CI6z
zx$?>t=oGSwsntb`5A2rIId{dgw})t!o2ye`%h9J}-N=2$kISG4`r%>=CYTd&NUcST`)5iDYg`?Jde*
z{zqQ9
zU21$QzWS6NHkHY;!Yhp+!$Q@|SFc2a{yE%TljzQ9dDotEYo)Rbw#O20rIs!zTb6tJ
z{6#TuPY7!D2PFLXIw}-D3B`;)xtnW1$pehPefJ*nm#+z5-&+Bu4JbjQ`gP&)WIyU>
zPY_U2o8!Bz@fK~AN2r%AzG!9qgVo|_ns4;w4bmu(yXfRIin4|lLENV_$R1ywIlw|$
z5G*E(owIAF!bQ7P>0Q~5D&E0`mCI!vp*{EpI4wyz|hUY%c2FnvbtHO2=+)>PJ&8DMrBGX*pss1Uz?_@UX>Z>#NFS+rjUUxvv3l(lfBn)4fdlUJb>jX9p#rqE
zomv^RQaE$jTc~;BK<^}0i#Oe}?w7?$E`fKBpS`kj(=Mz#bP-FIEw`&E0_Frp3BU#<
zynYC>k|SKtan5|yTEBSlqN3y>y@^I05>T0@@dT);C4KPVK~S@rKyHi}G5jUtJwpJR
z)HNCxJ^9KBiu2mDR?ipWw@S^k1L?$*)G_DIom1^$Kk2^#cplQV&>BFawb3eSElx|L
z-_rRzHS5_1{(>UDzP`$GfSTxfj`^UJrRA}A9)JD%bw$5pj;cmGhqhJodeLy_m=;q8uZ@>Q9&mlyfc9U1_(!n`@VlMg
z;q(DR{&-F6oyWJHk%BkU7vJgB2R3PGqlekwIH%u(vENKkF7=enAz}{C3}#qq&v
z&4H`AJn~Lf6K8XAwuW)iXs@}TxpPK}B$@%4Gnk7~VWtZJ@xXv6gRr|iFr
z-vwYdCtRG|-Nrv?eqAhgVu*t%{RE$s8WfsxnhN=^GbHwJJ-~KQ>;G#=W|t_KyX}bG
zMLWk82#w=PI)$e~50)aHo88xlwE~>?ueII@TpVfl-n@Q;$B!SIbx0YwQVglo9OEmf
zC-FT=${0(+#UH(pW+te@d9jD9f1f~FS)@!My-QGqAhRXssE%rbwuo@*;#C|ya}oQG
zpT(xxqk>xHAcOHdAvxSLr%e&)bF=Q*y;27{eLx8vH3zY!99}<~7QAWws5P%~ft$t!
z<(TKP1mg0ZF$KWg-QA6+u~tvhg#=?J3p>=_)_C(xlkhw_bLKP+R>#)j%w-?|9TX5+
zUzR;!zP=p&zPuMb7O%7J`;rdjCb(Zk^@iD-snZ+K-a=dF&8b|+?Jr}*AUMAJ%fn3@V?3-^ell9
z_Jzie9+>pj(%4gHtz2@mYaH@gm?O491><>c;yzwNO&nNZpJomuwq{x;aUxsJ#r(Xd
zcyBpH#>&|V&K@xB0a7~z9U?X@&3PPFdqCc+{y;f<`hk9Q{{>t!v{9d#jfqcT*s7ZdI!?Fg*=W_o%(WSFza{#$kmbNXDB8F&I
zgwqcFZO!@OcrV%_(!aSM2kZ&Om8l&K21>f6NokqX_xF3=eQb7
zb=i;R#wx^)t2p3`AH)9wk-gW7x%|%i*IcjI9+n_xB;_QB9-9yQ8$b9aSf1Iia@))VHGpp6di`
z9Z*c@?PzDqoIMbP9^gHtSKbOj5uqlF6Oa!zL+QAN*rOruxLVj=5>1zA8WoUJ51NBr
z7H1<|!3~Hu$M#w{b7(!_$d%8pr6I4?^;>S~%#<4LE{Uw0T)uz*LFjZYBekc8@K$~u
zO@?pT?vQjYTeQjoYS)t&uWUB`3#d%SzZA1qN|Xqn7Y<@aw*E>d1(+s*{kY(B(iUx(
zxMBrG`W%`fYR}haa{RM=(CB@EX!T(FTjESeFHCPQm%Y9`{eDE@c6|EcwLU0F`_6HN
zI3^2y<4XD~(t3Ptao)d<>0xSDy?bZBxwCraJ4yx8F-`j_rI31tyzHq&Yd^AB2wD?c
z(poFW*0{04SUF(C-nQX#m1k3RhsvGV5A^5bySrbS1fFp4tN7wCJmuLO!{ae?@dA?xL1JtMp
ziF!g!mRwQ}0e=@~AJ&9UI!aGDja7})JH+7xC%hB(+{OlRL`cu+S;UcH-;(Y|J~-35
zkQ;x@PG9U5$E);$T|&HGcM!@6ZIy86)o03E_nAAd6}MZ{Jc*2+s>waKP&km+o_mA`
z!e`ICb;QAJU^wJW>+!XZ3Hb3^a)cPJNyn7si2B%52A#PUFO*+7akW;?)$2aURiYAh
zu3733Iu>51bpHDH`ptOq>=|-q&5S?#_ducq@y%9a_x?i|IC4DBUA&AmDLhc3NI^7j
z@GD|OkEW=I#!s7#*^8DdZR2d2GN4`4hA3JvFV0@LjAk7T4y#NH&YIB+Uwj~wHsAHkFSCAxcE_e9l7~R9JCrfBv0M8A80lm4Dy5z+kVS5SRE%9YMBkqy!^$b
z7XKiW^Wq3QTT+9mR$oYs2R`g5#LwBb?a0-=%9Y@VU992#r
zsvIJdkk>t1wrtqDcQ5kh&1`jL(R#z%e^-
zV$Zc_h%VHdCB@lkheEDKd`3HI{|(-F^5!EZRQpU*cIM72bYF8-OJdxvXq3UZxM1uJ
z|2qbp$bVQ^J#qr_VRhuKw3`nw{{#Y75(lHG$<k=!LGJhzR--NbB@M~F5;tbP5fRanLN-xq3?$4G`(YZT>aa&
z9j9?)H@2F_wr#tOjmBo|{d-$!#1u+fg5
zGp{i_c8#r*FqR#UY4GUV{V0%TNt!$nWihsWcWPdbl*Z+>AgA#k7~LJttP@t~moXJL
zx=eY6B=u#YzxJ^mr`8fE5qXYka@w?eMfwyW*Qh3kw9!3O=tSRHC~Hd_t2}MbphL%k
zB2ghddKM^{VWRM2AP&m*&NdlFl`1>`>bPFysYbyJlf1p;`(6g7R2$^0fP^^7qSdjb
z-iPF}q8Q&{@GsQz83PMbi>qJf}hlmGV(2|eu=RA27v1H={tiqUqEqsG9?j^=&6^fk6ht#5?H0pt(F%)Xr7PT_rwu$hxACV9ju{0(H||bC21Cyv
zi{ig+dhB&ivSeDEPt>gGfukPOUPV(cRsmCQ`B-R2^)7*6uT-`j{o{eIiqvBCz7t%&
zmuweba<&RXGrH`3GPnl3P;l%?Zyk&n?#8N*|1LsKKKRT)zBQwY4H!~+;UJ?=n~oJy
zN#Za-SpJNT0W!K=4F`^c3g=)j69$ePPY>_%NlHQ2h+r4RHVd8q$P`rapuWb9SJ5##nARq{FiXr)>WSM;5pm=7SRcuhRUbv4
z{eBDSzO2JSQS{YJ;R7qf_eDZrK{pY8?&BPg*#d+@rbUKqn}V@ts}0IRhACEr)FGJQGt~18EJX%^x$zbM>^d;Q5yB#gKI5J%A3(jjdS%r`f<7*k&6hLLkxUA
zP$Bz!sZETBWQV(sIN|L3Q#tbxom`xv%Ukr}beEU#61uZC8dWD98Y{ah3Mn;>t8ucQ
zACl2{Z+u_wP+p?5{O@^CNF4$=OEgN!ldIPJg|gC#!!@n%33<@-JsVIJHwj&M-lKW*
zXnb#aG0!;d32KZ*MV5;4v~+5Tzw!E|nlr0bEb~?uU_tt<4
z$-^pM5DGK2waV3a#us4hUQ6>g%czB=xUXF;!(Qm<_m^Qi9(%+KDKz80ia2eoGhS!3
zxs)m-Ouu&EuJXilQs~1zl8~_XHz(C45FbL60@>a4Un3EE?fNf(8A$>_?5nN=T+zgY
z6#!;Wdw#Pw6k`G)-9ia@8p8n?G&jJIzXaaj#$N!U3NY4W8DREDlVmUfaSY8kP~8;?
zL59wnNKXJ%PdmeLWZ}jVM8MR^OEeNk>7ejd)pBYol^GJC#5W$*tAhmUFL+M)9UX^t
zx`acg5H}nlbr|9K$rjt#>d3ndiiz@E;aq#n@(f@6*M>#CD9wyn4IG6mx>WKr(~tVi
zR1Q@p5NlvQ2g=Z`)xKhW)k
zh^>?ZfAb4Hhz(ujN;tT_QQbxgJmOXa9lI^54tK_~8)3}H-EvXb^L5+FM-h_Ra|zqg
zQV|n*46zS0>3q_vn+__OlQI@JV``Hc@S*@idcTA6B|7PJQ(;&Uywp|(?woq__%5GYEMRAT5HKN+
z|4wQE2!8U%em4_qO~WZHZ8h`?t{e>t@}4S!oz~Z+oK3zF5PR;~9U;*H80?jJ3o7o}
zVI%Z3cC620)PkfE$)|T^O;|tIqZ8P(Fjq5v)luU21-wS*yHTykI}taLttv>AFo`;u
z2yM)VVq8!$hrKB&b$pF2eqaE9;ekXX^56yefQA@uDb)oo5^e@rvv-7&tZSFus&v?O
zV-MAP&36W9tldnMydKiB%dO%H%o*o(mqb1Jqvi+Je=ceILH;ISri`ZU(^bjbv#ip9
z!2Val?4ZiEXn9UxX4PS$lX?6)*|;7&A8h8M=m^urdk!1b_twRcZxMCl_Vxyi_lQ&V
zkC+gDK{08=^Nv-&opm|`6~my#A_Z*oc&5PAT{g2x*&;7x90Xat)fvcNX(3|+phf_i
z;7(tV2z&Ju<;S-wMz_}sKCg>)2|N@I-+`FpjYUa55ysaUmXxw5VJ@WGXfG|6k{~Ds
zbj&YRTdn0-1oeMo+g;`f{+RUF_SXLAUinXg@t<)E@Nv!q#)o?;06wDD6uWx>5c4fT
zIZuGO0h4;_4p3#O%`*WavL34%jD=#ha}}Kn*x4<>`3Md$rN3h(kcz>#yWhe9yWoPW
z^}%9vCIx4)r=7sB{fAR+bw19E5e62&m@Vh2#oSc@R_KEMCOXh9eN17jAmlvZ&#Ojj
zNy;?Nh4p|}8=FN(Q529M<;5KPY@}l>{jf}$$_W*zbTr1o2w1n8fHtn!1X($}El6Pxg}H8LasNhz*0sNTS;iNDfmYZ~%@
zDFL)S#LZ;+cEq=*>ky(q!jQN8B_grgTkzE}TFW6}DYdKwI7L3~@i=-ZX$Sb)o2EO<
zdxpAc#`#old5d_7uSl{o!tkkE9IvTS525&|Z9&1}OuDFr{aA{+=m83GVls)&tVO-B
zK8QOMbJbsFMI(eILZQxB$UhJlSu)SpZ;lvTXJ|nY+Ag;3Px0LUx;_fm35&vP>F2?D
zjFdz^disV^>)fC{T&!*8VIF1w1#bmKY{VDHB|)rx>Q_{NIhu}y?y8;c_9j_9iBX;=
z|GeABJz_U3SvgY87WS^Ui|FFKnNa~K?1hbQ*aos~`Jw3L5-}j0zWS`?8smU2vm0j6
zcRBV#1|lWaeSnJ-+%5Jv<{-S=LS_!`iL-hoVZhwu2!r&Kwr`F0c~u-KEpXL;QQlR|{^A`~+T22}D#VgfDYE8o
zvJe2l_Usf^#9L^+vu-)&TP^CI>+(56Eb;!-c^=B_gZF9Aa}^qr^qMWWw#CM
zv~PKT>WjVEeH|KiS!9Vs>hJHLB8v>4Ro@_C$j23OK%+{!nmA2k*jB%GjATyf<&oVH
zvW93ujR_GY>-ZAiG!bhKpg`P<*z}rfrc0s|oUlm=9TTc%bbd
z=kuaR%|a6nObr6PS1qPCs}>q&Xj&tWN|zkizZWI
zsktd&T_TS|G=_3z*L)NJv6g-yRdK@gILwE*?hJYG2an2s`U#54C`D6eqOq__dW>XQ
zFUg>ByY|rqS|6Jxb_hpGqIeO^-SMx}tfYgS@%D)it-J<0#ox2}TGg3=P@<)-`=H;w
zN5WzzwELw6Z?{ujA9;W~i>g1$BnSB#d&l)FoAl4|3|_pD7?3go66Ps;Pg+ODOp^=`
zMwKGM$d5x2$EE_}ZH4YlhEgGXVLGdS0l%J}j>%dZZK|MyLMtH-2ny5&QH=WtFP@4%
z^;ZcYIbL&K))AhhCKFaD;3-wFo!`QQmBI)RqQg3N2{b%Cr)!=G*{Y%|lkR7M%=i8S
zPL>{272OQ#S`d~sT>Yt#Ue4e23;Bi!{_v%OSDD;Zz1)x2-*%6HyryU5b?%TUj=J-i
z3+7ZqnDuCHAk3n%Z06^tA?@X+Z*0qX$}LJ!zKFU^@2%|768KtBp6Nm|Um0PC&H@Le
zL7RyHH*$7P^HExr(tYh)ITU%S&
zH2?%0%UNRtE=4EN1z>1dcZWFEd0y-vI+89(|Q3J5qjp^TNRq$+k=JHZaKTA
z$MrC@neHK%v{cnHziRSw74JC@#=hBG4*Oo9QxSi*FlBoB{4dI5^GKu4bFY%J)*7j&
z-2buIk8-<+%wWqWR-WC}9S+l)U`v_Owv!)#;(J409Qk-4A*JikDJl2VRO^8^QtdPK
z^B@U~Un3FM_p}9^cj2V_NmUukeYcv=Bg9rrZ-%=I{E3`OCl`*Y|Lv4H9bS?s^G|BZ
zLvKTqzVz5mc~APo;)T;kiqG8-B{B(Sn8B!bExHYylC?05XzYL*tV$z{uo@KF*keTvi
z(^XVxvAoQ!#-5Y|VmSZ#Z?^VQ
zU$MK?=mLY)WXMF)V4idY-H~F61kAWvV^RYEKA-W=81-HnBFBoWyK}at%l}=+%o;41ri&Jy2p4BgA1#gXEN>G&7NaTS2WRn+`%B#9^)+`Z3f;{%qb
z>lfjD)l7_y`7*~YORF6e5`N8mXQ*&p=}3(KnWjRI5b#0vhgpL?dV|nF%=@!fX{y2k
zCbX)91(fX+UW2S1$IR|m_sSF|UU%BAYWDb$eg6EDEp|uX$Q}9id9j}oxVx|K5OPxq)0Ch}DRA*&2|?f0B@&A!LhG;e+v=j_Nd+z#|k#nOK4)trj*
zz$e_*G#~642&d_bMeAjAu&E^PPzk$c{kr
zXD(Jw>)H{JsPYQNKzc4<3>J%fmx!i;tvb8Nr<&w!U!n@To#z)3Ph;RH4BPpQ(5C@f#x`F9LC=*X%FA{sL(9`6l_A)1*`xXr5P
zpqk7U30z%>JVyD(69(jM#e5?c;c6!>B8~t`ivmXED%kihmT`|Zk^@@|QvuqCLqwY9A5a;*skuor2(I)|T_rvXfEz`^*d
zaXFDjh1On|==k8E5E%%Kk(4JC0n#B-5e0R13gG6;*J-kSA%-3t3Plw;pg4NFJ61VJ
zR0mNT_5NvgpL=_(;MX(XbOjE^_ZgS#IeRK~A=s+Kgx$HRnIEUygotY%ebzN2L8sm(
zW8)vYCHv;xW{45RQwxbTk|GR`gQdQYHmh?qraxod{
zb)OR3m^SIg{QvYsb_g0=UUpncW*&Jq4Q$@Nf}mG)(pkFdU^|sjrlm-PPD^jD<-c;b
z`z?7d8Mr>J)WS<{j>P0|)?cLx)0oiCJ2GXlv;QON2x%WNCbk&=5-uFa#+kQmyqSHEtH?KkJ}BJTL)pFj9wDGOnhJ;?iTcHCm9RKKJC
zB;Cku+5cFeeZiN*PY?fZJ2&xvuG-w$9^uS2#Sy=P>P*P5(;BZPz)JDT>g
z{|ygU*8+pzUhUcwA6KTRvrpLQ`H!45hE~A*>(-m
zrww7uPAFZ3Vs6`B9%?#WggNe4D1AZ~5)p`Lp(i@Q4)K;IwE6ozHFJ%9)%XX2+8ZKr
zn43VLuNH99;XKd$g5Do(q{u!KlEj#b0FZcvmUa@D-3_S@{9JlY7|hk0YMHen_{o4j
z!8e8z+V4Iitv1wJ#IWmVRI1TtD4Ghg6G{ph;v|^{E
z{$7m}mGz@wI;%x2ts3pY74W$1HX3IM3_l1qX`mKfA2011*OzOMz_R#jfd+CYE(;~M
z^N}Aei)r`T8a{BN9MowGh?!xqF<$PDP4fLO
zPc$LVbi2C~Lxj`mnSb>xB+ah4x_Z6J!^RKbk+NXfb%QMg6!9c`NNQ_37DdzfEIZ&g
z)!)-j<{HQ0F56?~qeO65|5W&cqA-Eeu&~!Gq_>JXnYOXl?8WA4utb_>o9t=a@mg=0
z)3y5yvOtXA{gtBi%^OYLsdgt^R`BY_k=rJbYq5GJplb!*}7RD5}cJ(7_6SxY@PRtXbp8RoI87`9n#B|ADtevX4=nAf1Y^FR)j#NmdQtHjCLP`#cczlB
z=VxbU+}GY1OOm*Cu3yBo)*5`EJy9DSXPKE^keb~$#7
zPO70Ds|`|GBsPD}FqhQh8>i=9ds5uFLpH6ePQg!7Nf#wU&j(0hbf^isAkPXu{d&2Y
z`ur~T5_luSnO1VNA8DR5kk^8md~(ps{v$3(KVECi
zWj=if9EzKq(<9`lW^h8>iDFu7E@#db^aQaZ&0ojOVPn|6*=@Y@G{rErKjebPa_Gi9
zOt}-PMLT>IUPk`VJXQWb7Qo@YOftPtUoesqV7(EIz(P_hQ&oC9K_$Zz)fwEIV9%@LkHT4AOlK+Vu&~YL^o@?0*Dh}#R>`h%YI^MY
zUvE_V#vebfuoMl^j?3m~?3Qd#Vc}|Ts`%ZzmI!=ijuu_vf{(M=KJhJOlOz#2frG#O
zSKPcM-;ec|LSa6p~>iyk%3~8UpQ?)RL)M&4~Fd&2IiO7(|=(|5S
z&?ZIVnA<1%Now^Oxh&f=nG?GP8Fc$kP-UuVxY{Y_r9-C=rHQPQX`3efChm`Cq<%#y
zvzGfmE!~SO6|k4fu0~NCS4Uunh?3=8pfMyzTX+VF^-wf9(|SKbfs6@$E=DU^EL*el
z=aCj~a$1c2+^u)aUZC&*%Y!cG8K1RV&5BCM&x?!YX6IziuPzP6snyvj?Z*HvRD@Om
zWhV$@(zS=gWJ#Puj*E8x_xv8CQu|qzU*B8^
zFy{uuVRpi(aOC9XSoG@34h{Uniu=Rp6q$4&+EINm@?AH?x6pu*H4Dkd@GeG`Kg>UQ#@3)z*|UAEl#pH*6!hg6vN{MT_VT<
zc|f^eH5p8|p!k9Fl9|!;fXEkZ6QbhwV~m<8b6n={x*BWi=opJ(mgWSX`z^|(>r@fg
z>*_62c+xhCc^XHY;^MpCh@-(jNj%Jmlft-xw>9xmyiF6{OSLqa`AIYdhAS4E!aIlW
zMe(o|3oG^4LLCNL@+Y(dd$D4CoQi6il(wP4Y40}@Cv5euh&uo8>3ejlFf;rBdFEBIjp
zM%Hhfpi1NMg|IufCA$qZR<w&>6P0uTcPFoJTBtHN86jVAmIMf_rEGBz+KV5+X(o*PS%XlgUTbV`0
zwB%?vDW#tiBzA^)-&86*ADSi%f6wrIe-?4V3z}NT@DhK!w0Igf;;Q0(XRW%MK$pEr
za|MRKW{~{1+E-~`Os~`Y24^+?;&LZa$Uu~aW*N7@a$Zhd4Joy|KW;kaeMh@J&~U`D
zXs$~mDTi%_;8FF3@32d=@spl`&68#1h;Qfe$Y5>zweVB8OQCG{(%3`7LvX&JslEQ=?J-+g$m
z%%qHm!r?wLfwz%p8g*b^$J{0=^Uxvt&FW`($XVXIz7kly(VEvUk_&WWu@4ZK3rv)I
zz&(a#6Z%go;_4A)@bwnIv<)^|JjT{Y7-9c~y3}$2hGTqskZ&mz>&F{3Nl~nNtC}uK
z5yVZt)SNK-QWEKRm;RNPkjl^?jkV@`WpGb?z#<>i8OLx-skc?>WKnk|%$e1%(T$R=TXC4E*r&Po2HoGv5VVFBxDG9TCnH
zOPr$#HIhv;&YE9jl`ie$=mXXf+AkY*J&NGni4l(4tBaN-bF26^MGCA|JPy)YnbN3}
zyv9E^sI$V06;it<5w{(3+BlZ=zINZU)y&=jGa#BE9Rv{}i7?du#A$&wyjKjD%IX?qKTEK6
zOI69AHcY1!Pefs@J)~;Gpee7$V7QVF7_UKgIWO-tDHG@q-50OP_rDvEKs;a_;~Q%E
zerf!{XFrFe3i6h(7x7i}yZ=+>pRpWL63k79pXX^fzU_0g==&vHqm<0~y8tj&lTD>k
zWs-->;1irs)|r?t@s~$>JaJ;m5xhAHFU#xj)7K5nm#sJnJ1G7}vkI&1L(usQUr;&o
zr<3x_pvsY5fb#I1AhR&bXeh)eYUi=fa+=^WORPrui_d^e$hVG*?~EnpZuBAmkh=L@V8~~qUP
z*X=sOaj^ts;4t-!XBzQp_OQINJBHGD>{MCq9!k^1c=X#~kEy7B?}MW5+^_ew1asNr
zDBahQA}n|(_nQ8SPpWiec35#v>E||blSU9Gt~nBYlD&J$oETM3zBo2-)#OBQ#&B>Z
zb41A`bYE34BYk2GMdwvKW4XMmAL6Xkv3W!SnN$>agheMk@-N9itv*k$@;RQVy)%Eq
z1;_%YZh!`rk)`3EO`O~lxmq|a`@
z!@c>s)f|<-atiw-UKdg7O2z_}iN7ZIXzl3e74|zm9|8Xktm=0@`Xt=0j%w^1UyGrh
z@DAY-exc#)u|0MGpKl36nrs_4O(v)f
zxq);FC3^Ib1<)Vm2!d%3cSct2^w@8r31z|@!bG@>Cj>HJ;vM!if;!))LqZAvMI1XG?t>MKCSjzrla5LYK%cQh6}>F6@CsUvjEGtS5wkj^U;|F=Vn$C4Avqe0G&4TG-EyBAo)W%Ih*a7EHRC`IGP-w8bZraO3;QvR+pM
zbyOWkQZztP{*OQ{6|Z
z1Sp^Y6{Eu@`Ywws<4)Kgt)hKpl(=0#_;1DkF$nNnp6h8$RXk
zYq2`=|NM9aP2gp1)m{el;u9>tG-&(auJk87Zaa=ZW*n3kkolf%Gw|OSeIh>U`}nn`
z3N~O3xgN|7z3+N$Ea{?u-9}}682uHgTb#d4-Z1+6xV
zoOgZ7#tGw(O~-#yA=B6_u`D|Jf^0+`j2|k0KI9$X2-vV+=zZr*cve4b_?->(+4g4k
zibq$<`3bpLSTO0MI`vj_wS#`Vlo-}2oSW%8ZOKUPIOrVl9u3EJmB(!6J%P+Hnpg=@
z_|T)DhSqI>};1#2oqXV|jt_
z>)VgL=*UNwH<}JML&KFY`4XWALA?w7ZspbV^zAB{7xiGvo3q=pkr?b7i>phmR*qAS
zLV=D~gE9?rBz^*0&*|om^raQE&wO-~o1XbtNV{TiC{`|%a?$#h<6JQ+K6sD7IAdsouPeZHMQ~;sN`>l?;-OK~u8b;>SJKJ?y@)%HY
zK_?XHSF*DIO3RIrZI1}7AH%cXrmfz0-yh;M^uea|D;S2m#VG;BwE3yHVa}lr`%gB?
ziw9`JQAht=|9-adwB?Gf?(ONxYAslaT~UL&?5#X&U^>F!vEq(whO)d2+40BMX5m<(
zB#tYab15ne=p5!0(qsKZqXR!D%M`2Pd&Ieo*?fE?rV8Ei7{J`eWO-0Z;&ssDU;kY8
zL5Ir|6RhKfnbg8B-;=#K^B27D20CMZE#FNE_S
zb;i&SBMn6F&l#59DQ9FPSdxNvCL$#!MbZg}I}g@B^5Ty1WED#iiE{fm;{6@P8+b&1
z^W5gku&8B#{8$(68!0dBZzOM=Zho2}W1@gF6kuXp%{g>0la8h{#MY=OKMnB*Lz3@j
zTnK`UQw*;lu&%-W{BuW>`?s!57r$+&_$SQFpU;kZp4l*#2hZ9POudUOD94x^SB{?E6kvT1!<508mSNBWTY{A%FuiWFQ
zSQeqKb}5tyTxKew1*;$+I9m&Nogy8$9yRI8dx@V8{cUyuy1)+efkqh|23`RJLvZvR
z7>Y;O2h9yRps+h=tzVgqpU4j*Wn@IgTm#QKF~+v=e;!cn7Av7ksxlZzKKO4X9dJ`5
z31tW$wJxJ%P&R45E!HIQp-O-i(!AQzO{j3=>N;J-X)Uav1^mz9B$8&kfFn&Rx-IN&78DMG8QJ|SN2
zDVM{5H!^7GBEQWbwc5V>x1P6}k~#readi4wvYLj*abF?QoRv`7C6GKHLhVQK?WSg-
zK+wi~I8lCRx1qZJy2)U%GL^+)2Ma^%*Pzm@H)g}4HwaQWWTfU!1KPiNLY0=ZRlnyN
zhhrI$+U=eU$0j?5Q~8m6E@yC64(a<}DtW$2WELWyENJ`dmOezTbg~DnOb$WPlNR{YySSU4u!Xu2KTlS7L*75`Nf)YBk3;mankc^9ojy|+8M2sfhLj~L?D-}
zFFuB;p>`d&-@@@Xkw;m|XVyWptjhAdyYXdN1>I
zp4|27>R6L;xQ1mly-YUjo6y_}QOEF4m^nQUSTlG3AN{EsE*)@*_+M*Zt*_EH|5grbac+p_@xCZWwWRuftHPG78;+e*LweIL*#pE5yU)aS8
ztw12xHUzv-lF9VAN6ihQVQ6I(K$)cHXePhL`}KE!1l6eA^R3Czbgs0C7Z7+a*IS^&
zW6_5K4azX|{i|@Z?Ay6GGKpQltd+C)8cQOgss(haMgFrlnoWHZ^WD3+s2VN$-NKay
zSTsy0GsA!fWq%96+T~v$g_Ioh|HqkFO!|>z649FVSM81m@W22nXv^nyBH1|0qR$@1
zy7LJq0?L!q;}$x8^etCM$+hr)f{|pJSIs`jdvgk7Fz+gDcLeb%QWLfjt@6QR1s#t6
z_UqEOF0qk$2SpA~;X8d#xvMk|Nf0OPpqkWi$kaL-5y#~~OyV{R#BZl)J%#$Aj6=3A
zL)OD-!k{T$68;3yt2E`2er=6DvrXfdk2RpEsWcAu6Z=TFe}d#`9N3Z(MZKxgf-V7G
zJ}G{n7v^|jJ;uBKMxage?&62d`NyN%|sPpdtm7qyZzkY
z8%cA|WEt%G|7B`EN5l!ef0?(1-_#%vA2PZeiE_;NhoRo&kH{W
zhKMlt`|d8w;h&0&Xb49LyY8he(qb@JZnd@0a+(thl@`CYe|p!&%o0pcM1Ude2lX#Z
z@}1X~V2dZEnI)R^!*HSP{zy5rZ*CvknZ}no}RL~CAD*6iqjplXfn!^
zo=siJU%Tte?I>&0RR0v94l-M?uhXWb}NXM8R#
zO&S}i(+7?n(1qu@ipYx9K23X)=fpSQMC9u@6kw_@y}jw_g?`L=v2&Pj&gOQ3ccH}^
zO5O_iOHDkriQZ&DXG0Z5DsDxy<|c#zC8?MH?6#f9E#zeNn2~I-c7YWUu)6z{{6T&s
z5pvGH5V@LoHFa*DvQfZb?;k1r@kdem*gADq=ExDG+aF
zj5>6;`AwV+KGAv1^{R(QN?p1I4Zo0u8DBNEr1O*OL(U5g{D8to*@M@dzeD%Pf4Hdz
zSSG_2_|H4iMy&U%9gzZ=6y)__c6UHsjz%RP3DnTbUO|R;xgsH|I=~SYIbW(0D6gwi
zZ|)B`v`v6k61+9wA!GoK&iw9I;CRgiY=lp*mt79$t8Cv50LOy;9AN3(?pUv|;2K-?~65Kl~c
ztp$FLGA_~I-s|#hZ7Qr=J{ow};{%F-vb!7IW0Cm6?e(67XgG#cL&|&4YrgeLZCS@l
zd#(k@djcBX_B<|iCeP4!4KWmVq)UL@a}*KI9h0BsjhNPLK*g<(7JN}*-6O1y&xy{9
z@^hIC#!KAqwrSVd2dDxb(RI7+tuqEpwTs8W4Iw>68h)pbc>2?8#%n@@#|NA~EL7b@
zm?}B9Wn=J98c(u|e>fSNhqRS43AAXSEyo;c!iFLf*??(iuOP5N%d2L+FOLOz8?!z@
z$Mw1Surkjfb#Pn^#q^?~G?1D2!pP-7luuVE%Epya3U&Xjjmfz;ErduJ0baXh-p`k@
z>!$;0DRfXDKa=9Ge->|)iL6}_gI;TxDXLy-$1zzNqG>AMt7vie=z^d=3M0_$G|J}r&eDF>AwoHMyaXGLba!0?Sm!F@fp^Lb75*X53CvE0eWhu6(
z*6k6fb!VhixItaC1GUd~Of1OOW67_{2lGwTOUpr5aSGTL*WwRo2``6Z6ERzmujDtz
zmwDmoU67S?gSbhL#dlx7(4kIXXD?citqh1LiVX7Fi>9k{4kkjaynCsg9rJplh(T8Z
z|EZ0;st);}Nz5fsi+}jR>&^-60&%~&h^g!Ns(Sao^)+quk3D|orI?@9
zU3k!tU|^?DlXsq;SdHrXq?$Kz57hAeby#EUk0L~vz<&$>f=c>DFdNf~!%l;$g{s>r
zNO_>fdXfB>vO$Zr6um)*zC~A)xu!&+R0=o6?`oZ*U7(W>Z#mDD$&Q@XG26q^!h_d%
zff&%2es{U}O;4LcB}8ezFL%u^x0w>}!NVv^61UEL)s0Mohl?CG8Yd(U(z8c9>;AwV
z)XhKkgQ|n4L8z*3l=Ubb&fwB}O8A6+CWVRQYt{(FGCm#r3>l}lqH
zRYj1_Y5rf%s`jD)OJ>x}DniC8GgNL$r(X8-fV+B_dI`=cxX9_W@5dM8VZcp}%q1cM
zKi^-QPbJ`~K)*14IE#X>EQK-B$HWXi^+T2ZERRepUR2tIU#2y&Z1%sEbv*DDxM3^d
zZa$zVwm+GPyrBIkRJV$Gi^o@$h|Pe;fShSIu_W-h
zoIO`L>`mK&n3~`E1VWWIn+Rd(PGzSM;h(DCyXDSrs%k!9)1}{paEFZWB({^xt_f0>a7
ztMAo^daI5!m*sX?MqV#PyslB||N3}f%3(no6Qq9yrRh;+^_qR8=v<|hUNRp}>p>4&
zS%gkSZn&#lrE#X}ule161hqh=3f_rR@b=OyJtZCQcazfIw-qQ--3G^WM-Ylk;!
zf-+Jk0p_Dao=m@d_4u|yRY2q@h(St{^BN;Adsw821enb6)))oUob-)VRJ${ut+cQ2
zv4@gXxaNXuzX2WnTx&rQW2f-3z(k_t1v3zOF@Cy+
zt`fW`JXoONM4jzH`JX0F(4^^$f7?@ASc)1|rEZw%#xa)%pDgtCNXCv$?n@y-P
z+Xw#+(m{G8IORs2p+y%BV#OvLD
zaouOs$YSzX6nJ>HOE#Ls$2xcGT(4m7OMH$7^9DYLf0?VcTXjUhNLJ;QzsWXmQq}Hh
zGoXR{l}-}*#o0Cj&{LTV`mC
z+F_-!#|v54W!O$4^MK&mt@G??v%PAZ(6}@^?Oj5F{>>A?lu;)HXekQ{X4kW&i(;QL
z1Y9R_)*xcl4BVP(3;_*SVREfuUz+ZA%JDHSmH{do!k
zaeQv1G{eYODat900%X1dpZU=y)&C6&;sip*BRUHPfv83aaWVhp?!P)x@gUpGUuj%&
znRH427|ith?*3=GnHnWIldhv%&PNej6h+7^ybG~#3?V=;p;V?yQ}`?&OyQ{#q|N4f$CR(eCwJTLx<2wmPuaX?3>fkFm>fpAB)^!f5A+;&7EM%moZZ`3`%7
zmdyU!ysy^l%??M4mH72OEGbL4SnQrTspdI!UZ#r{1Kf(5SCPm;x&nJF3o0fp=AeY$
zpye-Yy;3^)DHXe+NqInF;5zKu94+N3(r-RJb(h6m^ws;H&}Z@}2b-68iKR#NyVob~
zRaz4~C^w$mfsj@n3oEj;-fu(SM>=b@rMV$|>dU*CB8Bu52)a$WvH~vo%ee#JPSh5|
z_J_CDup86zq=DQINr8c{LDP>VoaY#Wp}+hA6ojQ+NVFF)R&tIs3AKf8X!^pKkROS>
zc{J@j{STjWmL}sz(5S#ev!#I#_~{58Q}UL<7x#pFw3U5LqM-$u4*Q1tJ%JAdilG!r
zR=A~tvI^w6G?HJ2X+PopwrDZ9=Aej)x?UTv8hn7U8V%*{+X^Y~p&bI#=q>=d9sxmB
z-n8Hd)gAZhD3pc!XV{jI9@9r-!4&+ph~udFu|pa<^^~`p>R@=hkJVspQ<}hXp2DYC
zZ>XO;n-)>%TRrfP(DS?JjAJ6B`j>L-@os}t_FSNxB$?F<`Rw>hVkkLkrd7wlH%OF)
z82~UXOBLZIH>fh*TgqAQnB=e?
z#T7*klY84UkQxJ}ia*_hVUkYv1YgIIErzX_`N$V=f?RURWZr8+G{1T??66#=dm$=5
zc{SF9y);aIvW>b#(Hy2X%L!Z^O2bg7fG8dKKd*cW+9*8HOU
zBm4(=i{|~%UBhi<(0l%K|Bmj&eA%|n5)~(iDSmp)+Tc5KorX17!vPN>UW|Tz(a!9m
zn6TL$`j150oH7G)1+0P350ECHs=ozS3Bwz>(PgV5GOs>(rgn-%@wxrVaP}CPBL(oXO9cI;PXg;`DfwF_w^Fhl$
zsQ!EUe5DTd@f<>ReG>Oj*q(t%as>2hfWLY)pAI8wtcU^nCQ-!_uo%z-jQ|X%?<1Yc
z@4>vcD{Rwe{H8)1y%ctQKE3B$2#(`bcD_D~6Ddg{d|?;y60f!|N-_$cGEel@ZugN<
zhwAY9YCVVnawl0mOpi?#2PRVQ@wEOsS>hpOP}?IU3XIR!ZC>w6W9II>dtz8#|HdA3
zuR;}+}>;rbiJR%JLP&yeaLVWaus@oYlmxqvX+j*fh^Ie$M)J(
zqlgwPcZP{CiekcRqZh`kmc^rSY{;Sx2FOd9>-)i&AKqq$_D|OI`pf1U_g?fwz}J)Q
zRmDnDOz^epHj&b=MI1j?FO#ZPAH#<_8ZF9&8>xkYuHGqTT}63j8Yu3W612ZU
zhrC&96%tZnu>AYO%7LG5_C{KEP&>v+gYs~XmNIw=ollYH6kiV~V6}S-EhreS4sk7t
z?mJK#0v~y)wvNOAJ?uKVbn4%ePeq%aerys_9yVMuGB;zgZLizm1G!f`sh;^r-lX=
zy;1}yuLASY?2&afk8^|TG8c!lecbcutwxy))1#Tr!mjU!Jx!PX>br()IKn
z0xcJj^Uw55jXU-pop1q32stZ#bIceDtY5K{dy`|<%zJnBK)KAqJ!z_<
zAQyGZ#h1s()XN7=GfmyaVx1Xcxe|hvenO
zZpItWR3FiBvnoeRMZ^1ZN2o(F%Jv07%1f}Pq
zco72mn-}diH+lQ!S$*2(txR6`08~nuA|`gH-GOkI`b4~|6gkun;J}HcvX^7zH8{~p
z5OUD8xv8b+xLTw`DG<6~;
z@!GDnypBa~RP!w90v%)#vpG%Lso*k~w{NjCk#Uk#a>mqD@kUTGi&-AanBe~k{6Ix~
zR?qPkusTh=l}$+2a@*o)sFT(=pFa*(^Hy_`gX+8NE0Uq9cDO?B`(!
z^44>zk{9yw`KqEscn&afKL~PBo2xQ~a8f^mMIIloewC&5zme&&wUsgRu%Ypii&f1$
z?0nk_e=#?I3LsLB&@mYy>0-90kpyLYS5M2E@+b4B>A+*%rYS!-QigZ5IY0HcT=cGz
zVBH@4dK@fuJi8U)g*W(@03oDrj8#9tt;zY3!A5ckcRTEC#`Z7wG9VkoQ#;E|tld5e
zM3LE(XrY@{0qVogl(>v5JJ(xPrRNxt_*`{{uU{|LAs-2O6$YKQSnF&#fE8M2z}{Bl
z%Dh~Q8S$*1a~f@`H7a^-&~Jai6!VX7P%Fi3_Qt(+BhzQcJW%IM=SdMoum!XJdvw_^
zG-MdE4MR)ou3w@$!#l{=jArn`3|-cL^(P-zl{)`|>_YwRN@%8*SEIr(D8$0kn81+5
z2>`GLp3mAj%-Z(;aT-U4SI6^K?ccDlsyV?*p5*_qaSw(h148l2mv0k6F;9{>GG~!o
z1)HW?y%YGUG6x)|DL7ABfUX{StZUl%=vc<}_)|U0&L$_Mo|8U2-&$~SETbWL%4A=8
z>V$&me^Q#G;2~_pwGz!SvjLpJ-+b+$oM*e!4PmdhJJ@32qY@m}9~n^kLvz!r2zc!g
zlV=N@cU=Cy$OtsL9M7?@$C}=+K6uk>OQsf4syEM=i|a_ZHIk
z{R?lE%auarMMdzCbY4j8eNRiC$x$tR`A^bw31h5J-R+MEDSwT~+1-N0TFPc-hvyWq
z9>y_xM9GxzDZGrEnukc6jWJB$0u%PLxG7KNa0kq<
z*){4eHyvP^(OhGk@(E<$bf|^Bf+p>{{Zj7E;aX_+C4(lDu7Mk7fg>`9_w~SgQb6t!
z&0b|D4bC&_PtuMUCYmX*3)XIil#n
zVdJ5czV6ib2oxBh9SVE7*qMVa-tlB2iI4(2hhYm?-zmP)Z41z`pRS6Y6=InAv6~^y#Vtd;FMp%hyTxki))Nnt%iYb)-Sruut`zW_u|pa_1zcF<;CM79{|ZlI#mOPyxY+m@inZuwszk>DI!p4(o71O-g$aDMF{%T@sI?;N$#@Ggen|NgJjQrLM#<|fcX_><5
zn-A*yAq`R0AGeh(lu;7GeJ!?ip%oR
zntjFDFOZJK+0U1u1ut4XDa~QJN{*IN%o
zSFE0$0&{VH;2RPe=SavZ~-A
z*4DU4{Wk)Jnx~2-yn@$$yPm_BWdh{lF=~>?fiK%tY*%B%EpGcbii<{uA1>nPk2TG+K#uV8cR!z?i}=UUeE3y#p85Mq
zPtE&HJ#HM3P`sgwl6^JMFPT(uCyxkjtYlM@K9It$-~FD{4MuY9NA8PUYnpgtYG+V5
zdkJ~O0Cw|g)3K$^xXt-X)x`ZU^(i}Vp2C!R~Krs
z8df#CVWkBz^T_0mo%6xZA~J~7n5O*MH_R=58w4W8_{K-(j2{C{;W@!Lrr(TsddKk5mjTVJg-pysp9M2;?=kO!Na~4%E%B#Mf`IEYo$n%f{<8eEv;BfC^)_TT
zUDPL08Tdc(IwahhXIK5aSD0i7(w53lZASb!PTG848Qi~rLxR^At2nFRHr5aK@jAzI
zA@j`Cxr*olAA@A%(ig!7zCz6D{`t#yx)vZyZ98sj5ar9J{;4sNAnZ>uXh>3+B$Aj%
ztzBVS6W*CbMtH`i^-
z?!-Vh6NTi|@lDMqFapwTIB*)>Ny)K5S3Lcy@7J>vP`ItRNCI`8ye@z_e*kGcUU4Ck
zTyTTmZjcMD$K>EupmhtB&qRZmie#t2sEhS@uU&{bW_ZoOx
za6Xscv3ant8LjSTYYWDVMy95MtkXfaK@v-+*|QEjs^43ytnTk1FAg0sKQ9wPwyZ6G
zWnMC30nxi<6f&3mzcBEfRZP(T3YV>%8SNW;)%0)tUs>x+5PwKhR+fAOjFNQn%5Z;{
z3t^lnE;oD<6f0w=1W!M3Q)ZD7kZVTlUV6)sj5r8os+tn^p$DPh}zTVZ%5I7n;9W^vKq$IA)W?vCxSG`w#f
zN6ybM3A!i>@_dzPTTbEc2#w*Cw^}srTeY?#3|aA2hQKfbsiYLAqi94aj^|i(*|SDb
z$&*M7F6h8wwJDRkn1kTi556+u*4A?u5l2)9n#(!)|A7SId`T#(%1LVaBv{uE4GkrZ
zxi53}XnK8v@Xtin`WLvQp^BoVU^i64sC-j^F^Btd4O6J+OE^u9;>zYH)9f`UXvz)s
zs~mX@@}N&e{!lshhaB>;VoH|=jkUt)0p!|kb6Tw#{di)HF;`k!1wWw#z693Bjg{F&
zT)OJ)BJ_ajCMkZoDGqhmI|$*JQ=%$9T5WqAm
zvE4E4U6pqp=_(cZBFp@p%$L5$bh{}lA@qqNaPsh=cK!Ez5l&SkR7|z}FAMlg-1Aw8
zdKF~|=0Y&^s7S8$QhL!J5VA=vM67S5=uXq#$n|w4`t=SPDf3nxirtLIe4W30bLImZ
zmjhezW0JdB6jkm`Xd4{=z!O&nv1qVDoci1wFHFF3$i2_w`b+7@hSeCkFh}8Vl?WZ9
z+#Q`H55ckzv>_rUC31b&*8J4Nl1~yR-Rv
z>meg3v4U0`uWjqp7WM)%puAs`@nc9ToFbHVmWd=&`o<7GhvXVb`?nuyV2f^uK<}Mz
zh0o&!``RSqp?QTyR$5t%@kys_>_FXRm{(X+E6ufM!Xh^2WaTPbL1T*uPY@QsEyOHC=%f
z?ANb#p*cF8KGKx38EDRdDwF{FjCDj?yT?q^%2&?&o2PF?Wv>)=Whv=Rln(<
z#bNyU0@C=~*50SFL&mry`Uu{o$W3*Z3OIY|_`d3E?1CltYn|HDMqqnO!;flivTO4j
zZfJ15Is9gbuKF$1VV@mC%lU$NT!26*5$fnBYFW)X?q9DtjUPSd8P>4YBz@u+Ih6w^
zn~4Y$-_p4704zDI3*4U{ZToKmP=0VM9Q+txw#+pdDxdOO7p#0~_7nb59AB@KZrVNA
zA!__NmM-^Y>J9aZPd+T-cr5HF{3Gee&!fUSks*MB-l!oreKjy)??W(
zm!xSIYzu@3-;1nv(G~9^3pKG$Cv+@T=JWnT4
z`_6`IHkw~=#uhu*{=t~X#!v3MFYrUJ&-WJb49EJb<^2(|HOtW7#Z$IJd@IX~Cb-#g
zq4Q-_>&@oXmQ9O=ZlJrKK^z7mUxAY+FyI^EgZ#CqGD1Aa%!;D+R{Uvx3X@X@ukkX&
zrw-=OKq3ZT_jwD!M2-^$jmplu5YuNOe`axUBMvumtzJ_UhM*RxYGIm+^R?ybPeiQ
zIBYg^J|Qdbc4HQu
zDr1XYKmkPwz}{?bZk7keE0_q2{cQkOHVdqMVAn1I3<7c}%zCS9VW#hg`e>qUNjy;K
z220rh-2O=Ij(ex+Pr#oLBi`k6f5BFGa=KX07K9g<7#OE0Ec$AF`pOwNsAhPJ-9`pY
z?fd7)7flp!dsMX3Y<6=bN9-y_ttkyx%DbjEMT+H8>>
z4AA-e*tXg`x{Q!H&5qMciIYb4hi98;ywtBPQ0h0Sb1%xMfar;>)I+{#Gu>G;4}A9t
zUQI`Cp&jpQ2y)QNEhE#QJ$L{lQCheZkDII;_2z;|qG$8wf#CLBH7A5|Ox7D>Lsj7X
zy8tC&RDFUiG2_N8u?uJ8{*9sw-c1DA2=v^iYpqw^>k7+Bm6jnkaeeU}@~l>9~fA7142?hb%)L4y^sc
z#M84^!dAs^Jk{^T-n0Sr?@65>1>~g*v+JRQkRYUQ{OSwfHzbJm3>6%ZWkt+b*F?76
zML6!db2&p
zK}Z42o&m+y*HU$RbEGadFCm|`3ZY*M4$$G3>?~FypXXKUT`ASnq`B=PtE!ZO-)YoH
zM-s
zLq8gKfsx$0!+tkR8K)6OT$>t4UCNj7F224sHbZHF_?)uowOgPqWZW_o%7yVmk~4+R
z^SPzy7aYLnO=iF3>ruH?HD9<1S8BJqreeE&GRLZa_WqL1ee`KMsWOIwl($gJ
z7Lwo#q*9I)>QyLZ&H?k=t%?$G{8fvp0tX>isai>pFPoVG7_P;MiHTXwrij1>AmTX4
zlU|_^aOCXlnC?#(BNYl|q+=c53h%C
z>e!jX5|~T>J?62A3)ggXN0S@gBB3&WU~pt_Uz13ksIHRz3d1Jba)Ky@iclaf9FFTq
z{EFc=<#{s4;zC_^;02h-hRG&gROYqQ@>Zy
z2~yHmK(TVlYvHFm?<#zhue3bOuU*SQSSLWiyaZ)`f5B4o@4azcczj-lyFnP=RbI$#
zlYl7+h-o|b6s*UKR$;`fg*?peI;<#RUInT1GCsDVtAjbQO|=jWz_iPiAa!*AbDUh3kp~))AHx8wV{|IRb%YS;8+J8*x1nwpHgs-_NTp#xaP)CzcnKcX*D0-&DHMUQ{cZ
zns491&K0PWPWPJXPZmR;o?pB}?`1GXrhA;L&HsaThj01v56N%O&H;9?H0DidC8(v$
zU7EZg4OMXMQRKT^{#tR+J@?bW{uFUq9fpfEgWp54C}GvSy#;+!@C(Jlq)57|ZsEl#
z>jE|iOmxnHMT{}vHNojJ=EY!=6ePu20(oU4HZW(83ZE!Y)MSf2KO;&VvlICRWzfU@fv(95Ic3R2r~Ysugv+p$vG(1vSE$8AJK0EcweknA%axTF
zSqiPNxkO*~gtiw(Ew(tG7vnK)mX~Mgpj~+5@Xz_wIm;)dmq7XnnM9_qVic$)2f8N*j6WKc_kU%JeO@0*w+W;t>_4|&K61NK@f&j)ja
z7hLn~MQ-8QoMQb-(p(pS(JcnBTVLRBxSd%S&r&&Te>-#0jq^C4{2WbWEcE?+PmF2=
zO8e$=MBw#(np`3d_$scjhpFax43i)OLrxjdq
z5u<8$7+8cH1D^p{sRf5)E$OTP_%@x#pJwM%4155keOC{>w{wC4!eV}bbn47TyC>w7
zZJ|x+{D1;NosNUJoHAl5zCPL)+N;5~TU!@+*8$TW^NZ`eVE
z%`Fyd-nx{dCSF0mMeBd9e>v6uMQ)(?J&kF0<4rX?VU!#bEOUx!`Ug;^GVtr4?Z@je
zRyAmBvD8LdxFDVGvQ#b~?XYSN$Xdr_reI9V33fB(ejcQv-Q4i)k4J@VUT?sz?6ZGe
zD$TJFobiJjl1!0{)Ijih9dD%2X{}q%@rsG2=Lf@htSkXr-mIRd#A|k@iC4GFchPZI
z_M_7tZTeTf=8NAXx$IUdFDjEb>aC;SHn!Th4-V4VK^u^?hG)($XgOTjA)gLMJ^LnMP5H~(i$l}?H7GQ+jKrGll
zWj6q4f_Q!~eWu}<uS)UoQnpG^*2lYNLB(#ABwO19^t%Q~#i{-yO5`^aMx_d{rW{y%2lG
zrv~qI)pC@8@qtLtY4%J`_@&#HjMICmYTMFtN02oH>Vj?_mB9yFIzPdXry#%VTmFcp
z6nOIfrqY5UMm3+DKKlB7Ug}%mN7X`*2%pR4&qk|@J?!G)8L4GR9UDW8%cH}Ei3~Kn
z2OZ9-y0~1ePU5j3lt*%B9a7fSN(uJuRZ!dI1$OLlTUv7JqQzIi-M~|9es!i#*v=2p
zUv+vPQpN7*o!-sZtsw=DEbm%FR|?S%da7E%C=6(0hT!=qMdpIrf$Ox%Zp7j`P!&l@WT_R|cR>N)9y#;Bpof1LM+M+VinV;xCgdKs-tP@T>y!i^X
zm|t%TArIfTK5nl%ixh@QqW$i+qD1$6r4vzK-{NtVh+JKm|CxXNiv)E(on73t?kCzvw?P4^6eir0$}NV=Xid8mInxEMKjkVBU;~Z-o$niRx?bR+)2Ye*l+NHyet9^j1zD-ZO{ViCN+i(zNBYi
zm)wtk<6CQUSDYHtvjWQMz6j!b3%(wF5#%RRt$?vZpD-uhiU{8K*228LDC!E==Dp;)
z`V3UK{fpoHnc*SmtK^6}X}=U7YF9W@0iJo>m&Ki7+LACTVJqO3J^UI7IWQB_?bN&m
z!Qbl^rUf8G$NlOC8mGz&ZHv3xZnYLdTH=Oe1ei0Zy^2u;k!XNBhcAZPM%kp`D?ED<#Ca%kG4POo+YT81&g|l?-
zxV|tf6>Iy;aXlhf$?rws_v^jXgpyGeVd-eFG@K{7LxeDjV=)+T167M0bBsA_sac`+
zAhc6B)-!V5GocV6N&MZ34TD&O_VBR$6dS7#z2ki=42&eJ`mH;19`Fy$S4{2qQOirp
z(oQ7$K@$~|P<#X2DmTAh*Mb#wZ9SjkJ}QNjnDpslLHudslxV||L>v`z`DTlvnm6V(
zomgtaWe!w*$!v_~}i$2f8oBn&^o~MoWEH0^w
zUIksYYLQ*(5KyKcPR52Z;IGp|(sQdAZnY~h+omuXMm%N7VP_gBMKte?!#a_vcQ0Db
zM|-(v0sgjt}7b$Xy_+2P?Fb7=s4p0A|aiXSwW~`
zAQM+a$}mo=Ls8|H$CxltntjS;auDzP3Ma*YF#%`t55#z#lR%xVAF4~oo~(ftNY2N3
zdQP0(#sbMIoYfRC)QkN@{=MY8{U=|_3==raccsW|%Y*eo#<~NqA~n{rDBkRB18#db
zsrVvUUPCIo%7x>!0RP{xow);SG*c(%uTROlr~-~;c?hkD^8Q-O#diL@0SEa$@!#tV
zN)A5jf#qJ^#v{==r&CNzot2)wtlVQ$^JtO${Y{0Co}MSuI3
z4O|N%-3covpHmWHahK5BTcj*`mc*#boZ&d$;)@BzRyK>^fAJnr@+Aa3U#AObuDzB4
z)l{R}Am{z|`)lEcb1t
z-cp3w@6G-8ARkB>Od$!pWhtU}z4Cj%9QBF&gG0hi#nZM7a44iyla{Sjb5;RS(4M9)
zN+e6~y722{6wcq$X`IaE8ub!tue*H4(yID6{``2l3EGTP_;1Ob3@p_iu8>DFgzv@9$!t#FQT%E2CnZ$j$R(*dcr%nFV&rezf(*W6ZK*hbbe?)3lE-&}?NioY
zlYNCYI>o{t7b4g-AD)mtq*wm|9g&fBlP@t9(GH!|^Q-rGsY(W6B!!#>0aqZ%X1!U`
z^qtKK!i}ruS6Ao9&AerR^K+#y;v$B06cN}uOj|jEBxQYxN?zD;dVbafv$Dh~SbB79
z0eq>i+?7(fT*8lg;W4mI8MT(jqa~R_A4PuwpL2?|rHb{oaDWOblK;;W6@vj!Ij)YT
zH%*h{G@=B3_8nYVG~|rBubqKQKuN`@RL6Y@j|hp#M^QAAYMsS~MR(WapyyFP0d%tza*OYHs!2=8|0#P1
zJ#j$444I2&9Nu(Dt0!)L?p8j$nPY>l(Y0^_{XU`CyrZ+{xfBWIG*5TMQKS)3(_)l|
z%z`05)eh}+aRUJO>K|KH9|>nh7j2J*YY*o<$G%!WqA*R_?N0Dnt8o|0cYHCnTi`TG
z*vZQ!w)CJLCm1|r)Kvt03Pq-`6iPK>KUyBQs`*kv61e|&elpCk>*B-fY=9W&Sh6rc
z{lrw+1C{^KhcKTjimv)i_}vHwYz~T}^=Y!<0h^cVFF*OV3$-v7-?6cw1BJ2=XT`>jK!kCNQ4D|t7^L4^RWm>+p!w-!PNTkmOW-8*-i8k
z({()qCD(r|fOQ%6i9uSZBK=-@>Wl)HLa4S^hiWFQpwH^Q+z)MP|FJDj-BGWj&kP@;
zDm1%l(f*=uPHapp>K|g)3O@(XLh==Zxjz(`anVhM
z^XM^!&5s^5mBQY?c$moiK!7*?_DjAVz5w%GdnELm-3?sEGa|V*zQxyvi!cjf!fbs&
zqL>QucuR=}Mp#-zu7lBF9l7PrVwciUEH~Q;}qVAFK#XKGew>1{=h3fiW^^
z2}wvp1PjI~6UL=)5*tnKp6|IdD6&QU&L%(B5b-mbzHYu3ik{unm
z@BhYbb&EbB`(qraZ$f(CXh(ft^Zi@BE%MJ71K@C9Jcs_J|J=4EAyDcXzK|!<&2uo5
z73anZgLr#%q=-be)>rL1zQr_~|3hzt7=`5i^cTEhJxjt^61LPsZNacR%Proc&U%qQ
z7;4XcoyM3r@-_XwN-3|no@)n<(W(d7`_$&tbeUH`WL+*@C
z@GRl(2}m{m4Hp8xF0s@+f!HxyrJ{&Zr~RY8#rvw=$Rx(CaVSQPE{~UN`!AW=rsJx_
z%BSSOBS>737|H~o+0>QjhV$O+#RaU8uJjHff0@v`rSk*n=EtJjC45XOM
z1@gGlPag@IzXtEz;{JQ#WF##FXFrU_Q40eRGR(D0Q2Dj_d`bLJi_g1<{tu+myPN${
z_)=#}y4`Il^p;pNUAf}5k~W}uaPRmSyI9io33VvgM
z4cBNdm4c~&OUta-h757zxNLv=)ZQIXQ9v9_eEOCN!=S8tAe)E>cVQ}lD|~~$IkvCI
z)27n-{*=jX$zH(j))QgyX_2Oi8!Q4+8$KNsKG3M-d{6VKV{dB@#jG@=nEx}L&WrFV
zHvZ8}Hc(aM40hlh@l>oYsrcy7vf6sVj1?(BQkI3|-U%o!z4;Kg-ydphtWEK!g$N8G
zK%H*;-1_}G9Zs+#CneX8Ox_nHBk`oITXMl{uf2n^MV^~Kpn}88kG9XkBv3ecy>C?@
zIu6Gwo_84(bFjkS6l)PMG|fb%(ta)1Sggd1Y;!tpE8y^&)bL7pW<+wC$)5eQ)5w=*
zvPT&}T!+K>`}9s^SrB%7%%HiI9Q%qyAfJwbe>By=rqtplAHB|1J`i&;yNGG*=-j8&
zqM*Wd)QLV>>Vj`GiQvEE16vrSnG}kTG{drld>BEizmuWKYUzNZz!zKwP)J
z;xm=8q1HS67y9hSs_m?B$76!A&Qr^nKVLY$?uNm}%kfzj3Qv%MMqCmq`H7qm(#!3?
zbbW}1^$4Z=7Q(e4)M+tT4`gT?{=$AKeef;x54A0&rth6@{fX{D+<5vn)v=ZPGo^8C
z&IT3}5ox}qNg_{wk{qu1@nJp0i(|-$I7Nec)hscR_meiqvGFY#
z=>E_rK3*@ettbw%tj_4bbJxp_`9F>ab96%8G~Pg0Z4PHfcQ67j1qQ_3DTFVdpq(%{
zT(V*zt6xF3ldwI|)bx#&qsl5ch=+`mbEycuqYaTxuB)UI7%2lp3C%y>ZX?|o6&tjd|c9hW5IwV
z0}R*31<*}S-Nqx5>V{e8JPDtd;q~6L%KgD1O44Kvf8P0}xBKX;jgaRbJmN#FM^sZh
zW7y2!T825Z;^0IG!scc(*M}@+gSD-x8O9UD%Y#`#M_vb@A{C)@o=K|2qOg4}($QLa
zULL89JZ9y2rRsaWCZWcZTk8^E`2uxs*<3iAUGRh7?*pP?BS-u-;MnsrtS+@tFMPv0
z*h6KCw`%wlyk7cJ0_RGRfH7T?Ly6kz;xFdXqSdwvyQzf^I>PL*4zVi^&QMpPmJG9O
zho(lM*!2zXjpNlc1ZVOhiEkX<*6?Y~qPtdORPcrd6On+=3!X-;IqrBSw@j%-q7=4U
z>m39*6@p#+Il75sV*+Unm|KTn*LI{yu#Lt7!hRwEe}-_w8<(L^hlT-RXu~Ly)=@
zCk!45FM@<`qy_FU^=NWVnE30^W+5?~gB%9bJLry0%Pkr4LrV>Q0Q>Zj(#PXVl<$QP
z=lkws6wS2_h4_}(!xf`Q(cw$#7g}A=)l*YA(*ziGe0W7Jiet`Ce2H*X{z|aO4Y;ia
zS@QKyE=jz69Rteabd~k2KN-hn7P)J6uP=@-`u%Jg^AqV?X>9ysr3#LKpn1xJMyC4V
z?X#6mnXaOJh)?yI8(r^MRa-U53YbQ&zp~aeq%I!_>3C%cw@Dh0?wcA6!kxr^X*Hdw
zntc^%`*s`0wV$v3N9?mzoS*}-r>GYWfiKWUt%Q^!5_DjN4tyajXq@R1U|h!@ROUJJ
zAWfx3d)Fhh)@5ghYQaGrCR<%M3
z;A3YC1;N3Vx`A#1gT#Wyt|iMYEZr2?0&a}cF)Ju7o7K=>?I)0fT}QIRzvF!p+l7UJ
z$pV9bnyT#lkT@#T0=!Hx$9x`l-{SbFf;*hgSLU-=G~Ab)94h86sXFY23^#t1f%TQY&BLoCQayn9NrcMake@`F%oPdXzZ6RQ`Ei*di
zj30sC#he$?$(2S2c5Nm9gzvZZ2Obu2Z{f_xcr{S5=xk|n{}p#f^a~bF1>t81t+~TH
zGeS+7;&q6Ux~v!6dF=)P0n&Ld^J1n?WU1ZM^)b@NXOqM~4h51VGgtF1Y7}sKe)=On
zTo?ujVfDRPN1Aq&kfR$?05nrgbu8sG*=$+bNufGU$ArvRSC*;IWckbA
zuw#h|J3;eav@Kjr(h~Kh%@n!2C3EuIFDU@;@ARBZt`_S0?o$nM1J&b8|^1kudrVmFb9x6yfmEP`+5i5&0CNmuTr!Su}
z;AHWs&*ZTh-+c>=W7`Pgg(1v0qG=4oY|-B>RLCU8IVA#Jl0up3*ktg-=bpDdo`z{#
z;5xFq*!iBU9LnBMDN6t3;Udo#CxwE8!==t7CFs9@xaT*t{KV^CS(w5yxkk))ky5=5
zl%a-$6qgnP%U{;h@VBUbT98@tlsVx(W
zS&@UY??vVM@K-yv_RnKMLMI}s*O?iSy28gM!IZGy?Fv+92Ach(N|JM3X7cu
z8Jy)b>D_GlU0%F_`_joWA~85p@UR_EnJPNS^5B{JFY&
zp3B_)jZxE0Yk{!S(ub-fr-Hq9B!Sd{k2~>*t-T6J!
zkV@lYud0>UT_3)uIN!=f&q66pPE6S7#>8T?640&?laAt)uV%oyHLLE3V0-`Mb13EW
zd{|v+nZ#d0E*1;P38Mc0o%6Y9Zy=;*d@tFniOh{gWlg9FmB?3{JWpbzr+yCVd}&mD
z>o6&<=C|T@&7%SOhCfCbs*hgH)Z~soeg)^b%81e)NMlB1$1~up+}OTkHqDv
zNaM;eH2&qJMi@+xK2dB>QU1A!JdrvY9z!YlM==<~zMcBFOuLB%NYr!7ZOrNW62C?a
zKBD-}gEKSg2b~~`p~MxaIX0bL2#2o8h^ADypxyYwShFAb5Zw^Yz=G(B`na7sD&w~=
ziL6V~M!l3t;9A-XchmhHAJ0yB6Gar_zg$Z%={_iV1D)$t-`DNfUW7;XNlOx!n3M5e
zFfn^On{*9so5rWEQ}D?s!;?L^s)WTyd$ru^gLCt|rDKQqp}k&#&NFy?R|&^TdXVO&
zW*5&53n+a1vdr}19ZzU9f{QMUg_4VbU!0}?8Ks5(M^Ok3jx8lxj1U^tg5R9GD#
znCYE=vg>6kv8v$z-!4{EWdlf;kT_QYfG4Nd2WxO+=?bl8n789^ea52+;ehB`NDoo9
zGC3)$Hb=R8I9q2vYcXpFzLEmKS!RGTncsn`YID8%7EZ8q5BmE)02*rA6%!NFdKPrH
zT4yfY>}ZKGT@R}(YcW^!;|VCG3Hr~vvBdQ)^Jnp?0Q<&
z#30?GRqxng!Ix~dxW*pt!G9VoO}^_&Azr2CKdHSy
z`sUHh{LOY4QKi^Ri`&93r-?PEC{hG+FSnwVQ}brkwj+Ldrf-a?mIa0gv#bpV6?9(k
z#fcMzJfhf3UYw`jiQ7CmQO5&YXB6*gX$8rY;32LPX|vFU9~sGbp{_o(*fZ?&NV)oPt|z!~LSQ)T>Q7L*IG<9>*)clbJXuepp27o1vv3cfH$b$^NbD
zo-Bl|lZ9F}wHGIVrAx)=S(0SC;i-T(9)6XGd;%ng$ouo2fhc6yyP!jGM5s
z_}lHD89CLCwCHklg#B=dwHtcRB3{QW%8!>Z;vh-Pdsg-J;!xFN_Sa?Vljun!e&G->
z4%^x4+Qg&T!^uA}0K22u^jrRkJ=MP261G+^@W3t%pC23)pe8_*eYm@)IYqSG98NMq
zq6;u+x5UwWZ^|GGIrt$D7uZS=AOb=6{RkcdB_JT6$nARPMVwn-SVoIcw?j%a0?&TP
z)GmEu^;=ilHiuqmA0?u@W6Q|IU5f*A$fOtK`Tq0%p-aE%5~Yj=``lRKk6WqMi0Tvt
zo8saxn$M{*j*mLi+7G9yOKc~9Y(ZP_$<&$BvXMuzm+^q3=y&poPMSootm20hZXuNE
zzFyo9yfQ1L@rxy#C!iwKoKK&}Mb@yvRm6SuDYP-uoYMAKc3=yd4ju`*BP+D+esyD}
zQsRLM;zl${^nt?j7K%*{9FjmOEF7ISQ^GzFmO${ngtg#=YOvR#Q)_sq?s7_=+jTpp
zYA-Q(Tv@BEs?csFg(~_QlEu*Z3)Ffx8*U?9mVlwhX(JR)scMLuS}&Epb4W*%(U@T5
z63Ke>&z(r9{dlb|!y;C~+%_Lk&Ot4s5j`H&T+#{JJNS>s=keQ_l81}B`1Hq!@>7bk
z1Jz66Q7KYefQD%1`n?Om4v**!@LEZ|KCW1i>U_=XrM4uQ
zV*Oc**M9#-%O5*Z@g&8HG-~E3vGem6ZU{Ln@_~JIUatkmc_o-=6;RR;msQj}NSpTU
zDoOt}Zsz-`r?zDXLVB1cEUx7%m?sjFksY3bpS!O>^{$5dQj&2eJ?+2Tlad#yF+SaXg!2F>|Bt4q~Po3W0ccKg%H
zx*nW8(FsjpGva2m`*15s%^g%=bJy$iWw)4u3ohPUhseejYT4|l%q-L=A+2fnXv!iE
z%8ie!o7N(a$35^eBM{Q2-K69;TP=(6sj~Z#vv8oUD11dC_|f+^hgL6L)meZ>6LPse
z2)`0VUb~%EATsM1tYD4tFIeQ<8zJ$5U3fTN;j5L!>TI4$U^Wm0MrU#Ea06$6G!&Ra
za#aUzbV#`*V;`cSb{g(U);n=>ni$GM+BQd-PSaY)*TQ44ew~V
zCAo@+Fp>_HrQ?ptPNi4xG}hnM+cd7xl(_NGDSkzI;IxOmF=#h-;Nk*I^=AUD+=~b$
z;I(Jo!)c(zc|WRiaJkEZ$>rUx%IiF$5NyR^r{ML|Q$MNHD%W?jxb7jDX%PQ%F_9J#
z!i=cnr|VIuh4StTb&x^$&>ubwgt8HH7b{Rb7MQh~QRRAzL@t>*Z}t0k>(6}1q90qJ
z@lIC-`b7wB16v8Zors2ifZH`j=LQPwfQp-eiggne$Qx0uzCjXoo42x=yita>WKfW~(P0vp#HTZ7_o{4{
z>7&jpeCwB$-?aq2e;#CSlbsfJqI~po*I7`EJGTC<^#J2R^WXA@h43ZjcceN@Er2)X
zG=*-(R!&_GhcyBCUi7)WO`77m?!JEv&0NF0u(5H@?HRrXirjLd6vBtLQ-pu*N@%|8
zHJ5m}!9_5HuSAERwJ*@om8`zR>}*k0=ok%f4kMl#YJ4cT|ELeaYoYiAxN&v3C$F#R
zGqA?UZV}9EFdk}D?W_gpT!{O2aUudDFicBREriXmE`0^Nj5dL
zm)f%@jmq=7z(#D}bpgKRe&U{J|Ef#hNONJ%RYCUYT)+YSPY_HnaV>o+CKb)e3_>(;2leXY`3*EJUaX^?`g8@p=#FQ|Y~Zvm)iMk&P!Hbx
z!`8&(qn%I%NY68^m6Sd)=pOvK3Dfaw-gr6E(pUI7TE4gB^;(Y7EU?~&LJI$|o8?zM
zL;!nn!vDY)8qcn+GiTNN`t?_0WKZ%^PB3lzZmPtXWSHg$tf12wBI*i>6mrbBS5Cy<
zR0Vq`u!04I$yZ7Ou`O_kx6u;c;bWAPJxI_8@KP3$3E5BI6BWWW4_1_c#xGWQ(CkZ&
zJcfqQYU>k`>@JLzRjGgbBvskmlKaF
z+oTenCso&drDt!1e4uu;{4=j;r_ptN2ifx%L9IKU%7g0Hm3Vkji}i_L#&^wME_pZF
z0_mMxzqPUZuIZJt^cbM9@(jGU%FK8JpIf|BFY#D6^UM$ydMWA@BZSpT6FTo~va6o{
z#I57BQI@$dFo`F?PtZ3`s`tEq`9p2Zb`#%mC~#huK3S#4;7U)
z83M^)HLsVPHk&ogzZDvbje96(I@c9urw+00ZoQ(@Yv2zU6QK=9-Bh=%oSO;4-HGvw
zqR0_FuDjc#jf`wJqiZXRocnYyFFkTHQQNfP9;Z~ci(V@8qh%xDAR6rqC>lBcUSLh<
zcF2cJ$Fl-M=b1VS6{x8UoxL0ZFHQl!$4qEwXd0e7yTHz;5F{x?wS^k$}YF^RIG`4L`w-fPPD|MFR{dECe3f8W`)|LCVUNmFM@Gi2FQV8y+FxzOj6#Rm)16FRGcR|Z5nXjR+Ca@YJDM|a}X*$tL&
zm8l+MLgyW)e@(uCCwQX-joRrO%{UjzN<8}2^@T;@iA~n+XQ!L26{N@}xIavLg?-x)
zqKWUOk4}_t?%#92qU}4eY5ZU6C!9Z`&u!F4W^Jm>P@KVxrONb*L}oFSkIaR6o5hL+
zEga_i<(J(K7PAqxK`OJqA=o4SxP_Ulo)1Rw#%1e{emsk
z?V`;j63Zw1R2tQBlC`0RlRAUXBYk)SV;@;QMA_Zg>~_C|DdKVfbGM#AM|smJ#iz7w
z*P|~cEZE?wgvH+J&mv%e9%xg2tEq7tPmLmapzI_0LwCXM0*8HhUikIr=LB>ZMJvimo1$Zj#7zCS8R5erD9olT%__9G-2;<-Ic@nJj4f*6l>tAH)hBKAFAT1(XeuBc+|~ju9~(+*ln=Ax>D5i5EFrDP
z!lwDbZ68h0)U4+roz7BKx|qIAZ=SH%y|~~9d%`V4kg&}MVu-g659vF-HM>(7w5I-!
zm2u$BbP}!#pbS(s$XqWqg+(ruzbYD`ZL!olk11-(--$sX4?p;rE!Lu0Zj+<0YOHF~8ck2M{-rs(%;$LNm)LM%2oSjc
zITpKM-&+4t3TBxU6y=
z=gzD)l0h|+N+C$2+h2m0M6abgQ{XFZzP@sYp2+mYsKgPonL(&0CU}(3n^vV*E|K#G
z*GzG0|5VR8LE;aOgJXJ`j0?Wob}2ZUke}0*B+m7i1`1Ti9g->p5^e4WC4|TwcNZaZ
zndn8iW+${_=fQyo!?8E}LO$D+TujQ7;`4NJZ*7+)l^6-_g5T&+S~2~g}{vqtj8BZCP#q3q-hjr@+6c^2D!_^@e}gFC5(tmYUNYm9=D
z%lmbSEn8d&{8i@PSdV)+8ZSCHH&55qzgfXxq^$a#nF92QlTqSJQO&BLJ2
z2rNltOdX5!*sk=a8iw-%zZ<#}{|U9se|t&a&+dN};aujj#$jB^#=2~O|7-k{k@!II
zj^4MA1q0QaT-O%n?S|5ej%TI^9h7kgIL^f{7?D3&+Yce_8*qeimm3hE@^4rT_HjhM
zKCpKgl_N(drB&9@E!~PWXr257_5H`R0O=(Of5eyp5({4?V6IIAT6?KU#!j8%BZ}o(
zKCQ}nCZ_t|kH`bxP|!(*Fm=L{h+O-l@I(GoceT4&?BnZ|h|4sXR959{B_R9uW{n8g
zgkQVT78dP&ebbS++q3&LxihRld*-%=P;s|pbngV2cz(2RZ#-m+@NBRXrHsUiI_z;NJrh7dEf`7DvpIf+lWs_Ejc91?Wa7{Uz2|tk
z-c|Q;lps4}Wb5ESMyJ6Vv!LOF!zq}n477Tj$P;t^`&$bM3F#$r2Zcn`r_J=pv|oU3
zCcnq`1-hSu>Hhw4FDW9jcSO!o`ys@_;#OmrJc=UJwnIFHq;2Oa%5H!6Tq`8$u6ypq
zzVljws-Olzx*vaVb}6^IoP3J7%_i4&0&BPMNWQ28`)9xnlcb(YHo|69N=IWBl(Vre
zdlQrNy_f&|OR{e9hmp;erg}a#lPW!g>rwSAFS#@o%3{Fp3UK1kEa%X+k56U`K_#N;
zm*jccqrkRPl8Ddq^1`JQaur0PKe?vUj6-O04Wdh8q-0n!TG6kt6sOZKBC%PjkBx9o
z_cdCr8BJ+J!pY_NJb-}B`NbeZKBq&tg3~&Wgm|e>qn?21>Id6ynYq#^r$i8bK3*ER
z7+j0vCBbZ^cABjKdL9(EUVik;4>(SdIutd%(Gi8Nw=taedwD8#>H*HYzxELPt4zD|
zBC7YP2@xgn7ASwLe^McETE{$EZvMRfPRW9HvC(S-j3FJG@~W
ze}8EC;-%t%FlM1srX?;_P7TXd
z%GCdl8Z~*6UJq1HoeiRit!OgI&Dv%*$RUkPTMt8i@)4yCyzmrswLv^yhqgzdtoBH
zC9~ERj~AEXKW937vOCDk%=8G7amit
z@r5Jh1&UA6>C^(1TYP(3bh~*O`3{jn%>Ekb8Pw6^F<rZK;%hF6Z
z`_XxgJ)U;`0;)PT#Z5l>W|DG_wCPTy%#Z$?ur3yBbb{s6
zd#UI-+~U;>m{?5(N0enMCB;Sb>ftuSuli&15vN__H5%#?*KYVcv2f&@Ge0BjzB4C%
ztgo#x+xz59%Sg9dJVf^1V9O>=W_p|&UGh?wj#?-mePTFz!fh*olV=6tPVTYGNcFq6
z+G2}Wsh%s(UaSs@I?yCRwTQuZ>&8?x980q+^(W%7skL?b3`}5x89R2HDei;svpF)`
z(h59a{(PRa{mqr!kA)w6Pl{Sw9csbbCSvLQV>8Vh7PlUlbGT)sKmp~jBn?R?9RP|w
z-H0B{^8n&HJ)isJb5j{<5#}dJYP%3+as_k(ZU=S_+vU`|YmNe{*8DXP6-p$c2_`d>
zGAx8OT>NMfbw=1auB9S;@ouAbzATC!5zOanW3$K5fHA^kN3(HJn3npauVs>i94v0z
zIB+iymFWKN76sSYt?5Jh3nZeA3sB4`%cD|UxOsS#goPoOdy_%HK$N0#3sAr$fCw|?
z#7|)`G$845zJ)o%c`I`;RTR0NbF%uK@8NuyIiH{5$m{C(OO6`qQfWX<4GWk^c%j!`
z`Lo&9#6`8$)#l)F8eZh{Mw-@kTZ39oax!x_-^=w*->G)K5yG4>3ofC_D6y+|XG0#-
zNRl=1_^tR&m&G-SYkm*-j2}y?SG80WUHgryzPUW4t~8c8S!gtTZCs%aD|%4>lQ!Qu
zEBPxBtrm&8K5xH-A7ETG1`a^1*nWB|Q%np{**&o^v4kO!qoWN!+V$B8WU?w*xeX^Hi
zs5!BuLJ8A1e-c0AidCG+=TqKiz>SJI|H$nU8!T~WJ(wuuG?l4t&0uJ`T^~I19(IDg
zNw$D`^V8{~uBAd)ITU(eq6U|Ja}Cw`_ch`R%W2hsHAPTZG@r;7VK*Bi>N|*z#YQFO
zO?>vu(PN-BXy*shCQ21Z1_lPuc-B&eTdbZJ=Ay*U6vIJKbbJMk8bdc^X*d{1KGWi2
z)#oZnC6^i(8bkEHhZ?=e<>(LXl2$z@G3vUO@XJi}%Cyh7wrwKYCD}NW@`xQpM)gjJ
zw*>F}mEC85O4j}AmX6n-pu~S~mu`?L2dacRb=!>^n>84k&B0QbLI0HH@7!Z2Yd
z^MYJ_>L>XfTwu>V@Wx^@c8j*#_Vmx#4G5Io!3i$;Wl54?VIf-+G2z1Z|Wt8+QVlZg`^=cV|IK3RNlqwa#{esy<{QW7&Nu!xTvqo66eRy+gH@7(;rvptbgTKfhK(qC9>r&6bs#|4oXM
zGO^fXzzDfeZaSf!FP{9E94<m4Wkob|d
zD;#<2Vp*?zEnlzfWxmGtF&_5*e9Z=5+InV17&{o=wu#q8(E7;d2vRTzo17`mR84&{-8&%ZQZST6
zProXm`{fik2nHU&&j=;>caOxOknpGCFlv9GktT$v%YR8JW8T9JCG@b-7r77S(i3S_
zs6^@=w+7?gFHx!r;%MMlG{9$KN50+%X~vi_G}37Ppx=hC_rB2SHhJ-OkEtJzsi>b|
zN|@$uM`z>8-?Eu6qt8r{L=H7XH(KR|(R^~AGdJq@yL$m_PNg};Ck*P-R6b^xGkmQD
zO&*TbI=H!PuXlgsg_B==f;
zvo%yvn86&X%)0ZE4Sw*dB0ziVvevieCLC{80QCn;ZpMdvyV#99%OH499Ah_teQ{zkB1(2A<
zQOrXkw~{DbDs>5D2b1aONCbZYLN18NusQ8MZ(LrFb+O(B(~1x4?&%rPNA=rZed?E~
zZZ`ezi@r?YHxj?)-&V`|T;{YKUq4wx_Pu!{oq(~kzcZTg8s$lAQhLeMTDO+m7r27br_X1c)UV=P<(xhdSl$9n|fe^w&UvdJ}a2Bwz}Ck$|tqQ
zOm5bcX*4q68W6g165V&s4@%%`#RYHoJ_}&$tWv*usPXmXu-1nu0NhEfuoI+mNnG5^
zMvD0bFil&=I+`?4zFQ9Wfx!LYmV={QBkX1Hi^<-_Dr4C=1}%Z;T>^;Y(!Epl-qg}e
zozu8)H%!s@7wCiama~QV*~jc2y*Y`mLwO301oZ2D@66XnGaA_)cK8?Dtb~XKz2Cjq
zJSzD3Ea_Gb2V~#qQ2iIb=w-}dz4VU6sd?cJ)Rt<836BLblveqa8IM-1XpNN1-Ai
z<)^FBD}^v1nhRrV1UWY%ah9kTS(2$+t%LDYc3T64qmDb!@#QX3cEL}5AFaU-gJ_vp
z4kZmkt6Z3?cZa4}+hS8X^&67zP8Lkk@km2iG2mVvK<$kxCx|DCZT3P@xtg~wj5tC9
z9uM6msPU{f{Bt6UmdwzhN*dL8x&pI)dqfFa9@>&N>biE?TQBl$daV$W4sTg%h1)X2
zwt5-WKmXZ%F_w@S+X{R#n}tg3q{nu?%N&-l<7GcQNy!Q{`W6lhqVrFWsoWa6I`xE{
z1ulnN?~w6g9l8&m>4MJdolta&xsV6IQS5G1bUqZeQJD2-B=CQA3V2iTVPWt`Kfg)p
zPESJh7mS~BIo{rj-=@~XCZ+iXJo4Q5cMJ6zv@q)IC@~TO1;MpmA!@PG<*3LDDks*n
z3?;@*)bU13ayw0C%SudCSa|s9UI!EN?IG;M6&@ask8+~nd#`wWu23oEAma-JCkqSi
zc@Bxur%woXEExS~h)%6$a`C5D%Us?wTS>d|T(OM3w}d=}EzW-|4!B5hn*|eDRZH5K
z9-OSs35Ts{R!`k2u6{0lRrW3O4Q4eUc#tIG*i77EFcuVI8-G|*cq$lQds8SIyEc6^
zVQ)E^|KgI6B@TqVJX#!>H7F=3cHevM#WoLQ4-c$d(c_za5U
z-#sGu;HL+b4*WPwzNI8C;w-9hX2)SJBIMQ9>DTn7+(4Z*qLT-#rr}t;YVY>P|L)>af;g*nzyiIP-;%CN<;i9%
zQjioEI9Faw{L#l|V*IEOFxW`2*|Dqh$b757zvN4i9pS(fD@O9^{z*=MOfj8?>@Hzw
zD8IL?{!F>!;*b?5*1La^CFx^%RBc04k`kT1rG|cvv#m72xgk~2F6utbg@1MRd3KkW
zE)27NN71(wyNQAcv%td^x14rXpg043&C-x!~ZYgDu_Y>i%DA9a^8k
z!}zQDbaye2qLxB-F{6j(7PpgZmh5K0YbO@+)wMsnI$p^rSF0$_;aKdpC*iZ5K{#Kj
zk3=$+mAuJ%HMOfiN)xRj+P-V`NNC?WFBEpj-Rt8*ULr7HH8S4;-_FUgd+x2A$u4YS
zQGA|P$*-pmKQ$5(FUfR!vu!ce+$KOP|AeRKkzYX^H6CXRQKgnm6khhaxXUL^PMKGj
zMkgGmkDWjHoJdtI%bPN_Wm?Eh#PQ^lZiVP!zY_ZW;a&Jb?ytpZ1!wYW{Ml0d=v#Cm
zQSnF|aU8WMbKw_T(KHg#oL`{0z4%=Yu=HA-!$B#5!hL6Z48IVbi-?@t_vT!j8@u}e
zkAZy)-ENBXt8-!_(}VQzCd&3V_Ay1D(j`R?kBr;h!Ka;3!Y;-IpzZeXA*-gg1+|ax
zASaL8YFOr`s0E|HZ5>E0B6KOGlUk7F;BWbk^M~2m+CukUoSoGYYP8K1b4XqsE!3Jg
zg9FypvG<#(%O{4_xGh1-gIcN)F#oxaep2*^Ji^vLGhndHHxL0neb@#|n)+dVYqZrlH
zQ%=t_Gf^>(syD5Ss=pbXrBDEOFV4aOcFvpo*Qlf(L_@MppWb4OXyv~P4;GU;)xo7#
zXINy|;1FPWc(}iDI=g?C$%9@<0~BAW7Z(FG=(4;C&SBkC&Cc&K6BV+nNSVhegyxxp
zYIHCD3{7wO#0_jNs}fA16Z5{?pDw|T#0HnK=JVpI_eFWcvQr@SQXRvi60l>8KPW#X
zPKC9Rc=&Cd^^XUSI_+?)QrUZ$9kJt;^8x
zVaa}2^mfuf0kMZBZ^}GnmaN4&AnXIth?7C|t{dP|*w^a%cz=T^$l@gaKPdl$d;i%7Xn6r35V)FNRudEA(S^ykdSaHf&h1k9a6&wuHN0J
zBk`REJ_0dBJPBpE3P@-!59}P3me0sa6Q$FTh*-}sSGz=>cv6F#qTl!n-I)#y=1$O=
z4*J2~QE339PQxb~Q*ik5u<6x%@mWadda5AHQnhcZBFdp8Cm}?vJy$^#$dx*rw#!mL
zjNz4Tam=ztz1=g0DE$YclTmT_Ke;Hf`D^#ace(FL`EbVkpcl8%p+)sK=m1*OXr|#d
z=GHK9D7j!#9oY{OIyK9!2IHUwvk(xhcV7z?HNf8){};5RLrgt@48;D?Qlo>1=S;0u
zE)bq=6qdG3F13K@Xq?7Pd^1z1wOo51O;_7mFEZqS8WRViwcHVKhTEGwC}h
zas%WT@kdLj{(ZXV6BTe>9&mbU57i%({}EiwYei8HzByE;Qq_N^VrQ{i^dY~gP!6NK&U4X~$&H@4q#a}wTqvrq3
zU5c9E9cUFupn?F7^8n@@kmhYW4ra_u?)KHc{CotJvyI@ImOeI1Ym95Ey~TQZm%}+m
zPG^lHkpKGFF$keNP)L>!k#U*cFw|*OYR0c0$^o(0XSv>-KeqEqJ!Rp*PauieAluu=
zNBUTIx(anbaU1An<)<(k^eY}(ek+H2ZAHeZI5+5(#bzeoJ*)t9bX0E%nvG|i{+nMv
zH4;GZfY+}JooXotD*oFy>iZ%uwoGQ7%d%#tJHgyV8L-;)tc@WKHj*_oB|oBQoSCl9I7mE4ehTNjY}#BX5GA;5<=o@&!F0ylpRdqr
zeAd{fg;mh!XnX=^UrlO%i0Y^3CQ!csLS^OTo_gy&vTVu>hwIKQV*2<_|mkTOzmZ}sQJqS%)(5ty|@ZFkH
zLB=E`RKToxdo|bTgk9Vp`#imSKO+g`h?rC8IBo~yz}EDO{pmX_4C@zLD!>1J>M{;M
zA??YQ0$-!#eY4~C9xzJH=7v3H`|qP+(84}z26I@>i$;nPUj<$k3sj|@dwzG%uTY)Bf0v74lkLNxFz}>oycKj~z!UQI_PoEY
z<#Ua`>qSU?w|;%L>3nspXF+cFH7>)K+MoRmKkmtqVsMqj{3G8n%$T&W~U0cYy
z19Vl1oE(@br00zx6%4kAy>5
z!9~La;jR8V;7rS;XH=`aib34|5f%?h!XNXL3f2(+X;#BP34BQeoED`5&%e|J9fV%f
zzb~?C<p1Cs6zS#mC+Ss7GF$7J~P95F3(gDQz@#mSv+sni0j;F_p
zG2s9FuJha4-?ljNY?eYGuP0q`4V3k!;wdC)+>V#CCv3tl;~zfoqG(rfzeI446eSVU
zxJxuE@_s`Fv)29Zo?HKm(B|<+f^JgIt&|>yym`Ip&}+d;z^$8aGgLh8mUS7$NeHObo}ghld3;O9|jV>J(OGJ_WBnWCcy3@!uy3MG(d}
z0LH;UoGWd)_;^4045_y=P9tu;poFZ({mht7>+72$`K(2i7VCzbT9bifWN)w}W>-od
z<2U3KBsz7l&({%+*}KWUfV%$M&Q|I_;dizKQ0jvj{h>~(2*AHGJQM~0bz{e`kRo6w
zT&}w&*e5AzZ(m~<)LGz02-^n24pdUoInqn-5C2&?pg;Xs>w2^Z+Yrg?&=0xET2WUGTY=OWV26-gZC_61k!2kHYd|Cf;$TIYGoY!D`*%P3!pQ`?zkmI
zAmCXya2mpADF9s=*GNL2!x(Oe56)HG(#IGQ0q{ZB40sZ7W2D3q!oS|6kgrE0dHDa3
z@kn`3YnzH``20M=Ca}3t)hZ=1%afcSdsQpLjTcG1gV?JJS*b^EK#I?BmGu@K42siDf@z7fgrEJ0Q68jA2E1Ee8i69fJnaa
z2TW6F>;k#gN5-@Z%=K6_AO7p1dIa~3VFHfy%0!KVuHF9}y6jJ#z|Q$ueQj%Zm)b{N
zG7@L>RhUu3_`R()4#}7^8>}JmTAPPc1c^XP#^uQ4vG5q|ix*13P>Y70edZCV8BU@z
z8sqdu2>-f#5?zw&P_0sneP*k-j90*lMjo^Um8YKQ*|
zz>uQhQUEx-hT~pG6@Y^`C7|ADyYhO=0h(g{2ZH;ZTXi~D_$%sSu&t)S&n(sY_VEH3
z(YtWwc+r4WyHf1u`MOkX(2=YT4XDBe0Fewa*2h?nPal2^X)AG>XuU;0Olfl1^qkK{
zk$@?>2ke|e1G%A?8$wm6n{h;9d!G92pTX|k`d)T*!K(g7~SFe%4{W6x*dO`TPviA6J_iA~K)r0f-
z39TP3R_}narb_rrN(3B!cr>D1P++rM4Q*7zqmV$c*uU85L$)I%h#i2zJXI$>53Qup
zYhh;u=m$i8Y9SVsKR4W0R!{KQau6}fT!%YP{&UQwfc(f^*#q}xQcsShlIr&2;s
zTue+u+U9H7&tbHi9VxMy^)sKFbO22xwKGx}WZ13zP(N?)vvVW)!gUwN)tj|4kN3KJ
z{y)3!I26LLNeGUb;t;cbTWW6Q-xnKhCI;(Y&0^0$e01KMAh%fl^^>Bu>I%Hb5fQg|?0IufC4u>W6NVWYDhr1lllS0AX)jWAqE3$FZvqdLa}L{e16D1B}Z*
z^ST|&Az)1cGPxw@Zj#F%$uNztvlQ^#^f>28PMWmUWL=UVvOgb7H!;Y`0$3obrXby~
zAHtWF_HBKfRsF747L*1ZziX32BG0Jarz=)!*4C9z9UY43Ha$Jw#exI;`>)S7+FR%y
zj8oGPetWQ9Nbhq5vw&K^D5xV9fnAaw6!~Wv3mnc?T9eFK#_ODN#XM+m`_BZ3|Fz%5
znB`K&;;&dVyNzPA*uHtJ^K~5hZ2_+_#}oX2vW^6Z7Ei~mCwo92wa)tI+Z^TopqzW#
z-_hR|z50XKs1vn7q&~|Q(PDjhotq7+7tYf#KYM!--07g(tK#6-auZ4UTq{u)b%9lM
zi_2kh$=7MKyHO;M98-bk$Wmt56?n0oVjwf@{?7Je1QGQ8Bqq=DHPaF1DP;tZ0e^#vXryRQ7@!l|n#3!101W*xxdl
zBs$e9T^_I@Q^b~j4dFvczV1Z2v3CZWS&Ep~nT%$eS+4~vyUWZr7T}1LZ1I@=byB~3
z5toCRJbbo~7`~lK1?Q?Y@D`7IHShxM8nUq7lcM@vvHv%++hap3_CKBWpkomc$LIOXx3~p89-AJ=gz{kw}RCquO80^U~tMG
z3Z@jq5a;tP=?A%XM;%yA|O$v9>L3YjvpEbofLtf
zqypOBBe)kEy598QCx-xwF?lf0cCitEKGB990Op&!okD4e0zlVd0z{!w!IoWPfqL5Pz)DM)v_AhEpcix|}So?%N>;8D}(cA-RMYUaRla@_kN@_*J
zP=$eNOT~lIAcI&zU58zm8W*)OJLL?FJp%}TD~#)l{m-hlEA2iR>R+_Q9X5I$-b*eQ
zs?pqwii$25axJ4ukP!ud2&&W*iuz-1O^AOIe*AeD^qB(W`*3@+zUMKR9RtZ`ZnqBp
z7~uyB1GxY{PH7nUhyzH&p%yj00voVrJB5B68iWZvV+37}m$l%0{1s%u8y$?NfCmDg
zlOU!)UmQKLK}$yCB%K)#%8Y!yLU0T9EJT^43i@E~i|ljHH7dzSGqn0Yn-LG{3x+w6
zfkOGfYccgEN24D`rOCNpvU-5MzP){9Ya%j`9@gr2u?Xo*X4Ikay}wp%vN!rtezE8I63Ax)P1Hu`
z`per(t5_0bx7B~{Bp$v!Dfi&;Zr%L=KJTNrJGv?4ok$)jA}80OcFRrVjd@)Ps{r27
z0@xU1Hxs694kQ(RQH;dZJYebsgT*(afE>F29{co?GU8i?>(bu^pXjBpp?v-S?12Fa
z|0`iUBi$zZ08q%Fca!(L&t)e7XOaUtp<)sX$^d=G5@iHvW@s}h)*KR;?AfQcf}`v)
zoEgw+$sp=_I2R3C2-Y`nOuw9;_owq<*7Q2?+AKAn%%2ik{c1YAU_1v^LoJ;U3<9~h
zQo+Gf)M}fDSljkEl)qCF4yK_SM?X2(mJYhyN&jZ8PC;11Ob0+TEQ`8w@tX;i`uZbd~Bv1c(36j7-xM$n^aY+*1W1YMz@mtTPxt22<2$>TJ14KQ>};beiNj@zpg
zM*`^E^mip3!>Bwjp_)p-ofDrCBpu(!8#77H?LRJ0fJTDF=V~oLv`$6g>5Mj>3y);L
z{UQJyu_2o#d5e)A0;8g~FxIHn;eXy&1_fY2!t;8Z7pk5*{LcL}VjSZAPL8{;9zpn5
zl-@{chTJW}8~M!C{#a6G7?qx;D1eyVz>BdH2z&c+L<(He^Fe%5e0qHN(h)DqOZ1G}
zh>UkR{H3PBW-NK1!8r+@91)3zYTz~GDG1cy(5+zSls;LH_98SvgMXbNN&MCUc&9q<
zVi;TgFDQu@$pz_mN`T0<@ax-E?m$r5$08uGU+yTnnfVzrwxy5t?Y90c|uZC@>_*c>I2OFjMwfX@0n2cCe^km@MMd2X7Lw0f)nh
zSYCV`eE+xh0Ggs&EYuhUcjn8Di9qIXY7#_Q-fIr$2>E4C70K>q_BUhaNJ83K@;-!`
zA_tsd3y&xisD)-H>RCYCJAO^4QI%=LY4AVSL2zzVrVFJD
z!r&6mK+Yhe*~BDDpe2H83#$ZJB$q}6g&sDc$7`@B3>~o+gn|`}@zd6uD9j0*05A)$
z1k4%Bzy$&pS*a;iv<;qW8i5~(B!WMsSh4SVVvq7!&j*pH`F@E5m2j6EDgboAWPNb6
z$)8om0V15@TJsiDWH*_Y1<+dy{^JYM^$i?}wzRjV2{?Z$x$2tP7pBWaeHLM2
z0)X)Da-l>yxiN9b((S3DzrAc0Z@ua#-bVQ4bY=?0W=GS-CI=Hd*iRN10Kn~Y?gQ+1
z9>gskfn{{#oiEmzs!RfP^Bh;X>{hn`DU;?Oq;$;<{T%3*o1LaXC58r+3tvWqq$NWM
z@No^qY)aA)EYp-8)$i}m>W`sS%r%jXAS>K*oncd~lfi<_A#S(Up46(zf?LbY{&55k
zzjH?9y{C9RG7dv-o>ohP>mMY9*?~v{H27EjUjO(NF=`j+A3YP%UA6WfWk&2mR^H7&bG!6J)3!oOzU`zCSP49ItKI
zw89`bqo5+fkTVQ`AoFMjmx8BHr+g#Ot5>Ww?yIc$B>OW2#L%O_aSyJZHDdKj;hOc}
zsYb9Qi9&w9fgx>tm4?sedl0*9VRvT-lQD7-aRvQmz
zD~G2J*D8U51V{9SNq>T%{=~M{Pl@_n@IAqxh0c#<&@5mO8vYhcn@Xs363D7~C!cSX{f%p|a=q$>dkL^lwUO>vD{fjIAe%gH~7T24_ww;g0ah_+9?GQyuo}I12QNaGypg8$GX^&3U!EmD~)TuvhGhNZ)1c3H>rqFox@f}
z8g}&E#7BVe)DSvf!MIugGs*P^{A_|$Ye~rQb
zwFz>)Fcj2lTqfN;9shU6b)6WIxJX9v8A)eI1m4bY(Du#H^X%Zfd;!3T9bj6@@K^~9
zJVEa!XPy^V(U$j$SiVJR0Lj0OVoTFkR?
z*j^11-cO+1%0x_O4QfHC_&{}p0RXyWs)Ek-WKadc<#2>6qld!#2#&ta02oB5I}{)2
z9xfqQV(;tIE-$!I3Zs9^Z7h+H%9dZdy5X?}Sp96!$G@I07M~7#KY&;Rl)e!VBv*+o
zNSMcFNuV^n>H6vH^*%G{@mj3dlzy0NHyy9WT*F+L49sFWaF=0G7x_&IRx}C0`{8ol
zsKbW8qV_Q4hb!zekfN5)%mp+qSuUs0ub8aPEzbKdfB6jlO}<8Rdahy;@iL9{z@Ur7
zqcA$x%rbo(NmZ>e6xn|?0euJgBTDELz#oeiJ=VuX>Ubk>lR@~$8}=o&{5yD8o4k(x
zSnA(j4X4nc+A`M*aNp*GI^&|+?>-KZWo_hK9-T}`_4^>QF8OPBZtvEzxL6X(~JFG(~LI?o<_e{#TGB=i>YUxjT|uGNyCM!}bDP&2IY2TtuuSpCTXxeS)pGSO>2|2A
zA_aPlwpod&KN2WE>x>Z1L}_S5hA6@g9Rpn2B}(E~Q`+Ksq`e@Sr2*3!Wb(WPI>RuV
zd>?pCaRc^#CBv!bvG{@tDOM7VQg@hA`2B?4N(<$tVOO-ixw(0;uy=fQ+Q|o41`5av
ze%FtNkAQX)2K0Q*Zm5`yumV60MG4f>3v(-|9-ngIsB9g3>qhAu7-IZx~BpsqFWGWGFSFXM_UkK$z-wnMvcwEe=uF~1|RI(oCYS*m5qNZD4ycZGOVOj-^cohP6AZV_R%S-E>4WAvndXGhD_(!|!n*q=EB8OYU&SLL5*6^g
zDjL;keAVEZ{E&}@O3JS|()i&#+lJ3|%1%sjO1d_LRk2B`OuOi1z1_;|-&Y;RvmNYy
zlg6UW%rZ%Zc7&X{o-T4a?`(cHWG%Y>mND!5YpfJd?5N`~`S
zIM7-dV>I)B-o6tE5VwGAb8nFyEoocRfk<7kn#gfVuV#v(=fn)L5
z>~@kn^
zvT=tm>mL}i2fBLtJQ_oE^W1^UY1NkAL_Wu3PErD{A$|w!+q%D4tCY}i
zaukVrvR(uV2hNkZ@Tlw_xF-Jl4F|J!YGKNCT$wO1AkxCc>q%3
z-x4g}oW0qMD=Z9Z0&&cz4<&n__wYK!;a~heuHHHOk}m0z5@~7a?ilF~
z=^iAd8>C}sknWW3F6r)Wh8Vu%^Sr%lR)!e&GE^00SG6b
zD?lxbQUNxm|85EzelPMWoPJoNyxt#(SN>x-r#V*5U*T6QQ!&_PV64g>#mR~sH{fg;%{=FU9
zSuM`FY{R}HrSoO^4AbUQt0!s5BKq=^nl8jDC(Sp@9IqJgNu|)pMUUow7d{=}DBt!>
z9vBQb4aI#xmk2M3cZ;zdve$w>GtR%SdC(h_YUbYCJ(bF&EMDRmQEf2SQphkLZF=zC
z(Q8s{>cfKf9usz#D8}i}cj`Pl_3rjC-lod*L>d2k?qS^lSGx4e;;9K<*!4-E_^cdu~(wHBpKkH&gd0>zxL(|NV-t
z1RL_YWODBvJ_0?2(7z+bKjgho_4TRWv-Nz9cU{%}_m|?4*E@@P?R<(p=WhJLQNBqJHr<1^0#oF81Yy9Hdlf_G*-=F&^g
zG{+JjfxbpWX8^er_YweocrwjnYM+ettkgeKA{3nS^!Qci`@KrIoyHI@7){Qc8d&R>
zA-9h@ltjilHvE#6sn%rV0g+cLRZsq8x5~);psF%#Cj7IyEpY0#7;v`%YyGx*Lgni^
zN<`i~+|rhH`RI34snsiHJqc^CjqqLtsz2FSO;_`x5pxH#>a~h{Tx}OzfTId<=oG`7
z8zwHnyCHS@6_a-=oy*MBGm47+*G$I!m
zi8qv4ND)bLOYB0xkMl(~xgyWL-CBkTpDp`al{LiWQQ>+UHd?u!j8{6F>t&Su?)-GJ
zlubUJBT|%nId1O8p$<2O8SUn&i3yD`uO>Gpk?5(x8Y
zyC5ii9}Lerr{qF2~NCNycj-Bb^LG#nVeOf
zUbdf_JQrFivM!8rTnb#lBUU><-w@-Ydva2z*PNrEd#WCrughDC^%e~^$kkF_h%OP8
zei@(I0rXc%Gmw|3@I~i+Ie;eXlBtIiLGEz4P-%ABV|6)MmDMp#eyM)9fuRPmGL@mJ
zM+J5G21JqZW3B%B%Q
z_^(y>q7vwpMg6;8*6NN-LUkL?COC3%Lqz%X7r7F|;eEMmt^mUCI;iqvhx-9X^<
zIL`glU6Uf~G-wc3tTccVIkRYrORu!P4MDfmggO6#tj2iz1*Z)mjwneMw6i4a~p#!l<3wSeqLn0foH@$>Wc=Kb(zy4rPHeK_lWw_a*0z
z@I9E5U~B>(PKICZjZ*7FFH61F^iXT{3g#c4W1aaMOto-nRCo`E}8kDSNd
z(p0Z3Dc>#i%-UZM-ormwBXa6+7_7c-sJt)NZYyC%0c&5G$rnnZ9R{tG*g0L+QZ%#R
z-YNMJF#ZB1{jg`CFOotdwyBP)9O^;#_IuMXIBhoZ5Sf<$@|LByMp$CPL9LCzo|NEtOZaX1B$ND#RNI|
zH;nmC&Q3!5;N$no#o%7G1n7X+5Y49XH#cPO8=yU`i25??o-wvvdO_yE2=ubAUMG2t
zr6Gwv?uS?x{DDMh{h3Qt
zbaIT_EiW^Pt`)ipnvZz#Fn^LI)AAJq_|2Z5lCr5LIqy1baZWY(CK{1(PJVM7`)aWj
z;+I?4HeZ<+;1)6!vWRqQ;@9TypquTv(6q&LRziCLj|5M
zid_Q)8_~S)2de!;p*0u^~LpD{|E7Xr;ZTGWZZ-
zEAxV21ZNSOjJZY(o@@QiJY1F`x~BJUcp-Hvnb5=3!5xeCW;3;AM(TA!q{?L{v@8-J
z`>$ekP}4jrY%hj$65T%BRO~bUs`IWz`(gmz(Z#XIIP{KBCAu&oc&EcZ&?$Ou02zV;
z1(Y$QEu3dptz4&q|&eh%zf!uMBnDuXe?^KhbQ
zPuHv9_xI;ku;=z??{#78QEGW1!uM#h29@FqFRWFZ(SkHb7b(<<{ujU0G!&LAguxLT4RpNGG)JzV2iBv&F`gpRlQCKP)JpR
zNG$S8|vE|@8}-*jpqb#U>KD(R(si%F;c16?kWD;pP=S~CuGZQhcFu>}(u_UsYJ_~Mi9yXL8S--oUVUpl<$
zebn1OY9J
zJO8jRMU@E(sUMK@zh=rYZs^vY0Bc!k;yabWPsRKQh#~K70dMmD^^-(n*KOZu8pp@7
z$u6o%m8IZ2Q*56|LQY97;mLUpP(p5*cJ+IFZjTH~AHD-&*haUMp(#3<5Y_ms??N{R
znnEyaE-Q)o7j;$fN;Z4`c#^D6X?F)7(y0UX2Sw}4vt=miTkb_yWb*s-v*_Lc!vMso
zijF5a&2oeI#O;yTNPJjL;33zOMg=egj~}vTE(-kFlyqG)?vbxFgv+q6>`>`wdHK_?
zqrm?pwN<)10+^&JUlZIF*?!;C?pvZNv|+GDV?L#y=1ahBe^?U$qF-orIpIr(K7#aL
z${gWAzm0?ljNrX-fje)N;eWkTVGNmKtGprqUaZq`kS9vfI%UO{)d}vWRi(lU}
zG{j84==Y?&IuXAsGoQ;L7q9~?rX5z&thRXvv>x!^1ZjOBJeHHebLDFkEZZ+6FT#8P
zX&d?I^8rCVh2AIL?Q~8|0}X-2V!ZvCL&-v6gP0V5|GYn{IpIrjbxiCFbDgcb>{c+^
z{$*k*7x0w_3g*j!9mOlr`Ue{jhn?mP!Qgo`p~pdTF6Yl1hVF;ggsIF=>(u*}fHWU1
zT&gixv^mNlw8rknvDy9MJ-6lZPg+HcTmW(_&sH2x3{DqXe#tVReW{!%n!n={$xVO~YT
zXpz7FdHBc_Zb_EUz*d%=fVW7R()P?~-9|!Jvp$*XI+wFlTTQR1*Smo$^-AG?p%_)Q
z3zEw*-kyJuW|5j0D0ESo{NStLJC4niIZuHkoKrTsYurnvW(C+Ic&6ZM?$4U;PKGlf
z@4U{Vib(=6X}-N#WG`P7}$*wn%!+y2It8<){ES~GMTO&`@mo;#@S>A{IoO?aPc
zhd94Ryo2WQ%VG7ct4D)UURL9kzvSGIcw!A*ZXZItbTU3_lt8woz02v>8C&?14q4>q
z-08Er8l?rmw`UObV~&gp1#KN#!Wc{0xQ+G!lbV4KRt
z<+S)=kxsAbkD`Egq!`5r9<1NcV+A~+^iU$nU7}w~yE7WEdpe{=4;Kb&aV5pWhV4w}
z%gX}kH1VDC)*7Xv5-0?DK#^7ot;P!iUukuD3OfCDz)k)r5;
z(}ufJZ2)>M!P7H@HT23u*9jG0r4ObJzHbA?y3Yk!>v^s0NjSFIPqaOetE>CWg!<8?}0R9@ttTim&yR+>%sEgb#@LMRTVZ)6h
zyBs^F-hGm4%2Xi>7C>)&D=^b2?Q~%7yOB)q}NN<6I
z%l(V6Zp+ThrS*JyenhB|@@c{%+m~v&n#e@n_(#?HyedmT#$3N*k;nBv2L!2rX|`FF^!X8ECqa-F^ZI&92;9}sCT{P7g<+183K><-uod>lp=WZ42_iS!x?7zu(Y
zS+EG#uGtFf@~K*%yM5W50RKf}-asyFhRf$DzW1p&^A;^}
zyx{P`XZs^cyHjL=+>D=^d%6rk->-~oIS#(id!gT2trTS?oo^l|
zLy{GX3eI=X?miPJkYe2##K=7Q`Mp|fFhP<-0`DsDUaqZ3bT*RIL+1;HxK(aLZ@UiGAWeQEo=R+_%AjUtAfK|zM@+;l!?
z|MoA=JViEpPK}UeA>%y@@^BJeo-)dB))*&?%|A5Q9;GV6CaHtJjyyj(cib^1GERhw
zc^zzT_&Bn^I_Yu~L~?p1Vrs81MAwgKkSXNp7lzO9;R>LcMNtDC$+i%Kf!{tq(pobZ
zVKFgi?Ax6-n$+4ybhOp%t3g|>
zPvH)1HEpWCZQD$ulvxxC&W5Cs_|EWmSlKbjXU@naFB?a^AS<7zk5OoQbr(D3RzoGz
z{+3Vs=NsK>MKKDNR_WdZ8{~$lAkp*L-X0H>e%z!s4ez
zPMhhZTP6H0Htp75zZ@$huoP#)3`pJ;Xdw!&|-Tmm;9=`tz^P
z_@9~^kX0eZ=7H|9M`H93M^Q7q(~9o*6fkxQ`R`j5^aMB~=sj%qnEJkHvkJ^j5TsxS
zjzTz6aDJ%|BoDl|3(fo_!VHlKsT>l@QX6=C^!Pq_@RHBq*aB+`gi3ySh6dLt`B;`7
z!bBwIc?H91d&wq0$J>ca1pMSbq&y2c^@1%N?D02GV6Njb6<<}MGX
z&xi?BnKzwYzfTB;xvXQS*pjIXl>W+&Xi~X(YFPGadI8b5<4EqDaLjkDs{iFe
z#u6k1LRy{B|3i$#CZd(~PfJDJxbq2UDKvJZPpA$E`w+<}OM`OCgzHG1ZmvhYw7)NF
zZIC%*KOVWTh<5w3cMYQel`4U|n0#YgM~2bACKxH1mn~?Rq~jM~kf7eaIDw3-Ry#5F
zQ1^ISsuEczp`J+B61pH-U8~a{l{K5HmR0r(tjy4e7N@5w$oVY^`<#E9n}Yl<%N2$k&r1c
z`FR7cb?Q<0U$~3r+m7D0&AXl9^b>lR;T>7_@ta_#kR$MSZge_!J?{i&&Jmx|3W=j?
zIv*)o_v_~)Pe2`n9G(d&UyAv2_X{9k!8;;TTmb?Au>&Vvi;x>3x
zGsT9OinT`MPN|2#?9`)O-!G<28E++c4-i*0=9&5!gR4F6_k}H>L$iRGRFozRFK4nl
zLMNv
zBKpR?Zx?+S`7Rm96U9T5LB9sM)P=r@?yN~nI#4#{m*~REONVveZi2pL&~zUptApz_
zL?9=N47WOCXaPsX&%u<_VTGIM%7wO>>@pa}dmE#JH_xETChRE@Fb^u!pixmdkbA8r
z3|7(c`V<&!+o~WRmY1!KcsjYIc~+!1=b$OzibjFWW8JArgJkz1)h$gIeK>>C70~z>jDe
z7ilH$(r0STH=pE_%z`q->aRqb#85LWltZS-Z(^Y(lHK)AQWrg`1{OJ2RRyHo90H(s
zSd*tZsF{3L3LI9(mZ*nsE
zk-NuAAFcN5`m`1ViZ`)gW%-}GFXnX5^5+O;mpi&$eBfFZZDV70L4%|KS5;LIS(F#-cP<-TZzatuEME&nWhO7Z{93+^&-o1Iqdg7+N8~OO(eFsTutzO6STZd2YQe#6sxE}h?gI>Ek?+c)
zWQg0%BDa4RDgHCR>q)4R+0b>0^ZEqVP^_SCpuqWKNRn-l`(#(GU3UQz^z_&I5Ojw=
zwx5ILl2iUa|K2LBkD$mLBmGo}yOI8pUKrN7FFeT&v2hatv3%Jaro3u-1
zjt2{bQoScF2qOQOE!?)q9p9Y9Z6`qnH5*+E{ZfQzCx8cA*@^*0SagBB)D8~wPa~$C
zVe`2vBKGZX`*(A}=BFLm2^Yp;*_G@ppKjIY58059pkFq8`>|lfEUxQLKahw;CzE&iI0^Z-acElXAaNP`sX|xY;V6z1
zwDB>S`!m2MV|UWXCl28SRr^v2`aT7ct+j{%y({ENLN>)&8}iU3{=R;Lad)NVj$VG3
zq}F*u55U_6%os;EYJ@e)FCecezxFojnIAzx9C)QF%2n=w?1PBtv-@X3jCNv;)0qQK
zwOktsof;$d=Syqhp%g)W1o>xvW^m6h;eUf>ohB2?9R{^@4HkWI?hNbbDfV`hQ9oXHEJOa$#3z#$S?qPOGmSwDoE3p$A)1rNe2ANJs8srZvUY
z-L)C3P%akZpOOhr?t1
zY^!gYt?iL;G^Fm~0D^dZ$!gHfTmj%b>?9zFj3bPghgHz~nx0wSXUgr*H_yutYC5+s
z_Lpid_-)?MI3naDSUpRcYRXjp^`%3V+skVrE|W&Rz@NmL1mTpTlss}j8T~)BSo!+k
zekbeF_`uy0Y^t;U2cGQ@j#ntuiqwDXKmWKR$=>|;uh;$&;x68IT>ot)b$A$UWiCjn
z(WR2m_|or*`*`
z?vPv?;S-w_7e*-vmEAfrb&=3)Es!tX?l}z3xK^gW8PyJ6(r#O6$m(#974p9S0`UHX
z8k+eiJ0pzS9rWr{BWQ}j*9SX3os^Ubi*l8-hw6-~mnMT0hC~*n=&kgQY`v|h_UyXIM=+%_EjXg5^=Q0Au
z_q`&&xBuHfjVND`IMOaAEg3C*BcxVj&cOpw^SiE`k+eAoUGP-g{tC$2g#0}X(}n>`
z(*nHr@&5{AZNcj`>RHQsZO|MQ&%a*D2r1mAY-~}60Ej^ZZ`=v0`&lxX>mFTSUtadx
z>`^SQhzfCC7n2wY7u=
z06N7?Oy+Aer?h$b=no4%yiG9|hid#Qk$ViqNBzXr~$xdW*r
zefT~A>RCC*pZZTRLs8r{YuNt^b_4;vkt+y-Pw)La7L%E+_)gN^RUe9ewY**rY3P_D1v)q$?pV?q&
z&piTgW{Xv+?B`?g?|~T50hN9lg5s~u?p(E!1p^v6gFg0wxiH`q
zFJr$wJ+x~~@L3&s(Z0}6k7yX+L
zs&_Nh3(mrRL4e~qH^KnuRb)Fc$8%>WplxT$dhNq)NE7b!y2C
z5RHrHczxNBhUGkY0O6
zTWJ0J+VeVyO(4ZyXH(fX^ToD<32Mwf-Vb-qAGx!@28=5xfA^CO%1d}1wV5|3^y9F9
zafYY~yij_#b~6fUlxfY>wT!ieCy_F0x6-5SPY+up0t87|%<0*g;*HgOuC%JT5|nLN
zrSIKA(2bYyf+qcu$nPb$TxTPZoUC*3+Ed8c&EfqKI6e<_zajQ^=z1zY>}cm~u_n3N
zWK4VmO4EIbrF477Fm0M~N;bn?HqH8Za!`PP>sUg5Ga!4m>nefSK&~sPw?%Pn;1>lY
z+nVay(19%xL#EK8)|{i`q?v#lzK+z($*l~3P2CB7C3=Pm^>8+0Pi5vz^p
zDU4`LKny_a`5Pe9Tw=E)hJo`CZ@x652`iu{wDn&xG2FNPln$1&4Sa24R(0fF
z_A&LielzZR`P(HeHKESM?AKWhnuFapo}DHba1B|lIVZ_k6D6vh&IglB{yD!zliDLQ
zQQKCdhI6jtmHrg|Bv);U-k^*hIi*>JwzjgRs@?GdGVfdx@w9#(B&qD#;v^2ZW*0IGu$ZY==*9UwnL2rY}ju
zl~!O<5Xv1kn$jGM{h5lZpztiNH1C9AF!-n3_{b#ZL)W3>)G0rF>BN24A+|7fD;311
zMWr|bPC&I)u6s&Ayj6)nhB*2Gc}M9Gp3m)*1!~eh@t7dK22{CsNYkhaE-5Kj`=GtY
zA|MOLkZ?@~$t3~@zV!J9f-a4!@6k
z&!NzT0UqI0SVgQQfA!rtNi`VVkP>m|Zq3>}hbzkkxiB#5O|WyNZ|`GkUEVA}&Mb~=
z((4hB`g~O-GJ_EP2&F;YSa^lv**N@@a0TjJ=SoJHG(Czq!n!oDe3!#^c%iWD!%ARl
zNNN8SIINWpi2im{bNZXXwm-65JAVF@(N_4&M)!wUs-0Me&Geo5;NH=M=2`xYo6*a@
zC{~sRV?)}gFkQV>E1lf-TsZ3lR!)&UU4p`FQ1x3Bh=(W^LKi+I5H}7C##E4whV&So
zEc;+yj$TM)HjyZ_kd>~s;UX3u6Az?OirAMdrE4yhR6WKcMEgco{5Q@wgQ8h40heS<
znZ-SaxO(khvS-pg<1#~ifB0&<1Vt@l+lJfH({S_^BDdI0SL0v^ul)wc;AlV5N08pg
z#*uLL>G0UQPmGlsZM?=6yYKCHO+3lClr`@#ej}vP7fet0iu-dSjT4ySUuoL=B=5L=hp5t?1_^4&|r}27z-d|-U2IiMh0m+
zFXua`DN?#12L{to3@C!Ne3+AA*@%r
z75nL4g3*(5hzI*e1-XD1)79)tO1#JJLs_{Puco`vaZ!SDl!wZLYS-t<-mW83KKCsK
z7J#8}MQPIi6mGq;PybwVYk^KJJI-OhU7By6Sl8&t3n=0oKmf@ku{~OunzG+f@3EsW
z`?D!Cw#z$mpwNw6NF}6~pxB^2PwnA~N3%yIJ(uE-(^nBuH+ZtDULFq)Ly98IRUc66
zGyTdmqvffuN&BJGe#u9AxP07)T#=)4W%CPp*QYVmLuNg+rQC@5idm=f;}ne5qW(8~
zB|CD=Yp-{`vpC_XS$v4~v91_2uIah#`dJ-$;^P3)b1N7t)F&9=}1?TyNJ}MPop7N}_Ek(j-
zGE%EwY!D&8gI)Ba1pWMM;J4b!D9!y^Eu?!(C{2!EG$XsrHbKBSVz1Y#GGY({)!+);
zZ=o%$U&~ppNTwvqLU|G7;h*%qnHxsl_DpK4G2Q4mC1~{9*(}(NzAs~Vg9-&bV&rPE
z8&9tP@X4=pnLlmIf5k;EMHNxzIfK+
zRAg;Hv%~6ZchG-UoYqs*JC14y`lDEUU%Nv~
zZTg>UFyarLu{x7flHu0i(hG~wNZGEXz?2(z$#y^+-}nM%22(+Uc!6p
zf|yY5bvwDCehH>)e5X2EeaxQ+vbkHtUN{k`4YKJse62zvl^`2rlNix%XTAP#TODb1
zuqx6rpT{^UA+Cu>__CqyvV*FiLJlSt8XAI}6;lEk;D_W)-*&%`8+jg}o9hL+10$_B*AZ;@$
zHFtXn9N!GND;ZUNMpQ8&Gr`fb5Mi%SLcVyFjA0rZ^U2VL64fZS`bd(&SbDcJIfIw?
z_e=ofH^*6?7WBooo15_$Eh{(P5520dVJxk}sHJ|UqNfX@qDF&3+PbIg8cEO+FIY)c
zRbs_ZTS2C}Ogp#K?|UnGT|Ho2st#N@mUMTuzr=I~1$RrIy;m3`(#{R4b4^(G*Lw%b2&xm?*OZOpCnnN4SQV%^tncW{`8Nx8!%=lSX*}*A={l#xwNpV1;o?coZZX
z?!wvDYdIL4zTN*&FW`b3+nl3@n9&Hik_{z&0}5Y(zYl=3Yjsflez74tM2(N|scxK3
zbKOt^q|!J|*zlaRFGj-2!uJW)cOWqty}uq8x)_ej`gQ43dH%l)y@>bhNE%mTc@6e!
zQ$RBq#GhJ*%VXCQL~4H9JH6Pod9IkVt6Akb$h02Mwafa4zx{=
zoF;L0U){xGopb-2YS8tkR6KYGageoSgLT3BX3JPI*vCX=oX@y*Y#CvU=JUNBlO4ki
z74GNs?EgZv1WCTxu
z#L%1F9Q7nGKz04&~9FrA{G9R}WdHZ^8HQ4%_(rgmwVN65@p@|mC{r9C>el)&(E
zMnzJlx@UYa5Hcb8hZ;~Q6I_t4QWIKESniezfvk5FM4JM!l0@lxcosGHm{EcUX27c+D5v+#>8}(%Z)#8>eeMQM(zL
zqGU_~_vWY)cz1s`U)jAc38z465L=T8a;!J;g_l*pO!&$Wkrm5I`?`YDsOLT@{=yofYi9KELq}7{i?z
zzR+GN>dQ_1-E!g&;>qOhhRjJ5tan(b2r2PB1*FNL!Kma<61}oi%$kTWH~Cuvw)pzE
zS}TGl!%I?lyVYh1^^OwR-Z0z(VX5eBz}2n{XhEidcHgoyonq~E*HG-CNoHb@Fo9G3
zh_2@*mO6g>)^zB%@$OHnF_$uJ4^ej7aUO~WJ|7th(WLB4M*>i*4<1jI$v4Q#+WzL=
z$u%x@)*HYdj+t$Zk$q)iXn4nOy2X8^-hd;AmiF7b?z&86-u6HbsNz;3S3n2PCT{2g>7{K`8}dBm&4<;pO+_OH%X8uArr%`eX-1gw9R;emcz
z4r>vvcz`}y6U#5V0Eprgi+%muD{ojQ(;7I_U+n0WZ!mRrl=4{d*Ki&_zZ;N#-*PON&-7{`2Q(S4D@@XUL&dmv6i@+77CHo~3S#>CCEww{~R~
z|Ko!Jo0+M0Ml)_$J4#j`uOgfuV;Ct4uDQ?b
z*XN}4D}Yrhv!|3vCk1`qS1L{>dFw;3L;Du9h@
zwo44!Q+a9y&eBA;ORs>{1)tzmsvA$pN-HI?uaP(q|#pRy8y5Ql$_Y
z-i}D!Avf-gMqWmX>$K}dGP{`PKS>Lta#_&`&y)S;wk4PG9Es=sc`UtZpd=c(+`&{q
zV(-Y!LM54QIgkB{0=bZPb{$i4Oh?D5gwaBndY29?f@StzPCH|y#Dh
zHzl8-d1dIZL%O+fNu+x#R6}IVAr@Ub4wWP*QMEK&QofQI|IAQHKK7V|8(W7TXqTg?2WQI0^7fixYg&doj;W^mQN81KJIv
zlqx$Bu&+sfj;ryh`i&k>{o|KT_LvgRaU@#1Z@XGvy9)Syd&OO6=o-d)Emt3-;a!(6
z@1p(RbL!P@*`^$A{*L#C-w&V@J7+nkm^RQ)3h|P=-^F(@B}AkW$Qzbf)mW|<;#X1p
z2dd6QsO$zfssV;3J2?e$EIE|No!hguXoJ&K1`S^q88K@{^`U(|hSnUo;;kU4tma;)
z8#$R-w$%7qsQLgs6@AI7r=QC5c2x@!w)-heR%eB;LC>};%Ydf@_U#I-CJtj>j6t
z8h38X-RU$SoV4st#Z${!PEPnJ!IzTe8Eg
z0^Ib^Q!V5^>X1~md&Lwoam6e_NKHSw%pG&V0|_SPMX$mQPUNc`&kK~4&Q40)%X@ay
z(Kv0!Pj08HVmqZXdE`qCIaS8TX^2=v^Yl4R4Ip~2Z+8D;TqL=5K=S_Xu-d(TJ;No%
z-yNzAS9@cs{X?xJ-RJ`p3-aN#1VT)pVLZ2cP9BLQlT2P-+Y$xcsVZY+MO@}r>ooHk
z{*RTd#!-{Qz(hT_ezhXOyq)puUvTYpS#<%z{{@tr2=@MSAOUj#r&O=gSN_^H)U98t
zoB9Oh)YQZjzf1$Z`!-$@zx8gr->2!t{C>rOlH9Y~UKep^K+vERZvS5OMpp(7QwYh)*;d2D;c}Ui%~o91
zjr4<&&iR~s-j~o#!BTi+3+BrV)ESSHkr99m8`Sq@RYJs!+jzacT0152Bnvf@z{paK
zK^&-%abRHWQO@=Ohpe?n?R=B5SDT}^p)lN7f`|LTxN+}hI!%gs`S>lhs3DA?u4q`@
z7#)NQEFjfVNcqpN+z{>XUpPkoSBUqF8!??(B@`-au`(&8TmJxL`l$urE;K%P^
z{Ioczka#usYcf}+;iHtegSrwURxM39Qu|@KvtHxhAE;!QWshk3NsZPWq_b@FBW`vE
zQquuuIAuHh&oddGw+D)V$0!fTo=}eR4lmn2l1=yt|0qRFQ4o`tx*5@1UKQy#}1NiRJ#&t_k3vGhH=70{Tpt(4FSEe
z+vRL*^`A<+|6e1o*}Ya#H(%CYgU>L%)2jU6c6CE7NFXy6D+(Ho6YaGPiEt36q9of-c3>@`^h9VSK=B9#Ae82hd
z+~mavd3WrI|
ze0~rr$u%lG&!>RkcZJsnQ*li~cYvZbNfVZKeusO=5BL*`-9|E5P*2vH)Y;I4t_fJR
zWq8;Q;HD`!i;>JwMJqLF_oyb6;S*WoSTm7&QU{yRh-+Q7>3e25&_%pw=3_jL#)2mR
z{@HT8u279x-@Am#UvVDuweD|9Cl2e1BKzfAU1#n4smD?~d@Oo39`fJC^ZLW?*q{aj
zZ#hnkWeASj`qfM{U<8xisgZo=VEpa~v&wJvti#%Qy-zia2ESrB)FZ(BgHf70H7L^X
zkLABXr`VtP?aVID)_%nBUvl8t(UK3^0+&}KRO09i-$uQ{1jjjD3f=U4MTJNBr8tSq
zlgGt;Gm9syo|ofI0tB6qE>_-FbSSix2g;Cd(Qxz>f|&B-gK1YkhX{mLm;;No|1s*o
z&!H8R)KSko9
zlu07QBTE=a{*&QZDGEzeNX+aa-&6xSz%tcSXF<$v=Mo9nr7+3D_jW=oB*w##TI$
zI>{2(KS{|0;XV*Y(_th{UQ`P}?C-pL%xg7S@Fw!Xn51(wqDiRqsMBL`Q{tZkL-K{M
z-r86z{cx^W0!Y}&$1;ZDYAxr1lUNIdUmmiwkiJSS)M&qm5ak%L$xjRlu#u+68ZvMw
z462l4MbzzdT{$W{>SZtoekQlY*PjqK6M5YMae7@vJXp;vip=8r{1N0bULEfSQu-fH
z3)6iV!j9S1_@IQoDf%z|dL8G?+wc{WH3AuhHpT*okJ0yhg9%GdN>{x`e_(1l#KmIp40vcxb@NvO#*(!fSpo%Rk?W!wz!s?iH0*y|DAznY1RN
zy{2Ue%g<)lwMXe5c+_}qU6$X~owcM6a=z-kY)FdL733I9AUlgZ_nVaQo+fL%(R;sQ
zmc84nGL6M
zLIc^^617~$E0o;3?#M40b|Z(FZ(<}a
z9OECkpNQC&+B88q-92=7_l%TuHv^2)J;cDk
zQ15u(&-4Fu`>c1Z^JQ4G4)Z&EU)RpF_j+Q+T?@z#R`-hI+o!>AJhc;^hjo7x0)q}A
zi}U7>H!M5(7dj|Mx73c(OEZlcOMh3NZ*4P`NgU>+sD6a$D|XSI=
zSjv}JyxDY;4|*+WJ+}wUCtV^ItD*K1#%?LhMq{h{Bw)s@LH+qxg3n3d6Iqihl+^H9
z=);UpRMi2UR4LxURU5FalEa!ipdYBr)0B+`o+Mmfo9tdQ;cV%7Dx={eseb3Y^w*UU
zAH1?em`Uw5H>g^;%IjmUlvoQb!glNM%qG1}fMa@M$&n$r+O4t0MP4H^PNi33*>l`^
z`fAGx_Yy^}*OoSqrD?6J6mIoC2EsvTZM$kUrt%?^j^?8)vr`U}O{bdBP}@B+?Y@Dy
zb5q+vy6Rui47%x6q7~n9G1+5ql^*SN{fH9vcXVj$L7&*U3$FN6!ayWmSAtdbx12#P
zIlh*2^-XzQ5}qSk5&;0@nw69%o}QezH?iQzCHY}nE8Avu=GRc<
z;mq??QRmbTb<%g*UArh!+)u9*MOyqwuLZKWmLN{x`=^%)_fHYhIwKhm0_r@f>5;$J
z<6ff#`W|53t3lr`@Rkh;c5-?5E$vPajFb;<78%-RbW$rhESBPg565!{usfr-?ku#s
zGAyWrQ4^!gMYmHP;&cG&%?<`HnlaI
z*YTR_O0&eN$4^A_7k#wf-`PZ
z)+zmB<>6bp55EwY+5JK{;V~v6?6Cc8A$o<9mAGXCadn=}KJf6Tr_LY&&$tBgmw~M`
zSDx4=_T=|ru1UVPne8WEqH?Tm&+onvzeDt(9Yx=+Kr9t)bEk*1C+PdbvzwBlIm$j;
zRMbax^&{DPQtIEazal9x#{XOD6FSBhfww0^)k(waHuhZet(W3HkuV-n(Gw7JZR6;W
zSQI-~oYqUpqhpja5l$)r!nqNgWznWoL~uEN1H8ne=q`1WXwOb@s^_WGYWvPn9j#Mq
z#b&Tr9JVH&viiR7G;?vxz;Gh-vRS#nF3WS#?>pSSy-)c_8H$*st+7}BGLui(1TA^X
z*%q&J<+Abd`EkuF+NZh~43IU$Q1VN59rMQ|?aYPdN3}vN4>`+S?rfa5m
zBB~X#`u+az6yt5mHfw3qF=7UI`Y=n8L!3~(c$`je*r($)8D6ETEiLAEp|}CEiY!{;
zx%=`w(>SrC4tDPXJe-Pnk9sBGx*&s$NB2d+WZmXp)0d5W?t;S^=?@~>p1LnbOQ>gYzVlAs?+LiYld=F5G49`Qo2D7SRR%`43s?k-PWZcqN5F^9fMy;
zs$vupE&Gq%@eT{xo7AYs<00_RD#JNRKZ;7vGB|l)Fheq;Cod8$o7)C*l~O-dkJ|`t
zTC*L?vvgonE4u17e88UXrEVizWO~zAPxb3F5hXp6{=gLg&3VeP{_BmiFxIZ=k+)WyjkksYPrc&Dz+xs5Lz<*cIGar0GhvaNGfj+g70AiuE)nbc=BmI74Y_4mx$IJ
zy1zBK-Pnt&D8FyT7=7{R1ohsMrYz>pOLtlDTe+?2rOOx84{#XnAibAusYiKx@dwfF
z-vvi`b(_TwE*b05>dcgL`aN4s8?vM4t0(xN0NLQ>y6m}-he*2HGrs^ogtEGEN)|)F
z2YS~f4`^fjk7K9?@PSwTVf=h)&4Byae))8Xx??G8HEP+==|{fQaySSADfg%WS`&`!
zOLQ3b#x)q*%E!6WK2?MJr60K??fJ@7pS#r?1_#yEG^HdewiB&sEtfj1p5H(Cz7=N#
zUQ|7JO^wD2wpx_Q!~xe)Q50Dr4YaJHB>Eh&Gt4U@
z>s4EdS@3CNrlY+{b)7UaYg-TyUgont`0qXnz59={bVeH>kg_WI}~Y
zZ+QJH4!6Al|NVW&9>2URpmJNdx6s(6rdY^kg%xO7ap}DPC{yt3WqqPs;v%eK#S!Hj
zIiV=7Hye^8SZPxhYtUbc%}6$)mKNE{*GIWvd9&@c(nrTN>dJt$(U;o`zVC^{SuZVCFxMlGwjxvt~ysooQ0dX=rY<^;K=|N#a?S?!;caSeM}ros{Kzc4=sMmH63-{aKVzAoFAU
ztjB|Wapy5{lD;u&Lb0`1bFH^TVZU_@hC}>s&E%b?MEp*
zA|yCoI9>KFjp4s5x7EfG+&By8*?6}ze{K3aWE$4i*6b@k!Aaq={@F$yDiJ~M`iOw$
zvuDipt=Os1aPoz?>h+=7LJLPsiFWbZuF96L0I%P=MDD@|N&o;&{q95n*zfFR%gw^?
zof;;;Yt=Y_cU-L7!F$LC6PBHB*gPyLUJc!rvUnS!CX`}GCwV3#v5qxU^H9zQs0^p8
zHJ)!Nz~+%27XAw@5n>LkijiW30XJ@$T2ba|AdTShdg
zEVFDxRn9oGJ79F_1ELmNdh8+!7Kj0|!6rVv&j^{DpwNk+a~xfXXz~(N5_3wt4E@xS
zlqtx=%IFlbAAezY98czGkzl;V(q@AAm%HMB512=Mki(x|DLk+Bdp{qQ+Vrl9@AWW$
zc#~Ig|Ic)XS?X-Ri*+cle9Lrj!#t>&%%5?r!}tu-HJ$Mx8SmG{mmJtVP*
zOwxOjo2c>WsZjDhmwqIhJE8l|)tWM>#fjy~Xs!#_Rf7(9%MShEk3e=q-Ju%~NfYQ~
z19d?78i9%Agi*Ok=BbY#d9sINhs1QKRfpl}x?=+en$NU2j6%0d#oMw-*~R_{0e#=z
z;)>bie7fTW^y}!9v|0nYkDG%y^M4r(fBWBA0HYGACE=T+sy-<;(-n$@QBe4L`x&j!
z8z>h#4sv_MqO>-DI6t4*gyWP=_G_J3cV3Klb+87cixx{#uofZ|dAHs~;yh>G3)9iJ
zh#Aycfhe1P?7OQcx59D^FXr`E_3AJanm$jZov(!tttWo$q>>(@V}U&_1xjHpD%>pd;N?1C24v#
z4+g3F48XIR&8pu+S_RZ(NDL^lb<_?*>*$`?;fx`g3mq-5X~33|@Zo`fEOG7Gl~`O9
zGmob!8Y;i_Gje*$fo8u+6n=9qCJi?pFB1f^^6-6&J#ReZbru$&QS*8#n20z;=}?IFGg2PWj-P8ucPaSeHDw$_9bE$PZx
zu%r8*@f(!(AX)}?5LZ`*g`~t>F>fn35zGA!VPSe5gdC4h9T?Y}_STaSB-rwtdfjqZ
ze^IS3uPwU`y@=sd3X@)G$=@`tmiwKvP=YWqDRyOw?-VgqT2fc9!f5#&i?jNFkT?zi
zyOwI%Il1+!HLS5I$C%#ihCypJQX?jbinL^~Nd}@AxJ4A7NO^W7jpM4?Z-2=6^l8`7
z?UbJODfc2AUfLuDu&9b^AKSK>^*(8jG~^FaxVBqoIa6c4$?SHxQZ=$~1SkH8j57SZ
z#GulRXk#}crq!;WTk5%-<uBXYgNKI5&>lkGvBoBXU=8;cbb#%mb1cg9W;K{~E#Np;jKJuvWb~91RcfW-I*}pl-
zDmYtX+LekY@cn5O(zE9=sVE(%M7IK`lsMyI3A35^R9>rPMnM?duopym^vi^wU*To#
zY!AFGzp*yYZF)1bHM!N|Aj?s5?*plfBi$w`u3+6RKxWajzn9fBbk$5@lnZiblnhlf
zXOk+U-Nag=uNjcXo*ppH7UwP-T^Z7Xnj@HLr~&{SRrimb6~P^v=j2yUD{bPYb%=5K
z>nNDhx3zqKh&y|V1D-YnbDAgF(Gxvd50pAhxTF}QUc59~NPOz$im+F>huxuwbq)6R
za-TscKK1|$`wFrR{^K*)=?7f_h{quu-Uru+s8L=d0m+gVEQrRUv5t##PI=f!x18sh
zPQop@I=nq7SK0gn{GX*=)g&3-s;V0OU=rMhlq@4(Ca)w9ASNO|P{x3S29<
zf(rM%j3(l|PWSAG86_n%-4Kce$;mnUeW^bz%+9BEVtp_pn)6#!^%m0k;T7k#C-E^F
z_39`$-dfo!K?gSK-R_AJT+s;uKn^pZ-f;W0@8iQ)pc(8)rtHvCiAo70$xO>7g(qA1
zZV?Q;JK`Cxxl4nS4o4Bpl60S%H|1#ND}VCdo=Q@!^;TSYIM*rlR!M|C;uJa3&YFGu
zm8XclnxNa=q;zL|pr|6lHD5<}=2Zub=}oIb?yHy?e{&8(cZh5*COpdUenId6hYb@R
z_3%q>jtKNTZrMQ{iaB4|YbQcAWnwTcE|NgX;<8YM{VJnea?=Jzn`ya4`LKRRl`?kD
z1C&~)!@IsN2#O$T$$I_AjScjo9`$YQq0YUUNkDP68~%i&uR8TS8<-FiRjazmii@Qj
zYA(M~Co1&|Iwu=1EDh^%#a+FjFAUl#6&DpRLOAHs_C1ad>~^0@`w_~+az>qk=8hJR
z{7hfU(!oTK>HFK$TCIoc4qf#3Jzo|*1uL%MrU6QvstDwBE==8A05@Ndeln3032Wy&
z*M;$F2E%^8Pc?G>=|9-Vpi2r1qKF=N7dfSm(PVPXM;r1)+1b)
zMK$GqlB>$*zuMjAwD`i78m>ef?nJa+zVp_q^;S6sBNy~t_76*~X9Op;D7cRstze-R
zC*nWD+w%KdQ;#5h9gjpgQU!C%cR6@p?qxK(@0G9xeI!)I`bho(v&4Z8^m^TJ_vp<6
z>#qcQxw^^N*W5YKINI)p0Q;4Jff$UzXXH{_F%)VE&Rx}`ExhNzCO
zq_A)v;$^{|5rS=^*I@OYld>w{cH_U+ninLFg!$iKOtX5O!(jsfe=YH3Is3vBq)i!M
zcw%$+l3A@W!iuS{Vt=S3b>>o6xIb#V!fH8LmaB{*SZHcFyS;wHwhAVb$qyz80=
zcWX49y8$WCET7%V;vXk65gd@KBA*}j?!TjAx_*7>iUxX9xPI~*BAtyg7vyDm8I;sK
z$E!fD+uJ)|AHla5+h?>#3f!!&9pG0F3Deyj=(GM|l&x#g_oy<>vjN?kshTU9#Gb3{
z$$YVR0iCR+!qJkk2n#jkH9>=PKwz>JrY1{8WPEyJ4}?0ZG?}6To4tRscM_`O
z{Ul2V1vbAa(D82h$&JZkJ;DIee;y%K>0dlMvDY{jgk6il7Uo0j8A0M1mMEADZmnEG
zx!+P#wVIvz+D}rWn-0QD{ZkO2}{By+V&fAFdMpUZO$c<0-;oM@&205&XH0fVa(#Nejv!
zcYj%cuMRt-zYYhHfOAPmMThrzHa`mo5jnP|sFvjGEyKx*SGc+5Z&Q4tEH{}o
zU!h#vn<`tsGC!6|Mh>QGtms<{iYYw@Z1F;wD){mJH
zR!$)tDaH=c7SpATh}W>kmJ?G@O^+J&pkssAsh}?S&&;yLUmn5Lz)kR8TDI!k!cq4D
zb8IiM&LyZShUWKqw2tAY>e_xsOyofOSwDGD;7xl7hdOg&Geaag)VO*zC0jKFZ09HK
z0aLSTA^ZjplBM_kZZgE~q!D4ICjbfH0Ye-40+-FYP!ZIn<9XQx!gJP-2`+uY9>)JF
zX=}c?=Q3@n&w6A*
zQkWb_nzw^?F(Zg``A&l8U3P~BKZa?QS`Z8j59bEFd|3zdreAU#1??=pJnQ20*EpSs
zy&fqn%*@+lhNxhM^9oN-swa&#Vnn%yRu2vvb-8R
zGtc3SI@s0HCtsB{T1I;6Uj*+_!}qgV#Il4vJ&1|merA@;#S+S2EtNC*LvX2OhvqGd
zb39Kobx*xEho}cKZ^|=DCgZMm$0fy#o%ZX><5M*2{AzSAhKo~NHr*9JB-}lO^m--S
zS?A^@vX*rUdib3N->bg&Uhf+$9CnD@?FD6rB{DOTHekPJ~k*^c~ExUfOpP9efMd){Bp?u<#IJ
zs~uug*`z&$N7~2&ZRASxh>p&!!AvcI1XM(H^@O8te%{IELEcGT)yO5Z*?=%~i1ph4
zZu>yrvh!WaB3p045}6a>svSOX{#?hqXfaFkmtoFwTOUO3GKk>h?)Yig&M>t`&Ntf%
zD*n<|rJ?YXT>89AW;2z;(b~>h`L*&f#aJd7bR)aoDQ6VScAqehGedMDZx)_23-^t>
zPuP~8D)172{IVZc)2dsQVS7WPv2-stH&@Z1x33+TZx(#UNSnKsIelBw@l~Udlz(IJ
zjXzGgxZ9NwZb^pP?AFe~-ZI)tR2h>cAE*JP)D1E?pxbDhN5B92Ri<`YsYWy=hpjb8
zn0+sF+jqj!6HYW-G&?!(iA3B?A~_{(|0F*WR5nSYSo~BAp1HmF#HEl7{)SJ}dqntt
zU)S~JEm?y9D%Y)ej<(zOWU3=9T!fo}B_0n)N$E$KpBcuHb~+xJlxKVVlJRM_?8cLX
zXPU(T%Rio2mrg>WdpNG|V*b}4u4UJWeHf;ACGbj`D7BhkUv;p6ANO5ps8geRY4Hwu
zf85F^>`=08T)svXDn}eFV4qlNj-Bh8Y9<%dns<(}Gxl;Jt{X~JOhWlHp}sv#7&Un8
zF#I4bimS+8NnF~fTsX&~54M3AX^rt{A%nL@^8ADW5H|4JOFRUnuWo~)CHkl!^FbYk
zrtpu2g@whsvbTy7c%$p}*qWMgcquQp>Req9v%v2!DQMpB+|@F$eF*+!z9$u6@OCX
zlzs;~_BIwkp@E8)5Kf@!_M}gdl5~wk8ZeZ0Mb*M$s+275hHN%+=VxGPNro|s->6LD
z?X;P6vGv>EC;b}j?k9HHVY{ReTy&g~L>Izagr?sB*V$lJh6nM_@9>^^FH4M?&e5H~
zmVnPEN+n?i+E&UwMhs|i-W{#G$40)E3z=Y5qs7IqV%&YJo&`_YGaPsuvcv5L2Y%c?
zN`%Y2eY-*ty0_xKcV~FIpzC%lC}gNe$KmqoliOYPim)4F(g&**aBY_nyE8
zvfzkL9G38{6JWEBO_oe}$=!Ky-}Af`P^E{Q62H=(yJd1fTVM1~(f9C45i=Wdyf
zeP&W6&Y7EgYOZ?5A2kNLK$x6m8dUd%%Lw1BReH}Is@Nezm{MMHO_YeVj{J^6PiYE&
zs5KShSi$XfggE3Y3>KC1n-SYriGyEhaoSjMDsr1QG|yY=+MdSNG&op#dQL|2+l{?}
zv7eP+WOBKEB~$=lBB)yTfoL-$-xKQ7?WHjKPL8`O3%->!OBowbn^TH~xVEGgmz^!X
zjP^MF2K!uSV&$heQT5APdSYrfgX+!)GlzqSnmQYDAP}u&vNR-V=ipc@mn
zt%|KL%gJ8GgV9$$rqj?k3?hHFY?%qm4Inr8;TM`I8D9USCEM_h7_CQflWbyLMe`n%GA%(7CxZmA_-{TI$tjJsFoon#@?4qDQ8aL
zFh)ERr}&WL+BR`MTh7uHk)9tKK-n|^X8Daqw8m+;wY}M6F?hAyjkkA7EL&!bfl3a;
zyzJuyPgg!kO;3SHJhsCJy3fz@cPsW*AKw&kHI5TQ_cUsEBv&I1xsn_|x6f8ub@M8%
zdOkbgYwz-1mt{Qs7N5NYrusjgs(M((*{FW
z&PRWQSPGF?^`0$YK~Lu=PUi~A^#j%gwfvTs8`wwVz9@Lb&{-W(k^a~-jRy<^)Zz(o
z24cUEUy3_qTAKGy!n`S`5h)s#HC`TzC`A_?RO$?#3iz<|WQbUAxsczC@_R(l8$y6y
zu&>FgUT@=!x;$9;Qj~H@W~b#SGD1M+lU%zOY;zF*(ypO(2-42a(NxG$?o-7uUGhv0
z!72Rfn)Yjy^XrmPE^F>#tKO45&a>r+PS3Fe2%vsSsd-e`>U<|yG`0*UDVaxp$-@uz
z|3C_^W2?=tpH+KS4RPNZ^Piy};?rv8&%L}sz7@QGVV9niZ$Di6e8k6H+X`>%b5CK~
zw4JxEw(xpt0iz@iBdzXZ*N|(}Ov$r%jV=}MXlw5%TTt6b8N#N%XJe|uW~++g6n*|r
zpe^o-ZL
zU0W{2Vp`1Ijx$;ehC1Gj#(QOCERFpm;F(4Fc26y{mKxXRUcsP5!PT3cLe<`krI}}&
zGu50T?%O-Ax)@s2HD?;7bbK@$sKXW_DT6oiDq#4L`$xdr3@tH#+$_Sw8r60MpLi0x
zDZy5@jKL
zHtiV$(?lmhz||3Nv&&Ll`Kx{UGq|_s$y-v+WBKd3(bfd0c{#+C>KU!uN~7W$@ndGD
zToh)~nf^Vnb?<#I?!z$u!6Et^6pz!xA~X5XrPSnRK)yzUMk8FcaoLSU{B-Y;tbI~a
zl6<+SOTMn;&4;z7(hPE5Xdw*rPzPo$t@Jyqaf3_Cds(QoX{aE{vKF5)tPf=aXxScd
z9@B;7JC3lnC(7wM5NL9D_0=bpG^za+m;-}b7$fGy}rikjq^#5eKH1OLc-0-d~?|wdY
z2%#hQRGMM<4#wE3VT`T8bkB5a432q%I}0ANG_OE7*?D~6t
zR}C1zVp=b_(Ffj$T(uilasA3VESYMf&KG-f8>Rfgo+KtMJPK`1#YcdT{xhO;JxFQ&qh8vEG0GX7U*3gtg+5M@+9x
z%U)XZ6Jqa)fMtm-5G%oN$qZdJwD#*>LM093-(d;CA_w
zpKosD*6F$fY_Gat=u=58;S^3xq)CRZs_p19bs<8=eg$OmGEq!ZEb`F&2g
zvx^FT2gyM4eUCF`J&!4jGL$t%Pxp+DC~MPjFm9|Or{#;xGRR6sYd6euCXyY9wjCl-
zR%%zLOZ^J}xLs#1`-Htzr^JCV@42Q{ZjOkHF4o+&SmshK-y}-`$QRVf>3&v?8_LLa
zLnx<}IP_yc7iRH4+-i`SnFv{##CsuaN0yFoGs$Yd8kk4u;(LLgaOAq%u4}*?|FBuo
zwAN(}yT$wge#=1^@Ri?@*_lLZTj=A)=9tr+8n}X3r&ufxtK)K4&~rJF3*+ZUCpVZ{
zozN4T-&?EA%I0}BdH;*=ls{zOVQ`3|GQ{whwPZk`gu?`FhnsDZVsBl3w$c!1cK;LKy4hoCxot#Ya~p`sU|
zCf!4(d4OR9wRi%Yfz&@{+hk?7^-aa;jQQwA;NVH&Ip}6?;B^Mg-y|M0kX%UQ4)CTC=N|Kd-4Ly1r+xCDDeZ!FEkf6A*Zlf7!cb<&}X{ZU4g
zGBaa&pn1Jp`lgI$!Jdr^-{!~tsQV4UVSBe76?UK8+L~O7NJ-zVEK9|C(4<^?eNvW4
z8f0+Z{Lt;NpH2=n$NSYJjJs9exSZeA)Tw*39M3N(m{e0!+t%M%S@|5gdo1k2ny@2a
zy{%#Rza_4A=Kn2mkvZeL=|0j9qSVE7D9pv^TQD`PMv7W5u+UD%LkTC(0sI4HRA|c9
zos;F37ViG=r{46ncoJajP!@CFwH^W?>Qk
zV)h{Hr%QRZ@9uX(PfX2z*4<-3#;B!_zm)4?U5+B>Or^f+et^dD*<+k=%fM6>i6#g75d7mEUMXS
zw+1REdtL2yvRA>XMxf;TCq+eM@iL#Rzl)~~@ZL8M)mN1D=?{Z<{d~vsBy*_2H^m&s
z_IYbffBUb=SvR<%n}je^2e!kN1t9^I+ubEM7%*swK}0XSH!wFP$eap?+|ae>i7kg(
zri3bL@s^HS$Rk=J`MBcLqsrmgvuEmdc6KQyMe5nu!zqH9Hd94XM^D#3d@inG9m3Sg
zK3sj@75$j#txQgJ>dA9?y<{@t#h&x|pt1GXj-kCTz?WoHDCUjO^c=03E=u=(1FaxU
zZq%)6AwOkeO!x{V69!9V6|u&@&$_AG2b67lMPzRQUll9Qntqc#`0^$2i|(cP%M488
zr8k85318&Pk*)$w=Y_?4_P0WxPKle2;X?9(a7r=E#kKsN$j%R$HI>!;lvqC@K(1kfCt{o_S
zVq$W;3kA=ub`O8)X>rHReCIdAhgyqao^mbBMv`sJTDK34?|`p)9OA;f)3pu?k_e^
zvH-EX>c!Hi>KTlvQS|_jmi^izZ2vb{*8t-)?=f5BoqE1Ys=DkScG7=7Os7FQGWR@|
zr!?`8Zd+G#h-q>duaQyKu_{H<}DYGas8OhPB*aCpo
zEb>6F?mnp<9Oi*HdM!ztC)MSUhY~U_8h6^8cr%Vp%l+Z6pU0l9<;7*`zKr
zg%qChDx>Y(#Ls?Cbk$k;NpR>_fB!T#nuOhsj#oqWsxgE1k(q=Bp|7G~m
z+eXLLpGX8cB3`pm^zV=vB*#~bP$u01Vb4V&R}Nn&=gMw~F`8VB#?J`f0S)=+PDT%I&(H}sIO{Ed9g
zz<5x+%78HKTQfv^)Pc9Lm{x^>ZsyeCzV&3mcq1>Q;6id#K+K`md}|y3V%&XERW+5b
z0MoYVQC(u-H)w@+?th}t_dK^A@)^?Nb
z3!4{5ZW{n?H&5ZQjfaXNiL(5wW~?tcq6GGobMGXf3RyBiV9dU*Px?de=P@0_UQg#z
zAyofg=WET1orQ2mMogRP??xck~bz@Oo<5mJbZx
z+nrp1%|!M$0tO}`3%P=QDX(QrTE~#gK6^mhUL-trCdA#ek>J=#`l;>>{Q=WVxqhJ!
zXJJ9XdjkV1(#46LwKktmMv%cIOr1aRtqin0BWV>4gt^G2|04#H{}BVv&dH2}Eo#hf
zw~?2Jr%12fW@ob6Ox|X)AR?i)U4+b&9b@J78tql%;9e2;0n4L(WvWR@YynrKnIIVk
z2gtbI!Okg#=IltuR?>WJS#Mp0m#V&o(!dMgEZVT$i(e!@J?|`FnvU!4b9Xq34ydM*
zSM^%?mjxRgG4K`KHy7t;j4FY)v(}f#9P~cl^rL9Ri)6z{rfMzKzK1%}c=R^xK^B4m
zVGYu%?w$IuGx>IONjsgq)I$(4^P;2%tiCX2tT
zMMb_wJ&oEl+Kg!ec%p`9shAwS#K;JaS>EMo1luQ#6=wP>1>h384dAk1nk6vW@E;eQ
z`U9cU7qqMiGUDMc#-M(FUdVCzV(l&q##|qN_CcL1=}9Q~xAyZ)Z^!pBUm
z;d{#<`%PXwZKm96vtr2xsuOkaA;XaPUQXZ8c@xFyRzaG5PA`o&`;*cT5r3=)U;gv@
zB8~l7cl8OuJ0XU{jWmaeXOE9B*EeY=^OKNHiO)9Qym@-=-J?X|dHJzr6YRZRTnNT~
z{vYS~_iztHuwVD#W(-7>9xgUJTQ4>_P8CawdaZTf7G3rjMhsF3-^2HxzOBe_Tuyt7
z*&?2MdxL_ytS9K)S6Nbj`q$eYe#62)29Ld6FvaO6L|bNpx|w$)A0gC0jmorbW>&
ze_+PoHVJ47R(-EU<9<@rt_^iSfYB#tpuDxXu>GL?ARIu`(2Jb4fz2QusjR1(dbNC3
z+19pay&SsTsy2H@NvwQHNR#*cwk)TpkF_NU*ZyZDwQ@{MjNbSmrn3R-zb`3yjJ&fn
zn-QK^Tr60I|aRK}TCxBH!bUI#B%sKhcAxE-quAi}fVri%OY~%})77@uYz`|E25y>Ikb+
zme081p8uQAPm;yG1^#sK02`M!1fSCgJ29@Wt&NrG)?O71v9kU}VKG|#=Ea?%Xs%EG
z<#aDTA1n&nOqVA!h0c9V%FEJQxT~z
zxKNjUx;v|hS(u@sC6N1wSjCk;mHPN^wg$_8f*vct$2#UGNGan-GPhZg+dgK!#zfwG
zsJrgsVIV0$-QE^*v5{8OBz}PR-~9ClNx!~g9xNx;?>axC2XfGzj$d4GNqq*Tzmpyb
z_Th`D#3#`>G`{C1UPnEoFO$Ef%gM?XPh$~^Bh}yRJWIg-t3vs)Ucp
z_uj!Fpkl*TH{{JFLexAk?ysDKg`KaA#k;1QBzUkPjlI3SJ=NyvtcQVSDifkhKSg3U
zFOP9&zLo_Dhh&s!m)oF^bH4obW>4^OjH3zfbKojcGbLNUdl%`XL(Hh$1liHCfA1aJE!I5HiU*wsYE&CHf|M~pTq~t>}Q9E)%IX)vpbuf`57o+uYO;-NTm_FZLqe`{z
z`du9bJ8hl}uvN0Uy?psnme`c=uL6AYV-x3B&Cksc$Gf?BL&@9$2W=ubaHmy&h)#Q-X_796!zDjjM2+C80X~umvDVDW#h}UeK%fqBZ%1P)_B)J
zv$5D*ySMxLOfO)Gn3#B<`k#*1zZoK0A`Ev2e~zg?!r|fJ*_tZW?BR(F3CY6jb(qMF
zlF-{TsG^CcYorAspW;#K#~3%Q9N_+6hxPh%Sn-*rxH~ukz=b^Q=AGKu9~I6UL)kC~
zsG%PjOvMN#t5-3gmrwdIdIG(@No#O`R~W?c
z+0KLyBsGTDr%U>QF)As4%+dSs;lr}a_k8>5cD9>KONlO#6-mwwIcm1v@BAEKF6e8_
zzlR_qZd@%9fa4=M3a{w)AwC(!q
zhySbB_V3>liQitGl%!J7mXA=d%r}Y`=8r>8t?SJe8gKM4w+Del1wfO!b2CKeSH363dI-e
z96|lq+}*V??;$Ap||9^#jcQo8>*KQ(-mLMWR5JZ9?y67!>)aae)y)zgs
zh~5c;Xwkxi(c2gzqC_2Kl+mJhf*HM>pXYtw?_2M49_P#-vzA%2X8qh{-`BqOzOGw{
zy>og31cHCAB0Y!O53KE2Whe(Y8iRR}5UMh4EOaMk@?@jNVXLW@eU|QrMd#ql5<~jX
z2P%t$p?)?zzxOWCi&X=_oo|`;QM$~h1qIqBx?*e+$Ttglnr|1LpVqEa^PuWPOF@dK
zqsDf?Pxog!oMTY6pv%-l&Zr9Ap&6a03?8<10UkV7MT@lvjwxo!kJ$7y6JCf7C#5>*2?FL1j
z&=q??v_7V%uUEldLq#R^UoH%d*i~dK{9}%K1^E#q;%~LZY_Pb38$IKq)-p|E)V_mSUphY|9xp`r8cLfpn1$gLpSr`I3@@wde9dnm--
zdIIWVOk@3Bf^Yum+uK@#9bPV56$`7nw(0lZsG9Vj5BnsJ5;;!-%}y+_h_h#d)vg#l
zz(KJlqhmN20V(Kq_?fWGMeo9FB@H+#iYJ!
z`d7pBPgRB*d!tG6%@+}QZlK}zowB@}G>-XoWHdQJLHmU4hE3d=G67YYf=vyfj-Q)j
zE*AarEDYk-A11kd7Cl}xYpV>sB2q$M-8a8|-7`2}ok@$#eSLie*dDe+)poYS(y#Iq
zN~?h)c{S#buB|R>k2nmV!VGdXe8U|An5SF1qJP>uUPPm=Kuh)*P<`T|r0
zohs>8UvXg8HS@l=U$^vcssP|fschbRe_6Hd$T^#Q!EOKUYQY_!HJ}Y??S#E{D2_62
zu+-RrkyDsPC`kdzic-N*UPjDR+cyST^{~q+<`ECqs=@ePat{*i2@w0~Ik~W@^yesr
zajjt>6oc=Vciq)HZM&^5>NaB4MjOsdCsg#43@dV^qMxF7i9P4Df0S5r+gxnEmjhG-y3JUhB{t?a9%+ly$ROq
z7@oD~zZ_^XPp%Oyh5KXzF`CNHewZ=4tFS!Yb)*6mjnPDgLIs%I_zn)@1JMi=
zMr~-1(9GYA#+5g}JZz4C8Z$wAIZR|^MF4+r=Zg{Ck-Y0aIo99t7i|un2JR~Aj`-5(
z1uU{`!_fLOZ;5;-a3Or!M%++v>S%7*>4@e8uW8$%kOHA#Q@iBngTbDcHa#xPoHMY*
zxD=W%ss6K8e7Z>`3DxMp1L^)PlVIPaE_{vVrm0psg7NQci6GuFSMkG^BdC)$>?;dR
zaaeuXZixFg63Jn*(gQ)e)#MlVG3y>};@H=&4<#TpCPDZxU8*I-lRy8cb@#d+=t!pW
zl-T#dJQuo^3HGKbAN5N?^#!KN0;kd^LSp>;%n$s&Zkg23P^wPx`SG5{4^-j(rf_ZL
zXD_q1rw{Y%TUjC_IAUJgDyV_iK0keB+P3;<(2TyJy6Y8!MU{5FCOom?)-}881{N_%yqowd-xZa(|Yz1EKFl^
zfrZ+sk+N9VMBcz0hTl_Ilq1%?s>kDB5d-0wh8LjJ7R>b-#j
zF{q_M&mmBCv{4+R(0gFkqD!VqJadW;Gm5}cg?8}rE@EFZME9-#+7U9-nGxGIuhMtF
zd2qT0G|k`Fv`FPLvETIHu7hV;E^zs+G*+PHvIT+YCr#?)e=EIHl^$fV|Kkjf9!17j
z8|xAFp`4?)H?`E4Do(u~!K>fVOi#cWFt!>IK%aMx_vG+ua>%K*c`5~uAI1cb{}ePU
z-8I}Dhc}Sn#2_Mwy~iO~Xxq3*pHANblEdv-jq8?Jt9VYmG!MD$maOoeO~j2Gw3ctA
zUTb)u+c&+6{X^ni(Q_Vst8>|)!SS>=4@sycB=y_fa>mT04Ho&i)llHliqu-AQu4$&
z*n?k=wuQ4pK{EcCa_3^>`D_~l{;M<_c6baK37N8YlZMtJM>R(p0$Jy6Bz90dOVEcW
z*3!(<8ZT4;-E#gx&2&o<%wl*5(2hLYLv-g|h8kgbgeMP@m4#XEr~%5Fe`G&ya|ODt
z=wdYlT^Q9J$#Yiw8nBc+wAmbwxGiA_zkGP=8JGVRasOb<$fyC_uzXny)Hh(vxBPS~
zllB~h65qUH33Yip&XFoD&>nhXDD2Zr3Y*;x8SpE;p}RTRJS{}I);nRp^){k8_e!%B
zL2--iy4)R?0;@yMI~ZGkO|$yF-mcY6GXy
zs{Anu%u&^BDs^XFr*qzICKW|&7_Qtzx{4GD0z`I)wjtNgmwb-b$KH4JCPUyvKIK_u)3dYr=k*Je;Yd1yxpe#3S2
z_-%?S%BDevp!{6$Tker=w%@7r$K(u^+@&6f?iwLU>hGToO;{-Z#g|twb8s_H^zBdF
zO!9=VXITiyGp_!)Q#Ho^o#?7-SDM6Mhn^jX}*O
zz1(T?3^K%_)Fz&vyfDq`)yv;)(rZsRo(%MAvVEdr0J;8n0T*nzqkP>MzQ55*>$E51
zeBt{msF4jgo}I?13j!u|blKTHXRT_-0X-hyY++l@41Sed$rm$8uQk6&7VDo!qBSnT
z4~g}%nswbiJlYWix>q&wp<~p!i1C~k2I8>-+&mG&!HzhWSL@OiZc28fX<*IRC1eU!{=mO`WJHn8H5&U
zClkUaAPpAMYOTtIv=Nm7r9xIfX+(@u>7i}TP_z0|KhTKPVNgHza?B-=GteUNzFR2Wu;qWo64!BFo)oi1I)2|cm2LtdYAHjP+1;Plbhht4*85zq
zGxWq~93h%-Asmr#M#U`v
zz@KKckQrWUEk^mvjY;%G7}>&M*8Bv=gEV1c%k)D_o_?;X+w!W`2+z|Ddbub>?$)JbQn{L6c(!w6uO<9&@V)Byu+Nvu{Vt9W9
zy(eB%pqR5n8*&>2XIi|e47n_B^)2|wYP{Gf?nkCp9_tr6d!oTg3~JlOnx0K`R?TOhc@VKq
zX1*S)g~%s`^07t8E*#lZDU
zfQp7~<@MbF607JSP3T3TjJ5JmWIK?irYDaGr1#AniX-<363{)DiS0RWp1!em@E>KJP^}mSM^q^mulkB4?
zGwM>Q@&$$SW^~hNRt;qn^({~(8sM$CTcdvGG63r4iSq+)#XU+hXmmtrY0lc%N7wQF
z&QWn1bpW1L6>5nj>ArLAMv2A!PP-Ex{U$c`=a`o_h-QC2$Zs=G=tFR=)|`0h$|}`>
z9$CN|mye;;y_HuJA$>rqf@nEaaVsCSCjC?M9C_g~>_YLqd10es@EpKA5sY(W8G%{%
zNn}JOtYG!s36FR8(wbg|_|Le04y}qiP0~Fk?6KD`?DsHR1MHV#(0b|o851Y2zU?#j
z1pC-P;EI9M5TT*!{PokJj^qib|KQ^z=?CMRdg7i
zxZ}O}cM^cJT4ZL6fbRB;&-&1SWcPooz>thw)lK-DVDBpaZm|HPghsGItoX=@IEv8$!nb#--9%XA@4sVqf
zLvXOY^L%3@q3wp)-X+@biyGcSzm0*0Du<6o(+y-p{|Kaj4{x&vy4sV!T;CM((*0Dq
zoNSh7(|%rB?3%w~#`x%vQ(z<)^isN9)rj4-3`>91t_ns@?6UE|D3=OH8nY?NYNT6=
z)e*a`mkD|@z`)rbCFUlp_MLF!@N-w;y#uD%=$JpXhi+Pyvfe4nF1rD<2c-n=yh5Fj
zNeQUY9QT=N9@^HcHU3)eLY(+d?+L|QO`@~zk+?Y8<8hM71x>Rl&q-7EmM+A7hzi(e
zrN(S<B~^!EWqN-Lxl$B=((|oIysxrLS_Dufrr4QaP*DHUZ26-M08xz`X&}paVD)
zeyD_(^bIfupN0+lT9y%Z1uWQ7ehbv^kLyMf-p$oZepzNGTzVXdBLoIVOrx#%SJxYZ
zi;x0L3Iv4wkXa`+@nE+&s!Nn*kr{KEcIR0gZHDecfj=cQt%TWnAGt
zKR9JRPr}Bz#|zwKYzP?3l}Sm^%l$@qPY%5hVTlt&#^kPG^W;#FwplODsPlNQy+-vl
zZxtB~e<*WE{mm>ioPEd$XPZe=22_KJSIA-CR+Xn#R%tX~HwW6Ev*IE?=TBpwFMBeY
z3t=lMNyPqQl*##B;F^T(<~|8PKP7E^HE3{Mb=Np({@hF7?!f|#0dO&)U_SlUmj6x+=o3cp?D&{L*OJ4<4ynb)8mM0I)OIUpG=Hy)*Yy4
z?i*v1NanE5O;{TqT;)I_Q!_(FdEF|ZeHR8)0YN#R*5Z3NMR)SWO%YA=0GIWGX*zv<
zv%$CVq^hs`EfbEFldue5Ktl|wd6#`aa8_ovRLE8dG8?bkU01=W!S3k`3vSj`w??P%
z+!8}c8O|6f9QMl7afYXazO)GTJ+F687btvXIMIl&R9{Vqv3{G->^e+)*}C_T0$G3^
zv!)k(!WWgdf-_ol%K#+8=ehp0@qVTk3*avHyIQmOVW$hOdHfQ<=^)=leDinq%&|~k
ze|!=-xf0Q90M!80VQ}W#*Dh#Wud7OSL!`b&R6g8hqI_n5vILz@v$+I&-UnsIIp}GhPLCg5Ont{5H9x+yV&eqn)pRz~I!zc}
zS>0^FTGwj%f;kD*q#LQNZ{f_l9c2U@V1-d0U}8g_=(rpyJ`7jA1?kDwb}brN%G&o;
z2GBA8o#FWp^AmahJ<%~8|CtjbPrk5WD|bf#%J*}B1uT|>w`pJ+9F$HkrIwLC2sFEr
z@Bl!|4=l;OWNTp!@jCJD!PD#FUbm>NhtUBS$ARJ(PgESO%h)z%g=5*$Dg&1tRMh=v
zG2Rio7X`D2Yec$rfREY`4od%UdV;*m10jX=UG$>Ikr3q_YXtj}3mTvE;#9^UX(&LDj+BAiiS=T<$`=oxgs_N4e*Q^#&GB;s_lPr4;%)(r
zHeGzwb}V$)4mg(`g;J_!9n#*gIqFcGrn1p-1=oWLoWc78w`^sO3m^8}^HPDAZK@39
zs!trde>Myj7Gj@>{M>m<4yzI+-(s1}^ps)bpd1w!8V1~b8KvA
z^M276()Ro~0yfxZ$?B|H$EfA{3h)k1vge6r`onnqx^RMx-1hpDZkr<&a5toS)UpTj{_^oIREYibdlAD8=k`-UNl?nE{xu$A3ySH
ze+mX%5f5$}rx{UA#jjV{k$6gRH}>R69Ac-i{9IDPCgR|MF6%xH3HevfbDP9VT*~)&
zgZ|u}x~6=%5N#7Z!z&M!Nxge~UAlMs(M|HjTjJSVQT|?&x6>OgzV5BeWD@Lfu^255
zCl`;rCm0Kewv*-B*EuT|+l@Mx?}>KVttpGRZI3*l&nb}`1Cy@(y1)R7yu#b(*Xt7X
z30Gyn_$2%GhA#lVP9#USIyMu{sZY-jr2NO76(28TH#AWLW}1^p3U3T@q{>;c9m4LO
zw_jH6eC*B(3B&e|bKJC;e}{nFTecT=4>f-;g(~ue&IF*IdCD7^b}U^R;NIE-bHV;<
zgdUb4#Af$=WzL}luo+RaCOi!bI9=};%|M0|m>F*fd<4s5Wh@U%Wtp(KGBBz<7AwW`
zmOSgvUVOakbXJjVIQ?@N5_O=j5;cIyB48Ohv?nKfR^rRRu3!>??nJ^%@!0L%%P1A%
z&=vK46(VG0OWswdezNKKax9o*n_*77l!V-;JmEdet-Akld4yAFR%WHfd@a9E@uPXC
zMbAwq6Ve%cmutA$1&KQp1bfu*{a37NnNp7S-9_Rp5{-L7+K;yL5)5q-Xk^&L6IC{h
zgPtzEbD=h)?cUKxAdfxjsQ*u7_PKdhGJChsr5Em$N%XadsmvW?P5`E$kmAPb{T3
zUFkmUmI)~i+mtwYCv983vPJx6@3Z#g3HMc^X8ZJIE0_^orlju0v6e?#Y)jwv
z^EH8{(8q^mXj@2;tF&JAmTu=Mpy}Ry;*5Q5WDEzmdvBH(16uCz!6{uAvI>0){WTrN
z6fKeQgxV(p5QDLwAM~+Gwp9kS9337_>3b=)hvV~j+EdEUJ;tc!E;02plouY^XAhKF
z9+VG?1^adPM4#w17xI|*J^~O|zq|(#|{H-WsT{8CQtw=#ySQ1W75wCMwTY+$`Gyp8>#Cq?wcLx{@z0
zGgg6_5&?65ht}x(qulQIzN&qzbaxBj+!%cbpbVnI$iqCx4LRZZ+*
z^FLi$Q%J~`N2M{~M5awhb#3?2%^2}H4UH|_W+mzGhGPtc%}50xfY>n~
zDA4=@&1$BYhF{J;AyXZ{BiRf=pX>u?z^2b7KhC!pdY8Zj#b{RtEVv$e6DV2l^T8WS
zi2AS1$oa0YyStN~v$Gp>6{=G^GIwb~Pqm<{(y+HY-74N{h4ihCocDv-{pLhB-?)Gy
zBf8y9F2<}|-y>N_h<6j%TP=FjD)~7aVV=B0X;l+Oq_@BXif}OU#s4bJE#X~>)dcx;{CU@9Z}YeX)rueCPXQ-=iUO-8)uKjD^wfLo=gm#i34ro5?^566
z6NpL#a&ISCfMa9fd|i$A)H49&$W@RfDhnTHKO-%LAQlU7Z1_$~vt5-)8fy)MgeFdT
zV*Z+pXUsMbZPo#RQX1I_jzM2}#RJB{HEnDyWIKvQ>{>Kh5UO)zuzXb@
zd=cCTp=6uzNZh`3NyVo+Jt${zCt2D3)va-|H~Wk^ezS7-iS^qkujCl=-xgvoe+tt^
zUyVlM`;C_~%
z_1x-qjM213i2HRR8!+#@Djj-O_2$>MqyHBpN@kXHiT?f{&zzdc!m%Cp|3Q~Em_M^V
zT&pa+U9;19DVR^Xe1ARPQP4o)2sav;{_HvTqDZn0L~Y`aQJ{Ie^x{>CaJ!boHm3jw
z3gUo$(qyz;3EjKyJgip%>Q#rj$#1%ct*W{dtBUXZAO=tNi9M-4NKowE)Kt<>Fh!mD}|0g)UicFck^&!|8xSYJyu``_jtV2y$4&Z?+Y
zO}0m5vJq1ZXRo6X5U~*wi|bmwUI+n=-s+&u=|(reGsCeV(1nFN_yK*Dm}dg>V90{N
z1N&Ek--I)}dOf^T12S^49W%=vF(vgrsldJ?I
z)=~Jo;sRCGKByPG#SjS-vlTu*lpzNPalbiFTj!&N)1m2Ln&=|6#KLo-%2pK|M9f=Dd}=kh#FpeKy-@
zwNZMi-`spf9Z2dNlQ9dVVXFsDo^K%(s_4#okiSWEYGkRIq5Jj_vQ$t|0kDA(WPu8;
zJn-UJzJU{hQ~*e1Th(beZ*EIjbzRJmolPp`;dRmQ4?B=sWEdE08*&FNZ?oAhzTBXT
zqD&0OU$U8;lbHPtl;8>?kyi!)M3k|EQFjjpxl_mT0F&7KlTE8BWAmA9`$pFzTxgUX
zU`C-|oMP{lN{+?F|BP-Ha+uBFiTY4?cyIs)_!Gu
zU#j(GCzYyau{L&{{BmmSiJRc+sZ#ZOxKDTU4swkmD`HJ0ElW}9$lA2S{24E|$qxwnKBW^8BKXRb
z+ctc~_rhCH>EbZKXPbF()PhpIw?4&zSeX3g?fc?4{`)P5?snJHon4MQSU2Gnp4kV~
ztD}Qf`Xbzub=mexPK~06zuM8C=^&0{Mfzjii}(1Jju0Oh(k+KQ4r!98&3|-LUe2$i
zRyA%D-r-{R>0AFJx6rqM2d^#CZ+-VC#cB~*-$=GR>?^#re>`t%GnrOrwH-i9}#%s6500z$N1q(wGNXWi*0R!UapET
z-cS+Fg+{%WX0Hn)!gZoCJ$EpDmNv+sh;}uob
z!)Q~3t@e=8fbwXN6HsJZYT{BscJEKfXk?-A+$JzT)E
z7+~3V!Ni-!(WmEv2NrQZ=E~RY{49KK+Fo{Lm$bA&t*CV}O}djW&!>0F*1iu{^wvwT
z{7MseB9e+=Z<$8eZcy%%Ed*>-CujRFw=^w=%oKfE5QW7f=D1rnHjhcTX6@Ovc3lnJ
zN-N^)%PKE)_usi0jxXKtQ`gCZvv4ty3
zv2LtxTDu(*F8K=fOV+{&9vRzIzTcYgQzZR)){b
znK(?GedT?n0CkM?5~Y_b2we(I4z;>GG&=#AD>Eu*YVlC8uBnCFz47MeGv2PZq*lE1
zkN=s%_es1gyg{r$otK5&D{R)_K5UQX%c@v(=PLn`x_4BJDGVGIy8iZQj>0@&f31Rh
zPC65vIx^B8j1{y)5Rh)tF^QE<{(qRpz~raj`4(CD`)}R4g}Uq)#h`slJyovwc2S!ABbXGWV^Q-x;vcCq|Bh
zx!d?^tw3WpB0BjtlE)6e=H&^qXVaJMyKwGSEF?o54~60^zmYB2E=#}u+XlS)Q)dEL
zQ`)c306pEvc|^dxU}#8`|UU0P2NQjVr+)M|Ni`czxH+uYcU>rfLkT{4+HzJX;i^2
mf4URF!GU+(Hv(vSlMVQubI#D-(*BfbNsWORoLH`Ht4SSdX

literal 223889
zcmaHTby!u=w=EqKBGTQU(%mTnl2Xzk-5>(e4N41mC`ke7lx~ocR=OMM?s|*+epim~
z-tYav2ObW`z1CiH&N0UrbAMD)kivLC^Z*71217>rxiSn45&;YhoF57jc!fIccNO>s
z>!>Ux4pY=mvJU?7+(bjhR9+tDDfk%$1~%9N2I1~4;EM=+!N9<$!NS0U@342TrNRC8
zZ;=Sn;Q#w)IKR6$^7FiohJg`JoX6dQ~$0g)3
z_H>1Ytb@xpa|f3=rD;^MFX9x{t21VQX5Hg-q~SjGtnl0ptT>yQnVB}Z+ud&W_X~dV
z_>MsLK*gz{TQ*hy3F?e
zd6!16PjE1@!WUjxMMcHwI+ur)m6buyt}l+CQAypLcttca&3&Sh^ef=sn5$Bqzl?WC
zv6gL6j@$M4ygx_2#O~W2$9|LiSh+J(?^?@b%H_fK&Dj|K?S-MvY<1XvrQVBMh}CeF
zr}yR>p|`B-^(pn#!IbO%B4RYXVnpl{6)OEk5BqoTVkP9D*H`l1S30c4VE^<3Y3BXL
zposap?9LjF7HGz^dGgQJdxla4qES0-sGjXDz+=H09
zp&1rQw*J}H$M#S~tp!uY`?UK#`
zuQijyIu>s3(fgY#z7PmQ;K53d(ZO=}+>>qz=sSes35(>4>FHQ0eSc`Ng5ZogBij8C
z(e@T6RDMK~i=ex=WuhRFD2g71C4Dc(x-xC3Fx3d>37gvh_JufeNx_Yy@G?ivR^ufj6_$6L=Mjzru-m2CHsi1l>;<
zRpYYal)ch+
z%2e5OXpxpdr3mg#XZfo+?MhoNKHK`G_4SY3+mqC~L2{GxqF(6NHUy1c*Om;*rEk@6
z7?p>Y%PkH~+Z>j{e|&swH00-MzdnpgF6_Z7{KMv-6(qnQ6SRh_o^s?np1^W@R;%21=
z_!9To;uwSpBQ=bWz00sXSG7C3E0HN*Yq7@~)>IklhET*J@-}1roVSj+4<-t01k?U_
z-y++h8*aV)einj^2E!uywc6u-*#@KwW;(eq0ky}(d7iFHGj(<%yvFaIQ`KeW=v43*
z>C|LIfHU?mFa)n{&T2GGJnQM7iMmL)alyC4yInC&Fwqm=#BJ7Uqq0I2QUnq4$Cx(+
z*W8XB{tOyQxL*X-%d?cjZSYuU_A}`e@J5gcWs|GV{mvc`orOdE#INh7wgXP^(>Wb|
z=j~blPffS!5d$e4_-~VAzR5|R=+vl@_CSh#>a$X*UC1)!Qy*O;p`)v6f4wE~uh<-2
zI#_wepqX93-#Il!Y%}E|dLftaiOGx^oq&@dRmA(G%8m}>y!>!~NlKc0txtbm5%sow
zuF1PtXqh}%BKRS-dVz-51BDcRTy_OC*<$jHXjl7(uW
zB&WCwG03H)q(rXS`~Qr#V{n*xZALl>S3%_+1}r={5(5-8r%D;9xX*(Fu^!ZUyvd-Q
z9oDqn5kV>|R1JYsOcfp>gA}(VR#;U8ci!eXLW*@hzTaQWvl(O79r0-^o|IS;c@h4D
zblcafn|)=NF#x3ZmYM5-le8u_V`uTkdRI2H9l@|cB!WaP3qJ+Y$)(g7`U>P-(oc+<
z%CS#$Uyf+1X}=bc3|Y3EIvZ$_2p*g$MDy_Q5pHl>Jl-5*(9dNj6LQN82_KmaadB{x
zczGFX3A3_Qb2tej<F8PGx``J-B9zzPkk7P0J$7El?Vs^-<)_w+-C(AsqniH5
z#r97;iey(eo#tD!<@C*!4{*97kJd?t;V)|Y`?b71l}cZ#73fqDRalJ-3s>x|#g|%k
zQ{?vQ%&Ah}T-9V9Ym@DLlZ(L&PvZ7J(w*GfH7ZI3J757D6Pc$NK{9cTQ?t6@i`^cY
zux!>F4~x&Nr@TE~?Vuu{N5ZZ#oFk*+q-bK_tRsjNu{{YKW*8>dQ&A_oR{k2Rrt)Vln1}z|xi@f=O;H@Lpu(;>53$mA
zW<#aXgDRru^OLBu8E1ch8Ua>xDRs=+c)u@_WYkh~YaG6gOzZ6C|ouxQslCyl$(
z6LKfmsC8TS<9BJTE)~A0bWG+KP8F$FF?sqrZ#jnF$uiZYxMAdGCwa2i4>bzrTh^K|1(J~CezAdCY-wk@92Q#-x$@$vC9Y3&O>x68xT
zUhb*5o9l}U8d-?k=Qx|1)zG#6oaRbU%U*=|v?umti$<*CUq
zL~@%T|48ofG%uq|=iLL9IIqSV-K>`;sTjoE8ppS$t`1N|c0qz7?}TOXUu0wns~OQH
zEE%$^XB2oa_lAQ+`0aCT@imm6pP53vQ?Enekspe==M?xILRenotqtfm$Fb-uYpq7`{+-zVTc5BDNVAi`S5
zGQNVFO!IsCncGGyQ#MYa0oMC)RleTi99uC(*ta)tFoqi@<3jnnwBcuyY5az;n?ekw
z=I(Bp?a7JB+})SBPjl>CmqeVm#`i`S5fO$VdZ{;_8^abAHCCs#6*Ql=rW-w_m_E_i
z?QNEeU2P`ylCA$LL^e*KPjJn^*hy3GBeGQYneCYOh3{onr0H~A?MlZ-JE~$#r|Hor
zq_^>rWQsZ3Tc{$We)4^8C?SSE4TC+$q{Mgv+>5m|yVTi^Z@Gj-7jKm=U~^@oxea`$
zYaGRhX~ZLUX0VvFI4g`H2>xWe#waJd6NRa7P6Q+mmfxT4rB4|B3z^LCAQNZB1l~Pd
z0g$QQ^{`{lXSNB`+%ALK{lwbRUJOE;r(eaK;_hnL*`K_TUdbH_^AVfYZ!FZ#*H?vx
zhqUDE;y7PsUN2)HFjk>-){f-Sc()bX3pmdot#eow
zUjNE^nYRscV}5YcTdzq@3pFlm1`WlriZe{TNrUG>N9)IoF}n-YFY>B}Y2TNuZ6{ft
zw5~E~e-~}?B9}}0ayHlhIsY}r1bd9=?8ZPMEgA{QuiUFQlA+?-*fO4`a!GmZ&eB1g
zxhll{s1f05l`Cm|Wl_tcHRb7c$pTJ(mFS-~^hG_1_&f-{XXQmPTCfinY!|?3JR}i(
zP7G6R!&{*f__9RC*y@DZ0oyO}t?@uaXdH7c;nTN8y{exszX)locfrK=at^|q_w3_O
zQiSwUJhGdU*@%93vc=byevORTJj520CuJ%uEr+`FZu5j`d@2c3hmd@%idM8jm{@#Z
ziRb=Q0iZ8_2lUhXxuMdwxSFVT1ES@GrFtQ4Z!AuaTw7&
z8ysf+VHO;wqDXitLheCg5S_kmbHZ_tRZ!F+T(?r8S%n
zxEv7vwXPwfC|bAPELz#y)(Z^c;RCtkncBRpxbIGv6;&HC8SiqHHLA_o)C+X7j<2z`
zZYGHUj)B*_VBl3Tn8dESJSCq`tW;I)BN`d`rQC3N;?EdI1OHS$v8owTs9iybz7MyR
zC@31&Jji~ybTQHsrAf$P))yr4LaJhV*cTpo-zpyhFiP6Qz$u*D<=Ff3?C=p7Kf$vm
zhb1Jprp?k5z>}+WXbZY9^lpN`bU&$rBZF?YMOjhp
zKV};E8FNABUVTwhxGsr<49SojSQjtv)G%du)J-PpXc
z_$;b+Wt+0N>uirRMe$$1<}{y1pzQ&_RLOdoWy;p&>Uy~9iw8s0r(O9ALKpH)D84xT
zC&6uZ0tYgHndq8CPh_HL;@K@SwEAU!_NMUZ9%rSiH__xeE}l7bWj8pnz*9trT~bY5
zGcrU*yYF?G_jgfT3#-Ls=0s-54B8nk6
z7YRsbG|l#0wZmg@9CADphFF~rXt)Sy2?boXxf$hYo07vfM>8zL6T1@Vi=VReypbV|
zrjsWKuCklUgQXPO@yDYfW=`hMTe_0CM~R}odlXnQZ(aolN0dcE?KIBWD_?+J7JJuK
z;<-%@BGnhx1YH?C&JS>pk=!AS3b4#AG`EwUI-$=fV81%9{IpO`9X`G~cdYP;+C{!W
z#hx-Wxb^xoo({l8O$Z5JS-u$N(gT_s|6U>hv}iL~R^KoX#N8%r`%FgD`qQlXAv8c}
z9uBeSROfgYp9Jr2UJf-%HymJz&_wI_YxFcnOtZ|%KxCrdKfsn1i!7e63u|P-E0a_H
z8GB;c@iC>?+uhyW>a-u8ki!(^sNiu`(M(h~8;6dL&XMJrJoIW=AI+l&u^O6vg<5O+
z6N-oJ{+A6zuI{>IC&=90#Fu~z;MCUHh^GC}7-wuMxiekUi}xw`q|92yJkw|Lpn)?+
z7>%a-5>MG5?U!Or5NNK?rc4qXs8XA0G&f{SbI;<8SsC;j?KnuTo-GOJd-XK>hg>qZ
zdH&Nu2EiDy1_ZHBK^=P(72jq?M4W8WD}^P%_*hP4@8T?^qtH?K3tnKWJlgu@S_5_A
zcy!H#=y=W&pPE+GGON+>$8+9gUvhMLDyLMT=D0wIyG#z$rEb&Z@sY0mxnVU
z?^%B!iobcG@WCY!1B&U=$bvCr)mX7vk%TH91ZZ7TvyYC$?NDD_uZ=r&keF5p2^M`7
zU7+Iiya=0E66=c6Lgg|&KZq+zeyN^8v
z(C}COtGS`CQalq?kmS27gt&KY-ruWHDnKe()x^|e-fupvH{x=*oo3Z6(eH1Hse^XS
zqSoWps86=Hhsq=!uPUthZV0%PG4SN#IZ+W0)}%Tyw};YR8R4*~qth!UxAyDSed2YX
zI@qe*e#m9Q0?1ppX51Y|Yx~ITMk*ewAIV9SbaV^<>N&BJY>p9uaaZtgHkAR8C<#UHw$V^02?9($dh3_`MC+#D+
z>{yZfx#o2Bi2oKl6n(hSbA8Q)O|p_Is8dGCNMYyEQ+qrb&+p{i{zE3_Rqh)^fc$Yl
z=4`Fpp4#6!=Wwd&jHp}MsIaoU?45A@vo(r1W;ZPw>}*p(8%mbp@Z0qL;VOn+lb1gz
zzw>|#F-1Y6yI>r
z=VlJKJCO_RWP3Ky%k$CzG%b58J@`7cN7tgij5@=mV*jx9d6lhpt$Y*uxJJ?4KiY54
zXz?r4@}vdaeGJzQVzAg>NIZIZ3HTl7Wu8l`@}OIl>T8Z#i84mbqH32l0Ch1hr_<>n
ziGZ^Xz|GT}d@sUCvf5)9@{ssE#!N6SM+K=niFO_q`~GEMp3}fGf5Q2+!(IIO{xv07
zE6=6NjS(wLpER9`vLFHNMjbwMQts)JU3UqH^sl(kXib=C<4S+e2kvu{7?)~KU;CkB
z#mjPY?_V1?-R?1+9jpzc%120s6RCxM-MbY-zE10{j%Rhi#@wDW!*&-FVdh|>Njtm|
zs(YoAguAf5p1a5=trM)pWtCxgePK@$i+TrBzxBrRp#dad)Y2;XXSmc~FMPy9`!I|o
zzec7W>u;1`i6H=5Nz9g$CFXtw#&;{wh9E}cu;>Z4ZGRT+nv*Rx>f|OFMJ7Tgys_rx
z;}aA1m6zFy&dI2o(aL7trnFR-f$2Zi+^b$zqZbv(Xn67TKZFMfYs=Yxe
zGj$gJ#4^JUggQ~NPt~5P7O6gxh1_JFT%Mder%Vpra9T`DGa)*yY^VwK-
z**_BLldcbpXEPFqaQuoP`_8AW-UOSuIhKz$*3ZnKTwQXJ>0+-`&S)&6`OEm}tRExt
z@j@GN13&17Uialy+b>Z4jLr8coT?;@=tf5-9E?gE8;vgj*B5>oGWPs$K`YP&S+d|7
z^YG&DDThHUP^XZ^qvz%4f3tOlDjod;lkLq9AMG*Q7O#ebH1Do7U(-$bFTACxSG*33
z%{2-(UCNT!C49P*@BR$sS8Z5zgb}KpkJ?O8%5Mx4ds6pmCYoly@&%G;Lm;(ZENrE4N;?$T_h#$?}a*&
z#<yDp(eMIEh`uV8Gf+L$!fT8PKoI=&BGJg6VZbB
z(%}^%{i^*+lHt>1yCp%V18;9Yf@7Q`(8_}>ewT5=wdTkUrmV_Qoo%Q57+z)MSOUf~
zd~2c%G4H1`t?ZjKJBg8IUxW$>%|70Rm2aL2tAwALY?S#oYrbJNO?Jy6zjNtS;fBeo
zN|CosVY>+xy5z-BQUb*l%JW~@V;2FbJ7wRXQX~#aZS&FkZ-1O}ltwXSdSJcEc1F3B*ckHsyZO)+NfiVg&~u;IVyJa=bqiOb-;+Z9_Y10b#d1B|
z=pRy$f`j1Vggg!Mf=KvXj$8grp*9cKR5S0MqsB`<^L-q@XA|S&X$CEDH|n5;;rEW*
z|66?j>!Y!*;Ls09uCfb0{f$xp6w>KfEw0{oWiFTDKF#pR`YCM?14BGogj#Vjfpwo%
z!6#ukAoBWCF3!5^@%?@M11sDoYN^UeuFb=>f#7R*=lm>RV@-;L5v=KT&^1}sWC867
z$z`Nq7a4)(#_uXS0qR&YV2#3ERsAYygttz;zaoksh=9G%n>~2AAJkgnt5NzUuxs8=
zNBL`Z-`?dPk+!{!EA9rsZ2b5@imzQ|!$ESiwSn;V>7&n@xN@`?-%VB$xt8L$L
zt^ez*nCuc+2plFMDVymQZ@u^Jti52pbypbfx%@v29cZTgZjZA|8mTH3|Hk6orD{|(
zFLg$~Di)*mxH#(GOVh0k#bNna?XZ+#x2qTR9-~bvLcAPOXuq{b2}BL_5v!4)5xK3WKM+-HP&h;z
zcj@P_R_=})=DeEiup8QVo%wRprylV^hSe!qP}RQObhWRzWT+v?%sCo0Pk&_(pa76;
zzy6ChHeX4mVL?Aj3q#QL;Q0uT^Yr^_=k*X@|6qz36?Nm@FNo1=`ib10L6#g&wMRK~
zC69Jy8>|?fav<^7zI*YLb6r(ec@5B6=$+oq60@^*ll8O@a2d^&N4jIa4O6rxsy$m+
zog`#RswF=$9CvSg8AB+(i+nWD7dQ!V#n{Qjxa0ufC}>$t81>
zmYWW|x^}-jNp>rl1-eV7^e2*E>BBXv@Qncb+BP+%{n=(JojRhf
z&gjp)7=v0Fg4eAPC*-E0S5YXM(jW20vo|Hp6H)v?%V<4m>
zP)BgVV=n5APQ)uJ^ds`Ih%fhe6r47@vg(LQJbIaz9UN|}-dk!Iti;xhT
z)5W{1$-J!eUVxiM$|U-oHVj5W?EH~N^3;8eAl4e?&ZJ*9hdU?>$w>vizh3LEiAZ2i
z2*;p=$7TOu$ZXR8g)pu^_1N*xpCKcHgDKXni5dx}Kw}+1Q=%RW#h_Pq6$G+p$x_Qg
z7^@u)=@gRr!vWT$sN(ja;t(~a>qs1V_A1|0{+iPZN_+#!Nel}BJggFp*njqfR;fWt
z4dhf3lV_^WfpYq)Mp+VlYwjU{5GuMFweKnMn6<(Z%FUJ%@utt_TO>f)sFpbm!Ff_}
zVzSpfI*RxCsX8l(Vj|~9w?9Q@nF`5T7|dObu`IfIv5CiNS+Tb;)geg|Z@1jJ1xNccSn0N{FIv%TOsTkm?55w!dC^D|0-6oWjMFcQM&1-cJj
zU{t{Lm3{xD6SrR`*I|+sNy_ik0s4W#Ob2SY1kP8aUVHQDTH=o;AjYhAt1ez;)*)7)fiWknP^-xmBZD$?QLz|C;n^ZKH~
zunqw8u9R(U?>nqq?|NXcl-)-lVo3EO;(`=ywNJ7>G)OiO
z#_JxXp#O390f6LWfn-BfETHF0`>ah>?dF`JVf!zDL*A8K3}ZTw5DH+<{Q2P;@4A-`
z(FiF|lP7)!St|!^xBFxkx}y#}BX1M=5OMKhsh@mUT9spwX;k2E+ou#P)bPl5EIj#3f>D
zVn{aswPLYeL+i)Kx8|~}d{o+_1y2o}*59sFoIR>3_1d1ZyLw?0_!?B_w}_MjdTeWOW0b@JVLnYStihqCfQd
z?AY|&$?}x0U~^T-Cm=MX3^X0$`H;R>6)$mt`lf}L$PBu-M4|n{;i@L~9tkAD%v8`F
zw2ib&BGy1e$7U)hCjOC8(wZZIUuL0m`aKk{4~m~^+S{y`IU4D3THi0!JdL|E0gIY%
z;@#+Rp0UU@Y&9r-Oeqzy^=Gb_<;C?)Fj4GT#mX%_2eFItn`eJyD#M*elCCy+tbb&7
zt6&May&KNIM7uayFAO10<#Atq$2*oE85G{c)p?fl^ORnBf+a2wUw&rH^?UP>?1FfeMI-0u
zaIUU4M!*@LFX6u_I}L&A-290Wt>y>l{IsBIu-coz>v(~aEuTQLQzYt#=k7M^FPIZa
z$dN5`yAuBcHwvaNIz`Ymp71TXg9bnYVIp=VE;}S-?dJw0mC!U4IUg4)&W*
zX3!Gx*hC0lU*>0f?s}i?!RppJ1*=q&e2AlX^meT$@^ijAaqaoei~#Njt%y#=c!F*I
zcRzs9W!ORU!mRa&OjpV?o~`ICQ20L-mE-*=e!xh-!$xx@DXzpXV!bbaJ%yaiZ=681h>*}Aesxh*UIq`ti~*6
zBVhdW-DQC6&un3Sx>d&|bCIpa#Tff{>mibrp3m*AlCpy(4v%rVQXg9U&X)YvUo>8A
z!BJaV(+aX(P&mOyb(_gIpe9+`rvo3(pV11m5d3E$Bi_YgeUts;Rqpsen71<@XbK1Y
z_?ymtT}y+7pP!$-zrAUY87g1vU)a%AL?;#Oq7iXB`qLF7z_Z-a;4t~!6)l2Pq{vTA
zzvexed{UK9)#>*t*j#$!Hp9hFA5xFTx2TK#NZFvnNQ9*KXg+sy9gpO|#7SRcE$b
za*%&8reS+o)9_8M%C{llWzngBhYDa!dZ@P9aKS4HQnYoEsmk5KNyqlCg*H_c$xxgY
zAl9(D?+9{Q{USiaV{XmE?U~p47QvswCh-FBfRD=1hcakoqx6YC5O7>PbSm(ud8<+7
z1)nXK6eNkbiadTC^G(5Ub!+16Xgy}4jBKo!`gQ*o^Qr8L)FoP3h>yq7yPGS-BVtyp
za)MX=n5agc(1$^r+QM;;AcFgT|Qu*8!w)N_qF(hSU
ze|Xx=XK3T|(7cS5X4<|pOf^(@SKV;e*Vx7Q`Hp|ujJn~m9UoM>xHz`!A(xEA!#MEK
zR|ZCpsFu&BlW%2jzX*N=l$xmNqxY^2KK`re^n^MeAJbzpTlqirm8syWdfXZDrl$kL
zG*rKC)i-XIWj~)AP$~LEyzf!;8a>v(ZH*P{4`qA0<1i}0gcD!iYF*)=e?^x)6AB*6
zD%XTM4``23rfz?hv7$ReweijTh|3E#f}EH=G!OGkl*A&Rl$dO)OAGqZX*CKiE9}nu
z(4yft_f>FQoNk5!AZoN-dn~qd&^31}8q4r%-s_0f?cHwB$g?uN#$3xHt?iBa+Q$
zDoGQotg{Xrfr_bHcRGJ_eQD`^2w{4Eu>7&x;HA>VQeDCSWK1%_0BqWj%ObjFPfa=oud6dlSMr8E
z5Y`TANfYw|v~uxSp?GIaM^Q(oQqc%*Wc=2=9TCKlpkSwgjw_WSDn|Rkljk1YO^&KsYG6u$Zg?MLgPH@1Z^l
zJgltik~Y%AW#8-nt42uD$S0~=NW+-R@Yv4!GPHfVMUTA;I?@8IJoLUcn?
zf3c>?4h&)g08q^*GfPtf6S*vN-lH88d1x0&&tAoe4t)LY(XN=pbDwlJ|;l3(Yg)cFdKL?zV~Q@I()cD4eg=;M&j$_i~9HbfUik%7*(|C
zvGxaU!dx*VLUUGyqGVmw=|B@%ysu(Qbinah%l`$?idPM91KkE$=zvE3BH;jcxcOzC
zYHo=}=%YtGwmNDj>rINKpnC9*o#RQgRC_s2b)4#KXxxlpgRTliVd8Dh;@fL%fAK{j
z>5J1jZ-@0?CN>7y36oXBB9J5ZN>%EU(EF&Tc!vP|I*x(cZ@#BHCPPC*3OF>mRAExH
zfiE3LxhjlW3#)xe8)EJo!?|iAsQ_0m;&;ix&vgUyCy+YG29sogadOm;MqdiN
zMek=cOTAk3gkc}4Q(N7U+{ROQl{r8&IRH7FN&UGh*C7>X`su!PI|1!MqDaRp&dzeO
z9&xhDCNx7bH2%xu*ktW0+g#c_
zFg)-Q3b!#}X~6aypcLTrKmoS^ZCG&lM+>WtmgnaUezo5r@Gauqt~CD|U*v``#->b`
zmdrK8u~SN$b2@EpZO$=>LG-e_NSd>;X2#Y;6JiDTYi@A*qat_T0hhb2r9V9A-wr()
zq%UtUexAMV206XNV1Y`SV#8P_=KM9|@ByJe2f5YPgw=t0_%{vRQFRos
zvL9QpltkLv_Qes==4-|r3v{dWoW@SQr1c~kJzO2kZyS{&Blm8u-5K?3l>6w;XPc#^
z_39jEz*s>+3lxjtTY>|0`jTO@DJU$HPKEVq_6*Ry{I)ZVg>q=i0}qLblMGIEE8l&T
zNp3uq$3!}eXpH?Oa{DD<42?z2y{62>0W_{Lo=f9sEkl6;sE_Z!YRfCNlHo^srppLD
z?Pb_dSRoW|9vvyPQD_T_dKI-t=|T^h_xeLZa{XY3-?T3w)RN0Gyc2QtA2^}E0I*d=`UWLqj?=1moXRlqM=`pm90F*a=HrmxB%4eU&xxa(u!;S
z$x>hGEkYv
zLtQM`?jLsy`;U(k_o*?>hZd!aLgKk}jpJHtX$SJ;4X2V3<6^&myc3|y3P2+#G8#}nQHG_VI*{-0tn^Ca4Edys6z*6jE3^1T<+Kj
zqmFQith^EX<#|{}&F@&mJT~H>J;<||yIj36vT=R1a%ydJ8a{R}5EaM+25Zz=%2f#J
z?t39nwQeDb;1`;7S;M%pvOEE<>y4?i$l%IM<)SZqLzVJYOV$<1;>0L_7FR91Zb8!&
zg2UoVG?H&L*bE!wam($zHI0a>)nL`hE%9l_{p3wY!Gxuc;YC=&-oZg@yQ3M_T}q+@
zT6q-m8jtOCGtes7E>2(l+4?0z8cHgFw=rt>LATE3t6PO-5I^bNt47D`F*?Etccdd&JT+E*}7Y+Hsbzbd)r4NDYdFl2*dif%d)Uvvb%9-74q7G
zZbt;*|9$~D>%I2K0+QqKIKHqjP9Wsdemf#tH
z82cZ3$}+||K>k7R1r`cnBR9!-owusB=+;3nhY|2SKzxATp2TA($J`FGW!sA|)pArY
z-eK~#w}~`iT?vq~PG=+eugyXB6BH6!xR?t=AhZT_-$No!OLWaVi)`+4F;nRR?IyB_
zhy;orPA2_Q<-XP^cOLHrakVikeb2*UPpkeoO)`+7N>^oKH5A_fNY%jenb}CnSFWUtaFj
zhGg9L35}TBwl&N`WvIx8*lE%q`QgZX3(&V6eqHn0y6FU>>s@hsAS`x9M*3b~8jfr<
zeWFq$8vmw{-yiq*cb0s9e`J^Uaus8f#Xpl`0Tp*sAs3h>P);9TEc~uy0iK)lH5%gt
zvg}*5qwU|sfU)B?2J$n?tBmI@cXmnFGMs@Q-6~4uFPuPE)fGh^9}e(bTBW1za_=Yj
zHHB2pkH=BK7kXzQ17KA}t$SX4Pa=x1(NYu|5`p6MIYL+mla(RUUJhWHS&rYD|@_A%L;28@En{2bzj1FE;Y7Ruh
zf0eC+XgzasOVL{v@Cm&&^LrH!E2)s19Fxr*ww2H`;KpOo3?XZ9TZ()|UFvYHg}XO#
zR{!1N*ZlrcCtWEUFi)Y?yv8<4ldn68GU)jF0$r-9(bN8LmC30#M#Xf=l4|MU({31GaLJFFZ
z(h|E{ukX9deqd;85r2&zDH}{#P-%TPX8?RXrQ#eI7NQNZkR?ixEu&Fx4L}7R_Xoh4
zQ?o$LhDXz_EQ6ht4myo5LhhJL^l|~;)qo54M+hz&K8vpSw|?8aFvfDupp>pr}#saJr?&uI7`EElegPNAqH1}sK6`6TZ4oKMtu)Xv?Ekj4i3
zz-Q_C%@uDZv0GpN`BpfbY!_qymF^CbD0I
zUEwM#=g4$|-zBAMdIK4OQuYstfCqzg-)yauER$?3o6!p{ZORgQyYcE*NlqmXLCc4+
z^~CFE12Cz-F4B<%u&Nh{xiPcPSQ^;45>X934m#wMl*%hPW*bgk04jXmeXhEI%UT3z
zq&0`Q&Aka6ud-g^T2<5N!QUrVA6yjsk2KG!Kh{G{PWm>Zs^%!c>ee|)L@%H0ot$7!
z%PbTOFLf%idUnWx9Lq2$q!`CcT4vmho!Nqz>v4yXM`_GU2|ZkXc61;iqT#U$0IznI
zepzp{Lo{s+x7~zg*Fl3FuK_JFNQ*D_l}+Xs7h78ppVfS;NB_G+227ykEP)7gLpj87
zP+~<-cV-q=SF_2+e`+s`#{LrE%}Idid0jLN#~^~bFDtO_MFMg@=DxEBPYGbqbQ<3x
zmRTzUlHvROO!j4CPw?6`z?sxutv}2M(_r0BcSuuM0zIKD+$G|`4rCn2^EnPGwPTeu
z{~7wqX7Uq^c`U0b8oy1-IZ<`7Z
zJvB(y1cvta79S;wMbvutQ_OPHzSgky-}|fE+BnRrC_n~N7mIvM|I!%dd}GPK>l0P3
zCzutbwr~}GF;}*w3}#*KT7}On^k>XQU+AZli-%VP^i8f#JFDdk4zj|$*8+ywa$h&8
zPn;x+*NNO5w>Nna#PKj2)fdJMHLVaG+sw^fxOU7b>)l>=RZcshz5}`I-Fx5>0(KTG
ziK(5zQtHV{n@}(|D$Dk7AY|TUo`cE~6{u}7q_hWi0YLX+gQ=)?P&J}>9o|Ur9El-x
z4_2)Ryj|`B0|n3S<`_LMPZoxARsQ_S3MDd*T8chmzo-IJx|>+W_L%{_4Tg=I;-m)2$yWYCgUiyj4FETt3JX@TN}493LNBD|qfj=}ofxjYj1aRb#uXw)}@N!f*s
zjSx_~4Z#jeCvxVtpYjPhmw>DOfX75c2jo(+x&dLRM6A@j0eo8qPF1n^&__)R4nQTJ
zP`tK<{$v5brUv0L);>4qBBh+#KR6Hv%0{Wp6!$XtojB2626S^{AOrV_DVJjl?LMM`
z>fqlDoDgj9{wU!vD8ZQYK-vM4<*mkgz{70V8gO=fI!DPtc(n3OJYADp4Ene|IHLy8
z%iACx+U*PL2=I}$Z%oW%X8pzjBs{<}`+~BN`0Qsr7MA%?M&LmThnWJ(aQ{>23lsHC
z1BoRw&AgQH{?dFj@1~p=z#OM(W?(GF972MEQBPFpwx)n2D+?JeKmTXX
z9uT0~Ggm!A;(^)|FF8?DdeHc{+Ls{+C`3`fa$AoZbcK?OG#U!jTW`%a5GV6H`F;IL
zt&l1l&S~*00BnP&NH98K#!e3~W?Yd?O&v|XJaQug$d@V!NF+8AyDF?MX38ry{Y
z-_HUVj^;iXJqMG97_Psw*JQ~byqT%D?XUZjDAMBlO
zO_XogIJu?~W6O{|=M4U^>}
zoajMg`eE7%Jm;Vb9R+o`@S$^p_FEa#HBResEN5qD0=72@cQgudzIvf@d%jwJ_Go*Y?*rFELEhiMB0iKZfjmAnl`j?nXkrAj4)xCVLgG}9)hdl
zj{%;K&<4y3luq3g;Du+ZZ$VKo&EaZ}yFX#qiv-8Y|MZED<9DMg?vf-s;}|1sH!+A(y|`2rQm=4
zmT=|!8%E91h>Jfy`h&WEIn{ngT}QaOx)N4`Qg1VEJNZ3VtFBKn3cPz3g)>-_TaUBOCM%~htuj}maP)*;od7u7u1;C8GGk9x7IiiC4ijaGIQ*XtZ5hW>92Ex+60^PPjqzgO=Qh}LC|
zu+-u&8^=uOwlzPWR%s$bqWAMFT8n+q
zVdPQ-3Cl>L5o{rIYN8fH8In=3g(}|^#E>XyY-Valwk)vcHBD)|c9n!dm#n_ln<7Mf
zu{+$?hf(u(MR_e%*psLyj^#B6Nw0dB++W~!=i2)S<1eV)9lmVxS0nM=V~C=a#VgQn
zBJ;kz<^j`^u$uiZ6lG%QBmtE?jLluv*nlqFrM&=BsyaNpTTPlr54?${f|TDrh(?u$
z{r;A~G!!k{2_nF}2475y7)y+hB97p?E`I|!F&mQX_080|DlVvo)
zqf~GhHIzQH*@`0l1OvoAAnWB2P&Oj}4(oXfwW&UE|LbtaV_3qdsQD(4Ox1hlx~A4H
zwAbknYmC(3u)j%|)^*9nJypxo7^&F5&nAGGGq8Zmb{u9ZblH0NH|c}I?jdwOgckprx36uE>#+X--7GRWkZW|&nx&&y_OZ)eTTkOz(*BS{1bOKV{+f$%O5CjSe
z+2O9Y(gy~b5Q(4Q@WYtxl~gG{c*8CGYgPi$5-09m-}3m->)ef{uT>FueiwRyiMuL7oJKT#oqqa6;MeUe?9Ym(3~
zlY$cg(3+gF6G->#7A-;3X>e03EWcguehly`JfI~oPq;M<&+vdbqCBykqKAf0Kp=QG
z)NmIHwlEm@0-uP1FHfK@l%}+&KM$_t&r}r)@c8+GJiG)T1c}aG>N9cjk3hE#X&=gx
z#Rp9fow)YiBJ{^$VFNIYf~q!pKMlZuMFqDxTBy=P>FVaT_$y1+-pNT402c8~CUr|v
zIKb3MfrQa)Ffi&M;))BkWe{3WJQ+8=N_e1NG(`Zm@LY&QtV5+
z#)O5yq4wtX=DI7I77e5$Ur+&;fHSS6yqv=u+Y$OIMSWrJh+$DGyn6tNgYwK5OgR~K
zMNxLev5>!vV-5!vY=3_=LnoeAyE-uG_=Ac0SX{MKA`3xgSlAJ
zE3{;>!;Q5hY0#!=_EMzz==)1*l
zqJtVS-_%EE2P^xxYG%uiq{^4*HfK9Ak(S+2}1r-;WHYo+zX;u--5!(KTYcZC!oY%#AYUSqGG}6s;Qv(?X9)CYX)EDI$FlRkF
z89i+Md-z5FJUH4Mdx(rl!ivt^^_B}PTg5I$fR8U-x6V1Lc>riLm@^Tj6)3a@W@f>_
zssB~IP|K2a6GSNxFuVfT=i8yOsuD%@T_&og(fm*5ruc%Y)8s{Zdv*8_aGLO%kLQ3n
z`0goO^(M?RjV3}KGG*M2`)r<5r&X~d;PF=dagI$@HadDg33rK8cQ1iVN56C!c6hw)
z6LD!PD>85w2A`}1+<$Vl9Y~DJ-igx~cic;;${QcETPuf$$NEo^UJw5Q42+H-b@k3
z5XOfOO<-Eb{_m!H&mT3|EFH`}qwn;H`#)4&V|ZoDw(U6SsAJo<*|BYR(y?vZwrzE6
zCmq|iZNJ_3o^$Sbf7h2Ud#$QjV~#oIsH#M69?#2^1gN8CTitt3Pj3h%(wyC%>FDly
z1i4fn>Q3jiM~2V5p>Q*oD-*53W~)D+W7}a@Mg0Cq;l3F(
zlW|0h{vw?I^m1-H<3=y*Urh8&Wv)v@WmBpldZZj@>U%V
zH?{nVt4@CO*a=h;|M>J-?Yi+Tl^l7JapUT|eK#NK^5EF)GU%BOvc-96{GhQsU3J3)
zUP-o#{doN(fBEc+aU)*$c%nN)T-iE#H)i+R1_zjTeQ-No^R!y*Hosoc{SQ<4n-Kt0
z(A=VYM*iQ2TZf>Hma9QgmYRJO;xG>6S>RwZs~=i{-08sEVq5){#wlJ~a1(qk{-$*Y
zIcD=^+a-@lv&yOGbDiVW&34z^ZVOa_oFw+8Kt!X*1n!6%puAo$5iAfBdS@pfHosFa
zm=4DT)0ja>Rgou4@4TkjA@QCPZy=0J`r4VCjGZ!U*
zkbH@7Ww(Ew!K}I6eV^3nkdCrFd<6zTzUaV>;dG+}Q1b#sDUJDnjnyXF?sN|1Zw2-E
zMeP1NY0+Avb#P=_Eq3zax_`+!`~b1Thah5W`LAdA^`L`q42V3hcSFwz^RYsjSwW$|
z2-;iAr|>15*Z3#QCSeS_WFihx{Mub@ZCrqpe?m;xPx9qk-h4TO1Ie$5fLEh*1>zB)
zw*ebn0MgY*ETs$FzXqH9&u1dnE+hZ<8Gbz&sdR=XAug`z@W~RJo8V4#^!d$nSUMYUam6+
z6cTV$T^c&L|NJ84<{!qg`54{tSH5_nssWU)%w?KbTy2S4V_@Eiy_3F=*_~S>eaXL5
zn0a^lLomHKt$KJociwSP2JyP{49hXJ7P4ROb_SsL#~Ee-bUBemjNU&kZ*G&T$qTiuCn?qCnPe#ivnMw01uhedVP0PbF&E=P
zaCj(*Q*J|G4{e2SB@D12ldRqv;N;kB*(F
zP}&BXF#S7SE_vcMkT65p%VOOo+>h|U)u_Jt7G%VM=-opXIYw5T7N6irojmFLyzd<2
zc_URT?q8YMuzM5RqGQ!=OH_!lz?sV<)EM%5hK3+^RnB?_i4L#VUSV&z3VW^==3}{Q
zBbUtq5Qii}y%BVNk<5%{{ieXkwsw*v(7M-1y06>m8mv%-B*!i4CS_M{%&yY*N&TB)
zcdq=0s}g0(0Q+doy61~qzfAm}eVQzLw*LR)CEdS3(No8NT|l`1)nA81^#$mv1h(~)
zsDR^T&`9C!lg;2eQ~hXTCRk}%9jW{+D3Cy1ZZeIn>0J*|T)N(o^_F93ZFI1dOgIi9
zEp`XG_eh^(zo+|7J1qRbesIxmJa(Vl)E|ybH(`GN#*N$63t@Q>WZUZU305GL-+L^@
zJKf;Z*HrJDs7rb&I`=!4zV$n<(r6;$Zlm@fUOFAR>W^RAq>_vU5oj2Ie8(_+2zT{|
z7j{Ue!xON4s&lnA#!OUdbGG9*
zHXLkvZTuP;a>QM!b?VUp77E6RgMXdg=mQ2Ngpo%AeFES1cTa?$9@4>N_Ukz`smKfQ
z_1MCKmV39mw!4cavjL)`7VZ~o_I*Qq?ax;j=LdMAQA$wwlFx6Q$AoUwe4&BCR$H5@cyjFiI7R2J
zVO>${gZXHeHJ2j>%+v`JvYWhS@cBViT;WyT@A)fn5QFLpOZYk!CQk(&)4ck-uF97w
zG-wSFbF#5a$^yhuSRglpUVX<9^G?u*E#nUOl!ED-J-#
zm)yaTsPTlA4^uibCrHlqPLn@=uhcDh-!t%fzlWVsAPL|`HKllhiti3Fs*N-(WYx=d
zJf{nqe`cr*8rE_JRq{n=qT!m|iztqwKew>cSfb6
zQhxB9Lg-vhdEcAvNhYIPZ+JgEx>@si=D)RXc=X4f%5E3TTNm@JW#>IVe!jwd%$==S
zf0VBaa`i`IJh45fPhopTp7QGfU!%#=`obn)Je}7P6Y;fOD0Amvu{*o=
z10Q^RH6qCHmJlGzy~g8F944(zd>nnr?u-z8Vj0x6q(erNmP|P=z((vx
zI46jasr4?OZ}1^fyW1nsKz-BRy~E%xt`=!U|E;RE?b919d!cmI1eg6g?n<+Up
zQ=4v(^qSi51j^<~^sx$$M__$KHg$5rWRJ7=GP-9Q;B9&833v3mTRf@eenl;#
zb1O7)_W*>##z6}LC9B&$_%W;<%OTBpzQInO3}RfJuN|ph!sbG&p+XUnwLec^8i
z$nKXN{W=Db_v-pm5j2OKD?va=V7PXHX?r31O;rN6h+P4#_r@bA?lA>VT
zhwrp)TwF|CubXX;sT}qbhhfRc460@PxGB@0WHsZfHs7e6G6Ui}XnrXYgNjPkXg52l
zW@^58q9x=y3G>*M!HX*Q|9sQ30_Ev%$e~v%Q{pHxQ=kP6@cH%)Sb!fA5{Tdm7fKAY
z&pm)H8TIcctdQJ*4GYE-S)z_~eZvIr=7!8{ePQ{F$_QLv*d3=iU#<%1R?7_7^rT%s
zvbntjgSyUq_@uYs&9%^RD4T!aO7Fm_*-6c}Nl*AWl3ABRoo6tMTgk@1zaOt%`@EtH
zm0D7|{Oq%?kNdz&>!v#}lSO(_`7J=QY>mE(w^%1oE4*>Ea30p6Gk
za2kdh>$FT5w1`JVlXhM73#Qe0OUXp;?lJ2b`;ZlGm5RpCaLe=DUB;Ku$Z~#KPki62
zpA7i8Z||Xck!gYw8ZNiWezq|rPpTRAOhNH
z$RS-~pU*(3L7T3qrlitXwLSKH-mnK5mm~@cqvRft_wDg9(L}ChW(aSCLS5b~f3-IDK>`HN2!m%7Ft`b>_NLKi=P?8QxQ($^C@d~(YNiy5;Ftw+Xda|ac&DlUpNTw~Hp7c(Jv}`P%Fpv0)Gg-Mvp
z+iPM_Ie|AC*5cr|Egp`5KyX+>ny21j8?lw!?LkeO(6)BS+Et(-Nt6
z%193b|F^90K{t?q*1JGAX9xCw{dc9|g9h;5AOZG*f`&D75WJ_)7Zg>=MP|TL4TpY1
z{zy=%+fQvz^JqlqH`aut@ow2OCcInk*L}D4SM-
zx5nSJGoM6==9L?Iy^74zlLpKk`l*bw9ze!LH}YN+JLm$7vLRLSSbZA#{T{lk)3x%&
zDNJ8)*WtTWy_=Uz(+%Z~AK6BE^tvBWeO$Jc7zgoz15NQo@q6vp5BpQ6qF81SLjpC6{GCUJa&_E{9Sum7maYJO~T-Uy|a>oXmQvD
zbB}S6%7Znh!3tFGq8`JsE3)Ag8;8o$wfqnnS7)wq?`HbTo*akBG)$3{H}}eO`6F(w
z39X>7iWh1+$fb!gV*UOH?bH*adJ4r5kT>6xtL1wu>equQ;i)TINxPz&TT>k2(3)C*
zQqo;xHRft0fB`9p)A6KV>D-@+&-MjI;vqUMgB83x>Pwx&vCY+*vc~yE01;YssrbI<
zp2Lws7x#lgg222!F0%ADU$BDEsbJDai2AD1*6p=7l8?uDJMue=3VYJ!serpwSHaj+
z>@V%WS-0;^<=UA0Uk`+D%Ld4S^+A9Id#6iv)IIo|aYkk+U825CHJvd(ub242A5`%9
zz|AcrVuN;&QY-JS@g7UE@OsIX-#vrXnK<$teq_%)A}KxZpn0%f(_wdh^Woe}G106`
zIBAsIINO5syQp8`>s+YGfIomUjdtiKtun`Y9ES6e7Nsq@ws#|P*Pkvj5>J!~A)+<&
zy2mOcnc)bIT>X{Q!-fOX)%DW5mXTL-xwqepfgGmGYb^ukk3SE_Wy@>;0m}vMhbLkp
z9INRhRl|}?4r_ILGz>U@>OIAd96%9J&n%ALb_v6*$f2Dq&{t7fH<y>Hz5D-_w(vm}P=Rde3n~r?@)7<@Trod^oiEmW
z<0ENl^qOH;Y{{XTxN-DNdSmgwp;eSg#=I~4
zJpnFs(v%|esU3@h_RQ3gGg{D|K7p**->Bu6hP=?Az2k4cE^5j#)6Aut8PLahDEYhw
zY|rc2R(nWDuL6}vdmKw1n_C4}`fQc!%fb*ynfw0u`}g0mJ%h)wPg&|pYD_W+`=XDC
z@wRv?#OhX=oU(s#BIgiV7+thi5Gt^vAVY^0yE74_rst{ZGkF<5ts~4L
zK*GbTBDFRqtdNuL+GBsWxq4GhJPEoze>7JXEbMWWp3KO{5^=KhkwhNomIr<*jaZ|=
zffohWDmnrlXW#M3gtp+mzLc7VhzxrBE)(!pqWtR%ctdIF%!SGQZreVxFl#EcQW$CF
zIMNJyvjBzn5<3eb7w_Ol3x<
z!n%5SRJ9eOAv%X@REX-))v(JbKMjqDCWv}T$_=z)pFWfsQ*EsYMOtad2;QWDBfTRu
z)Ir9xZ}8abF46-IJZYHwB%4*9XrXW-)Knnd@muQzug@EDqTIy#cMfKkUJWLvk{S^G
zzGF=nfeDZg=KMQpi(KNoF&Su?o_uxj#fq)p;VEOjGh)YBvLkj5{Y+^g(cukVTpQ8H
zUXiD@g2&Yxi4^zqjwJORDS6-7!wlqQ`;?clg?r_x&zMgXmNOM7Tgondy}Y63^?n-}
zgkmMcNo)!bW|iJ4z8GGwf5v6tzAVIH^H;$4`c1u2f4rAre&m%TQThHDff|s9!Hz^o
z-sl(IQ`m79l@i?o2v07;E%&>>FiE%>lJme^Zkok9u}2u_jiyQ8+Oic?r^8%%y-&hu
zVWaF-H2p7husXUkTE*(5h$Y%>J*=tmefQ6YJj8>4keD)6C?de6S^2B5RaqY851IL;
z%>tRKtc;?(Od!du_YF>~931C+iPpa`SPF_o_blCA67eoBtl)LHXmg-_jk6Mp9z>CE
zD^M+{@>MA&<#AM~v^IV(CuH(EnHCauw84XOS}3EEIaIoq?gtW;vss~}TRft8
zz-Bx>X)6Bx%_&YJ&6xiblO)l9o0(@fFdd%AxN%G(*pI8zg+{J0tc=a(YtrYGv?tq=
zZ9yvZG1FWQWP(9w5fNj(Tl>E4wX=fMA6^e&8~u*`{Sb_p)Aa*dSH`?50}@G7QrFixpF7#-O}SZM@gVn!(=rzQv*
zlw0%pHST$BOcM_M$Yz#uyN}b^7x_fhO!CHDL%whI6+b0nlc!xvxXDl<$7>gbT932q
z`#DmpvH7Xnu9T&y1$RzkOs$aq2>k2zU7mme%CFd;k3~WE?Ocf3vZnN1Lcjh-QwtQ}R^o;M7pfZTPP%L=vMj+1!
z3u$`m_(2mAfCs8Q2NTPWc=P;Dz!IH~LOsrE&y6*oj8TectIu$qpx`ylIA1M#~D
zFY>q-|5-Pn8C`-brL`rAfQsGG9gCQ;V2XdtFovW~$jtd_*wx-?)%~Y#aU2Ac{
z`c=EwP2%y#L@S61lYUj$*4FSFoTCI$bo>QdErYNFHd(E);zE<5>ViS&baHZg`4uSSZ+t5|<>~#Jex?AVpzCrY@Zc}!A
z{VSebtU>h=Uzu^`Sm?;oOgandK>%4qb?O$h=5n20DIb~B0(-feaGcMYJvbNEbM7at
zEUzko{MuQ4+2#^^;sl}H*Q+wyEECpz@e#clSMu-WFUu~7kF?Syb#ip(Oh;ItUA|^phz-qEx
z5OEsc$3CZznxV3#%#|)!y#kS5!>axjZNbkq8m*<82JZTsfidjF?rZi+fo*qri-7^o
z%U-S*TkVBCDDWx*V!l&Rs59!Zj->vYfd;bj%~dl5}v)xA2e?uv%j
zL?l0)Z~16eN`755Wynm!@te?D=uP>d`#}Yz-K{{A$7}$pW8E4_6gz|H`*zNh&CgMb
zC(T}}(`C3!_%GFQ*5WU)??n~e=n@07jVj_6>@K){hZV~4*BK+Fk96MhUr8h-#OH8l
zKi9+Y&@4q}M}((p+Tp1w1Wq5`=}8tc=!{tlS$Za-%?57D>m~VOiAY$(&sPS|XX39I
zb>sPf3Y_V!`@W;#FnS%UICcQ-aUaPQ!>}hNQ0#_!80@+mZjAY6@~0URPoIJBL&xd}
z|0sNv3shj-w{0Av*T4AwSQ^Unm|72WMvOVvh$_tQLS@_?jhx&2#&JGxJ?dQYLUMDc
z5CvN4y@8uEr*1_H2M!k~u9U5)ka#s#*$IzUwa)S_cfq>43r|!o%S(9~4fuZu{>^0w
zw7=&n%6D}CZ)#D13~acDtwfnIQ*Sr~T)euCa+WN{-^6aY{5AHW7k)fkA5{8_eHsSq
zCAJg#W>LUOF$aGdzG~YEJl8jHFRgEUagcX;@2bk`P+=lR%0yi&)$&sq&=1eKBPrv}
z9Hp*Gf-AesQi|f#t+NrHM~votE;G%9p=|aJVA0ki#fD0*%duEWT)3C@QHBY;J`@lj
zvJN*-OKyvI#M?Xc5t}wTfGtV=-PXvC-Eat~i65I8{^T4U9cgFQ+6&6kUeV8w)P+(b
zK{E-g!{HftXJ58M-E^j_)q`)W*qxb_I0Tk3dLI~|sJ-Q>4wy;7LUkWoAe1$TWkT-^
z%Js{vt+24nc_)j4HC3+;sz2UQNynjOPVYEEcG22=G@`$+F01IyE(7iL)K9en+{<-cA
zjy-?%IDrH;6TC&M?vk#E>-gzf+k>JumWl(j>&^xu8vjQU$~-jjQ#NoDDl%HJ?M2(z
z!`C+`?zIRzJ7RPqNP|x=9nl)5sMRwOwYxaZNGqcc;^VtF6OdUue2T#4r2DPjzYv-w
z{TPIzZ$dEeuZaH!T{Cj&lnd^z{`4+T}QrI
zMAzzkiJN_5+Bxz5`Y7Mshoky|on-Mi573F2GSM^H$;`y(gDkoNA-t
zH)!Q}UkI559YMaZ!#GkWYO#@it-h0A&AR(HIKDI~F9se`n1~h9*Q^m7w|iX&QX%J<9-qlon~JTd*vH*)1p(;h_D`|R
zO<08{JCKP_DXK#Z>|QU}8JYR!g8juEbAMOn{~@C;?x4he+vl~3S%1%5I~p)|>>Fq5
zBK(z*H$|93NcAE#JL1lI(q}5s(btPt9LFJ~Ix;YX^5$2FME)
z1R^rK-!hSr{tee+uAqpGYsCAJx+27XQFDGsbcC(YpA)*YNjH#7n(5{BN)*;V8410t
zpNIf2*R
zQ#O3Cx3^Ix>G$?G_p%y5-*HQzTYGQv9Fj^hygr0*el1>zz93DM1O(-u&*J3=-qCyd
zg4eqC>%Z?+U=p
zrrY1D*d#`^ohqSw=Q&Gg)5w$=|F+;il)_Iq-$nAt(`er$z~DM6
zeAbFy;y}!)YA2dJ@|-~L*fk(34c9~-`n@is5k1sc64bwEk$GYL~^=%Dne*;o&Xazqi{%u0`&AbiK5)LYypQ4%@eZG=FOH6G!;d%XQ
zFB{Y1R`VrnA(vvt%-8p&4u+fkv7$0-9ybK0#f!e-QMT>GA&ic66_H8>os+dXG&%;4
z2@MM_C#WWWyxz!q#FvJ$sG`L1v{t1sy7)(?0X{;EoPVb73J56%Ssd@GFIn@y8Qr&D
z+=s08$y@)Lh(6z5{jMa5Z(_t1r=qq|_ufPU)|NFQx3@nMjUG_U)YF`X;c23ZehEco
z1BkfmP+{yr6^qip7VZV!}2Z@X8hru6=*gPD~1+!pSf3p?d69_ROlsyLp7tg7pfJL0Lz5x}tC!=L-li
zrbD=XBYDt;5?G2qDvxZbZR}l`Q*o?%B)i&1;x76aX+2fQyee&9u26
ziz5+3B*f$_igmJ!QneW^XBso2sMHJS9bE0_5pB6o7+MD!k=UEh6E!h~35U@8-R7!c(H2_XWFdDd#Ytka^n+fnXcqx5dG+;^7P9w4PkF?8o2e4`}G(3nPkUe)dYmk{inXJ^BjZz7F+YUx@&J
z4DX*0mxK+&-_^PiTYkTeUCEGD52ODJn6^>vURe1mY#FgieR}6n`>F#A96D_-e7MM}
zEdZfhCY}3y`=@`$<^l`*5F7|A#3}i<8lcmO;`#0EDD}HG@77ByQ))SbDH%(!(C_@i
z1f64qi8aUpR6F{oy|IYfJ^mC)d06|l#GxT!hw|(eZFUY(QaO#HDrN2%B1zIps%1o=
zsF!W!SA9<4QWn4M4SyTz6%SDVm4*IAOY`#d;bbB#fac3FC_4HAcItH^e5uNrKple>
z*_9e4xd?KcIr+^XefaaNzh>E7I%`t)I1Ek}@25@FJ@OLvSq0(f@-v`$2iybi7!~~+
zseU(gRot_lQO23secelmA`TwRj>|t2c7Pj(D08W2%YB!7COuXDPJpD6V&Z92q&iem
zykrI2(mW}GT2`71u_q=iD=9+xADQB5>W`Ky=)Pmu_1_VXIe@yNw3ybL;xXb2lX4L*
zuPJxo6ybd=1Iy5Nq)*5#0hVUxP*qtce)A7Gw3)Eb@JQvVQlA1_)}Lb9niCxyO#5m<
z7pW3bOjaD!Is4uGG2!Ia^C*28%huT4+#bcW$5F*w9mS9$&KE3KebP4*>IE)HI6>if
zV07^*PX*|`>Ss2j%^Z9+FWM}ESS12Xvh_y3^FB)%YUGS^sS34s1spi4jnN^nCN7zF
z!nb0wFmaF6(|J+`L#J;k0LVC6A*F5|dzc}YAnj1+1S$5n^#*dj*e&1u?1@6^d@68*
zDqQb&%#v><_Da5w#DP}7$42*%cB**m)JI#>!cI_gykL7Q-lSzYlLQqsbCF;a`A3P|
z5+uQg3xdBN9`%F<`ty`OjRYS7Mm}0a()gz#2Fq}^GMyQInbDjM>1~Nv;FFp9o^vCZ
z_l|j5Gh;Cu0hF
z4A7mTIjA7*5cP=&Tvk^ZXy)@)$|Y=Hx?~U$s`|%|5Ab-N{n6tPoGvd@n+cv^5b3Mp
zA$1AWf@R*cH{;t0ByE09a%ODUuB;%!XFI)X=gc31X)rG;B}sX^Tia+Wp6A9
zfJk|Bf(v^#lF`+lYKdpgYAdTK{;P|u%hesy2@`%QJtH%~k(9`T8TSx!g(VO-?5hS#
z72fk|EVYNc@kzco!~g{B-I(>-GJ&aTyN<~O}DFon1CLS~t9l+YfU|39v{u)Po
zFr=+HQo@@Gcm9k`50Trc`HFR|B*2?(
zs;IYQ1uQH^cbX-l{NQFEOO0p?uCJUgM=0kJ&;T?YvihvmK|IxD|27X7W%?nyqvRo9
zrlJw`nAbyF_i#?HxF-fYklyN@gT!q7{4~DVev?zDy*z@5`0GK9phyR@+zQ9?%_F~1
z)_oTQC+-VLsX?|QB#r1%TeT#@&nsT6j;R%b%`PK(
zcMT{xW?h;4P|8quYmZB#t1g^~1)_tJk>$FTAM+`=l)%2wYQ{aQ%4p*@dpnNG#&dp=
zX62v8eUly6w#$C#Q>~<=Zw;{V;s(C?M1v0|5j)TJyH|ll^lagcxZg03EZM(Apw$SClc<}#eZ
zVc)<91;x3MSB&URQH%KiVVNvA2I_>;d&eU+q1Yix~@8tUY^8th+%viLkF=
z7D;V8RdCpH&`SyPvM&!%OW>@s@XpzTBxiqbSCed!ESQ
zF2rm1xR)sWdNZ~Ux}Z{%2VV^L;xeVQrp8ODQI`8?%SXNT_OyHp_$h_KEBYNM+Rf%*
z0jp)jhCu->7;dhJ!BD2dtsO+3OeHPLeZaW||I~j0-K&2P+&a1X?th-Lc2QvL9)OV$
zLn88Lk;>Dp#9r=ahSD@RMNdn&i58B-PW5$9|Dz(Y^H?lmV}eL&fce6c1BoLb8xVIR
zzz`i&sg}leo~uRYwfOWT5QY-LML5E2pnXwI-g6f{uVBpk>C1h2sX$3dXP^tp3?BcMB2LQkujbnDx>h~Ni*eD3TI
z<4gp}Amvx?9R-kN4rdABn`&RjHGPaLZp+
z6iq`I0S413U@@y(=E-rPzs~{ya4B;{s|lIBG5ykb9{~3!`E(|7oS8CJD3COk@WkKW
zL^%Ccf%k}!#qI73Lwq(N4?Ap`Tm{j~84qU1{!`|ns*n!}^r@|+I{Zgo7&6pjl$u6>6|8p2^paSVsergaWWF7>CKqkhJ8I@`hX#x=V;
z+G}1{pssc5yQ3>$Q_U{4ED@VcTKy}EO)x=9aSD`?X7WnHu47BC7~$|*106NRdP7o=
zX3z2sKJnk*>)moT(if{|PtHl6Ivuzbnv^s}!kda~XDrvU$#>7guCrMr=~tVkeHzF{
zyjg(+SLXjH??iMxGD*|WZ2Gg?x6Sr^FaW{7FPWyCusR$>12#KyD>$0)^yISQxt6R#
z9G|Oj>2&e4bh9LZ5f>x)-yz4346uC&)yDAr3)BkCfC08aU+*X#(bHd5B|FCf0Nd1*
z!N&J}>ff33U)|{-2+AA4PM&^7w@+ddqmQ$C4jIDex_A4Kht?>81HQQbSHR3%EjPQ_cOX^R`W_K~8ulbZx)of33+*v7rJ=+=eO#l>w`)^n{9AtlCFF#~~2{6{=
z+F7)S>o)Lw(2d}Fwo}DBMgay@vv>_F#O^UTF!HqoTDuD7M*@{r~r?6lSp
z$}AU%g+7&CAGByq-l;0gpHBuRxZ;JZ;B$e4X2tF8xmE4aAMWF=48|mJxZGYYZ(=YU
zusY*(#aL$4%=aK*gSmWTpdM%1I?2>?z-mI*=A{kZ=aStTsl>kmbw7+A-8nTaxA|Dn
zMkXLQGJWR-+^d`NHYszp*Fh6XUcKb0P?>eu=$WR<;<{S5aM9N*9jq+GPeG^pHq`)R
z$8iF>T)hUFXSx(ZO@z!2@=IjFB#-uNmjkqzXl5dxWld7WVhAh7WZS#4)nI`
z899;)RpoHqe##n{!Q*3xgma$Ce9DNB18phEYZL4MV~-CBcI*F@!U6mdpv2P_5hi;9
z{{(PM92B?58#d1&t`gPFa&33~Jxlmkx0T96mlw6mI%vf&dP2lU>
zJPl8glE6@P;8yv7=(a^NLZF?|V1>4--2KovrLlYZFW1^+Y$U!)p5hZ-0tisq`s(64
zflR8Fe4DB<^x6|<4z?;0!TyPm9%S(t9(&Te9Av>mD;}W^sMT*WyZWbHd+1RRYeV>U
zlmKZnXp`J>noqP!_%s;WMc7sfZx`HS^6bh)hthekw7jSFjts+u@b-ChvH8199Ehr4
z$hen7y8V|?I543F!>qG{=`qF*j0);8o76j!{s4Xn{BJKuP^};JkZ?{aPgMpyKl6rP
z0BLyZM?aF=p>A$p$uqm%b($S8bH$c_D)TWPpw-55HzcdO1)zdQx2KHFKN|w00R}SN
zI@Lmy5&rn&bMxduD?6zOof`4^!{KL)&x^j5cVmD+;=1pF&gxEsYQc#N3&uqRE~OoM
z2IurtVQQFh-xZmr`yNy?B&)nm=l!aw?JoApS)#Or(?z`+WylpJ3iw!kx6PKd1QzN`
z|IDt3flZ<|VXkv$Dz(J@99!7T7I2;LwXEm=1_!j}ZzR-ozD|Q!fe!bZ}wDBzqg%*L`?j$wu
zOC%&~Yf%|Me|Z?`?KZ$cg950E1vl6bUBBSNvKJ`wmE~i<>?T_Y
z?|uzr@IERoYqv7BPI%gseRAH!U?vnD0t#z9Qt4_p8?wCJ|81mTk%fbD)=zKvY|pWC
zb@#nUWHH6e*7E}tBU1dag#SFo-$yDr-vuVqlw{6smxMxCymAuxY~8RaU~Dwa9NDTCT&&Tn
z=3}Rs{+_mU$SnS%5;lWYdo7gLk;qZjC0^Zcn1M-w&+g}(xIY#r&VlyibE=vn?iS@X
z)nty$(2aoHvn_iKE6iXrcpP0dgafI_G}IbuqI721O>aIa>&nZvuag6^6AP$o!J7rwGciWG9#kLgR7COE|3XpyNl
zj*YbB%vUQB;~B}#`FDb?!RA7S6+;r%?tPzC5y@f%+zEgU_9K265$88S@ANl_No`?(
z%Dz?NpNGpqbZW-caGYU)igiAvXKr5js?9_mv5QYHgy#82JdHS?MxP|7QP>=!s-nDFwX0o=P?owXqNq;*Zd=-CIR
zzfMNpWCeGM`}d;|WeJB?Hc}@-=qx*$ii)t%iBKf{(z{*eDj3)b7f+|tV>IQdr`$E0
z^7b;;ydonQTs&irCuJN?y)iro9WnOsaZ&|;x(pPNb+L!z?qTCrX&nx
zJNE^P3(pMsvHx_jIv>`T?CpM>Gk@E?(t9FHf#DeI6!nU@qcf1#4cis{`p$LF3*bYO
zo0=EKi9VkO6&Bz8(U*HG5$+q9P<(G5$qJa-g+rXlHJK9wxs(79h?!iy-8+)0EkjUQkSwvySJ^keLS=fdf-V!KpvC
z!lbbXzUq;nVmjNM25#svWxw&gkfzp_9~OBmc6p2YamX~fNbzTI(!&vEWN@S+VJ$vu
z^gXnlR1KJ-TLBXrZVpXh4k8jTh3^+=Zr$SfhfT*fHBVY?c%i~cGX0N1f4###|Er-T
zmS-wt+sWLZw*8|(V%XP{bG`inG0|yA3wPPq9;)cN@7Yxot>h)Yz_jO6o`|^!y@+{0
zP+bp>d^k=Y@~uZth{?6#5M;A+=bh_%;$H^gz@ws`078uaM}xly2TtwoB3(nDi={o+
zF=4)1MH&m)l1nyCx=ulV*m^N}qT-JMxy%H}K^f?~bE2D-?12~a
z+*gm==IoL$Fd1>;o9sM_18-53Eb#yliqh-{qA*Yerx(b$#DU6jI%@0`Gc@I68gUsMyX)Jr3f3^3uOv;Oi;}XF(e=9N2SMS&tDS5Lcns63^HDhun=BRb~68{&tkQpW-(shS!Se;
z>W)HQeloHt3QRVLw{4+>vf7ViTiL^vT-W(rt1{<)o*snnC9p>}y7i8|u!-@8i{-aS
z9pTlH_6;&M(c3Q?rWDHt4-VMLV-nISXP=s}=1N03q&j}#Ay4y(%%>1(BF9W63~hk(
z(sEBCq^T;qTe@t`bsa~@Vkg%>kRU7BHzM%SQZlLVu!#q>pj9H@i_}3%tDG%`
zdSLQB{>f3&okAL6LGy^&C>Il7HV58B8wn}dQc)%pt|31l_5RXF5yEuv)h$>8{>Qto
z{gc7Hj@#k*RJ{Viow>5Ma#A13C|$wwFf1gt|QQfNTQl@T+uKLC?9F0YLW|
zneIlBa61@094{+reeq_GMD&i+f7vxI?mg={pEf)bV8&xrbUZz=hz1kniu+wp>$tMP
z3&k7km;EI;Blusto`iPOH?4(Oh9|g2^
zK=nOfnx`|1HFc3ZSy+5W%5lzAnui~TT2l}zkgg^a*UJ=Wn(|aDQw%OGh2J>4+
z9zi3G;N5M7w^%`6+dmljeoLJ5>)@jM-CdxBVgDh(H*Pk*F`P6sIE@B}To{v?0Xc|@
zcrA5JoiIy35NF#_@5KE-D@oRZ6^;C({pW`#@|Dd{B402`+54&mGQYqWo4??^5YOq?
zBdqLQav&MmepF;kIOsA1^Ma0?&B3`3v@XD%P(E!-oVD<@T5aKHaH+A&oCnT{_qb|m
zF92>|n6Ky2aWy1sJN5|vHD&@BTDePd7ZT8_xugIxemsK8CrlpWp?Q&oEp=g0qlX(Ki#R!+^roz#puyr81!4Pwxp*hF}UAI`35HQUC^Ug
zaQa#-^)yCH`EAhMKwX63JzFSeTySPs#$;CsqiQE2!2adpI7O4_~4`
zeWm=4L@xOaVeYau7I94oE;qxQ;F{-E#7cmdUoOZP__-k-z2moMS@u|
zvid+<^6RqOpM%+k@}%-5vL0+pQ0=xIC{crf$qa)_p^bA%q(Y5`s)Bm?yK*=jA&7Aq
znk46_J)Wh|_;ypH!xGtVDpC~c-iKE@0}HmW7H28s(H2@PMd~=9Mh}_mV%X#^d%Z-@
z=>VY?nZ5Ag{dwA_Bd@OQQt4NZyB_Q%&?U9Mr{}-FcD9=FxV89rAd+(v?!HN2*F=}0C{y
z>NS$^ku*%sL^rFxTGgEze6w+>?VU2bJLg>f~r#|H=``C{9Mo+}%R
z^&h7&pxrv+Z6i%Pg&LOAm)uy{&F#0{c4sDLt0If0tVbXHGs3MVi)ea;G|^&az%_#T
z;F_cbbap12NWEm_fDenX1c%L$OEU18T9_#~R$>m!jwzl`>Zvl)Bjk>-Pw=y9~hL}j5{d;cG`)wiaXL+Bj8-3e*6EiGx@rzy(%
z^m=NMR>w4QdG_sN<$y*OL92J9!dKH|ffr4`tvNn?UgZetbt!AQzUZXXAVH}jYQ)jl
z#mN?^v!{VwcwZFx42-6B4TV&~3{k@f8;d?9WdDh~fr)%ju5M5ws~RW)JCAb#5(rS>
z;yZD;Ph+}X9;WqS>LtS{JWE4o397$`7PfQ85E`@M?sZHETbRa}7kfTji8|!72qK(c
zio0h`nfu@2CKgPwJ5&C0J5a~GR0zp
zam%$vNlYK50+mVXb&L`2zuH!&`6yytF1Tk^;Cuaae7a-1S$*-%p2sZi_M0^r>7GfS
zspHh75f2tj5B@mPJ*;9*BsG`yr7_NbrqP4_4UJU!tYa)u@OVkkCM;v9@FN
zAeW45SqEs{pWa>f@%ip+Ca7B{_k`8^{381%DzOO;?py9)@Ot!W8A55lO4X!HkiZ|^
zjlF$H?|Kib8`02W%2MCdw|
zW|(3qgWEi<3q6&&cQ+S&9g-}rHexXvpa8C%7vl1w^mk1m=rnnfIOH=qWo0jgac!OA
zjAK=Qoy2wUlF;u$^C83*FNaLa`-El;`}Q#rYw!%
z(sE^&WmaBOuSw9xGSKz|{<^w1_ig2}dd(%Ha}=9E<7?>B%MEBA($ZDAd~9*`4y9G6
z-VlSdN#W9)+8>M{AJz?WnWN=)*-%IpZF;jCgmQtqF-?9>VhE?d{+Z{U+!ddDJ^#$}
zFw3h|P8>IOlbGPfNWeJO&&$>D93fhQH5b$y)-mSZ!@bHMaD%nHLXPP;8_xUU1tN|6
zh+SwJ=@dwX?fzsIOb#i-p0J5;+_oXqCMo{IiOQ4t4b^ns*lV{QHWzDce;u`;IFRQu
z>?EUjlqp$!5az7D=^jm%im`EKX
z?|jRK!8;mRL0H4!yqus!9lvffR!{=#q{$wv&*;&%Pv)DR5s9x|He*Q0i8S3>0vg;7
z^4;hLM`_2Ns5hK`lX~j7rph?oD(^G)^3@NH@F*E_gs|jlzA`Ec0_|E?h9mXaUdt4;
zahQrxrS-OP;ejFdMbg3Qm2Q>)xi+`{xS2%aVrNrk@b-OZ9tiVw0@i5WTQ-nbf1
zdq_gZ9y+Ti5MFuOQRje3z|c>rwWArm_!=ASQs{O|>QC=%2gE*LaRi4S*PRYiR`Tzm#SA?zx#>Dd?g>)Y9P
zH_miCbCd)L8=@6R<2gvHL${3;>~}}0#p(Y{Z=Q&Dbr}xONv|`}%Z@Q2uzCx%_5C_;
zBf>|m>PJP&panXEFyk>hLL*T~{!vvX8Og8ea&rx=TKfT2`G0&3Egn2{cUCxBg@!69rD4kOp
zAhc+t8(!q>$b-a-bY<7(y1H><3KD1eTiKqqPnAU?j4x?qH~*fO?3DtqyEjCM>#=$J
z;12FvT(>#PXP#WiL@}YEhQ8EHc%YODKY&^n5=u{RiIe=Ggx#vu`hKDZ4(T)hvq)?aWf*8siL!Alm$5A3oQ0dDq}x
z-?1o`q0H+Iw}-dz=QIJ($$CH+Y*cP1Ma_maCA*!YRmyN=JmcU+jE3l8B^z^c9ioH7
z>P^Ox{}q~(U|Jnz@hK;qv&@~8D68#HEU9$kz5&X*z;Gu779%FpK_GFLttD`R1-vsb
zsdAs?8ojQXZADSY>E>8j?pd*lHKF2%Q=<56r~GVTyN4br>L+>3R@G(6mxF=b*i~ta
zzHQ$ox_r?ybXkvxm=fktZTDKP6@p;1ax(Vv%Dp=6(Jzin(?V@8=D&Xwe1Y@BDC+Or
zltlDHV(JBPD5BJ39KHPT8D{KnI^M^;(7IafgFvCcADEVG%gjPwH2bb)S0GaY12t~!
z{yXfQVKT*JsJy4H&v%;wyB6QBove!*3FME;657SMoJ>tQCn778DeV+*wM#d5tZquK
zSegoUvlqf=s!QKQ`kXo{V2sKtg3?So!FtZt#^>E;0zWXj_ic4kWQz5N%r{K|_STE*
zN$G>3XxvV!G5x(%8mX4c>l{&K_Cm9UZHf23-LrWa^<&|S8;B;IY>^_R)iKVUJJxH;
zu>sT`#}t{hxb$;y?Tc@_OZ^F0-6=zsK$dbO`fBM!(tOU5<_xKs#{4$bsJ{6et+i!k
zDOQ5VVA$OET|DiRZd$f?2}7}5!MhJxtK8dvlU5j~J$X=HCwkevk4CrLQ=?1@fvI=c
z;d2s|$*Nakd7{;y-#KE_2Op+U17~Fs*Aeh>{h?d^#71XD21Nm1Hk+l!bN3rWCNE6=
zstj>_KNG^HJOQZ{thC2m5td28&(Pg&&(e2}@15Q{$A%@;6Fum@xD2@Q&s3@W>=&81
z+!`FOGeA%HH>DZ;g42WUtwtP1E9fRmT_l6ZG{c#EUTP~9A7fivvniPYRWPw2EHm?0
z4tR0yVaT(7iB|1V;~fma9`qUA>1kH}E|3%$@`Vum
zVps-Q%ILvK2$qF!VM>U0cwhA-23(KgLtmQhKwt??76`ktC)eYKz~K*nNBLY%wj8Pn
zuHuQ-ieTK(;Pqa$Q+yu*{Yy>f;1+ilZ%iF$JcF`u%PW`xo3J&MJp!G`W~#3a)*rMv
z4xaOxA46kq8i2Yxn6Xzm@4wY)xRvL8quX>9Nr2d;_T-tzj!c7WtLGxMQyIagE~3qI
zOggki(HSa*eHpy)8_5Bsu|tVB{aMxKq6&zaSe7CdjXy+A`>_<=5K_8mm9ZBmE+93h
zIy&CY##(E?z*>jGxuy=|BQJ}
zgK@^DcN#h=Q}^xW!EWq@8G5mn*E9PScT)gUd#kC^4Z6!1ohGb}%NQw^))J$K{6M2c
z$>D)3Oa8ae*C7^+SDyG&Z|Bh1tY*In86fV;!-V#u&EPo(P7lJmDB%zO
ziE_Fqf+=a$OVdGDaCKrAG1vzP@D$xa;gVKCj9lKFjEK-wV?HT}iM$yb;HI*jmIzX&
z;-24i1ikv7sf3NUUKQTvFjpULhI!WTmEzK#Tult+{@!g_GnFqo{k#FL40zWRZjA44
zYs<)UT-)y5`@VP5)ZP)pcAbv{`ng<3IkbO3dmGP3$y+F0pW!@Dj#rd#3CysAH{
zY!MgoE1f>qRIM2|yD2t+J~7u-VP)lukn?8HYiu)ru5;G6;v0q)fF?5Mts6#9{8{*?
zuj|UVO2$?)DGcJh>5E*GvJnURf5TJJ#jVn3Yr%CA5O}W6IN|7~L#>^iubviW<^LrE
z9;Sc}ygp|%gmTAPJRYBm7T5zRIV;gE80YOQIaJKxmh+*EZYErAWJ6}z=B#Y}kV0L<
zA-k&_kQYC29L@IBaOQY|$wQ#frSrM%RT1A$14Wi<&#(#gs5&S(CHz?%jMCB4PxW?o
z6-HLtR}^f?P5!$~8=4za(uLP$LNf=wc?)QCJKYCA+nTx@0^SR16#aEqhEE3?>Z|$=yBO6+O{z%yG>%&b0z+x%cq#ZE
zCM&cgwc+{;S>9va2xRo{MbuT$7v}6d+kUm7)m*2mPx_8cR+7c^920X+{=*p6X+b!Caj*fPaj=ke-HIc~{QEgcWDGB;${R%m
zJR~t-y>NUsA!iaA8dE4dM(r;wX^AG`Dy3ZWxUh>CI&Y8yOH3BbrK*iXyD$+lEpF}g
z?Kb=p@+``|I;zX>pv94G`?BEqtS_1zJmV(K_s#il8jNX*e!(lQ%t_^ip2`25iz)#b
zh^VOR*#FyOXdd~2c3jGRSb0z4fA5RW7UH#JVgU=XC+?`t?lVgiZ+7<
ziLpCyrHp>KZ4vRf-cb*QHn^PPfk~k5PHvP#Q7Z%m_pTaNm4m4gCj38YWAS*q^29QgJVJ~xu6PbHsW
z5UOe3(oyZCLW-Mu6(%Dv=r*;5=uj{jU%gv%aY4P~;DV(#5ZsXEr-z#IzU}&W4LITz
z_Xd3GsF%5^GNU5a30DQNc!2>i7RnIudjE
zR7c~j2#esM68MnaoijB^8aRGmPDeFnVHckI2@^XG(pHXW%AT^hMf?J==5(~%I=G;P
z?VA#~j
zyw@yCbA;$klPBLaa$Df_fCZO|r!}|O&^S~7z#YtN4)pM)xUP@XL!4Axlz2KAaXBxK
z21TGU^3{V>rQx=QVe%8s5y&1^ykoL9sz00HKDXQO_^_2Lo*9xK8HV-yVMy9J^+Xgz
z^F}-Dbg3=~xM^cA$ZIM7Z@w0EWim{!Wjq?Kc_M&JOk3eBEW?z>^La=C!MG9>ka3q)
zm&sp$mfYc>hk@K?x8=%$$cp~@#XCxtofI-#tIqbCl$H|0M+rYA*Au%wY(tTg7+wJi
zI*E9KZar1TBEJd@3}c|7naGpF^a_7S;Ri_bH>pEo5)4adIcmqb(X&d`t$AV2cQb0c
zQ+xbBR|{pX++iU0{PZ`^Q;-M{MJb=Z2$dm!$bW+dE!Y}&;(A#fv3gCe+^@YuC-6q~
z&KqQe*A4UZ4Gf7Z
zth$!g3ac@%v!eNLvJ@y;!HTVEr-RjFi{}dJ_6PY}Y|E~{&_^#*tVSF2VWz*dbX_46
zc*;{XlkrH(`J6saZS)8aH?<}cu5(oPTE%;J-+&2I1eJHCw$Cf1YLOx=j5$JFZ8-_r
z7r!`D&KkcwAZ}z!qW@8QSxjD{kP^9e2JzKksuXm
zaIfM0gOCH7mx5NM89j|PFL`x6l_8hx_AX(GSZbjo8aX*)b9hi(xfxB4zhU7CR&`Yg
z<7%35&)DBh#(n~2s8RRtB(TE9g8%gkKC!|rot@A=OLY_HRHn*#HZR#Qd`60k*t?2&
zzgIkWDU%CDG(g+9&=fczL%iCq*~%2nI9VIK7fvBJ)+>PSXHK$_-J+
z`wA?nFzoew5$Is`X)R1Ss?QclkaqhmR{dpm*g6(jJlzJQxEj$&kqMk);_gLa!RoHs9~ho
zAH5?1!1L%0zA?jgLw`@(lVw(AhP^LXHVi}Vn&tOlKn-y22D)LSAM6Z5-5+pnmu#RSQVZ)_UnBhjjL%xFq
z(_x*QXBEmIyNXY#J)*_gFgE7d+xVZc6Cx=@1=GCdYm$AQkALbSOaktaKA>6+-^Jd8
zk9mpea&&`vMaP%_NyZbDraXIp+J^h3{XF9C+|`GD2>VyVCVQO
zoXB-$A1jIx_8~;FoX?uG4L-qgIZ6tz4!r%iwDvrE^Q&x6+Dr6lPb$>Z%j5ReR?=S2
z>}+H_b_<`0>*3sNEvb>le?#!}Pm$c9#_}L+CgwkmVY}9HeRO(0mVFKTrI1N<
z+6oqq*%4VxHE4Y(#+2cpv$)?(kx(CP^RCb$s_Wy>Wx_qA17iWFEfME&&GNV$e`h?M
zmU7)U7sJ?h|HlgKe*l|e{;6^8W5cV@dzr6lX-x@D=Uw$pmWg$`pisfu%J1^ij>qi^
z0$+BaZFqY)-AMGk?bXXFSNGe4%@a=-!SC_mR`+WpeoFOHsj}0KXUKP)ZoEgCo0KZ+nJY*jOhDD~fgC3EzDdyA-mPaxiGA
zz~pefBzI1$^ELl|*7EARm2s?KjQby$u(>QM+sMhC>SMke
zM|QO+)NdysrIac?*5Ph=ydb3FX@hN=E}sEbV@rzJRzDs~mQ(RTBq;F{-px_4f%N_9h!BeiR~+6m|U~O
zic5$bt8@QXI{M1T8PsoM?7Z!hT_9$M7di(Irpy2UZkby$la7#(aI{j3Jxsd>x7ll+
zC4BzvjSUT9AbOPftQaP|ky#*@y2x_BN^^47kR4R}PVp?&AQb(y+sO-}C^2C}OWXPx
zsTESSP#cwF_|=h%bghpaj&bj1!!#?P3Hf%PsHv%W9Cl)FIc$!-qE#<>f`rfJmy(ha
zVre#8MOS7TDJA&%osIgl|AtFoeUv4DXw5rF?~Qs)h<1p!3=AMDF`Sd
z2PXJ?N;3OP&_@5o5iMEe6u
zZ1~en!C@mah}K`xeVPbx0(?1N(@=bLmaUScXfZFiy7k0Eh^G|~K&<-~n{}R8<#ypb
z{ho4J3HAF~vQUha=-A{S?%d#SIL|OCL=?jEX%fdb;gn(yyEQe-of
zkqDX$lAl|NMk9T$Y{#k2DbmUGqquMYMPeZEvwDJ1Xn1G3gx5v)IfNOioT}I{kdK#Q$_8H@t=T^kS0^GJfyz
zJvA^uOysMyM;R<+9#mVPU0qyWK`F?uw9D~0u<7Mjp&}Ke2aP=nD&wpMpxSy$$6b#25OHqk
z-c1(ZI&`Akym=t6daal^UfGd4r-r#_@USkzyP-z>+&&{?d`bH=V0$H=(0|SWhXcFj
zcocyx!zuieWSuOiI*ukItV8a7b06`7KuP7m4o&$0ica>%y?$}sMPw4gt*A=hTTcr~Wrr~?uk<@StS6FtSFdY(!R&&l>5n|;macdIG|%c{-=m>{>jp5r%s4%&#vokQ
zAtkyHeuHvPa)idga8XPR_7x?rJKNkuQLm>07*rdlh>_oV&#gWuTxA6A5P|zEV9YD&aWI)O#xl
zBScd9T>o%)_;Ydx_HA?5*J7C5$ypbM6+Xue)I>JxpgrQxUD$*9Edy^RsT)7TM2j(8
zU7gs*QqR3mUhnDr{{4A5eto@`r6sLA2I5ZF+3h{EX`_pV
zvxdB~;qPMun`8$$kKy^Xd1h0SyY{efaVCdtxSd~(*nq6MV@S`4^J{J{4&hbAPgG`;
zH!SadNOWqIR#qYntbG#y(;xHDEQAUr(uxWz@k#VCTdYpfviIW7d;$XZxwvqBMqboZ
zD>bd7^c4vdgy`bHu4#f(2l{8?3;o}}F24cdR%eBi&~&rm4JnMvpBxaFm1aEL0-h|`@?+9XwO$v0>DmNux|yn0ATh3Us-7}MZE@H
zuW)s}-wcS+zlbSi=pT*xKVJjPiu($Zk^Y&-T>8+`9Z6iOf9Elf@EVOlk4Jo-Wl{o-
z{q?-}O_*P4#uaG`Tg2+JW^mQ;Z0_niDwB$yAY9fmsi|B$lcs>-+a520DV*Fqct_$Z
zH>X_B_ugzN2=H)k)!zx*Rxp3%3L8J^zVcjg-Wc~SD_)zZy0X6{2t6ZYA;9J1+>@yI
zreM2&ik;kei&47LC0-w)Gg78P(-DLwoThT-{hR0&BiK*syC%Ez
z&MF|z6E|v^!jrZ=?NsKxa;x;T>faqX3$NDHm~q`rs7@?xU^~+ShXf`X&I5j@s&kyLgrJWr_c6N5{I!8D_CRbQ#HEQ$b@uIyl+N~DF
zXQv}hV`rVu8QSdD+>VZyZ)QLi_L`NL*h`js;v)v(@tph0?XHBwo7diQ6xU`CtB4Ex(zLWLKugijn8iV6)+gh
ze$|v-4PF!?MfW}ScL8mS8Uh#OONxP#%A)g_Fxn@vJ=stQ!(oB|oHWt>`Nq+*x6%NR
zcl%zM@b?xDp9hD_DXont_(x!`p;l8-c#S-+&#cyj%rB4Dq?p=NkqF9{)M_l?d6l!W
z|B{j=VEDaU!{jCBWfR{(h|
z29TH|l@@9hRdUvBKhwfNw7W#0v~RT%89ouf0h|cml!;7AhmUL6q>y}
zMF0&~w-+eJl(H|&r
z+E%_vr#(7-53+-`er#IJ3RnQh>jt`#(d)(~nb#d_;=9^8(hyb2pEvv-;sI926ogF7
zS8yknBM}WDlf$&pbtT4YlquYit%Iq0zIT8=N>({p^KJq_0n5Y
z#l%H=%@JUsOMo3mDgX7!0~b@H+$3*hu|TfxImDm0DIh^kVpBuv>X>Eq_-DRfT7dU&
z*(2mgSOb{366-mxow<5!!2Lw{otVX;1jW$>U7z_J7-A9+)F^kpoyc;43D1#D$zNtI
zac?CPkN9?93SinT0Z8~Q@SbRY)+hk$_ZA{UAGSY}#`8}#IJ^aTT}Jy=3Cyfdo?_+p
z27rj$lf+)k_}`E8`t+omCSSeug=&#@Fc2kq(?|rAGhJQJtY#_{RDJn1{I&lUw5XC`
zdyv|zXDR$yQD6%wqcG~XJXwS91@K@1I#e$=K`JvD#N)w~pnCk@{5}vv*e!Pm0?NAW
zV$0K-y1M-PR$%H>0sR{W8S0)N>o4acpiW8xA5il8(-_L*0=XdO=jS)(8|&fuE&=Qi
zGs?dA(<|zPa{8n{eO0Is%~1kJtG%Q3?`;6LO|0rJfmKdhdu(#9WmWww&!7z
zW8}f|rn{bG?LPj(FjyBiw+ui&e4zIooerbpfprA{fv+IT`Cng<7ljYFzt0e`qx<~F
zN(|at=n)7B2;>VjS>Nhh07&t|Pl7z@zubMR5O_|juLAtz@(xVlYU=B0-Xj9WrO?gA
zVQY5`g`8NKaqpdFdzLdl)Qf-jUOx&3Cn|(N!iQhS1}?VV85RTpsZ?vPUS)|(#N+Y-
zz()=4ZZ0JfS*SFtEQ0~f_amfV=wnR(ATwA{qu76qFiMU6xzxl2w!~%~uu?Mwd?5O|
zkHOn#1msS24UI3gLbCUg5IhDa3*Ps)z2p*e;HBJ%LD&!9i%o-!H&CkS2JHdU<)$b_
zx{VR5meX)TK8uzjh;5cJ;fvob&qAxBYJ?%_xW?b-}}}JtNWDdzt8T61ffx8
z+*)I|qTCR`WYo3T=8qte#70N0S`ZXRqqYq0Dm2K<97A~AbV4LQ(1GRnMl7`X@AQ=)
z48#ssIzW|E4+#O(`_j^q55TRH0JFN3emy4s_!pPJsoFKL=mez96|gD#QpP~XU-V1K
z=PXt;7{G^POt~)|C6T5)E(g^fh%M69F#igX$Gt-gu~YBncW00!0KA$!u*qwDy>oMfAchwhPUSVQun3)RbYB6iQv>`DC;v@i
zz(zkv&=-NCc(4AK0A2d>1m-Dlr962>Np7`+CBU=+(u`66x3)qyYQK-McyB11g3Lp;2@=8wSGd%?WT`_oQG
z0&C0IEgFgi9)b9PaPEpEj&xWb%5yqoOT?F{j1WpHX5R8=?cES&&
zaAO06s#0sR%-Uy%M-u?t3nO_;C>luC`9LFJQ6fBZ&Z~JhLH$xX9uW}1A-Ww5X(a!)
zxlj~f6GBnQP(2=b0aVxq_jB1=CIA|4|J%TaVxtIAefm2o=>_+|=5^Zo(F#7T?e-+y
z>R=K!^+VzK~--I`!;p_w>&LQ><;
zOAz271J>7;ZE@`1F}Di`C;&{&QAOe6;8^@Un8e|AQ)9!p&Ob$+r}4O5QDGS#x>^Il
zZhzn`_wkbPN0wxyJ8q7tH@Gm(H+kag%1iM7**iY^FTgYOA*ebZcQK=Q%j_)vP6UZe
zB8$t_2^AoE|B&WsUj5sqEy|%%DrNF!cqo;8ZyAx5m8Euu
zq`}8yw+;hF`c}EtzVuJ=71jAciS9%cp8Wga?ogkmiU2DW8}fZ5M+V<u}+Zd
zyG{9w4!*vhH8A*rdxtTT-V`lNnE{Bn5GW`p51i-oy*c5bWKM;AWvYk7SFv96U8=^}
zqZotrv6X(X_s$#HkD3z$VrMx(EjyLRb!_1+fl@lK{L}&&K*<0Qa3*;KIE(uX#mTVV
zP%fRhd6S4)MT}p07B>yChFs{Ic+W3)&{=o3Ypi=w_5a`oYR`@{l
zzWDan50HBJ0gj8=v;R6A7l<9XdLe)KJ~=+EmuB`4j)6uE`5(sstxEggW`OzsQP!gh
z+cGqS+yMVWeYz70=I%j+kXWJt;bj?~d(y?=L#dGQ6}VdUW-ndOF(Nnm`6gvY|jMHBenAu=Ne8+?SvZS3FV#>EYOb2LvnGL48H=knO?>)W(Ldw
zEqavu{A;pav@5uhZ1V0zduQioS`M08S|2|Of!LF1hdK;fcf((Y`gDJR7^IejK;ZuX
z)AlUe`ENuoJ@x~XD{?j@#o)g;cf=2Ut`w;9?lf$6l^skpR}7<|lA5
zhz10poCg#Z((!p-WgXeR37D<6Ap!u
zUeZp-nt@H1r
zsbQhAv=@`8_EVEZEZY!;e4Lq9XgMU%3d>4^
zhiuG;F^WgUiE|KLo9EJ64^~of3?)&wE9+Hy%*5jt?rvA}h;LJ
zyA}*q9EVJkDJa)Dit}DyHk3lGrLwF29%zs?
z7uVlCb1&8FQ7Q`WlJh*0i};HdP>NVQ@m&^TH)buU=W2AiTxoTNFCk;r-~TYmBaENX
zPbM0%v(4JlVu+2UuS`%3_n%}X{hgU<_(px-HL9u6=aM~>U~DR
zq5_*;=2vwm*T2+m?ex!1-&$N>6?Tum55i=9zT6cQW)@ABcT`KXd)>P*XQQXk;Bk&S
znjzMD%YNV5c?@mXE&ainu>R{?5RAK(odBVW!)?M+W%<+Ba&tHOXDgOovqyka+dj?N
zgfS+{H2bC8`4MsS5VO&)sC0k7t9k`C!%%LD65I*ew1Xz@jML2IFIT0tBi7$X7HeN-
z5^4ICb@@z-A{IGp@)4DG$B)dlJnJhQ8rIMzBD0+%j*ngGwPLy%bV(JnC~Z$`LPMvl
zGk~(%n5d6TO!#rs!)A%*@OEE@Z;$jbRV{!7-AIAM`
zzNt>5*5BPto6{w<4Y}?qEeNI~gTr7VQ_fG&7s%?-k%)Tx~)gaYQN7uxW71YcOSsst&)K
zVa>`oxlPNwy&p}*y=cbFwa{u3o1MBx`7vZych}&>Uo!@>jGi8rw9Q;8+vUnI>r8w6
zY2(uReI?*S*JZQewCg&8-a>#fSBzP9m(iJ-oG=R7GcC;z?W`Adv%H!hAIed6u
zd$r!8{WjbSvG9m^tC@7qKxFcLG?gfifBeJUlLztpbVD&46}BIWKJqxHkLu*)~EgC{MA6KPPxDqtWx<1$}Y&ovD~D+K@pC4H)9^Q+^;zrVz4!2l=%Ea1Vf=}x6y8GLR5GITW0Qwzhkq>aCG
zH=6EkM#-$k<`+~(;l)o=%mbV|K1O?TOy)C)pybm1MwauRY~V%XgYR>9m9`Qhd4rz7
zV+TPVf$KxXI`-8T&vgT3{3BfoLR5=gW2A
z3ecz52|+-D(k`w|x69?D)TtsMm-~`W%P%BcpB+vb+Op_#Tk3d6F!1gZOsx_HvPO8!
zB&-j=z@qOubA^sQo)dm`OV6m$od80X)IYI`9>_u+^b1D-e|I4Wf!-A+G~9G+XD7g<
z-W`1QF}vrj5Xlaz%kN`2mEXzYgLUCd27MJJ<4owh25QfC@T*L~MGjo~Nx=m6Qm|0r
zEylTnIGi>}wY@Hsdn&#G^dhS!??rXEoH{CS@0a~&aXVr!@JJ1{@07`=F;y?G3Q=jS
zA$I1f_vKiHw~(@-u1>ciMpr17CRe$F(8=CW?)tN6!9wx-R1@=MZpD7=NvwudEzrfk
z!ITJoaQr?bsEiyFh;7jS3)J-A=wJQFOJH^4DMF$r?SnSruqUvFvu%jhxnFJT$Qy5+
zen^+i=K7NHrk<_B6n
zY)1qUG7-H`=w$;n&`8FTesvG{oQLqgzaO;XBitseBogjxXD>!{Gqs)+spR%O}
z7f42{a~B6y7xgJ!sr}s+ch>@I^oXE7NVfLTjqSx9Q*IRVVEaWVT%Q)M?zhpLmvfZZ
zFksQDA8gaa;tBEkgyXXt9yWZL9?OO@6w4MEdST$((*AjW(e2zdVMAJCL*#@qI@W18
zr1Q80=%T2Rm*yYAI+kJ|*0I?lUF|PmiwXicni(a5^+Odxq1d&yA)3h2jT$vuN2XtK
zcGsKS$_ZZZpZA+fr_Z^E=M7e6*?B0`1V{yuf4PK+g$|Kk2MK@U9frOx<}aNtV9>i4
z9u4j;Sz8w7cm3I@P#^bsk>d&$KEk2v1Krfp)ncVhsyE7#uqcZc-N&l4^!>`s2g=~HdXVI8gY
ztF0Ez;clN39gl~dug+&J{CL%)@*VvXhD}gJrb+us>VmtNck7mpxI%`jUL3o_sq+T+
zTc#Ss_{cQA%&89leA*;KB!#aADBh=={diY$It?zX;UMk3E!Xb_`>Lf>-R*o2Iv3Xn
zjmBgEb-XWH$z-J`n3*((x4Dj#VpvyTQ&T3I#ex*0KpWf0bl_NN%{|0v%>BZ}m_`>m
z`1Pu2{k*X4cEtU9j#}{oiBdGY*{nrL+V&vf4!;{j>8^wRRBI-AdC9BxGVkt8AkAIw
zw8?Q~gg(WDJmjmD1=hXs*95igLNRihFhh~+)AvpweYVDTDQp-GS2%gREnEU%e~30S
zj^ELok1j4i_KaB&-ldjdU+e5A6^83T%l>r5%@(&3-@C&fF=V{yvvUlgB!m!C$--0+Z(H50Y!xD6;We+1-F~M#Ex#uUk_73}-82kw
z(}fP<+}P#|*?BLY09U22HkT|Mf`UmF1ALHakpWFanf5=S6>oW{JnlnsnTe+xjNNFm
zf-Z?c&yh*Ag;{S413%>Td3z7?DH@aEhI{g~!wuk|Apb%r%4YgfYL73gQ)fV>6(rMq
zeE^4LGZUDoBevX?axDe-)y8tl?zq%rmh{JWb;NQQrJmV!rY`PV;yE(`tUr&JZE&m&Fu??
z3WWIjBOmTiX`bk{!u%cZo`>Zj>*wMi7a>fMXUz
z5eg4^%lK}~5;28wuKGIVS=fj1nZa-OYM28_ED(UbA+}8f={a7R;vKuZWc@XEB{G#q
z^wzshX`bQf{P!lM_);;O=xiOGc_9d3M>xx!p%MqGYPUZv|Mg1(*>
z-Ool8pxXPPg2iBv=4`gGAx!p_q7nVXAo@6NOh;EoGjlteAfn3fq%;bXYcouPo~Hn5
zpr2XnFji-i3{Nrtez?;*5#{!i)RtmdtD+*vH%^7i7(z*PJ}U;
zPXQh~gNMwyW~nQ)+E#)9p%5Wg=yD+E=P|XHiIXpRkDN<`K%XBIV}-pNM@~rb3=Zlx
zg$0j+CM0PQhjB%y0<_sl-0M`S3
z_6ap*_~%bt1<7K09y!(WUCS|>=VjT_;acYnrTeSUU}UXzcgyJRYHr%~g!Xx7WN5mh
zc84!HinJj;OXa
ztWzBti+p`}(rKYG)FT_NpY)RH>Wyv~bJ3nH-W*bI0eWlyHC!S}vagir@nTf}+lCD1
zn6oGbxyKr@*?e-EWpL%sDuwSX
zaetiVY^g#+Ew3z%W9RQ;hH|{H;HnqUKIHAcv>;L!xZ(3YET(llZisw4i){Eaf#{p=
z>*l(iRyH_i0=x7(FEM2@Y>sO=9f7S?w&}C|bIf5LT?DorvD{}WCFdQk6~$VT;rQ6`
z{O)d++`k|~B{muTnhVFXou8!(2R@|Jlr&bgVM$t~x@%deY6i{YdndIU?UZB;
zBli{vMUS|buVhdvc4ml??u$aIWVZ4w-mz_eIC>TO)hUx}y6pOinCut-q-x&iJ~mFb
zN{iq+R_Z*D!TJ;72wGgq!BARs*_AdQy6L)g60Ab)$QAnaD6VX}fU}*TvQC6d`*ya)
zxC@8)f;Nh>uej5N;5tp6CTB>c-}J45>q-)qaj2q4oQ5K4^zt-&Q&wFOo`DKQB4}?C
zYM(ezAd5fkPEKeveuY&y^C5Rwmqc%G2Z#A9j3T*QejoBT9`Cp$PP+i|TQ!?`A2?yR
zq+WxolRB&V7QwjnH#iZ>REYQO^fkKe6_>IbQ?+zF4SNQgi8AV(rqAE0(td}ROyY`L
z?)JzK3(puD)N8_`$EGI}#&aLX+#EWZX5apW{cxG@p72l|#&~^-CY#7lE;=UT4SDb8
z;hf^A@}=;W9q(Kt!$C)r`azo6dyT~~a?x?al`q^imm92orq!b8i4A+zSBORYF2My<
z=!R}Fz7hme2Vq{l+qeN%9OCB1QRI@^6;xl)Fzs>fby)iGe>pkW~);Shyn;sp>_P|nhs8})fZrj~1Kf)NAa}B6qvs0ilm8zhL
zfdgfomN`?KQPT&n(&lg%JipS*;*?*kksoXU7uyM`Q;Hkmy`R~(S^9UAPpM?s{0^B_o)~*u-
z&=(yvvBapqbEvy)Jc}o;4>q;oZ;9R_K2o#JlECTLTcCI2YhOGy9sMegN1J`-hI;)1
z_me)0Oa?^|rpy}3x5&Waac0tq#u6DOs2>act75^Nr7UR<>TziaSTNAv&kO
zoHtZn;lF^kPw%HF`|-xG6mKs2WEyHV?3^V
z(O2`*6EodkUu{ikKIWr48x3c>e@o)X3LL8_zU_W7hd35WMyW-|FiK(<~4=c+oB
zOC<~giJM`*(^q&Zn#Re1E~y!m0g=u4-O1ffcGehVWd8|{y;yxls#g!x(4&ulNP@+9
zid^S<*{2ck9bf{wkWg=
zbw7;!)?vQnxpJ7{j8qIwv4|A1(v0m5`(>b0#@g_e+C=`6(4w=JX|ZM3!HJZ;Vir|2
z0>pYFQ{f=#q-w)w;6+}~{{0-tmX~zRH>~YP^_XE!$*b4vuyls6$=Df)m2dS=bpG>mRH4N$gGN`
zch_F~y%SK%Dqkk~!Xb95yqP+i|E>#a`5Uj~s+&vPc;q8AxKm<2|EA`E@1$Dai(xr2
zDix-7(FwbxcfzHp$Q5O<)8Q1{m;WO)7~A+O)17ZfYND-b7Y?pDMkhj40;Pe;2w3Ss
z;ny*e^6Nw5&QJpO!MjY-;nm)^>ms==>fy|CW*a`wlf!fu1?Gq^F3uw{`{R!0MV^h*
zlHJbPmcq*K_ix)5=VZ?UE3svHPWfhraM&?~T3mJc_Gay;xgh^19Q|KKW-C9Mwhy9<
z)|$MC*`;v03~<%3td=klB>kqc9qc_H
zAHR;9*Yl#pxYo<{))-vaE7QlJse~W)V-?Yx363yn44-=t+-CZ+{#l<&dA(Kx#${;}
zo%M*M;+ZzT`~xhpPyTZ^mXZUWC1_B?%AZy(pZS1+t_N%q89newyZypZpkUnolF^22
z!iqV0IpjdpJK!!$eY;z2hkn7&-BL>X^A&ohtZJe8d6QFZ?VF~E7VPj+mmg6W_!_e-
zq56vq_`BR)gn9btJcnLA6Kka0&d7>;$vj{LD(4m5fQ$}3y{8-$*+(J
zZSh_K%-9Yo!brU~VF1$DI?Az25`qd6^E|C@e3zfH$~Ty%NEkUqur_I}x-*<1&8IR<4{wQdHW
z+R16oYC{-+Q>sQ&K{Z}$A{FzRkI0nFn6F+o+kzsO)f!l7o#uZtNw(kpifyT~;IR=?
zZGR|Gcf^g@kS`>W$OJ_-sS!B=BTgm;B6PBt$NN*YArx17clkW7^_+>IooWB}Y((sU
zQmfz3vq%5IiLB%f`4dXR!p%h~a6s{RP3n7k9!390x04&xHC)xD>U@b**!JKkrEtGm
z+^TH(!G(^IAmHLU_9k4yPg0lDza9I3SMlY;|E%UW;jva1HPd-d6xBA0ih!81cwHwZ
z-62{PnXcYmm@2(?g@n_C(6%DeK}Xlj+_G6oiZEhK4J%AvYTt-2o8lj~Wd77K8RjNW
zf9Y{mzBZa+w9Pc>TvFk(eRkN3fP1dwfk6Wiu>q6inp@0Ur&2O&1k0S>5Cj*me_WqIyyI&@9Qk$OQ91cb
z>cR;RYc{(5=fdUz7JHx(C{+rlq89b0j=Lh;vak=$ClEQ{&RRhB)t1Z1Em)tEn_hA`
zj?BR7HXD2~1?!xk5pwx&{^iq4EUAWPI%&{AciHA9yJ|*L5Uv1ml7k;MI$uf_ov=(D
zebQI8;?DiSQ)OgSuT8fPZXL(9v#&LRej-O!62l3oyo_sNS`JWsh9xnX@*6w1^u4??
zLT~j{{R(lmm=6WD6hGtLZoS14u1w9beJ)0pRn})+TH_6ey2Mp%4g#pJLuT=eFKWN@
z*=XA;c&jxpX~pQ}eFCK1#;@>Hf_@!MUcKk3T_n!e3Us#Dck`qt>Cv?nSB!g_&P|wB
zVX+C)O~>~E_pKg?jH9(I}d`YW?g^5WlZ*T)k?UHg{H#yws(w2CDe
zRBxQK^>$3|;ye~}3fQ4#_&T&;aGIHhr#xJJ9Kg_9{#!$l|0f76_vo)w*V6`&HC!y3
ze|sG)dnTV9`L0`$36LaWM_OxBkn)Li4|=ElgHXD$u|HFM%Qa127IuY@FQYD3nY02*
zYl&rtLIE-&M+p|L4B;TrS(#7OmTzn7N;`FpGZ~CYp=QqMPSNn@qT^<|G0G+!m=<5u
z=}Q7A4if6&t8_~f`w6tng5Gq)F{jFfqzXPS+jVQz^U~U{NU450vN9wE0`PdM?8Sbx
z{4x#k%#RH#{~kIj#~aqqcBDhK`+x*H-(*Y7ZMADC7hqM&pc=SG=!#WVY)cfniec%+
z-DV4M$=(%1ke^{B1%A@)pOpHh8eYAcO~uAAU5xf**5mUqKRD3;^!;W}Pz$eGW$uQGh<0IMkLs6YmBOQsDWQ+~28>lFVBiaR
z!&>zVs^D=uZ8Ph~^C5OsEJu5nOQ+ob8iPA%hB$OhhT$dXxTLXtx?D@6R+&!A%I1^q
z$Q>ZZ9-(dde=9iRck9;CKXX*$lForq{
zxC{)v-qOkOp`~5tWKL4|sI;2@!
z_ecl{#0H(MJHGNApq~5YhhgTrN3r0s8Q~)vPx^wMAm)$rY0srH=e1>7V&pB^s>?lo
zrK?S`xK*39W&S~g_dV+;eqeRH&GzQhCZCfVDEsSqmX5xM$Fb7kj%4f(YM7;~CR=qy
zwSrzI_31zopq%O%pFcT+8k3JYwT~KDC4|a$C_I1mAGd{P(Y<*Eo#N7OVNBJXI(!b%
zW+e;t=0KN~Jr3K2I=p(PqG2{}d=|t@TdnsscoZ=u`kkb!e&oIG8rEP*lfzk4Nz$(g
zYd6{;1~)Eo|B(biGz&IMG!GC1T^76Io6B64hL~)md}=$o$_POR(B5#+|0RrH0$uR?
zg3Yp|y?{Wr*{9H48SnYo3?PRv0@a=~Ggw2S^LaDiJ#OJn-yY9L3vj!C2+N(01&S#Bq)gCI)
z4`o`VHADA-DE*UYZn?5*&o|Y!Jdbjz6b!tDR@M=ahepb@X&>ya^c=Gmt~Os&zdQjD
zwB!#FEB_?~5^Dp)5}pd>gsC!~WY8bo)(?x=S_(|?M}@1Q+GpY3fznlySi
zz~_xTG7V&iRhIWcSvruCBcb3^7004#_6^0`fsIKLrSSQj>Z9d|0Q+`YLgv|RPuZk<
zkB|Ij_!g30a_gk0T_n|h=Tz-yAoo2;^WN6{b%OQxC>$8?vE!eK&Yzi~L_R2<#PP^T
z{)JNS*9^1~9dvnRMhmU6%gQ)*ZmFoX%0z@$Im1eb46eKE!&UbZgo9ZQyY%x4X64WL
zkBjemNn$^+k8k`J`jtIKf64m225=e?YVgmI;|dn$yC$9fQ=hm@85Ozzmpc{;&vJnTM)_F?npsdFsZxf;dXrx0P;M#Nm+4qs#d%p
zIGRam{KVp%k^X;D<5xPrAwex)ZHVWp&?Cjx+1ZzqC8&?U>NElzRfbEwubSQCP=4|=
z$nKfG?=h;g*QMw7t+xKuPqoLd?}zoQd~;XFUN{h~1m+4|&Czxwbv^F)&ax5ur0#I4
z3+q=|rm`_{;&bHUmz(z2wUg$*|M1+a*2KnA?b}WxyUTWG%QsR!
zh@LWp?(@k0b<}e6PYt4Ovi-u@ws&cN{m9H@Gf8h!VRyA*DNs=U*-a^d4v;)U>)jB#
z)o|8&pIA=vSqgQ6koq^peoN>LxcXxvZF324;_;0xmM7=8@J{X8K%%kAwy-J&v#uC&
zEW4IBUPSXd!J(y?P1-R7R~a;U?M@z6v$-J<-KbeUJpCTX7#IJK@)Vd>X>KyI==a
zY(a4>CmNd~ZMEe%!=O&qJKRjlQCvozES@gkLk_3O%XUnhz4riw5uMnAj=da}w^#5)
zJePsCgEf{K0QYf;CvIEdFzg@7BL5>6YG!jJChiwi;nP1KC>6ek^gs$6)D!>xJ;TuT
zr?vdBJEOxY0H@@!$g>07#*C
z*6Obe>mqXI)&=M9cn}T5z4?1PPhAfQ2sXXd6FWq*yQ?zHF!ru{-_mAF(&MZ#ps~D7
zz0C)z=fu>3U;Or;@v?%r&i=Nunua)^_dSsgLew=q0At<5rqL$P{_GfMbe5q-
zT9r_i=N=x&Ms#X5omP%WuviM8*0;Gj&X8v{W*0?B-JqHX@w`mB{@bCS0+zw!xYFZu
zyZws;z{`|gnxE_a8JyHx9+|>bhbG|t8tCE`9PeNh;XurOzWG#2*8p2X!Km8-N+!?H
z;!w58Ydc4H|Ih**8>v6}n1vpQ@wZ;XvZ6cy#d6g<
zgrnW>c&iQ;H`q_j;oiG^5L8)=pXWBH%%lg$=&NPw}~Jd`G%uaTwT3C+&TUVOjjZEs#b}txopL)
zosIatn0=8^V`h)@_)VCxjVUGxm544?hu``{L$~?&%;C0|%Zsx0&iB33MktpsBVYvX
z{pYR)rsO^mfIz7Jz(j*b#ne_m(Ve
z7V!H5W@r@(hw7+GXltt7UY4^zoP~PsYAMKOO^pm1h3HPte=?RMPD{)D?wZ}xx|t(w
zRh0T+|49aEkkU-7(rKw5we|Wb1dG6n!`qd(4><`73&vlq|3>r(A(g30c!7XUo1s(U
zaVjE~su_?eq754PiwWtpt#CZ|B=p$1Ky_qbkF;CCXR%01g}G
zvH`TyQ}?eEl61st9Q=~Rt=bhM2FBEDx|u2b$-<7dGmWJG?Zc(unm+yt%NDdX&(d+I
z0f*`c?bl0ZzpL9T|2meue27XSesd1T^oDv#Gh`>S_o-ad&^W=Od_T^;z}Vz95Oy;J
ztZ?eBL)}p`G|Nm&pZ@iZsjvHE`&pX{^;7GN!8YEPHgHDO_HBCoXUxNb>wO_}Fphp1
ze0yKMq)@wALc+6S;l17rMH7|nrkAU=a+L_Pv5Q1O{8Bg#ZS#|X0i3YH-V?A>3KiA2
zyq^J4X8pOvg~*w75tKC-=(Ea$UE&3}8y+F>UHzurQ^TN6kl^p_%%KxGA`*f^+8l=m0w^jC@2?s2IM{rVb*_-(P<{cBLOF*h+Z
zB82%GaGe>|b{xTC473UOS(fzv&_tMVUdQ=JIy@zoZ}{Qj
zOP2AX=x@nK+2*c8!+}u00K#`n;UllI!TEgCZKV2HP>l&uHb9g_U=ly=TBn6tu
za!m*mX|nb7blhr<@%uYT?&f>~V5}&$1?mFjEH&mi;QNDS1GA9EhEYl*jK2Q;4R5f>
zS^mg$H<7n*s>bzudhs?Mi=bSBnKJqL(sg5QBnoTLZL4kVV=io$<^-XF52CGHPq<3U
z6_sAJT_zJskgvlF+{bJ;g)NL(ac=uk-1+R;CKzw1Xe^(C*g6b0p6IYNPT48lv6=&H
zJ#b#rH3#hAVW-3o{w#~fPh=4xF7n$|n9F^K6LVW@qwx;bUFugJFn<3StgeScY5hej
zVUw@El)Ho7nHFo(2RS=bgZ+#&s`=cu^R(gBNEbD~4VC6=b7(<;BMpt|OZL<7@Bs0m
zysnw#K;4@C#(4}1NnB`=tv~3OFPkTSlu3giXbd_I(j|%&+xm~zW&tpE?>nJK2=}n{K~cb-wbJQkXZY9Y1Q
zJlv)LNWl}aU5ogBRxd{N7A@z~C{b*x+DWagPiCz558+r3>ql?)PUSfgAiRswZ7Lo%
z3FQw55Lb16E{9;Q+dTPC&l8Vr^^(@MWlezWjNaCOLS!e=edUfo^!nSy`yZua?{=EgzLe&rHH-k!oes
z&aQ6pe3ioKg(oSN-pfK|ub&Xh&VG`4DYhVBb^*ri{g^(QJ4*EaHu%a^O{|+#a~%$B
zdi#ZK3@UP&KcNBa{Tyyhiwc9FxY6_h+pow~vcrhCIEe#G_gf5Z(2|K7Qe^p5+{|3T
zKS&Th&``M#i{xwE%@k|2WP=#^vhmQGPn@Rpc1%B}a+xJ;AH0H!vVLG9l*Ci2nznxz
zxcc(l>(9MLIp(u7ao1s@2@C*N+{r3`P5Fwuo$X8zpJ7k@_;QDt)+R`RKHJNM^egYe
z!)BcCwbV)$DLTC-+mGk0&0&GvYFl!*%|Jx+QasjkqY!yoR$M`>3-%dH~98pDKZibam
z<-TvyNLA)Oa=IMkV&k#dln&qLJ4!_DuBtvs8l9ln^#1c+_oU}5WwQu~+%z=G2eZ@l
zUA7%YS*5TPTehF`E36uGEd0BYcNE;OaUlKPUj`7up-80+OeVD)aPq$HBn^rfg;UFK
zl#d!anzdM#vC$YNpa)GQ)>EOxCm2Kv%@r$3<`6Thzw`@75+|(YmdGMMLu^VBmsHSE
zABoR`>$H6D9nPKDK_yIFnYow4}i)zU}L
zeDx#wD{A#6>zPN$Azt~3fQ=rse#?vjX0Irqy
zSe>CZ{H043@y?W}D;;K0zg&OI+b;W{wSiw01qfOR#*%Rg
z(zy=gJMTBHL(>3qi18tlBO#M739v6PMu0Vjq<4E(4Kgb*5ey+>yQrRQq@<3
zf7^3~&sBVVn&%qJrpva^V1^+Z$vXwhHOMO=jRQ}#n(d|u#12iN?pn^k)WYP;)9-{D
z*5ghnQAforCMGiStJ)7UW%yIU%{2h?aQ0318g1Q06uTc8*BaFJE1Du4rgb;v)%DkA
zsruEF=`MR@(e*JGRbCUkJd8e3T+3xv3PPW&ZEVX8LdvC0di$0L>(W>M$h2OYBBGTo_sMf03xE2OpG9_@?7C{kXT}$L6m+$IjieaEp}31
z8K&|LHU?4NUvP*%oC6o_NUGsA^P_Cf-r#DJe$nGaVQ|k;=M&F!l6;vVo1NF|+panm
zRKzCs1JvP{>gId29JbGFlKssmE%oA>Wf;o
zoJ^{HS2gHo7w|PjcFUc+d)CRxS~V>XdvPu_ALCLr!ndE}!c~$HkRSfaqlf*bInU8C
z_V9O?23BkYcccRPSIpnm)Wi|RIg~rBqH&3zZ+q&vV=)!jD~n~3JMLh^mpE1lbWV}1OWAv!6GO}(3klNsg|?8hq~;V7tc&N8?2M;_
z*xXiCX=%J6+;wT!fK^Q_z>C1SQZ)(ruf!D&$Hhv_j2KF>(9|EA*z2(xQ5lX%HJFF*
z0#H}G1B#WA=QE|hGjAsoCltPGZ8Yh+_U;x<>c3gTb*d_WQk6}c#cGh^DQgDEG`oFX
z$K3Zul9#jQVGoTS#QG%xJvh*O-I*!nI0G_SqV^;%IW~7vd!>M_>&T=rh#GA0RuP7|
z#mBXLrS98Ja)t2n%6MF|zhR?|_Rr=hf~5+6wKI9hG>_K~9yNYk713S-i7(rd_qElG
z`~JH?bO!LS>;#xhS6D6@ESsL=-Rc%^HFGK^AwPorGhEn!=;tn8jsAlyxuiiiIFL@R
z;dCQL17K62S)`}!eb?@Gz%ZcEtSj*{rL@x+DV_?NB4r=n$kt%5s
z9kNQ7yD>*s0f^Uu@K0S=1cm1g1;Z~97AzkE07N&8HuZsgEdAvR-iQ*OkHW6Wg#U)p
zlQtOx@zT_~Dp?Q|UgH=>K74V^G6W(%16tNvG%JWU6{%v%@j(r}RAs(;N%v+68;5>ULDMU!fEh+(M8!SU-OInSKj)`gf2PbBn883d$%asya{t+J|yN^8&x))
z-%#}F45eLDr^u$t^^v7svnR%4(nTmVTpJoXJ!fx!u4`8bm5||6U-F&n;kUU^*Q&x^
zY)S+h@cwG@Mzkcu{A|?qMsgn{Qb~AoCQ2w9KAxTnd9Aa^;x-}KM`%sl344~(_;OIE
z%~EyRdA;?8>g++m$Fq_Qz{Gh5)yr%yOxvQokg@
z!)J}mzf*sDM-mzG?NWYE1z>4r(q=!}JS)oGSU3WH6w|_xP>Lp#n
zXFn%Iwg=jkFr1#a+!yp|L0P+xk=Q#IgVp
z%(fCgvCE$}RaRBrP=cLRZG5QcHheM0Uq}<$>6ecEsCP6TUrfeGvRmfGs05nQG2ClQ
zI)COzZa007;5bw9E9G6NGTaI8%9{Du027w?legp|*_Uj%2h#R2q_C?-o|UM#<$r4C
z&6uP`L6+k?J1*Q|s3YPVmk%PcHKNq%WvIT-Xv%)S#b1eQrItp9_wrswYlmfn6P9sNK-vQB%x|{
zwA@})vB(*h&1zFA;}@<}LhwDMiKu42V>L=OU$lq5)DmqZ4Cp2cMZy0;aWdlZqpkC#0iPq?D
z4d4f^ZasFvT)qDedSKD2Jr;YhW<4mFB!3avQp1?~;J9QBz2{mjwj*fT@T#O*bob9F
z0HJC*(b~E#K88@GYsE+8uf1`W27PfhFW{0Vidpf5f}2@+OwOy<@#ru(;eqsuM0St&
zMSa4dy8IHClN3gGj;5;>H19rqU^NJJ4E;C%KQ90~p*=A+z+_LtGbZ7l+6#oQBQC)d
zCq4uGV;wZmKY@NXa^;;JrVXH{q`UK@F8dfs&}81?LOmUjOpyF?llPR~YsUXk%+o)4
zNPfZv|Boyz>muUwYU#D3Oqls*xv1MFh}i2%;~yh}$}eCzYDA2IDzgBQHA751``f4VG(4BgZ@
z99-ORzcZKHs>8{1C#hflj|hqV;j3Lt$c>98S&`82KPZ(>y9pfuC$cBd0;-DZixdJ^
z*HY9^fw~nzS}Ooe%{yhcGPkXH&{uh>qT1*RI@nwHCc5ZdeTE^%kK1*sMPXGe3Ho!0
zXVS{utNsu}@DiBh8vtp^pOof0k?tFpou-t-MLzC5uFu^zp+gdu5?tv48zW1|{?=r<
zXAOn?kZe2xfY@zE0Pc#J`Usv_Xc2U8CUx#-0}@=)>}q<=26L$4*6v`xhlP@NqBE)6wypxL1gb
z#>&4Nix;nk4Uul~_Wvm4Gv(s-!kWh3Ldo{m$3?X0iME%^IMoJ^lKJbD+5B718rPS1
z^8LmI4oO3mZ(3h2_k6TLOt31t?RPoYBOidWPaVl`m8~A`yV(66L$N$!u_Jq6E^*P%
znkAFh&u6?2M?W4~8Qy)Rm~O6Eoh`*2@QJXrA}+Es(4y%F`
zHGllWmumcm%;*q6j=_TTnTQKrQJ`I%7+IM}yXXDwLa99X7WMw!5c6+oZ$LTvIlo(I
z8+R?~bEAji=tKAp&t-eJaK0>(-sl1xpYUYQ0_ybecwmXpNczM>TH4?@?8_7*3cN!U
z3wQ0p8oN)wR1sf8?Z(1UDi;&W$dtSeZcom?hZ0eO!O4&754U|9C|AeQ#*z3ZYTGyD
zUP+ccIDTQ6)I0YZlz7_YKdW6&uTCCl^%>+f30pt54_km(*?-p&?S%|*?sF!epjiP{
zYwFP@12$enR*&3k7p;9K4_S%1zSuCQ9K$obG4{)C4p};|H7C^VuQJ4snhoOmSszwW
z=+wbSo0$P>#Mu*dZ(N+Eg|^x#4cbh}S!a2w-#abs%3*-W(PI|`wt}7m+z!Bg>$))9
zTT?-XKby1Anx6{htu0319DE+bu|$zz?kc-+uuj8)*={(6$J>sRo8A9&RWXO0x1ZAf#_E)3wv*Cg
zjZ_wOL65(kpm#q3oJ+y(U+xNxAJ4ws>BBrZ-BSIha8ML|dA>!zw^bU2!{oC&7nbTP(Q*vW8~nf~`?R-0==!x3&AP|L#8We6tMpH=Y#XN#A(xw9&^pZTXg~
zWFHXbXD1FMUJ*NWt)~+?{{;&79e_SfsIX
z7{oxtwvyNfDM{N<;ax2XXXrC_AHQgRM)2CmI)c%&ByttfYbbg1rF!&fh|%BgGnHTm_)MFK?zGgKS}b?FRqunQB(
z9d@sgeJ?7kP69Cy7y7yoM^{|)82}40!K$j(M?9bkF5ZC%CCmKzv@EdFl=PY&q*Bf`
zM=TlMj4#N{R7naTt?>^8q|4IWYOvvrng-Kr*tWqRnN$D4cux{n%0n9c+Q}ERWU$n4
zCQNjcV14*!lyDT;PNe_XOn!%UV@ZM7(S}-Fo$Z^6Erv;uwfmDpHxHptXg?m00?)MSeEM2
z1(U|e-z9I
zxYlT5Mut*O^APIEtTe0Ha}|_rmlF=2d*X9OuZvVM+RNI`z)es8MS)TNx3rv)_~Xx=
zlsWk|r1>f1G(wJHGFXW%a+#mB_Jo;nMxrwBjHMTeFN?NsRqWq~H(_w$U
z^x-|L*-+&84ixUaI}*z*c_aO-^dSGlA)@F0j2|%>_%w&9b#^=r2V>LAjq)R@FvAF#z>Wndj`RJd2Nopon#2kiNvat*FN>MB
z**C!s$#GC6T=Tg49M-dU@xn~@hDnY*4R^DD9Wc7&bqFR8
z!IG@@d6%%eJlcw_L0K^11K>dp0se_tb-!<4>uJ_4K220+igcU24cI@)Kxs(V{|ixP
z&dc+dO%25bQVlR@hp37c?oQ=S7i+Vjd79PNQ}GF4;#doXQ>bDsaj=MwYen?u>*LK-
zc7%Unc3#Fg<+uONpp1)PgL>41^BJ=DA{~|c)OqC`aT+X-Ays1~_;GVB74&V(j>b}o
zUiQ)Au{)yz@D2>ux-unF$ngIIKrq+|w|D&L-~ByH7Z6r;Yxp
zYq$J{87+BJ8A-Y|(&+xs8<5LF(F)hT)ubry8Cg=8kc>*w7OlE!qc|M93S2u4n?{GzG6Oa
z&h!xFyN^s^)fHJ&;vVYXg8E`ZwVcG$KI8Zx8Bfq)rE8Ybv%!BT1Z!$xccrxJwBh;J
z(_tRoV-U@yOsvhrBH^|pq3zDVnWTm5^F=sEkMgQr&lNM}>dj3Isz4^)yGL(lx1;xq5Rt
zPiU3GjJ`a(T2}Y42QE2ja&L0lZ2*f;>2)Q4r8}XzQ%Q&dx)O;GdoEAa{Wf=g-#wf(
zV|U}J^jH&JaB?};mUu_c>Va4~ezsPcgd?Sp8!&zLMkC_IUcD#!ah=a$1}cJAwO)%>
zd}}G`8kjv`6w4&im~N2dXER%b3A#BLXdY$LaNdK(pbCu<_@=deLA!r-Ijr=Vh!tBY
z_(AmG^FhEQf{PQ?gl0A!dXsLloBdAqe=}3B4}({glJYD=RGzFJwcia($82d>5yA6a
zgs!Lk*x+CK=87M-f=|mV{wBM*K3~}_u&VFG?#J+_z{2!`Pa-3JYqO*#6e7C36z~Bb
z%-auB9gZ|mP`8INkyAR6ALoTOr
zEjXlekN`?4_sTY;qAg9IX6w$1t9;Q&sX0VW#gVQ$-Usi*s))=ZwbmTY@GFmJjL2V2
zaDLy(U34ti8WK%Lr7Krfw?CzVpBoCRFXBrC8GZtwlPT#R`X5VjGfMfNWA&oA3muio~
z;y5yrKh}JkYzSrQZt}qBa{-?i#d%i!9laD~f)_iJeK_ejHh5PC?Aj%LwW3$`jn~6<
ziDLiWErd;$Q!Oq!3A5|Q4d*q_F$zkN6liXP1R4QNw3Hmx3&A2Jk7Njtv333=bf)Lf
z8gRag+qJtP%Kr3mIgG+a%%;+><0f5$8=cweuV(A1th8c;U0#etB*z**&o-G443c*SGJ+
zaTtW?L9lC8&wQPG+=k*8E?|8-(qrow1%gCp%>LS-_$-Q!Txb=&lQVgBvn0~%;&Cjn
z9W4|PyAV4J)c#>q*_F>~{lO4;Ui>DlR+%^271!iEh*Ckg==YE}Xtq|<2u8iUat|S=
zp4{5E?6nQDpR9jbom=3lddQ-SNA3{UN%(OH>cqpbj#e?T
z*_`C@>^g(0<1G0Z{$qFEV;=xn5-Hr=<-_CVA^n4PTO%-SZZC=^F3oRPoW%HR6fGMb
zejQ~KUAx1C5SjoLO|L_AHE4qN+<$VAk?Z_Ebc6Nt1@!=X$}|1!1pcAd51a6&Ht4+(
z2}}>m-W{4~Bm=tBOn}8+=#XqUy6ma-PU!=2RgtY?B18`A&HQT=PCJ|HYwtpq(nad&
zb68LkN~@%OO~x8?4cKtQ??i%#n&*9TL&f)}z4O^aFN|a?9uBTQHG29M16;ULITpQG
z#2*oRda-a1dw?p$1qlf~SBL}(1B2-OBfT|c+y=~Sn6{sJm@?+f?4IFkEMh)6?9A->eqwM^1*#kM$EO|I$NJ>F
ze-nN<2Oku!A~lkvKqIRoMe#=>dg}yCdR)C+u@6M!Jlu;V<9B=F$n}DiX1w;uS`$^xk@Uo
z`dIr@uuZ#cSdPMZ8%A~yil4wUm0L@JI;d|PFT@$kL8r)DGXol$?qfw=
zhMB9%851LDMh<@YknZ$+2g##?@$v$U`s;;&GEj*q$h<9;#ig4^J;5@cn7jg;>ta!E
zyc)tv)e-W@>Ns8dP8b#s0gYBBnBYQ!i48V9Ae
zxSt7#;)s{fT6;*!SU=*>7fAkA1&SP
zh+)*=U89Bz`r0F@$YOpkV)YKmk2zbliC>=4sXnoXPjsl;Bm@lTwluWx~$GFg@qO?{c9Xa^=zIl#u
zX`wml2Y-XhkSW27p|~tKx|(Ss_gs`!U#c=D7sZjb+B3JtT#5<|Vq%SMR(OVNq0DF%
z;J6X^+Ec~1Ypy~LAlW01%LXFG4UzJyWNqr_L4P5u)-FRp<>K(sK7Lp0)AG2)9k_qq
z4P2Jp1PK8klhk+wk5{dvJnYO{?$s&-mgl~h!yc>B!b#WCoDfIR?Xg7fB^b8kedckud~ozKAJ>zz;5EEY3+*l^zMdAmi?6}7`j
z4j?5IMoj4_J7v%S^S5_-IzRs5(6QnEK*i(x;6{^F`j_y>M+8G4BqAYIAi6vy_r0;v
zX66$E%-K}Ax3!|vbm^CWtcFDI%T3?fc^Xwpi?&zkyU8CwR
zIb*M4=`j0T#kXk|3)|8J8Q@?rxqyVh<>ShedAL7%(iw5CO@eDwYoM!1P7rkrEwJva
z6t9<8e1irH0jC=hkSpA)5xYG!Q*BEj<3!zKo(MN@t=#s1aCGq73$)0Gq)@V=pG@%e
zfsG!f>RW(3V}j5LRw>i9s$UKIYhM52iraPsQtgEUI6&qF7q&-q^mrWwW%P%%U#A#h
z4J##%79aP_D4;Q*_{Dq$``km~u(s}alX%H@%N1Mwo#?kL4NdZE0zzF=DSl`V9s9)u
zhP&sp_^%Uo3#&v3LNC2xXes=_Rn=vxxk*FIisg?u-h=`2v&6-BFN`7n3=Qx**`G}M
zayE+Hhh~D3eSNF6hOKX5#*2x0zj&IlaNc$
ztCE>(y4`OlxvjTQtB{PQ>CbG|I2{qr6n4og7GoEiMrUVT&*_Z>!g&3mHtB)oV+=Yix890jB~!`Q+)0*5wBvr9E&|pI6rM+
z^pA19Zs)ASR;DtWYG*^#w&J!fd`PFkJgDvc0Z5NmSEab5H6ILXq^Yger`>BIBst4R
z7*`J_L-LBTAO3^sNufM8%>OmQ_R><>sRqM9l_W5Jp7p=G?_!E2l8pTILNGolz=DBM
z@~C?g-zu)=2230t*P4fa1`I77+qPS##?zV*^f%3EYx-dOOrF}5LwqG!ok*r8ooV#Z
z+Ylijazui$Py8(KXTjjwl;pq9;umO8cPvNKs4gY+(RaUFum?B0F9V*rpWVb^5{c7+
zK-FUirN`o^+>^mf$6xQ%W^(EXqNQ1Ci$fSI1qjb*D$m)Qd5i;{B#rhQh=ME7#q(7n-3@2V<$BM
zI;9QWADeIWpGxV)wLR*H;kvLi=^o(6WA&++IG)&xKYAPiXu3ypJPo}`;jA@XZfB2@A@pH%!!zU0aB4awkU@duGvZc<^pvh7q0qNDp9L1WAz$EW*
zx)NVVM0%SCC
zt?b@T+({{?r`VDb0-!55JBs84?ay-&aVc?Sr%ihA*O!`%(7x+%l}~_~)BsW)u`0=A
zY4586zrI)`#8VunbXCT|N2Uh)wR`_)KhDf8{@A7N0p|&I6zfb$5t`d4w_43Y-`!3Z
z)#*=l$idwdbHHVIT*{#HI5IG{Q8&_YJK$nA>dh~gV*)_*ZsQ-+VXP5KbXG8ZaMlV(HRw#Y8S6LIKLlfNDq
zTX&COs%k>qzkENwNS{>G@EJ>xk$jclhlRYfch$-KlCTue@Zr8FSHxTz#)JO50{Ufs
z<#oj$FVU^T0$OOAx5Ua?4>rSV1Pqah(aXp=2KK;LM-Jc6zvKF%;jzrT7hyHD#*;9d
zn%rOBY|<`C1Ip>sb8Jc(3$*X{?@5n+m4}%Ved3
zw%M^UVcxhjRmK=ft$UR3a4!99h6_^6s??VA&|8zchuu_FaKT?)`@*C-yBF~0M-tJE
z$5XB6>GVJ}Xn(19@a^U1Dz^$#=&^v@VHCgbQ$L2@KyqAh4n_yZXFD*7>td2|hx-zkWrZE(f*dij{F@|K4|~
zZe$3}*)5Ey;aM$vpS;K8L*{_H-fev651R2kWO6>os54(dKV;PFr3$Q6N7fUUslMgAOFdpt)tpQyq81mYTm*NNqkbV|DKs3A{JR0wC+}x
zJQ0!=dn~(>_-n(J2v|Ow?&{m;usyfX4YzY5#pDdB?^>j64vpO-ySl|Ysc0Mr13e~R
zV^l~z&O47km*4DlReh1pmJ9PAlH3G#+O~vt@+kYF#VOd5;7Y7p;^4^sKce0-uG0St
z7oM6l89Tcs+qP}nHQ6=Uwr!h}ZF90~vhB0`{m*%xm-S+Q?tSks*0nCIdengjDPCZ@
z1pF^_I^TxEN_GL`sjP|?C?YeLdFIP)>r|wi%z`3{f5>>1rIn%=`Gw-ZMf6qRtP&HT
z!dm#i?L++p1y(#l4)^g3Fg;zO=#5li9~n0JYbTcaY7!l_-F~4t4JN8_@xOix%)v%<
z5%c`C;FJtk2!>83*Qbmgh;&t)?{gF0pTt$uMq0c%m=Q!)V>{NH#{e$MLX474`1$oyp&s~DKznpvVey+nN
z-QYm8t(7y@S)h=Pk%S@pHj+^1ntuC-<_D)n;)EOy!sj~=7G_w8-@#SuDfvmE;fN+(NsB5kpX
zsva@07DK;y;oP*f*M7N?-WIuAye-ZE9I^QLmbX`;FtI4I&#=PW&hp}Q
zBW-V&pO6KoI+30KPm|qSv}r`Wy0KX?F024EKtaj0K)7irpXZ;0I!55wv4yiQ7p3$uK_VB5Mt9A`yXCcQIjue{Rd`-DqBBrCmPJ8{X2YKPU}7
zHlG?_4o}P<9}SeT|0pGziS*43k%+h%o-*#f8{b4%2>$2|JHxVgal4@5QyvthZ
z`8nMomD6p~VnaenOacwkwRy9G6%+#db*zC*Vyh{%>{eBC5AJQNYt=g<{G$
z)6tSuH{I#HptMr5%U#4}j>h|;A3y86N=%|@_q^A!i0{Ncv}x-pGSauETW+Rul@%0B
z-?rZW#Cev$G!SSHWDerM@hqiri5O6YRke>*wEUZ3fC?()$}Jr95HJHAtl@+Z_MJx2
z`qPpe2m0hO1Q(?|yUPU<2;hi96jV9uckULsO4p14!0s7pMMd612fy_C8!#gY;4J+d
z)SbJXQn=toc||{t`gu#hVjo&j;Rn>>v*3v{Iejo$w44O5vK$3>V*f(@etp?HUvnitIl;M<&jte;D15$6-0TMj
z$Osq&A=;8efyTUnUL|MKP3`APB!Foa0HYF0CbhzA<3G0xMNayN8WEt&nUcBHgw?(DqDb#ZjcSd8_RRR{&g(?Y$?oc(bKcm=)l-lx+
zH9^(XXWRW=oJD8^JX(B(34)NsUsY-V*|}D_z|#;|&M<>^G?@2ozK%#9<~e5#$P&85
z4WL23|~`T@fFFS$8KA8&ljZ|<;IpNwd&
z()C93IsLE>)eeflxA*O8X?dq_QFMbOY(i#@{B)f4~c6{;>qo
znCTqLK3@;zDs=3?=rHr&d*>(8kA)J4fe7fR1b&uZqL{BMh?#um?_y8a-&+Cg&azhL
z8lo+PW76$!?yj;95o+=+wV@+&g%gT%WT2ko#-Vpo8N6nQEWfhFeqUc}wuPcR(BP1*
z2O1Y|Oj<3*rr+%BiLG~B!l1W<(v!N8i327|wrutku|fiyr+K#zc(UXhyQ&;^v5DiU&eqc1&l;TnFbziw
zI0s?pPuxYt5p7}0362sGq1+t)MUbjlo{ifABhWA47aOht6
z)5g<9Ri>$&=5dGQvI#qfl^-HQ8%a!skh~JoS@-^;1QygOCKAG3syBf=xqf4j#I^vn
z#R#0;!w`CW%{_-5m7vhvc)fV~_!C2g6QMsF+-O^)+&o>^=V6f7X6N@AkNc09oNoSc
z&LoChBc;y6Iei#_P^;_Q9b~M3*6sahuGuBH=s;_{=G=7(VX68MZo^QFbQSI1S;&uE
zpC@<3i=34hE39eWf8I|W6FiQt11}q=g
zRZSY_M6KI^9WvL{S0j@uC!v_DNB7?zd~A1?M2BQZPpXLGt)KE+(EbkD|E;cVY$5%D
zX0Bz2gCDDI0De?-CaVsN(CpLdfGYONu&3vResucrIB@3(;b%FMKJ#Rgk~jMIZ6RjY
z>5#T2n!3verrO;=S0fh6?BCu;-E~r=B~(cA#Ps?n$Y&h(NZpN=U}kHesU1!zzoRm!
zsJRz{jK6l6_#Y_xy!#a?H3pwSO>P+Yu)3KA*L}A$PCI~CyK`Vf0WRs3cIZWbH#!>r
zvCEvp_Sdua5+Gg?fr{Dk4LwM8ePOQ4&^rw!*$o)~G<((sD$GGTC4#w=U1ACnm$;jq
z{6|M%F^|9!{%@(0vj6|?q*!2EA6y99pDH66cvQfC2hj!7bL2%t_tfoUm2(l;*Y1pM
zOPjMZILb}h4i=xK#6x<9Qyp=tGF+5!?aeLwhb&k-?2M(#k
zXnm)>h#w~=_0L``C!dIFFa)+S!8$yp-r&;e-dB}&n7WMIE$_+=OHf7*%-(9FhftBP
zL5i#$4V1s51172*y8FCnc#oxEo~C+;eK6i2-Ywa4tm2$7}!8ks2QImMY0K8Ql+Md8D?Q
zlSF57L&Ds0Ur`_WUH894;J!;O*$jOUlPMGlVp40gOT2wRK0G1e^LY(K22owTIJ2)Z
zTZh^!XBTN4Zsu>G9MbL%tVGl{(E3^wgyec!Ya)yz>ekFC}7PMO!IHc{~(?$tB+h{QKR5Hd&W#ZG6
z+ua+>ZyKI3S(VJXdw&Exu%Oxeb{0-|PNRWfn3mTk)NmA}yG&VB1A||I?vOUb1V3>|
z2exX@;}MR80#4et4_8H;;N4edJ=7gf{AxknR&rki$#G^IaLsFq
z%*crv<-g+1SNNO6Cjx;$B5>aT0c#}x(OS$=(*x)E$aJJZquUf+MV6|r07g;6@r7JP
z^=|wnRUVkR`~>v4+`L#{>o>oVHG~}dK|TNwsQM>ar~)Q!{fim{x`*^i;z*Pr>c08J
z15lJ6cmN0Z+z~yD<=wBUh@5J*qV|fsksAj=O^tT=I_j7VhHESiBKK*#*UqK#6I>+4g6
z!XNQMrjnTA{T)^vDD$|uRC_&V;j_&c)SzTo!a
zDU{R4IbCYPjlNS&^GG%g3=n)^=ItN}DqQu?9Jko@M*tvYW(t41Q?ECokBis!?BW9=
zamA*0rjxW3N)#!CW3}RPL??%OETrVa07w_r(L|Rc&!_os7K+qwoYrM6DG)Bdfc51aFJZ2C~X{hF39LQV0H^rQ9V{V7fuUc|w=H^lal%bwV$MT&=<{UE_36a2_&`
zPh^c3oA^K|JnWmDHQVS-Gf^Vqqe@!)1j*am+zH8Gy@|X~qX3Rn$b7coO8yd+$sZm%
zN3POAUOc7a-%mCjenU2`C-}KVHj}^pX<&DmgZS|e?&QyTaLKg4NY}rdZgGL$bHIcd
z2{pLQdKcvMP{3KnH7%z
z3O8J2`G>wz&b3TizThYDCs+_7xSNE$Q+EJ0CDmZaFKbt?yrjQ`@-p*8+``mB!|NXu
z>phZ@^cB5e-4^*&A?q0Uw&nd4IL_OPQF}st^KUHL@NcWd(ua-%RU=B;F;@Wkw%n#i
z_$XI(*v(^3u;GuzEw4ocopLnfu@JtG&OjEj%i&adD&pWOJ~>a=mDvVUgWCLp&3zs>
znLR_ymKRvs*_Pn=*L+D@)l?aq2-X}AKmo3-_!(0r;pA;kXt4UDR20U!SFQ}+2O$!>
zBOg?x%#z4(-A?2QS2zszJSk1}xL6!$tI2RR#Dk=wXu-peM&pQ;2)PHPqErvdlggxV+*c!?hkquAzi}$=EHU_?F*O)?Ix}OS}@}|4fjV;jf%noaW-QE{W
z&9AN+hirQEcdbJ<;3kktixF6`|6~J&oYoy&1I`l2vv*c@qKZs*rrL{XwY+A2LY;6-
z>+8xL%g+PE@i%0){l@;p<%!8zV^&5mIb5H$Yoa4R`9pLd{Fr=C-|Xtq5$zVSvH$A`
zB92XsPD?P+jPclq6dthuJE|gCq|UUnE}R-7{cmZmO$-fqG3^>PHPmvvZ;`hNLQB(8
z0It~7?9>T$T{H(7(#S7pDoXE+_ZZW7PVG>rB0B5LqG-n+F8RTzxu=vm!;ea4jqXGC
zXuXag5~J5&Nl8dy{J;6Pd+L;gH$ULBG+Bg->Wjc~dM=ercSaiyYw5?oLuz3g%*wYF
zj{S<>fVAFngz9#(sL9^Uc9xP~PD{KV$eHPo!oNFNK-&vJ=&@83zTkEPEJ<+47u1jK
z4)<4K{GIqzTHo7RtopGec@N|WRbk(YGTL~W!BUw$2sNMnn|C8fk}lqX4Y$V8-D*P$`p#ch3y
z@Cz1gB=L*TWW4qed`d1Vkt41>CeyR7=%FI3+C~qRINcIAfkgaH9XotxT9XJzgzSX)
zK9PmJHd$2;XJomCeE%ejwgHYB3fcErVKvoC3-8_?|L=AP8eYn@i7AP6({+3GbT&`j
zJx}I1uyw3NO#k-=BM|u_0|9<2tdw873>vh<&Q_#7T$8pdw0z36bmvzzNE42}gax(Z
z8UUT^PAv=0m#r0IH|rJEyjNRkIS6t5^Qej!K|_%IB8h#nj~&L7nC2V1E`b_rp7u6<
z%Av^b=HD}bEp%HX(PZe}9sH}c2d)!p7oR&{ZDLHb$)AnzU-ZJk(b7p}h=h}eNMJ-H-6dadEgt2Fl)SXNE32yb{WkD_O!5^JWjeh&nKBL-
zaU#O1ip-_dZtQgqWkkDO2lVaY^#ywZg^S-x>bRjjMbQAZS8pcd5Hfn@l7-!v67Giw
z=;C{pn1`v0sO0h~)Kt`y%40H%dFM8O_HuZ
zKJ<<{rfh3$3T=SV9W^-k0pmQW!#zk)6p!je22QW+;--Ac6X^=NhF3)5J7omORY4>P
zZTWl*1g@#~!GZih>CbtdY$Vp9Y4_fUv7P4f=#uhX;TJDD0@c|*F!lF
zJm@Csgtv#O25xfD%Q+!Yz3D!#+mC*HC7iR49p96|fu&G1e-<*hfe5tUKObEFCZcEg
z26GTtD|x}f>xDzt>w(|~grIQ(7b&NQUZx#V36x~UXiA_-fGa?|`fn~PppJeLB4@ia
zO8RU!Gzo@UI~)CY#Bc|-L+uz6rS7TV<;0X-Pp8T7-;Y=*R!Bc$%=O6RUjOFBv`
zGipQ{2(ACaU`SwwdOVVG7sdtu-9iLjggsohu5a3+cJ#F@-`h#B{+#mC!af4fK$1#w
zkWmZ|gjk5UKl-$83c`NSJXwa@|6TU-Pd8c8Ju^!Y@-&BTw^sb=zfMmvm#@qVn5tNP
zXTRW5>%9bqTscw`MTcsH;3F*wF*H(U=~2*5(W(w@6EoZdOV@MNbc8z=N*`-&2#^C(
z9sPqOvVAHp^KGK}L^Tt~V)h>OdLr1#mUr_I?^ALB6z5?PyCL=9Js1}q*V13IgHe75
z!X9YffQVruaC_YUJvoO5s?BAp9M?HxU#AKFtjeT=O{vNzP4Kis0uE&a_T+VHydfd0
z0Uo5x^VN02HtUSE2P#^VI;N~mBEUkcn=+hp6}m{o%1AX4M-sQW63A5MN7fKiV7M9h
zwUVx~lu`m_UedlegMS##$dGHD#Kdc2P|7GMAaBI0eni7}|HBW{88
znp*H@dJI7XB)>5su_35cQe(DNhG~(>H#_cJ%aaTM`#9XN5W#&08fX
zN;I1S-z_P&-Fvvvk;#j_;FqwurnK`}e
zz|;}ctQ6DG>^9VD5#pi1P!Sd2*
zL7OkYAhkmqHa}oPRe$1!gjB@5<3s_30Q|nA@m~0R{rG%;P=+_eKuVXrV+p)W5Bei!
z>@qc3P)T4oLN6Y2y5b9ME}y~5Px**kh5e=l<(GRkvxY!O(@o3d>uWJkLRPHj0!6S5
z5~e}zquk{WFI!tQ(X7@37;cx!_tHVkuocg$GV{}fRZ+`zqP^vE3R)tvMs7>!BTVp)
zi>1F2eil-JrDiU(5g;0cJfI9`y&@%!3cIYys}x?NKO4j-)iwd(o|bzh@1#pm8q
ze`+APumP_2DZ*RKwYKV1zDpG`77e7bPlSisGyUr;*5?%{#aT;b(V?;LbRS=35$!bdoD@&R(7ff^OH@PDzP0@P
zG|RV0cG#~E8y@JhNhm(=pcT(vk(q@iGPn>*29rj{z5XCxefPqELD&*-0{^ZhCcW;h
z(unjPlaNFKq-7C#IHuVDxL%18pt;-&PmRcEaN{L=4ha3n(|xV8Ie3sVeQ~ZclWE}D
z#?Q%5@csM-R|p5(YG^GMZRopRXMf1_XNaJC)M7HYv_mGHnJSSHJb*S#5}AW3fzT(a
z_OxJ@>oSk)8po>b4TCK#$3S1^e1jitKRZl^jKQ6e?y7xG-4yp>=)qF%x>HM6J=K^n
zBH~+Zkx(}l{D`}G^}9o+2v>f@n>d(w9~YdyG)vbV-DL$t
zexFJ5)ahIURD?WXjQhu5iU3o7FNgU;3RTlq`Ne>@mITXU!>MNDCj9`5@z243-%Do8
zv&okG2%8x!oBMZeNT3ab|ESI;h^$kWG~dLaoT^Ypy$N4&d(T1`PiAVgpry9{;?lOd
zEl5qMnsPDD7v|4F(l>dVv7e8N(8gt{($hj@yEWvcduiB>J|Lt3y*X?e05)9gUiwfI
zDHW8zp06XdCRgG7#EyWZ&otd?+9gX@PisHuxcb(sD*gGNV*U=hT`)}cUrpHqz
zHRO%wyDNkj_FLC8(Qm|%{$j_?7R)QqiJe@EF|M8{~<=p&(LSPJ$7nza`wErp&KwT-kEu_?@S`X50+9qrASI;xjMq1)#)$VI3m@7C5Qtej(H`n6Hd2QUVqMr0OkJt4_AwlcX0
z0u%wc3D^%lfgB0Hv!1i*DjQ{oHYq_6qdc=D6CsJ)LM>EWgmLb_9F>4{hXrgPG})|v
zg9}rB22Sqsx6RX;_A}HRghB7S>f)h4xEwU}+e!q)q(9bFw>o}&exN(Y+*W_$dw-~?
z|5Trg{?yp>5vm&ra-aM>zb%OtIK6nE?RwOorPflKz+lQugqyN+b_sQ!m>u5M?SOiE
z!crJ2$ZUpaPbr^zQ
z6)75unLx5M=1gHD?#rT5XMrVJ=4=;t8#QIx4)p5XA_F~|Swiv`
zBmC`SAH~tkijw{jrj5IP|AXIDQ{Jh5ifY6RMU$;+3>+dTS9O}xg?JG}maBsK~Vmfq*y;bx;mkR$Gmh9fxbS84A+WBymI4^3nY2J|=^
zcgB63lC>b8kgT@Vz@5lx0Ar?(1a{9d7{cyeXAXfS(ZP)xK$`Ox(CHVL@
zdG|hrI+fYoLDHFUO{N4&I)@jxD@4^mc|QKw2_g)?89{@B+xr=!$UKkt#TDZ4+5;Me
zxq%zOgK8`_c2UUCxTl<{1?36aBlvIs#&At6pDTP*2BFS5kbd4}fg+I&791!*?CidB
z#kXG}Q{UG(kY49XsQy4sp!&gM_W!Z~NV}LuUivbrvE1+!^AS@!_BbafoBBeLi1U$6
zzQj{N>KPQ&p+b_*GA?kEZT#2Pxz|2@wBXxkt!n%TqTWJ-Y0)l+!Z857ONq;;T@aA
zfKmwaMapI`ElhVK
zA2LQ#hhxfW%0X-G^Frb5rz!D#6Jg0nDg4#_%iFsf?`j0MQHHN;ZZw2Hankmti9##>
zzk(duuZG_j6|}X%>?1tA{)_Jh>Jw5A82uEQ2hK^QPZ`EPzOY0fv3@og+s2;68h`B7
zrm@6-m@&2ks#cHyT?0+(i)jC!rbCi|fCfVmRAr?Y_)%9fL|13`9^5|fAR@OuxTDLd
zLqYLye;i|UcQ=OV_Vw<4$YpV0QbNVkiufX9iCbfe
zR8xWD79=Uv8-Sk+>pn8m&^JDInmSM`HNyOmpaOZeK^}#Mf8Cw-hjN&K+VZy`WcA&%
z05iA89h7=xHXVDcX_3preA0G%hw%OrLchrp^}32KLO0ZF-A278JKpC=DTYFmFXFN1
zPpGdly=R1)&XfjGEG*PjUyb(LB80<65I;5VaYRNzxusblFfSR7?dqZ5Y&lM=yHQxL7KXC@Pji
zAaxFM;IF{fx3#x^M4NyviRVj0mA+6B#t`u8zrjW#3ZKOu
zK$`X2;C7CK8D4Hew+(t&HfJ~?_Ylg%P9s*3thjR^@!<2~CR-NPa7$KWv0;BL%~sH?
z4LEL+n%fD|wfK*uc@ltX$>Uj~gRq*-!{MB{PDZ0K-3yA)g;Y6!ER~8Q-Wh^EM
zmUQ2jjECnR`m%ROXU_e|U2irS+O>FYAZfhdgL@;I1IJePn&O}-gw=S!ip1D0qR~#S
z*AgKfb~7wXcX!t)!tuJk>`n}orKa1PjAycKgVJ83S`|#uHx_nbS-rDo_CZtTaM(=a
z_x>j*Z+@&|HUe8X^U#9)G0GF7@J!Buj^>H!`tD#2L0C-kjX18shR*>W-Nwz}V~D3z
zr(}+@8lxjAwaB4uW`~|L84-?4{LC%jbq5Ga<
zRDZYhDhTyN_7J(O?#PCeYQvB4CSzx${QI>npSz!&_qgFFNn-}HclP>3$9%<;0v{c!
zM_ZjJYI3n)JP4xgy4|~b!*ai91Cts`^bh@r%{PnW{8z;#2|xbOT^<(_n725qOe6Pa
z<>XCN%54}vwYnhHs(c%7);onFI+@oMU`Pykt%2O0E6}em{JS|)2m|vWyW%`3r!MmU
z1OqCt)1hZ=r@MQ1xYr_Nbh*`U&pMPwTh2aCsjkXq8NspU;#*k3c0BtyKS8&ULE{5<
zwAxl3s=uIS2h>~b86!>h);#6#b+%tYTR_SJ1Itex=l)r
zpy>hNpQdza*~Jgfi0NN=ug^`hXr^6^Kebvc@=Z2sFj@TCKMiR@X-aFuVh*E-{MlIEERWwVuU
zvhbkRYS$>NTB}EiT7Bdf_(R!8sece$byA=12Q>HR4a%)X996a5_s@;Ev{0u|sCARz
zjQ>T3x%T#xW6ltty#iA7b?%3}noox&Zq^)bZ;b?vo9*5d>1^c@J6;~Szk=%3JAMVY
z^s*bi@jVTD_bW_w3gZ+RCx2ak>rjnpj!+#`dP?k@USVVPC#pSCcN+H4Bq8q+=_oD+
z`$I6C-Au{2ogvY;G3KlRQ!vqni(#)Pji=s(<_x#r!r@jTiu49H$5w+@epy
zT*ls-&Ph+Ox~w{BPu
zcs$(aV6Bemzg~NAHzE?>@^RMk79~K5NR8|0lh+Ab|FNj1cmv)<1DVW}-?ds^@0#rL
znA=qt7@hxmw<3`Ik!W_WuQ)XwPXt*x{=WUu0OhRd>DL|e(vw0>EXv{(R#Pf)aM}(N
zH{e@kJMBvp&3Xmz2C7O#UVph@lG$#La1n%S=jR63_RbJ8vBP8}eTg5t_7n0&TDw$<
zVf*z7vwFj$kePT&?ecd-)+f#m_=lfT3XlxP^tM=nV(ah>bReeIcE-0*DHOMQq%9#+
zm3F|`85pD3)z1zb82i3Owr&4_#fGr;nBxl!7$8ja>^p6*jrjR7DKV=DrudR3-x!O6
z_Y&*Ds|b9g$EzQ8^WDl}`RqOioZe@AI2sMcLr(_paa1J~-gvAKZGi7xAThnrFfy$`
zZ{8bm5|j^(wx2gKS*|`W)!q4DBsQC^fJ0Ynqz6vzhH%cz%tv-MPLDqE%#|jPcMm7U
z$2>6Z@swVmrX-P^*XIM)R*%nkD$}n2&H`9%$LW0|{p3~Oko^V8D6W=K2It;(9J7_=
zwiiw)xv9N!jX(jCI4NNpI;?+SPq&yahA)Q@g~i2WIi}HblU{oZXo&yCeVeu?c$KRTRxh*9~ALkt-GdidlTuD>>&^
zF_jOX0M$%%l#k!zX2QJpLxLMrDCD=vGBLYs2j4oZVJLP#t{Gd>80AL>GAW*VjIj|E
zC4@-gW>7!hWSFyLq~X8towx>Am*=x#Z1MTz#;so~AQd}T7D)=H?G?v1j5=+;f71&&
z6GG^e(~kp2ed@Vmz^0!2ND>DuEN(>%&yTPAF;PG77Hje3>c?jMdB5eyHs4w=rc4FFERbLfG0pjv48rBRv2C{8Yxu|OW<
zgOELhSMPSTnXl_l_u~(bGv+x&;Y$Ts1={n8+JmoF1aE%KLsW%j)8JU>mg#2!kbQ+-
zLG3AuYDhFM&%}=h@2?8+9*Lu&>KpAdI;r=m3$kNx`i152bI!-39G1Nv1*oO%=Ad|q
z_@@f^p1c=QxD6Iv$5U*wfY2vPums7M4k`tCv71Y6UW`LKI)uM6*jjAXtHlGfa
zOjrvi$M$L*t|YU@RMe?3k5MZ9%b|f!&*!z$8;^yu|Ek%N)Sl_Y;0N5eekSp+27Z4^
zGNU}YIBc@`I*FA;KZ2f1$+k*YScBg*W2;8zWsn27R~XJx>U-N!{4Y~au(D>GOk6NV
zu(rGKe@3!C%<#bs&^7R!Diqii=5tyPAsFKoy|-#1w!uAy8vhOGsJoouog|W%F8NSl13|hf5T%EF9X3S;n95;@hRrP#k
zpu|uI{lrebl!9OI?7EIR2}djvPK&9$h6Gf&
z`$AIiqzb3o(~o8PRdVm@mAZ*kr~Qs=FS`Jmn1^W5L3O(p7lV3>C4j1s&c>=!cf;f4
zcXpQ@QpyxcqID9u*Rr6ayY`cV(D@#mJ+$iK0$&2{u3P@XTQOI|eT8kN)$7mTkt87o
z^A|;PN|~nQ2~b)iqa72
znZg9`$EqP2Kg(Pc{*H$!u8&Us6#-Xt{&bV%5bT-!*C5!_`CuJb&O`yB-O=NHTtNcV
z5e?`|vj<13uVKPjG081731|ileg}4Bp0hH{`59Iz(5#tX@g+3)XIPw|Ox;{Ze+hEg
zXR_10jH>|Fc}AtU%9=u;t-3qL;gKwDt{})lh%9KTA5tgox7g1c?c^
z^1SA`8vz`nj{nxwHW1$*r*3=47i>?6P%${IGc%WJp!!5(lConW&(UhR*N&GKluQ?h
z-bTB9?gW&w8t%g4aJZw*iz3K~GmR<+Zw2qut{MByZDT3$NGV^A2eWc~Oqu^P2!S)}
zKoZE5gtUb)Tn3t6((`P2bNi00fxdBZZvmUkR2f7%i))?dDYyH7JYHsRF)vSr!L^YA
z;L?$X7f;i1EKHiDGhT6RbzoqrLVi`~$u2eVzH>jQ`u3;KAWrr>37E6H5b$}+x(bGD
zS|j&dbNX2Hrr2+Xo9@Kb#g1DN8{wsf(dFQEh+?OZfXWpoZ;*p4VAevA@3PK-EFPSE
zSINd3jO>%)O&4UJ*e|9(9frvnBMv(x6h9swX<~L8N2VN}bZgngK2hF_vRzLaP6DtJtnN;H)
zdQ-KTDf;#!t+VTewlWa$Ze_>&pfe9l6TM7l)!3W!tJ=l$LB7N3
z3=0rQx#Gc-1)B}O%J&$+P7=xLjFKc&FO#N-#9fsGe1*rUjuhvX(-q<69RBBPb#CW$
zv~L_f&N^j@Amixxw}mU%N^d{F|E2oRU!~Gr1`l99=^zCk8?0bc80p0mf#D!4uPQ`K
zesoeqVqSkKe}w7J(u=5vPxZNiC$;Vl|8r+OQyH0+dL{~-Qh|955)$f@+aiM(=MoA)
zZ2O5~t2if*ca`66y8d?^Y+7*3P_-od}+As&>X{eUdkY9SJ#KIftZ&ef%_E#t8WYUBrBJYODsf9bxVdE^s
zBt!HLa?IAB-y0uBV!Zz37`@rh-Bs=6u0%pHQ82ZWgv(4g46H{1v@~S-Wd44)D%(Cb
z4t`&#DOM(_h>U}4bTz-hIiGB>ENZYUso&onnP86~2xz7s^(Qrc?(Or04;j5Nk_d_K
zlN4(3ay`x57K$SvYSk4u(BM#@!g>r;9EzO^-bYu_t4duq{F3w5J@-&de7>9APm|?{
z;DLdG6JM>sGwfMw!VKDgj$1&*O
zGRGAC`Dh+>W}Y!Y!W`7raROR@Uortf4SJ__WrmU3t1FC0b}gI^Pt@SA9!tV+mh0C<
za&W(-AdYes>IjvLKLP0b5%G(C`y4Q3-JSWDM@%*UGv7vMQ}3F#7JO_SQ(vKmaXsP-
zTxb`M$ldC6JEydP6Yt9Jw61qOegg*b2}`PfyTJr8zWtFwb36<5fzF3*9TP2aQDEds
z1w3K3`k!3`D{6JX2k(nGIVFyHoNnBM3C!-_y`@z95iOj|O_bTIo5;+@!ESw2Sw`9HPwW=6J<8FEyv@Nb*6+AzfEE+4JF{29kHXA53@whxX}^i*Y_;7cY|B^4I$BJV&4}
zJ71Nev1)>|5S>*AiS&vPQw;{JH~i~R>=0TJXcGXYl1Q-iznA*~=N=W}s?AFFvnY76
zS@!;!alW>nHC0%x;sY8hCRROA3O+l9v9s%EJsM83P4Z1$xtQKWjHetQHiTt%IqYg9
zR~u*Y>QQW7x_ffkZ9)UEt;EkCtI(TX^)T5TS?(&6I;LxZp01Y1pUUKHiQb8hBd4y*
zKE&UmBv;5K^nkoCAvEnQb~qj1Z+D?k?72U|gxUxkkU;7o@QZqoQq!7JA&QNJx{@6n
zcmkIhs#Yp0`(VNC2`Q#G>qFD@HFMaHyi?Gn|MP2jRxYVBF3IJvw#d)WPhYa#?;-Xy
zpMMv7B%6fT!EC9Sar;H7o6f)2IIZXB=!AQ2W66q{WMJAM0fROqmp5$%3v2_9Ms2Ya
zMF#=X#qfStnYwpUJ9*KJ?R}Pg^hizA?mVqLcr&~^V^Or?;`X#vuRgHR%**$EDjvJ=
zNU=B-#>O@$B;(P`K(0F8%^+)#j9&bTHB*;IMki=AC8jEsni{X(Y&@KMmKyGCRA#Qs
zr+Bw(3&mj`nUES!I;sE=dJ?S2ZfAZRh{2;wp#y7uqT(0eoG2*VAnzrIaMgQjnhmX*
zS=)}`h5kTgP5v0&W%S;(CpBr-SM*)W5VoPxqfYN*{-%}sFf#U`{0}yDraSvkmN^(ku)h|-n87;
z`*qV}FT{*Os@BY-PKC|b5E&%1s=D6%H1!+=$}@P*#yvPia%CQ)%Ckw1y|L_N_4xyK
zfkKKWE~D2P=N%z5*fY>F{cIefcFTLg5z)&9Qc;bW8~{9ZF<_Ie5(IRVMbTw-JAo@?
zY)sR5E+-O`QU3+ufT7S{!i~>^W>s-jX{>NRtqL!jX2#ov;$4L@5uIkFuWyq`k}!W?
zgzpkdk?wcAYGxIm;b{oWP)f~Ws-H0Iu~t7nZCGY7ked+T|Li93PLm=eqeK$dgSv{2
zU3Q5mtV#Rn-Pmlp3KRBkXqE0hou+=5fOZEM1Ocw3eb7aE$xU-}NH{2(*D;j;Qam0L
z7R!M!wsQE9{?Bk>R45#RH-9{nIA*3y@C^og_SD|x{Hu|VMBet=rJ-WBSNx^tLUs0CwI@~5`
zuhM&`7#%O}Cf6x$d86F^d)vQB9JrF2_10Ur=L?(4}F2@j&*bq=~%tyx<24Ndr)E6YC%Bs&x}Q2EA|nmsMrRYa5>^)^Cf7B-vh}o1p};sX9C0W@gNsNyE)7yS{!YnlL|Fr
zS>R04uti5tb&Gy@)x=UUQA8(Ew7np8
zA!uhSgNp8ol(kv~r+x+$yxV!snKNZ?OB5JDLu2o|8PfkD?60G${Jt+>RNw%IknT>U
zJER*VBn70qq)WO}I;2YkM5K{!B$aNYLt47~-RJZD-FMvk*L4gXLph#jKYOn|*IaY0
zz4lLh{F67N*`Prxjo+s6>mxA|vzjsx+z~~L8g-|%EJsqrJNg%)3azguoeU)D6s1R2
zwpu<))eOw}uF^3zH~r*_f8Qm}Eymhy*VXx-MiuD$g6`iuQL=@Mh07neogQNSc4xHl
z4HR4YOZ6f}yx#Hh(gb>S0b-wD=>l@fIlW8B%f~E764HMd?=l!3Li>A+zJ9~vaKk{7
zY8)v#V2w<;75QLvq2XthpuV?^aIEz7k3JRS#}A)Mp6q7NJvQElzHmModCcUy
zD8bOdl9}
zdy7`TD-57PW1}6+|4{NSODGV*us1B3Ry}`Zg(U11)pFG)Bk&^Y5nO4gmAyo|n^Y)_
z@JJ#gsh!N2^upuMS7;g1_)ddESHe{E&I1d&zd3cM`tN1^Mh%7cS>yZcdwV4*wPvaY
z<(~T3A69x!uPs8~{o8iJUe7Y7u{FHcUS~A%;;fFXU(g=U2zuZwbbOA`!s8-{0~aC(
z5|!wWID;q(5OpLt!`BIyXhsnS-(BI}x%?*N>sjL{e&gJzxBZsi?cN_c&)yNUNxq~y
zIW3t(q1`F;3clc}C88A0E$Ku-iQG6BHhCuNtx5WREB7k
zTa$w@aZp5H7*nKYq4(CEv)kr~FkPN*3R9l?C*253rt=G*)TDQio%Y03v5$yvDt~*B
zi;y}lx4(=FX?@k|MpLoq75S(^&`-1ha?zdl&rwp=xV2Yr202)WC$->s|P*z2YyLkm&i*V{UuCUg*6S%uGY^Yq;LOX
zjm+#vKSW~_vzd|XLa6qgp&S2V*}8NU>wZp?41`HaIDeJ(?T?=-rHnyd>7Cm(o83?>
z=TX1H->BC`F@F*7zi&KtC&s0ihgKa`{sd0ny33x#8P>^@*lP
z93-91YQ?Mv?XN>yVF~Byu`kK(*Jj0p-rn=54&gcc!DC*V_@K!nQknQ__lF)h&ObXR
zh1AF8GjTsf+h~tj#SoC(?4rQYV%)UjK0nH9WJELbs~O!i9!2=bWG}hXOWJLxwQ#~M
z$HkC&{KYAIj_UK2;UDAfQ+D7CGTeU|W=0dri@k7khGW*Jt%+_6p2B_Zo$Wl1&vI6*
zb0N9N1qPKyCnp=l(oOrO(UJd`XZ9o@`%+}S{QsM0mdrTruN4@`n?-$S@qlN>Rw7&K
zef6r0c$P$8FotO2h@Sjs<&{)wM|`2dm-QL1Dvx)5kQ3+W1tam~@f+Z2&zZdrYL0V7_`sA@8#9bs^ixtov4OmhoZa%qIC*&?j`n
z05wJzaOXe-j<$AEkTfmTQJ%Ea3+DeycLD)-qeFAB`|3B{y`sFsQOn+xiv?ZOQp_3^VAa*xb|-#dGy(|h5zakgkJyOSdq@btBRJx
zGLIuM)0sVTnn{MF9Ze7UjNj0Km2Cef4D_4Z{sm$|laueyCA&T=^Na;jKUt!mCl&p!{d+{?WzsI&3T2M>Sxb(y
zN|?YE+wf)=3!C_>zI$BWuTXRQkMy1dK|4<(mgdHJ%RfO-q+J|Y#j2c!`)U#Egtx7<
z7zH_G#u)t-%@H+1d!x&H(05O^lY>lRpR?1&8rD8!wL1dy`dI!7gGegJRbgcRs6h0X
zem_|4xdF~}EQ$_GjJ-az0@Q(Dq;CY#yV6N)EEQLVp%wipkD^A_5wuodXnIpKoNSQu
ziB$v1d^AaIxbE{vLXf#cphv9V7lZw9sYazBL7PuPVMH21%x+LQgT`G-HR9xai2Xt_*^Vn0M&)+^KIYt7wJub
zk5C@BhAxoEt3FI4sjlboSz)~v4T$9&R_tf!CU`EAnk~&-!*=X8OiC-3kMa-I=C)!V
zoC~Kr-HN)sd~NNg!)5~Yy*`%_)x$+fVW9`bc68jZwsk{+zqdfO11fx9U*9!+_n+9D
z)Y*eCN6boM$AO+0QMAachc9LI%VN5+QKGnlSn;)TFXSxxg&Cy197TO)<9TPtt9>Qj
zIt}0!?Ebt_atijO%nkj1UqI1%M+OpfSotO+WCck|*WQqUVy8H=c`Z7lm3Q+qLpjVZ
zmjkZT>qEvumSI^lFBfOL<8uMS@Gw}`Os$Ul#wnn5VY`X~UBCb)BV%&VKAo$d6T)d|
zXsB}#E%zaOaO835(`U&N!0r=B@=z;t&e;og4XWI;Rag4(;t?4z`SBJ-pNF=wUpzIV
zwbnM09mL?w#}4$^2dA1tS|#-aPjVG>hg4eddLDp_DH&&%VRp{tynr9t{t&p*!H!SM
z&WVy-Z-ifw_(hMG?(1_MK7WX4ay8kL7^onIokdV5g`GvP@39ELfkTj>0u4tz2^@%o
zswYpAT4}Ci+hz>qJl4e-%+mHtBi2bzk?d3Te`2lcSnDsM4o1>tTKtV`Sr;y=NCrAC
z_qTf5<#A=gc=>)c8>Da80RmL9WLOC}dou`znt5btFLwz3tK2A}T{8Ir5=P_mo
zXjSYyye%N=#}=xTSmAm8K5*BZqs0-4dUaYT+xr^l&30bD&-{2lmqR?Evan|m=if&L
z(I#x7@Onk|MB4(T>*60m?+NR%G(iV}POeEk_!ArX`)ogVxAKAkdT)ykS5;K_V`TcVyW=w%vp
znK&Qo2u|;Ux)!JNbFV4u<;1f3x6!$@FGe~OldU-I&&TL6ZMMFo5D7YvkW7_DiS?u7
zW~0kH@R0L39Zv9`?~ra%Fp;C{U~r~jty;-<$c<{0G&r_o8Px8K(v7`5Ld-uo_j`ZU
zVAi1DchaJ=cJVjmdW7`hh#qbVq**-`4@H~)vAaioc0k#Oz%?EpldG$m_jnW)2@OEt
zQbQ%vTJua$m!$r$pVxa>x4DwoMzAEtRM5X8Ii$VO1YOvbSQo+dK}xKe#a(Ps-r98<
zSHvfe+!ePGUB74|2r7e{#cRc#u6=ln+^~DkvW)2&31fc9ip+3Z-REg7=uaPQ>ZK>$
z{D`n;lSpqq4#o6Sx~IH-+d#H
zj|MiPOAWL(A{AZq^%*HK^cNPAPZ)1hi@)rdu~9KV{@oS?7U}4s|1!Q?ynOm+far-Y
zM<{~l0e&$a9_kk2F&@^wR?HV4Uq94ZTatva$@^J|T-@`TeKb_G!EAvaQOPH$#UqB|&2N8&{7g#Vv}ba&JNePCn_Du+jU4|N
zJ@qSz#_#=Tm!b&M4Hc))i3_X#&Q*Wqjxk9;vH|4xcPnJ@J`1U=tWx4j1))E#=doZD
zX_%bt-lgAbY_ShTc`y$m>oVcJ#wOoa@tV10`otvIoN{p=-1Ky6Mi+}cqT%0Bx1FHa
zWEDk0{Rsn!oZFP)?QfQf*D)d|SPaG421^T2v}LANHKIkCrb)&Js%@CY=X2
zmXyEGW94JRGF5GPd-JEQ@_FO?R_AmUcqF6V>3N+QI?m1~q$RIYoXPqenH*n6RT?vR
z+J%i*@;WrC$7Vfdul&w9Z&^fwEqf=vJsYbx2+VD}fPhN@WXI3f>-!O2iz&pcc*DKP
zYB(=gF46-_gav7p6^Ec3ZqOD52U4vf!q?f(%64=#n-wgGU|Kd3P8SWpLE7)|CY4WY
zgm=ylw%=89_*d%x3$5)56m`4!Six`n;1z!>yB;v?O5;7D+E+4cc42o?^{^=QR$B|}
zU*6#-+3BIc;}?q1tmpKTa!@-_SISu<{|L%}`*FF0>?W>^k8QW-G7S_4)wFU*Y36t`
zP(p&qyT{;;zG0bql7sUPGU${9$&c?t%uyD@+(@s;rn~ZTyx4QR94N*7dngMm_if*#
zCyumg9J<<0*0ZV?D~I7aSbe(exMF+E4v@*U0^M;UYHxJ)JKSFqm39fOizJ(qWh{GK
zH@?4ni|n!0BvEZ9i&+DYq}#}%*ML&25-pQdz2b6@6ZGOumg#Zl<6OX6#qiuSqxu9g
zIc$!VkTUz2nxi2~ni++!XQ0u9*~7J5ufy`+XyL*
zWe2s|m+H=6LZR@%fl|=Oz%J_a!Xzpx=$s@~r4`Zk=AhjuUpvrS~dN_l~36Bgn&P<
zRW{r+G3L9=IXz^j@B2p0Svsd9r(`(#u40>@9RKup&F;VQxX0UbO*;#MJ5UwhU7Szhfeentmwy$<4)kr3Qkz538XK1`e4MvkQV1@A8lS%0!X~vAv%u+l?B;
zbhC}ap+`kRCyHz?_wUZtbj4;X+hF5??>05-j|t+ReZ1hAJ_+$Q*{|@+J+;fJwK>PV
zUQr^%wByP{x;fvUb)2djM|rq=inSgWZ81POMN31xB?@v{`)Tqw{lxmu7f!w>L-@p>
z-sKz{8*utn64?xEP9Ck>J_-0p%_67Ym6N~NbUh($Dv~2lyVA={q+`dHvBy0}D?)lg
zyCH|KJ+%|qcoxp6owU{+an%y^9}EE-7hgRitRhS)d<_g90_Ow8V^oI(xE+PD94vH~
z%VA5S5Y`fXyPSC_zq!JjyvY-pRppan=~0(NthAm=Rs5FMUrqJi@M1$#hlk#JHGn&(
z{Q=zvRPqQ}{1LCo&DvL(=Yv<#T>bYuIWAu#O-!xT(|I8U-kGh=laNI}#`9#II%;C^
zXsYTxm0|3L#)|9EbIvc)J@K2-o;(UTV)b(K**hZs6n!37O?-ctdn2>zT$o1;+CI*W
z7X+SAe4uphZn>fyT0-7EPxmUVvYzxOd*wDq&uR|8zr|-g_N+giQflh0z{&l3qB!Hb
zCBIb;Q|aiay!UTUKahH3kc9n`y7SZQ6^#HVeAkv*u9{0O7K`EMt?2CkSxxn&01cRntZ!uBVAM3fI(MA6<3jF8N$4GzkiZ)ac+)e=zUB
zJy*AV{Qa=~dpK!yL%C_mGqi2{Xgw!NPo^gK&bW!q;Gzz-*}$epYszK}B6fcqqG(1P
zO$dPm1Qad~LL
z@0eWv4mA4R@gY6_fP_5hko0I>ZMwKvye=}JU|nf`8k8gz5VAs-PUZP!_sKKo0$k_F
zGGB#jA{illi}%rKZI9J2mtT>jCyJnVzfah
zwPJRva_t`>VM;^K!EHb8wzJasz2Zfw^$Ax{I<*bH^XUEAar4T
zoF9_d%oNL4>3D?y2+#%!4;Lj~^@NAGgKq8wfFH^QMJMGV3c0)Liu}%)V1!6cfCJIT
z4_YFm{Bp4FYQ1pc0XkH9qy?^R4QDZEe5b4V8X3{CFpXe_4Ml+$m09gAoh9P4MbtVJ
zj5`qW_~_JC{7g{I&)?K-=a_k>LKjp3Jj#A`v&pvpz=i)LxoXt#<69-X2_Z8$cZBwc
z=#52I+@EK_X?6;UDb2(EliY_M%>0QSe6!T*P71QMsGr#TY1`2D-v`+V{~LX5Ae0E+
z8`<1ff+~`_%3(>c*7}rNT@Evdz{`|2ZNi@r3?DHlM4Q6wK3C|K^EPka_iMV%-zlys
zMf>icmK3L>)-i%Uns6hn8DF(Som2WFL%XrT9?kXdUQz!`cy2AbzQ{k!YSpIn=o-2s
z>vGr3U2Zj&+*Tvf#c||n&DLe{Ckx$bh=04?!v|?tjhKPs?V!Me^zqT>83~`=
z4q^A}JhsqI(+h|33U@T$JRTBV0{J+lkJ1^x!s?AuZ&8A9wJA*(XU^vQ9^t~{$XGeQ
z|NRXEq+$u+gOz{ab74zR?NA+W4U;Nm2oo2CV|FFWVJ0x!xb?&zlN+}AMOzEXN^#V^
z^VndVS4BkYb4@8wPUrdeJ*M_zpjkH8M=;%C-<{<8cuh=LOZ0@;HqeX&Nq?)h&M`OZ
zmvDu^%4;8^M=$+?*bD2Nc$Gn&C%x^pC9=i-V{g{w(X}(ffF{{di>;gV$(g#$!GFbK
zSN22W-aM7@4+$~*8hM=}@v|D%_>DJxXYsGxHnC}>
z$rb(x7jV^IV=+m%tRyxod3xygL1=4iHd36a^|{%;gIlP^ID1}W*gfNN
z#g}_`ET8LRFlt6(wjijxZB`nm?Kl7a$TN4_OSk&^bdj8}Ab>*tu)}D-eTU{@`_I`n
z6rTytVW}-1K~9BX_MPJx$YU?m+M^*ofgR?H5JbpsX#6joGo9lDo@lFIEZHwp8s$)|
zJ@O}!pUi2&d2HGuQ*!>=zTQR}9=JJMiRa`ioY%`IyVfISq_!fdS!csGD@@#@K{k5k
zxs~NE`>kc}t+u1LaZ@Mp&aGJy9eUivwA_@8?
zBQR=2SymwJPv*^58kx;j7`$@d!yCyFiXh@N^~A$fS$9ZqX6HEd;^E*weHUT=FR0p4
zmQq+Sv%fu+-6(FVM1$6TzLvs($j`^zS7B~1P3Ce>Zy{^+;KHJ!`@BFdzRIKo3=mg=gyGehK@S+p_4nA6(d!wrP9W&Q?qx>HVp(xHHY2;Vj#h?T{56m9M}2`3_-U
z)aO!ZW#cBd|0-T7gUjpwGS=G#58{gp2kvcth?O9W1$BYap-^n$CFJWYexr`Qx?usL
z*#>8x?LXi9U0mA*4)iRx1UT%Yht4jdT`qT*h}utb)&^cJ4|g{sYYCTa8s_K?4w-@9
z+i`_A7_Baw3V#RFFa=zN!tlzYC`3SCLghJq8
z(V#3ux^p$Ag~k;JK%SA*DwwUE*-A>;c#7Dr5aa^?huijoVD`p5#v@ut!rJ^FmdnxO^`E>4u!lRhzL<>tB;(7sgVpzK0HFa3g2&+5h~+
zsFCfzTidK#_daEERO1B$^c5Ps`QNvdhU9n@&EJjVv>>gc59gCAjqdxYKkQF}F{EjE+aadLtu+ba6J^y)N4-rf)t@xIi-X^Z^uO#~UoZuhbH1J4PE7Tcz+|=R?
zioTc0&KkAWYF3$<9iv5nXzME0e>j-`E%tyMKxu@82YYdGtm7E*X1Coc<1To+#YXHF
zk0Ye=w{mf0D9w2Iu+@^Fs-mTW5ASRb;R=9#IA!20M@fVrQCC55*whQ5J4fj!M4_-Bi^AXgrLh}P3}^BtTRpu@Z~kx)g~e3G
zVBiWk1$bvF=qqG+VlMN}FEY_?+2!^N^)}JrfN`x2XYr44N@Rrn2T4Xj0Fv6_d2DC^
zs~L%SW;Ae3ZSYlz*iRqyf?x0XxZ#VNLJYo7Jdj$v&+KrqX&h|dXIUXC$QZaU(7_f9
zb>E-4sV$izgyqdhoIPRDQzyBzdQNtGb*W89ibe?TLDiG5zZTZ
zHw)=O7F2y^pO22bFZR|h4;F$f=DAd08%Z7VSE&1nKm6y3#bByQZxsH!mJh7zg|GAT
ztd5r3&K|l7<-o_VAoozXHSrtls)=$iZ`UI5wYtVJh+He)m;XxqzMo0WPk&TOji(fA
z7DTV{m*_-*YgW8_xLT=8eK%KSB65nL80<_mQ>H`PvENx^!le=*W84`CH`n69PQa%B
zZ3*sXe>KRQLsz2CdK|v){dAHidXV|3d1i3n4;rc!V*dti08^8aNV@A9)pQoWKc&BN
zrjZ_7^+;D71yiu<1B-;Osy^kyW{(n^ikn*Ft5T!OW>mfI^;!hwy|Wv!0(b>@H2Bdx
zvDpd@>!t)j_q|RodzG^oN-;`n|>45?Ztis=+Qwd8$&cAWRj5qW)#8zAJYJE#(Anthg*F6q&y^z
zRZq*d&d*kC!Vn9}1hrdfCok8nn@~z`blyNpvAFpoNJkNpC&5gP2DA3nmr&I)aX7&a
z4@Om$6ej$y$~jS2M=N)~JSR&2GxU{Un4u$z*f4{kJw|l`Z`x)2uEAmH=da|KYUR4T
z0pvzVxFDwL4D0-+R^W89E`v_W$Kv#6vCobk1NI(+7TB-+Zfon6XeR0XZqM1GLsSTr
z`IFFZ?eFujnjeMmC<>8gS0!ajYVcX*kDWmX$YTJ>57I>3
zcE*$ie=Hy6REglys+HVt>sX9{*%U7=XnGsrLQyIF3e0IL6uT~LK{V7)xzK^
zK+?EBjvSnFM4%A%db%D?VDyP
zK{zI5Tn4ul&HJfukKCSAl$#w-kI95X;_VS!50!F(vjJ`17qx-;>O~4hD~Y)Ec;K8(
zBO@1mfK(t!Chz+xao|42%JujINckBsq=PmWn_PL#-cNpgBjNn9<^}B0v@ooQ4}6Gn
zKqFv@1YXkgXDAA6FEUuKwi!U48^L|$!3Oi^2$KO<(DU73p<;V%XTAMGt;*(0mE6^p
zfG3$SRD53_=&lMFe=q~QTfv8?M?Z8T4yu9j1P2qMxm1S@fbZH;fewBHb0qu@*`=&Y
zDl??qmP+L(3W`{L^*H}I5#WilVbY6q2W1?&5CPy-Gb`gUYIJnqk->ohG{C-|r1kn>
zpJm7Pe7tYEL>pQSG3$%z0*<@)eV1ZSp(+@`5cJ<-6RuP^`nWS8*zhK7s5rZTT-hER
z0EZ6wcric&fs@1j-01|sP~x)TfB~c3KxLitn(xa*-hcZFE<(K8w71H3mMw-%K;7rr
zz$Zpr0#@C^uroM*W?~)i2z^92Y>vj?%7cQ4a4+Qo{ppBpVLoX7eUNBs*ney4gcsy<
zb(n8CoGCbG5DY;fvnSZu(=2<3U8-G);dYjaf(CCC-P!7ORy)y^$!m*6Eg4#QAUpp@
zLVTTb4oTgFIhn8v&djwPC9##ZA(&2!tr!WF%W4UD%6@&tY-Lhg?zR|KI>F>eg@0~osNs_6+{Ipvumha8j;|L0Q;4wH
z*e<^`ph6IJ2yk6tFpUXtmmoL9I29qVo1ny8bjsPU);wuT7@-hJh;p82EI^Unz!Y?S
z#=joWo+x}pG~%5v>R`>tM<#y!Z_R?ln_Mve{LwQfuSGF
z1mAvBg#!d>Tr%8YshJwA;Hjt$((4HrG`s;rS<6~Y?gM{W)qw151!|#9b-KH{N}a6t
z`pt&{-T^ZqTBozEH&f+$+P1F%SS?tnG8(lDYgOp)a_y&tH50(I8@4_(2;#MwkbY`6
zu)Tn5(dEjP3_~rd(?|I50;p%OjT#(iM6Ft)PERKE0ljtGV|M}(OcMMC8Os0oD=Tay
zdSl7;u#s>8|NpsH5`Y7z->q8=DZg?suV75>TNndmhNl$u?Zgzj59G6-56o<+kk&ss
z+2}txUX#++tuLEK#9BRSmW?GDNM+5k^lj`+%6b10@R{y1ogcLpB|TdPOEK8wTI{i*3qlS?`Tt4;Oo
zzx~quqlQg_DiHf!wJ9Q(`9L6mt&P9Gk{RAr89y6+9S+k+KL>WEFlJv@{rMmb&}
zhl#3q`<+*{v+Bcv?pRSw0u|OAwsFEGr&*tMX*Lv_26DLx`FJy_5EGo+@p#Mpx=fA$
z=7M2sAp*bRcSR2NR#C@F7QbU0K&v3hZQy#mR^GjbeMFHK>z|kmTWlbUA+Ty+e7IXk
z@l1EFh|lq;*i&F8rlf!TvkdOO=#?hd?ca<`$;3z|=mrIMGtCYa@jEWBZVslM=3c!SP%x9AI;H}&H^{2;{{2@xo}p|3
z+%}&p*@;5=3r--4`SWIRvK*l)d2}8!D*ohwn
zTgC*y1QDNIy04qJ%_6=DyDp&=CL2-AQCTHt855Y!I`#uImg=
zff?-N$4)pxUy)aQ=vA{WS(wDU96nK2ZQ*iabZQmwbs;Q^S(%?LlaXkcIA=JluA
zOm4B+T~UwK=>a%E80P3dN7Vu0QVq;H1|1#<(X8xp^8y};twwWYF7~F?eptQ%fjQFM
z)zQ-z2f*&muNYS}VSbxxhn5PQ*?*)MLGhuMeh=Qmk=^9l-~X!Rb|`b{qHfOv5r`3oPCjTHoZe6Vv6@SWSCKav?Uus@?I
z`EX`~L}CeoYu3J^E`W((Hu%#VY`3dRJ*_?~3BX6hZOxhw%k-OSv}geKv3{eIE(!*7
zZ)=ecD3afx>Vy{fU)f}E)y7Vz&y#jh;NfDQ)nQndiwD6YqFK@)WRaYqpVQ$TL&xrU
zOn!z6?l4<*r=zE_8Q|C_2Ck4~bZOIw6Ox_Z4X%Cxs`K@4!{86x@3CIKPI%VfxN?cb
z!Nm@OUkRVJnr88joq=!&99AP_|K{ry{U?Krks0-0Ijuee5g!sNDY^`@>TrvQ=ueAV0-q6E#i@p3y&hfZ%yw)w!F5l#WN`phMP?*OE^auCn9Ofy(fzi}d
z1n6w_Yc={3oyAbq$zx#GMo_{zOe3mqMaC*E!^LLu4~|k)T4nD9p$+3`Z!&uErp2Y=
zM&h0k$DLo;kCkW?kG1>x0PK+_Nc+j)H{|hRp&$3@9kW!i+2nN9eI&r+Dvc;AdhDM=Bi=3=M2jSRvz=Zk#KRau(cPo
zVJL+M3k~2v-6%}TvYSk~@eFykawQu__7Pa;?gKaEd8@|)>c8DL9_lPG(A5QJ^C~((qO*l&i}(=e{IKjDOk;P?
zyH76^5H0}*_w8LZO%xtuFu<4{;hPOI=oqg;3=!wKBKd>9>_?nlffp}+z!Ju=0MeKx
z2q20LzgVjSK&A@f9YY@1T^e21Y*D}Y4GMcC(0u_q2iB?&LY|acu~eVi
z_DGI|aDhBDmAD8K#(T-H=rG}bIqT5BuN;|*u<1SeP5Nzv=wq%LDo
zQrG;c`GJCP+{vMh08zUbmsDc&zY9WE?R9Z&!f{rI3&nmU7xv`#!(Cb;TYoYm+AiZ}
z>ZX_HXO^21M@6q!=150UY_{40o+yC9SP%e$1DDz3|MYOBt8v*+889_}(maOpcr4{{
zb-46d2*2-RBnb8Uc`eKO)4$$ia#!ZroVowHwak8J`J$lR!%LQO%Ufw-uwlAV`=b1T
zWE=f!Z{4ItRDrppa(kjp(8z5epH0#O1mT%~}(SN$wGwgj-GTpXEzowCBZ=3J-@*8>Wit|;(z
z`th=T@EdG^+Zi#_Ur*xsx)UK)6!N_=5Yq4PUy_O(mAGc%UM`67z&Jbm<@tfwUB~iG
z)_(fr)^?$Qz5S79MjhMeu-TlTRc=a|J%jjnogZK8T{c^g4C;t&f7J7il%MgO4$H~y
zRkhPz+?U4Tlb(6U)Vj0H@Xff|`kgfi*!TX_w>+H8=9{ff;X94t>r1@ZI-wBrFI#fr
zy3_`7R?cr;$_}&ZED`_v#>Zio!#<#vtC;7WWNkJNb?D`i?(>1%7kcZ^rjk&SqkXq4
zcg7vZHur&MhU2HKS$pz@mLv4ac|ooQN1G#jwi*9nAjA**k=kqwxD)#k$+=SOCl1XL
z?-P6Ve0sX^G5sgy?{P-DNF14p^!HA+Wb#!UIlLx#BzC@AZIVS1ka
zMfrB#_tfer9Pp@rMuI^h{B2gZ<%4~j5_jql$A&ct
z%zo|$-r|SSxPaL1IHl_ZqKPR9*r-wJ~I;MY-w1Ef@!o%J`|F&xH3leh$yyrHip*@4zmDn#$8GN@N
z@6N0r)cmCUZr^;TEb__JOdoBQy(xBNT^bqS7p%_ZUKtmTv;Y5R
zY?DpLTA}v7PH4P1XQc_pjGnw&S(}@VV9;pB^WUF0(0kQ5d->FX&L(2qx$>^vSAfgv
zbwrGf#zx5P=SY0+@P6}{Gz!}8>jaL|VN6Wgc=y-!6ymaCeFy`uw#T({$0M33CSfVe
zHBCR;jxARozYJhwms@{HZ|ZNWc*P2Go&Vm))ZPr#da-+rJkf3aa11$IZi!#rLwx8d
z;eO?^PRFEK7O0#PAk`N`9Lwb51`h$^*jF*hGL3w#Nf0hbbbTF5gO25i)O9~!i%|<<
zZ+DZYrGr8s@U?cru3It1`MX-xAPNIfMbIS28{s!&E|Oi)BN0%c^OM_xIQ^xqEL{vP
zIndK&rUnjmGwe+L;yJ`h-&tn=sJLme8fi%5o8>Pq_w&-UEFW||^Q<4qu%fpmEx6C|
zK5BNpqe!BUs?6(e~r?OPKg`)HiOJ8{COn#mdUWGP8c!L*^*CJS=@!KmX^qnP1Y4pqU6xJYJNVOe(>$LW9jmUgop{nZtKG
zcyOTOf3nGeZbIsJ5bBA8;6u2~$=8R)HlX&E_3U@VnR1V!|^5bF^w
zj-#sq@;lmQ{Md-9{KetrgSTmRgc7UMEtoqUH!tc}t}$)-9mX>y{oC9!H3=BiiF785
zyAliosJ@_Dvd}@V@nE_AFJ`0ubHxyg+P=+J+{o@BAj=+d_MSp{c~9d|dMN3vP
zuaJ`a;(IGQYOhRe^E9_bB2ZD>HYL$YKwaKV8A1sw^pya7j4Ke%6zJrWq{S46j|SUC
z+}k}|TWL<7q$(#qN;j@9JW~viz<@jLepK&BvDE3NHJGpc95^y94?(Z+NA#?I1o2LR
zar@$q5hvnfMZ#iqruyICP@VyNxutiwwh~hy*ZlxN&3sS5^6nXe&tp(ivh2+-qKT4y
zYOn(}&UY4-k^h-vaZx|XtpWvUeuYKuYrJ{$rAgo%$Wn=?|BXA68~-9xrw2u~dX3)s
z5*Qmle62nQBp6-!7ZDSXre`Izym)+7@2*6@(V>4*I
zDOuV+4l*4Uapo}-X1xhaCCA!2Yj?hAsS`67Sl_a^F>bYAe?@eA8|7{R4sP`KCrbJ(%*7os~pj
zUTF^n9s^8a!N9c%*8%>4U&P;c@m^pT2d9%Ppi8PsUX
zhvj3P!2>$5Kc~R+MltR%0nuY|agoDqM^(%NocG#a1XAzg;VdC(F{aF?*Fl39U*a@e
zSBcmRT5ifYd^PSCt|->8chNAe);>IPvFSJE@2|A_tjO7uQ~V`mYS-#&#eDEVVmEBR
z4L8(pu_=VEIlFJ;pc4y2_h&1w+&Oj(tAQSs&SR5Qc||@YByp|-~M
z;olrg$EjqC{n^a4Irzn-p5a>#ZN+f=P1zsUUCs(&^f{mPB`Zhg(;3$+*V+`X%R@AM
z?^V%lZV$uNtz`cC?}+b~4~iXqYjdAveMA4p3hzGUxE(*is8+3SBZdv{Ed=+ngo=|W
z+nhGro&?1C-PisGRfUGA<4}~(!+-qOIQNEyKbE*pe6eQouy6k1VPtyQwE%7r6*ZMb
z4?Ep8`qAT&Tl^XI{$7R{5!jC6PY#!cpc{36*)a=@=Y^vc4qQpmqcFHyArT>*Du}GR*nfyGv&X(nJ^OnVY9&Y
z_PQE7Rf#fwq8BwlZ3G>a;54-)93v+{?qI&oe&!@viEI7C;bKXK6Hp{%zP~LIo*7(W
z=S}r2oMGR#&UR(ns#P2Dm76fEv&u*rGSe9x+R)k07#+SC+68&eUAB)xv!fSDG4-+r
z8`*L=`#GY1B!J0E=vpLYvKxsOqs);f+$kuMAvJ+OnqMJ}t-O--d3vlE(pSDHVFS4t
z?=4refVqga+NSnIzCDs{;iPOYBLT7Z*9U)oAa9T6A!kg0NOd}upJl^w>K)-|Y;Coy
zYc4NfPORLwCZfMen-6(~pXz4#({u*`XQ*ytG81`^l{_w&W`IYaK4($dHjX*ixZ=yl
zLB(9|l=M=iB4E(`^EF{&brQ9o;QfQG)F&pa#@tHtU@OLzXXaz^M@uN6gVnL-+--ao
zN^=Wi0{4%G2aN1jqy<(iGw2qM%%!NCq#m;OuQ)mKRjlrLU6{VI8-_$L5C~388SUNs
zT++B`+62dPhtoP<#yoOuHGm3&et)I(7IZn~T$y=1`LTKl_m)}i0FRIt!(BDcg^pKG
zG<@^u$!|U0$9Yzp?l6*R`PCkYtVi#bjV
z*VT#Ezw`x5FPIC2iHHPIKX27JO;93|<#~KH(-9D0pD~!Op*S|kD2h(&
zD7(IVu0Go@(#FSe^?G>wH90d`=B8rcmy!aXz4M2NCsns82{)EhB7E+?QwE>ob@Jv)
z;rIvbT&lN`M?5MOmi4a&utwXt6oxab)P77;T2D9E$%$2bN^m-^s9nIQNTK{C(&FBg
ze(HQ4Q5G^>n8{A#babP~ljlKlG&Ou!Ak^`Oll@U@qsbu>pBm3QO)*YSVSOo}8(-M>
z24hKq{=xKIzQu99vMz1v4UcHM3Kit3j&JARh;pRk>2!z4GAI7RB33V7iA4qAgu
zOP0cjUpk8QKd0Y3ddE{;>`fiA!0)@K#)3iOeiViDoaFkq9K`7NJ?R3v8+`~1)~Um#}lKauJ&6VByihW_QaCo
z@{-i|kQV(CZFR(4^f|`ra61T-i_bHDw{mFo@4PzM;oEW*qQCt((Q{eN{wVh>)gNli
z|8##qnri$kLl@PQ(sbYG+Oonw`v1cwHbC(CAn_w^>3mRR^dae5f~zWkvdo|bLkCAr
zYHx@7_;k%xxT6iNORSB|@9pfPN7U|bTCd8zRp*f=VhbhG&A6{RqlP}II=}R1buB3t
zbVv0q*$sF#(EY?v{cMp97ne(B2RUQfV?Oi=#M7;xt{*ZtF3m?2VJ>g%q(m_2e0k-6
zG}ApR+ITK5dcA>uN20rSX1fM?KQnt`w_xtKG^ZCFR8sDO;E@VoYmiJnYk}0JAjf%x
z_80M0$|yD0J6ZHcQn~mX@mxI41fElmYp1)C2?x=jhOVXUk(F=NY!$t)No{$yy;3ga
zVJb+uImJo~m#u->bGvO#;(eL7`<3a`g6e(H=qEPEp0mL`PL2ltTazr~y_FVBO8iuI
z=bIu@KCzp~ywMe@uQ9}96&pw2bnCwV949xrTqEou;y5vtt{#lJi01SwRcSy=bH3Oj
zFU#bD%D7mal{<*s67!m2OF}RY69T&m#{#}u3IWC3ll{cGA|CL}!XC5RhxEJv!V8FU*zcPsy+f-lRM9^6jX%I*TI)oeCBy
zP?&l^2F#B2@%tIbhJ8N@5tC^=*4%*uX+^0|49ncEe;$A&>$KVl2g;bp?fULke|(JM
z1nC(}ApXMXE<4Os!2xIZq0;-zi05dnkUNjgw4u4X#TkanP|zg4N5GcL-ca~JOV+n|AXpf~S7qS2g_
zc>Df8^ihH$3IP6Gbsmpjy|*biI&t$E8qGe5%9&$fc(v5naQXr2&-xeh8)in{WKq<0
z{TkcOJ;I6;_i+B+OT1CSG&40;V_j8NJN0tKL-Ls6FkA^X<{!R|BfsTA_|U=4RXx%@
zOT3$VU5Qnd{cAL+o(DkS$?EFPkknb{C6|-o?MSULbv(
z@TjC1*0r*^OSIc9K8p--!)ueY-ItzVYuoI2_WC_~DTepXi%$?H(
zo|5Piicz0Fu6^mIy}tNj(Pq(EjsF&P!!zi}jc2%U{}ujFn3PWEzRqpuA7iargDn9A
z_a;mH2>i$UoeOX_mw{79Hop(
znwvOsNX0l^t2G!MZD+3@U%Env41ppA>}x61Y1$T{#D*ANoh*Cuh>0zEK&5zDK`(yKt-Rh$#3@Ry2`R<+CY34mE96xy)!^cn2
z<*{N!*H#04yLM79fuS*Z<~(fQvk!T4X4m~cefAs`OBct`K3@tz292J8|0yl|XV0Ib
zXr)@%ylkExf84ZLm^OQXW}@fLkre}abU~s7@zAk*AM8JL409)sLj1UKaOK(!)NAnz
z%9k#NEj#w$$TuiDa`%ny~zqfy&Vm^6F<(xpzR`yDf726pZ}pkq|bRLuKT>o+Tn
z{X!TtY62#W7=XNL&i2OE9XRmk)}6bUs^(t4R54^upVomEpR26fydCYDHdK4jNoVHG
zl^x61Y`~OR^N{cR95|#HciUGkRM@&RNsj${3RCAS0I?4jF6`%yRBHCZrFuX5cI)&>
z=#pFYGaoC~ZtxH=@OMJG6{48lrm{Tw60f8x8WeKUcW)#u`{r7*8x0w@=R;Tw`)*C_hF3H
zv1uJ`4RGe-RmI>=igCR=YDTbmlI*)x_n|tg>)?s_vrPj$RNClkw(Uj9{JAh>%3N(7
zkU4D%yW_Dj$86tc1P-1!jZ{e!p>l}=;Cw1I`yEf7srT}E%kWk)pt`oIuNluHr!T_R
zy@wG~y<5?`q1P{sHB;I=H(Tk23cE40mg2E`m*M=VP+-vJJ%_P&`(Df$-V;+7uEK_$
z`;|~5i%PEqncD3e{(yJN(q#Cw1$g@W1!|TrhF^br;k|v0Y9&LX{iTpWn=l(-aeu!mo-YTM
zuHUBiZn;eX!_cYo(6m-XG^<@ncNST>GAdrLy%tPVEma8TuUto_G%55o{^e-yIhU{B
zLaQD_G*&wg9=BJ`BfoXsD(F3SjIRh7Zr{F*?c2BOcugJy*htU-y
z5nU>gmMvSR8NKJuokNl&N$~sczoSl_I=YOOSkxxZn>R1UjT?ujPoL^o55tEK*UCJl`CsTGL?(0#6GTExuWA%(bXj1c^T)1{Tbxd8o8ESxpJXp
z%a*8Bs}>3rC;%!+kV#G}v_Jm%13&!m1L%5j>C&ZIYrasSLK+up&QG2^8M}7vLb7DZ
zG{&@suti?2X~v8hnn6y@{{aIAVCvMVI?ezs1Gz7JAEzrPuCuk$1SS)n?)R!ytA@UP
z`|33JwtDsI=+vo`ZksuCW(*oM$ktRm<#7)eFJ7$o%cnvwT8R>)DO0BCKIsl{(V|8A
zTCpWyAciekw9r_y0b`93Zw7AW+`0Gr_wV=8;?&dR)d>~(xJR@cj358=_q?^9Bj*f`
zo+Mf*WQ=E6`bMzVC)sIOw4IKc$o$
zC+xhAofjS-^vGyO^^32A;fnha$=Y#2cl=V9=!9*R@B72D#&7
z)`44~W;>b8v-aQA+Ud>nCsaI_@;H%J@lb|TNsRCn}u+G9jHh0|{(k|JEF;Cjw`m76NIpT6%Vp1=8^
zd0)NCrLe2LZ*R_Y71cyB&s!Eqh?)CuA)VserHvi_i+#ha<$I0B^)IB#4HCzaNjWVAWEGyMoh#&X{iM1%-B(iD
z=qRHE0o0c7q<)|zGPs*4?fk>npP{+ZxG$?Dx&Q3DrQHGP3YI@)v{zCL@mtT{psA`K
zMcp_WsrDOAKh!m?Y*AI{gg`}-Y8ou#y*7(
zBx@f!)6O5RkAe{jF>(&xk5SvM
z=7B3uwQy&L8ncicO#(`C6G!`O%FMkAry|{-yZ6fI(&IQu-F#0gzl43tC-=y^s2-Nj
z)Pt+T(sTGpxoR4?@d}=*8?AR#CPf%5U$V%%Y7UjJP%p7nRO|(>h>YlNeFx$6*$b#v
zp$xX{+=Cp&%4=6U^=ejCAM>yAlVa49xEEz>Ho(GVt5LaZN&SEDm`P~%>mRzDt`dIu
zxh3`;I->mYH%4VO_SmU2F=^&pT~0SXa~Cg1@4+MSV~xs65D*(=?9W@WLYJHW@B~f@tZgb!hra7bH}|
zix|k-am#*j}KhDlw
zysRbs-)H};-s^AuVqUy(0gpNZ
zm0;!4PY0-VKI*+a1KO%WO-mBLrp%}{-uHL{OE$bmkf;jkN@B~a>vIdok7f|a9
zfsCn>D;GNHReY@vsNKFN_8&Wic8cL&sYHH^SL>LfgaF7$x7LCP@9=#5^tqZ>AH;|nSqlrOiC?{SHzin67KLiu9V)iP
z6IF*op<;F0YQ}q?4oyId3o`Fp^>i2TcsGuosGUEI@j#(Uo@&j&d2dwignaUT#ct1C
zzK(nf8#5(Olt2r=%y~@Ty7xd?UcL5IKY3zz1|@}Mr1Ppj3YVI98-yy&yJ#0y#KfR?
z@7|++$3EI6m!p<5MbtHV{=x;Vp>6c@wMQ~U$&_Xgd3(hBHyPyA81LS_J1$gr0qZF!`01ygGy|7RWioE+9p7(aeIk|s^6_nb`n1q&AF
zxp97Els9S8MAx0XLN7j;|{UH_30PJsv4
zMR$)BLeO;`8R|7^)BqJKsJS0AW=szzKehK68^B7uj=fqFg#$~LEJ5SOjkQa?7%^ht
z`0?Y~-4KNjMU~q;3T!B(p}>Jch^9@OYDPYfB$n^nw@+IiU%GS&Ida&)m$N?WP&h
zig8Y!P-CTWepoS}n=9rw%c&eWOEH1HBMJ;!&UaG+fVU{2cuizr@4Ni@+GL;PA(6(1tNB4k}>+
ziTzZXctqpWwRAG2Mm5Xd!`^q&?Xi_J89B@vwWT~q5@`>Uzk=zsxJy&1A!XvlA>%%MJbwI&R$U
zPGPt8lOnB8euLys@p{RYAd<&W$&{k;E5++0kIzaM)qdwr;(1=lqKDIbFV-!aJ|@<-
z%#K<0Gg_N!f!5}>IZ!S`3|+?Siq~Rn<)+*auzbp5?D~&c^r*w6>5*}
zQAxQeJQ(*wYUS4~rY@%y!;sBa(L+gp=@x;^@ZX9j);Bm-*rz0hYfe2tBNaEmO1$CI
z%zm|!UXlU98&m0VM=@PVU`r@1e3E&5-dso&ZjApSjh6D(oOS|<)dOyP&D44g)iTG_
zH*^X<3`+=eAG!m}A2RV59KM5DDlP&aSk(CMRW`Yjc*oRINZ(L)JtDo;MHI|>~l7#b_f4O^5>&4qCs=t_wX!;R{p
zdtXW1>8s96l+F&{dJRA4j;CG{LVFZ3AkyhOFSStOg?b>2P?%GCy{=_eEk*aJXh~#K
zlI|rt_gQkk?)32CL+H?~zPF7tq)m-&D;MZ}y#Mea<}6x<*%L<~Tc!*kqq<@1j-YVh
z#Hlk%cyJVRCXV#*{m&1SUV}zpaIdbKNu5jy1$uS)RWT^tKh|g!Kud%+O&WO2W8K1;
z$e1oQ7$7)qx|~{2P^bE5<~b`C%9mT!Nu&HrzeTG?^|a+dzhR?woPhaL#^}DvDQ)X?
zd5UXcKz-#psOzt-(R08sJ>RvPwqok2L5LS8Hk#C{jX(Mf1hxK4C?-0Y>fRAuoy{6Q
z95t&{aA<66e{@eYecobBnlT4wFI>_V3v^e+SO>Jq2o=Y^#=)`eYdtJ0*e|uO+A8Y>
zgXrRj78v1`CKA^@bnGNVixOFDh8cAjqk(k)q5F%SkU$jisWG5a4sA
zdvRsPdiWmFr%Q(`>f^k2-Rovh1q^Ao)@i|J+wjdClroC|j;P7BonQe9AAd
zP{1HiGfNDli5W&|tX9Q>MGH~Bg6HS#x6X=w+_R`>d593fz3u9c7C&hwJl{8?XD_RS
z4T^a&p}%f->fB|VIDJ8<2Tq)W;K74q$GoxX`J>!Lsrq@VH^;=e%QbWT!-o&}Q@Ng@
z@PQOkJs(>t;l=$&pP3+C)L4JEZlKHQMrys%=(ker7}2y#oKec6;KxcOwU#|ya>sdw+&an~&HrqT<-fe>65tlr>2D(m&a&uJh^kDNXXk$eS$#
zYP9|n&FWM_?_ZkWnNqA+rPfKj^JUA3zbDK>m68Q9dsq+Mu597l$W*qzjxUf^2_s1M
z@T3(9g$MPjl-8CcbTQOJSuHSNwV7rwQ}ML!+{exZrFmvNw)V
zZa$KkKS>EG<}BZ!1xw|NvHG2in*VgcR<~|l&A=u@pUiR+^EYnXh>RICYVG(~v0{Pui{izLgK-M!bH8`*
z-nuR|(y38Tt#~TbP%y!A62~S3ox)5q?I{4D@Plz`C@7(jf#Wkxra$LI!3c>~3_W}H
z#L%HbZCYf~ohRM*rAUzi?b@}|%yKfTDeR!nf4+BmwMIS2r$4k3qs0LGB(@Zg(1MQ|
z`I$0h(y?+VgkXQ1bHj!WHRGSc25Qoikxm@Uud?+arYFeU=Aw?!0A
zaX-yOVTVc_7^e8d5W?JQPIhL+Z!2qorn&nTZfXap_sD
z?Y~O#59r}xO^a;EtoQ_^Yl>OETQT8B|BzOBq_~8fi6U$5a=Ly_5&_Fp
zuXr8_bX$9W#_qU=m=Qu^$1l0G%O$$C$us23=cl>-0_FB8r4kC{QtqT1%sPpNiXZYv
zsbq?e^Bx=eJisZX9Zzcvy2c^DYtko6gW}3lMgCz2bX-g-u96rBpi}WAI?ZPU1s<9T}`#XKFczz*u9qS(k*QUk5qN8`T5T#_A82G(J620FPg7jK*(b`0x
z)6!*od-rd9qXPL;QIycj9B(11x?V`UcC18!1cfJNN~&CWo^yL?&a#=NTX4flp{}yD
z+5Blg`&RHiMlFS9Gu9oj*>6zK^vr1t>Z_IYIfTGX6_hq^q$FO}<;{o^>p`2Y;Fay6E0!v%7A`P@Qfv~f2KFoMdlDs})bgjak*;+{*GR2yR(G`vBMJoesd~B9
z4cHcOyol;%)_=ogk77zL4XHFU&Ep%Ds$`1kaYLd@Cb}^){QL5lU73B+%EBO9eHdk3
zS6;xpLwEE8jPVRcRZr#V1{Om3QXg5({o}@HV?I*)X(kFQS}r)FT^Hq4HzmUw#Hd3I
zd#ML78Tj>;5a))%Z<4ZRpoPJb#+i(29>)$v6QSyq~=katAQcVi#v)+na>PvHV7&Zn2UXUeSp@NswT5(X09weO%VzkE$uo2U;0YRa?C
z%yGjoWBwAwSf7h?im|>!2^i=;$DnBHqm04@630xOAihl|xKZCYUc7QmyE(dd{{h;6
zvQl6fKf2DIzoe^mZqp1)R;SKM!}`OKvCVz7>!REg
zDECAmf`>q@sueJO>`)JZ3+vG*^FO19n45S?Qz&AK7|fH_A4A7Y(LtWKu9%O+3FB)k
zllSTaz)T#QL>E_9Vjp`B9!8_OHFZ3PwVSpnVTdwMo;FJhKI3n0sjDejA_O)WF44{6Uz0AD+AY4`OAWL6*UtDB@l
zk&2Zy!;g$F-y(f)h!xGD?s8{Muj}!-A6;p@)6I`Tbd7ZJn)|bcL3SgB52u*0lZ|4$
zQqR7#m#^t_2GgbQ`cD1F;F5APLoIm{1p@{ZGzC>o96z>Z?i*#^9jS#+>jQ(l8jpgJ
z@ZrLEFwNO!%j_?$C72&wDFI9y_y=bHZ#||gSgEw;uYm#vS`~OlM^2v8!WF*nkdi7v
zMdJAO-|Zqt46lU?tQ#YmJ0sk{JEFoioKg&Vsss(hkDB&(A3Rc4AfM+QTrmZRw;^D#
zhbOHm;>J`~)r~r$xQY`{vOq5U-ux%^-WUxG@@x>(k(3*uEXv)KK@_}@&MB8!v?d|h
z8%Gy@`g8r(T{P(254#VaK(a*f&{?^|s$02~62N@E4rxt7fsQSrMTwOd?3*s6%tWCX
zuTv|rF6oPRE>+*ant~R(_M*Ei(xmyz(XQ7B%>3UvEo2%!b1^o}9N}&31>M;6CyrVp
z8PVTL?8CeJ0JX#El86j`5}EJ}j!U8|q4MR+dwe(WH%(?OT^<<}O`gQ*FnJrg+m7A%a>0lX@Oz5
z2ck8CF@K{TkF=2BUQkHEIa8QH*HfWFef|cg)d~037SZLBm1u#&eK8Xisc3;=CBs^U
z*NmB1-z;=6mpPi|KKO&lz#^BfZ@*GUG$yNU9)QClgt<{Y?J&yVpgyDA}umB^bSe~T`M49cm58zh6rMQ;`q
zE|B=Ur;DbCN`cf$-p$z%Fwix)nOSe%*K;Z+0XVkGtWw`&c)erf6x9@(N5Ybk@fgw)>41r;n{CSS_2^&Va
z((#H`ser*vRS@EoF~jN`e)D_@6!ZDKrzTNNQp_ixTO_l)Hv-k>M%xrtKh
zVN4T9Gqov{(0qz&>H*YE-B8claIs{r$ZO}}WA~SMbN<()A3c7A&??}ZcN8h&=La$C
zMzo`~{`iv#FU`AY?oXooAZpP^R3DFzpFBmnyv21swd6|_DX2aQJU^7Eb#Ep<+#MlW
z1Q=v?ZB(~L70?2M#Ft~bLLxCfgiq*!ay?)s&eM2-c4TaG+m-7#X%|EdYFE(K7smXI
zI^op^K3zolgkC6SK-C7#RGf%|_`O{VbZ*mJ$5CJ(v17)-eHGZ5b=WuQf%09;7w}Zd
z0>w&8R<6|+72}8W(?!*l3#tQ0jwx4F%QbEkDs0`gM=_*odiZ(n(tCi4tFcR2s&un<
z=W$NP%mZZs<6FcJib`cmKeDWLVScSP-BS{;ySZzj5qvPXO$vr@zep7Ccn<2e{)#=B0%
z63CG9GxNqCp3L7gYXqh(TCHPZ(A84cAroED~W25r?h_vu|abm-7J&WCsP
zyeohBs5s?5q_FCHpWYP=Vhn+C{`A4HxIg%sY=2lc)FT=9isD@8{$ks)u7$Xl@zwehXQcn${9tAjz+v&2Rk
z2sGw^oGUaE~%V<>BM~*{gyP4rnUJuFW?~89c
zJ1;Sg&e_WUG4Ct+Mq;D$Uv
z0}IAQpwPhJkD2>oeXHvvwx=bIcXUhf*C-sL>!(~w(8K@5Qbk9}F)Ch-vf}Ydz(5=y
z_!2+NyLmcW`Io;cXPcqS$KMe!OuD@qc+&2hS}T52>ug#|OtZJ&mfLSZS41SbJ2IrQ
z)W~0^BDTkh7*a)Jzl%reCbqUk7R?xE-OVbtx7%=;-l1Q{J
z=q@N=K%x`M%&%7V&)VgOlrKUC`KhjAP?(7~(?3^TR6()h>i|wHV?7Eu;#q_;=5|c;
zITt}`#Q6qA2p>W9XXnZJ6$2zqoJhB7)3gDysKC&qBPUK`!P1q86DubEnY#!N6{CLt
z#^uPGF}*Hdr?lgJgOYt3n*oh`|KYCbM00AoXG)~B`AEaYPf_lVPUs-YiozjxFJ*4x
zDZP8`5~NA#`I`DsSrr^Od>nh$E>*E4(s-&*S5l;7r`)vF0n^)e??>ts$!s!(xku^K
ze*W#Vew^4baaC!Mvae39n=wlh-8mgNa$E^2k|BK>cR`L(W^_z@(YhS1MaOM~^vEalyC&+GeXc^7{ir@qi;?
z^S)z(w0L~q4r2eD)myanlwOJIc*r6EsW2dcBnQCnJ((wp|4*k3(S2U7*>Qm2kT}CI
zIUu!$>}PI;s07A5@I#zu?gmLf)O?*Gj`C!WZ~{pCELlR=kdh3QG`d$?vHs=DR~k2W
zMI=&eu~Y901kVA?S%tRv-17;b;>?AxY}0P4&QME|NP2q45JHlH{_;(4u^ZVKaz}^M
z#@MoO-W+s(%R*YbVJD59wVZ}eTTFF5%TSw!HHE<+^N<>C`qAZEn*R0s>D#U;XFdF(
zQ=FIei25*GLf=@PxpGTO;w-NW=%geZxO3CnS6eKRG{YeQXe5xA!&y+ZOfgXhPK59L
z0+1zRI&tkm04>RtTL~!m@iRh(k30)Q+Am)w<^URfW3qg|x$7`;>l=yn+k$%m*!+?J
zYae_7fE&U-DN8`w50YJEIYcGS&YJlg`UXD%1f2$ywPEJlD>i}>eQ(tvROgu3>Yc&4FT+x)TbzreSy3i5KwY~vjUJ!^7{?HCpcU51<*-=x&pkG
zBp@hyY-8g(Y&RLg
z5gghwUqNPuW*%Eh_r3%44s1gEIY?46;7U+KQh$b)h+>B=1mwt)00#ON+de?JY>glj
z0s^R{L7tjt00Kd$9`P3$1X1-3s3I7rr`69cuKm1|!8C5aFZpuzgIpv%0qbANsDNz>
zq8Wp6+nBtci2BhVGx|Ul2IK-~kmL%_ngB!*91rH$%RHDT;ZXNNw36YGCLi*r34G2q
zne*U}B6iH{kxDq6K*Bufpr1I7{#Q26me|X1QF4*4C5WA+><(nhz+MJOBYi=Aa9+Rx
zh2IF4I3I9fL?y80frNZ}2M_JBq|ApedRY{P`m#xy*ido2FN=H?;{7SBCm5v$%&PXsvkcQQ%v^|~dS%}uR%EQ?ZiNqKR{C$B2?RZF^+1b&zf`ieI^nj9%
z<-yQLLp`0epPrnG1wn?y@0?CnoYS(PT0k@1V8_F0#w}dT59Lh!NA1oSbx2ifwfc_8
zc98me72CJD>Ximg}SsjH}s$H$Jh?*Zf<}U%4wCCVaYT2a^?LB;qGZ>N!
zz5`VC^$(<{&z>`(@jNN}-p|{f1H2d?l#sfZ-y$VMPC)WbBhZt*uU%
zh^RVn^rRrDzJ2e$Aj$?=Ha?fHvqQzR5~b!S%6`Mg37(AYdk)abjhnf=qUZE6_RqvM6jjAuCptf7@&CiW@G`5Bl!>HNOe0D^S
z+~;qw>CMphT>n0$;}q2y^<2fou+`do{H%${^fA~l8im6vFd5Ye*EYvn(JU^!=aBzqEu5xMoiS3nQ}N(h2;10>O8=9)OmM(htv6fI?h
z8XIrpniT~(G@L7lwC_1~rVb>Ov360F!=PQRmFz^+>M-nYnzm#ErDT|;
zs02}>gD1}y?ccrsNI?u=^o9SsrSq5CH>g@(td5_vLJ&a12)}6KPO4a<5ZPynm{Uvj
zqgn#yXyyop&^fk=rm`dFzHbm!Yu%Uf*K8%OL8}HeXlU1#20tJN*Oz9sJ;m6EOYS?uc9Xoeba46KRQdYnoAsYc)5XeSwXGkPC9~?7hqDxF_-fiSG
zk+E>%!ZqqXdb;2#spnZvkeZL{)1LVkG@qIkOA;Ih*oO1Ol$h~)X{xz)pE>s|cnuJ}
z4*dL^H*X5VJ)+0~wbP_Y6T!s*s3HXFAfo^VU}OtKB`_`nIth*n5G;dCdCZtGBFY{S
z{D{bZ`t&Iw(_!DfeS!-CeIla(g5&4RnUiMBm?0pB5DgE3@f(I{7_)F?qD4d^R43#wG9B8Zyx>+5eLAPf8q$hLq(2Bh8Kb%65&4ksA<;cP(v
zhCYU%xWL~K5D*~9qJ5aS92p$QV30B)z|n#0iTNUa4-FbL5Mu_M(e2x}h5o?(Adx-m
zkJ;R9bM6DP`;tLio7+ZSFC`;LAbS5f8^#gz2#3I|&L2`vTk#f+u&}d@7>W2k&@b5)nhz
z)Z{AcC#76cv#sbcXCXXg+t+*Qz1f@QoCqSEdsG4*fRTG1(mDUv!nOtbSt@QPovYOT
zw>+BpC+u5ctTzOFuDbAu%4SO=oH0^8bC%YQ8OMNe9y|!k*+Bru0wVt>AM#@!iT`Q2
zH=-H~921l3xYM;kC1@NwlOD2DrQwq6REBvR@H;5Xu;BeFIT`$bA5>oefd$=-WMXL$
zOyhU7iEVhWH9`Pc)pFG>Iu%aNjjRd4ijQaKhCiPJ$i-iArbq?%r~}#xux-wc?R!%W
z`wM&6rNB3Y-xQ=`IFy#yP``9E5cM6QQ(fIqHK5S|$VxkIxJSjNUJ_CDjrr!Sz>%ku
z4srAkW90E%Ku6$Fp25&Q{kA<2WX#ypr;XBOXMP=X8mzhWSP%hqV9dSf3cvRVk(M=>Sg3y`2f_PLef~AE
z5006+Oq`7rMWhrSWXDadaHji)49^6T=SE9!2`D8*34kX8q}Y<26CPoR
z$7I+b6)G2=g@Keoz%x08d-0_{{_7?5FPC5b_luYS%gdzotXVSCvRPB8L+`;DX&d1U4Lwa`=j)noFCh`kwa_(FAcv9bnmeneiqlHY8>`|^d
zeF(ls3mDR;ahopmN5`I`I$(@IoD2w}aSF$Mh3pnPsOsO#0NNn
zOV@0qC9Bs{gId+7SEn`xS{N?6a^tp`XdgIwf*lPfbSiak--@P<8&2C+FQVGb+EKv@
zHN<{!P>dSbTdV`(Y4?F61SllP4AvPY)}#KKA@qZ?3S~+%Id@;I>Ii-h5Pf~|J^BiGXtxg$jKkJItMlVi8@~uwg^1dR4D${JgIC
zQ7o7{hguX$CkXhg`jJG@SiOIvP>_k;nCrCSYqVhQY??EFp^2I?ze_&%z%)df86Xn<
z21n;loVQARFp@fzOH;pgzYFN1{_UFyVq#=90CoxF(Nh+$qpB_Y(2hl#sAu`;ah#~4
zOQo@(o)pcDK9G@Y>Uh4#H_^=DJ*gk_LF_wmo_Q8hQs+PFQtO7A|Jp`1DzJlOi|BJ9
zXJmkH;m%WArOZ)Yg_&=BG%pKjAe(yDkTJ#$ys)18t%HXa79!SokAoU^FXH|wG0;zKQA%BSv
zAdcV+0Z}vz?-10_pVO889@K?~Oj$q+nV1=Z^Pe-ik61VKB|na#{V2orY1_D#_`s--
ztOl>;o9Mxj4aS-nEU(2zeq0L>jL+ou$WWL(sH@iC|BB&{;7q7rwVY@_V|WjCM1%;Z
z0}7Ctuxs&TkrjY3t(-E9M$KGG9<^GD`uLqLnK(#rMu>tq@YTP6e}XYuG8!XVeEIU_
zA{zikd|>;-@GME^&YiUm2JstYL%Jcs{k3Mz8ezx>cL1`WDpssWvuDrd4A|r%ihta=
zapG7hQ>G-41cPw8eEISO&KC%>7?8mLW49z(vt|{J4iG{YDpW`m!vGDqC1V2KgYDb5
z)6AJOITI_fFvd$+1m>Kh)Gy9gF4DHN1kVDD@^Gwx;2DPbR;^l5%a$#L0UcQw;Cs-Q
zsLsido~2_-eIc7;`t<37XdC3*5bz=_T&Q{eOZBCt)XvZnP6b>GIE28>07u2zwQI%F
zP<>-%AOgqrLv{o>H{gJQ!>C9RjYAJk1Y~-^!GPZfh_~U~0DlF<-rimied}l8Nar9e
zjUE5LESG*M;=e}P-~qvI(7l`kLoo*kbO+cB`c!nJ+3aL!u=u(#E>EiMM#Gt(K@cAE
z{2#{SW`*s<>C|D~M+8mcxjwx?c415Eu1CCXl#sdvx
zOqYfZwsRK`a8suPZ2CLwSw$2bvJfE1Ccr}z4!GNe22YZd@9N7#>7%w-a5gb%g
z<9Nf^y~ut*RsaO)%!tSrd=J@OE$S0}CF02{2%I5dOj!4VBO#p~H85DB(8QND$CkQ9
zA2=T+(%hv!r6u~ke)}%j+u4bX52@VP5}6Sf8B)rQ;T*C#IU1{Gc6solP%2xyDMw@l
zncc=%#kMV*lk1Dq)Us$~;yYt)%q{P~!-u12o>7Y;89AEd8*OJ|di;}bYL!bE?JLi{
zAs~`cFia8FA+3sL4AI#hW?kF3#D-f0?21r
zG3|35PK055{OL&7{Mslb)P>Ypd`mbv<}`HE6!3SwCAUOc^s#OlcmVAV1y)vH;taJE
zMQFvE^;Y$!URR>{cwOn+f}6YPZ(?8Cm=+?17*#u30#pXfs?>5D3p!hPQs`Q;$6TGW#&xBn2@NBhvOsL
z!p+G(67&o*E5y8m=RgoYh9U$8zjRK>W`IE(B*;+-BI!Rfgb}d*_2*qtwKcUL^O6Vt
z!StHPnlqcDS!0p+hjRkVPmz2%#kt7$$=9KayAJ~doD^V_^#2Ftb)&ofnWMoI%Lr&K
z=?CNUX42_=oFRd05|!Y*0o`SYs{w8CoA}84;Sfs5ydu_0uko{~Kp9U#vTRLXFlGae
z28Qpb1kgH&w0Cya7@y?z5q%HxUqq7|!|<2am#<@f1fwx9_CZ1{7m!K&`uYk892k*7
z^oxuJkn%<)FyO;60*3(%`o`{)_U+q?%#5>V&x)vc7_C#MP93RxYtB*L*&F7t6Ynx<
z!!w!q{J?EctXMIAKhnID$m`43%{?c18~t@0ABKB7em`(>a#HNx!`Pw=GlVgL^cxP3
zsC0ny!EXY`vlbdj(hbP7Cr_R%vNNQz=q%HwPC=)-=QE0M6y!jmsfuw*2kRQ^2k0mslP@s`mBx`{tX_O9w|=$|AGPw6pVR;G!_yqE6CT+5Mr(&~t_5Q%
zL=-F1jf{ySJ1gV@FSRMb1I3(&im5)}n24@(!dSo1O(MNNPGfbWQ=Z$-^dFPFPe3{0
z6J3ewn(Ft9KT!OcDIs`iH1F4F+0l|EjiPOWHV@pFpA2H!vK>u*+6w04hI#iTPynvNvP79?(^V?-ByaICh;(2OR
z*8Ka_+(W{ty5Z_m>Q&C3`jpo=mR#8RH|)Atc;kDg7(R2#i0vUcdcGm
z5Pb4zol$x*T
z9~1CB#xUs-5uPtcHf<9agJa6MfP+MJJRCm<8|Dp3D$d-L3mkEBLFR(F;^t&tFoe1N
zN%b%<$?g*UKop#l{&`7#%a_mvxCUZUqU&71m@vMg>?AQ)qB>`ZiH^YFXAI&7kVY^%O9JDwInD*thp}7Yc97beT4Efg4A$11
zqx`t~P84JVNE{3DdWP05Jtsqb;P@p=mNc}>s#aw-v+n;jx6TfMj
zcRNKnGz6T=Z4@x%W=a@=Lyv7^guePvLlWkEu;f^1
zk9#Vn=9l{8i*4JggCVwEkytRE|M9~UMkrNO4Max+_yd(28KcvO)CMn&>aS7KhPsSt
z4QPJ5mPIuf8&l|$r_7JyYHWA3%YB}I5@hRT7z~#$drHrX{9bVS{Rt&XQ?_Wb&vAir(nq(Bp*9@5F)`*hx6>G+!wpBIo
zGiX4mIhIoW#AThsiD7BhMO;2#b?lKfAl{Mro!eT$JG%niSUU{0y{dtDr2*ViO4-BW
znqpK3LtMLJKVZD(W-$DKp_qLLq>&*_4Lk#;1gUx6Ow0rCGcep#QCz+5|baQ2F)96J>80%XtH?0AsqJeOf7jNELoPue3beg?k
zoPG>Ro;bo0lGeubUwJZFDyX?H5qL&g5~Y)#$(jMzgp^R}+4_pNK@5+~1bspMAST_C
z3+h|T1tITGzMhmZLtnmmCohWDx_VuS3H;*Ubp_nocg~oLX;JTAWi*Vau#X?DPdis~
zzM{;T#gAS9xdcp|RZv`Av~JPHB?N*?aCaxc9fA|wU4y&3ySrO(cXw%my9IZGJ9q!*
zo?CSvyQ&|Ws=YRI&H2qSNcRY`RP_gN3rhwQ+Do1lABKETs5wTIDsq==Sa!)^#u{lL
z=|AVC2vVaI7$_S`=_tRIz4%mu=*RY0I9OkEX_Nc(Pq_a%=ZbMsH>#$YK-8yiJG?r>kK!+PfWa}}C>>`c$*(4V`1Q20Nk>;~RD*de+nF@&
zo|)&EE(DThaIb`{g2$>3bT@+0uMIyJ3!0Up)RP-Rjjxz=+sZVMBS$-&fvYj+Nqk<1
zcc-$0=1Ur{G%1Ix-oYFGwB>$^``rKx-}sg&RqpSIlL-?q_DKD;poaUwirwpg#Pz;X
z;b$B#esw;L_x7ngMa5!aAs;6-)JNdfQBnGh`P#+X?cW7DTHRKXx9^tBCg7x*ZRVI<
zehH%zx0$Mr&7_Hq#yZ2gnTk9Y~
zA*VqYIVzDNA=oG^^4#2~d?}flUsrAYWefu!yc~WYKcvrg_s&uFB@)TL3cbss$~@<3
z@xGvWX{mR>PoH+K2#HXWCs6teD~hMNl+N2O&MGRZxAf;U{g=^|i6mYVhvzCo8oWu-
z)nOl|#B`jeTyF`Q+hZhJt@%JpHCl$SoZ@b-GhW%7MQg00X^Xj{f>qh&?-dgG(iOcQ
z;Z9MxCgs43vE1ix!zS@JHxkKEU=dmh;30FmSES5ytSCOp*|GUa;u1h^2bXEJFrpCi
zuOEGzHOcVy{2pv}{Dc4e1;xgfL4SFeib~YSX
zzQ2EJ3~4Jnk!bQAt|M__hN)BT6GKnwcUN$oLE&92R46(!L0@Uq@=Wb!Rs=6+Fuq>B
zOOxF{B3_emi`&-Mp}d?P$jGzdwa)o_GL=9Pp#qn7o;fyaNtfzTiX-z|4NY(9IX7S41%FZN*I`?FJZ%FRahnWOw6vy%_e
z`!%TqtRuPr$Q)>5tMH}YU9hg}#FUcx%)a!R_iRs+Nspk5hvQ&74(fhL-c>SCPl9K=
z_s>yW5-BzI$^*^r*k6$H&WSm^dUIRB#tAHV$-*u=v3_@{d`#6?C21aTPdY)4URD9Pc0?s4w8rFIan@_
zQWlhI8cd#9Rq@3?S&7|Dwlrfmb#;U4d5rT5tvb`bEsOc871_v~h+9^wK?LlTe3w8Z
zJuVk%pSe-{Rkk&_avQ;hSbEjm-D}Tmdx7}e_0t4zv5Neh-yuX31&0Yf`qVw8IIxsa
z>D2#onO7YYyRWt{Wddr!jP3+AQlkBX`GRB*pEDog|Gs*uL-0#oJKEc40q+8gqLT`h
zq0~m%voNPV;Bi2ZVPax}8gl{A|FC0sKriA4=Oq9#AOSO~RwOlV-kmO$1Bf<&`7EFb
zNHqX|L?EDIl?0$_En&Zf*9^XCDxUBwe85$`C{9`>$l#n6o($30D6P)S1llOHcZw?l
zC&d@Emem;p_eau&)(_qjj}PsHzEga}^9zf4pyfnN922rUUvTSxtso;9
zaM91J)2wn@l3s<-?pqd@-GGj?^Ci%+o}Krir+@;=$|he*eh
zK~Dn8FFR)I?)#*12aSgQ4EQakv)ZG&C;koJXQ5C9DfV}?o~v0V6JvG0sMf*0tGiP*
zJWe0P^L5{_dS+y$-z6v(8!3eB_K$))^jC(vUY|JEU2Wf}I8OfrN@Ft#t5t6mh>XZ9
zuCkX~DAX{{m&yHz$7Qze$IRd{Am1Oe=t?vgKMN*>h8zhlGEOT&Y|8v|5hzjC_iMqK
zaV9>|>aU938<$7ZL3k-dY>deU(B)p^-2H$TCr
zIHS!v{qZRs95kDB{$q>Ytju$D!vPOAB1hGgM89Lf^62i*kPP
zdlv>i=qXmMTMTtq@K~3dtAkq0DQk
zSc7M?>>Mh0HkT{U`l0$N1X}JN
zkKLs5I#R3I(6ZtuZXC{={n5#>3I
z1-P5&8mv%hhDc=I=~l~
z1c=XAFZf)xvVeeavD_+}5dgXi|EGKaM+52Vuxh~nCJyZT=7)c#Ld38Vd1KOUjRWkn
zBqZ5;FKShqVgOkf%jPq=BuPE~7bacVwU%EW$ArikA55nxof$7m)y~d3)sV}1ltfpB
z(5@?>A&tEsNG1ggOS!YP6=3T?X}L%+j56?r2(x^-)61|k7|~ZUWeCnPqJ6?;P_lnc
z*Bj|G?nh=ODQ6iE74!-yS%Uz37goisOKfQwbQfTwaJ*kCSn92d;DpT0jmT$SPqodi
z?n#pb^5mWC`IejEDxKJbN^0v32?ame@v*mCZkPkQvaL$`KG{h&oTCZcEbg@<)!7o3
zNBfwKOVMmPpOc7-(qX`^(+m&6kTo*t^uoW?0=BLLRR_xLUjBKm+Mn`>kQ8GMe~6h0
zm#qnGs_=9$bbY*LgP(iynK@l>F_f7qpJyE8A7lRxr23Y(8i{6PvU!$%-F7b%Yexn3
zVCQnLKKiP}eFTw@r7+TX8|#Mks^yM`5^p}!I38UR?cND(YcG!-Ha&nHU5v@|D9+*}
zN4O8#misU=VR&J33Nc8pD0U@5A`e1j7DLZd{b?Vs`*fBYUAd!sM_W~a3PVp5YOrw7
zoYhsC#%iD<~I>@sgQL*vOWa9wr*$Z2%5ND
zvLkEk&?(U%g_AM+H$&t(o%Bxn-)|_m=gHi@=wv-RVTlVlGIpf46~RUk?nA_v^aLK+Qq)0LJ)%Y9^{HM_tijYcc{
zZ#?NHRUA#pSv8`EkM?apYV6r87MkDY^QCM&-k(vL@LyBYZf-5sM6(vhBA1STq=CB8
z#NVf0V(K6Tn3{fRW}4Ql)3d@>Br}L#@_UiH>v(6YvmUzRpbLj%P!^Fsk>(L<6P_Of
zM6W&K>okWp-eV^LZ*??=`A@ARB@8Ph_iEJD>4!-)S$f>!MdI+iS12Cj1k_6D?d*Ec
zc$KcFTWCpHztaVtm(?18gWBlzgww8@&5hk=}GZo
z9d9Jpg{sDnu0TGQztL=(=wz#>#|vAvT00#uUDA@IKtu-J40#_NAD1g%nyvqN_0^ar
zEOuZ{d%r_AFwa9#%TbsxOh2Y5l)rsE%&8S}W++6+6P_11V@1a~2S&=`L(81=0jEs7
zgu*&pnSotIwDn?-VcH+jK^`onwkBy-46^Sj!L
z-46$+_{_lmM=5Z4t99y#nKzeH`mWA4uw%g9!d=VI>!ksKfBcZ}H%sj;(^$s9Fua*-
zdssOG0myeV#w!e;j&surjiV7rLWeINxLVgZyp?#twGs~gT-XUI(XT-I#&jUQ(@$U)
z``fxQ+d=kvX1Eg%{HWrN-)JsH|C66X)}g-{ajmq@#4=K>dx!JGAZNF@s_W0FOTo4B
zWIAbanBipP5GwiHxohW-KY#c)8fTtu8Rghh+K{89E(k81+}3#CzrP9(A-Tty4IDTD
zTrBDS&yGf{HW!9?+#Use8$9bcrl<3e+1}fZenm27bSaPHqvda7=~w3~H`uGJ78F%B
zc>RZ1_(*3CSf3IfiwOQLb$Qb%Qv9kWaC#iHKt%AC)i`6F$qAG=K5oEHqErn^M8O)^
z>@NPNY^L1O%s^w6()#&4PlX-biG=b@T@jcP*PV?K&IBI!U!u{Mj$Wdr#
zKx3V7ICVN;5>8fbUrsF&&D<0^jXf#tsC;SRTkuHJDXs4OKTh<4g3X?B>NJ0x;IC7w
zn@L$U>U=4!^F$hOVYkHVMTDd-6IYr#5h*(3p@BB
zza56#iJo%hv{{p(9?U&-QbcI>iZF=C7%CopOq++O=PHxo-m2kkpI$&4cp8owB
zRev4sTl`l@fA-Z0SX7XfmR7bWI=U=6x6tiAG@KA11`OhLIg=_61vnY2gAt?SD0ws{Tc`@Ykg7L_aJwb)h{m!Uh-lD#O7AFL~_CPBFto
z5V(W@d5GInsJ*;_m631{OXAH`+Tc~J1Y>tl!8_i%!e*B*;uBSLIc4wb=pXr!w95&cGFzqw_yd@bR16F
z<}k_vVO)_P+GF}x_y4elAEo!DEtsIgVt?@9^cwg{@ei{62A#fBBm$x+r
z-a#PxXe*I+@hMvyF(uzx4iyFtM1U@bnn8_PUPJMlYc{n;&s;E8PDwFKofZovC~;_T
z>v_q1z!G0;Abt4kp>&w;;!llN)_<+upU_+b$qP#PklIMvpZ{z1Em|t9+)L}InMjzl
za{=v)Dq_d`zsD7WZqT`Lsm&!E8<8g#XWa
zszB!bB%dBJ>h=6`9_f3iAGN`F9|SvM6$6DK>l478VwZz`v7D-cOjefzP(#
zC=qPi`xpj+5%5XLm#H;NQ>p&w)>>pOAPuprLkB^KivH*B@*!#-8~)wU$&0
zJ*lB&#?^!izV_7dh4f<@UBc%+0p_srR0W07UnwDYjuHo
zCW`8C3~wAhv%7Y@xu|0Y;A87U^3mw6!q8?~D_By)cQ`$<%?Q`H0QV&9Csr|Ia>7lr
zdA1~h&F#TCAItR~ay+#ZuhuTeYhH*sfcyaL?qKXl?lqZ=Y%vFKc=&+aVLb2+$UU*3
ztN0@$wdpfD?VVwbAD@Wcr_U;9igT1Z58x&Bn9{8d%YnxFwwt^@qhx%`tw^PlQwF)3
z$c3~fCesLPE!S56Pm9hlP_BS4gbP^%Q7WF4w3XW1Bl_R8dhS?)Sm$nrCmN{?B(Wnb
z7#^Ju8e=S^q4=wvA-QZ?T%5tz)LLfBBokW@92^|O-hU}N*d&a4+-=$APIoqUH#c-{
zJd$WU4h(IS?Tp72LS*R;I5#u369H8p@7;m+MG%E2cEXv!FqnyEU`hHqGCjiCNDGbs
zK$;gnx&bJ&traqUCZN#54GraA;_4Iu4zhO97Y{$okTzK?0)1Zupc;&WN
z&pI8bxuuo)4A&B~ttCgVp~d7&oKpye_z
z==ZGXWFMcs+wi`uXJvEW@vv;P5<>oVzCMFh9El4w7+B11G&p5MZ1oXei`^^gop_9T
zk~zLfWnyRk#sjB$>x?AGAMRugL9l4w}y4|7E>Z2d{ff0wH8Z>ymouN*5;@tEzFyr1)KKRcC
zACogLT@nXRw864~jA;2O?S(fisTBimN5aFU$3dB<6K9ewE1a}-{LIViF>KEtj|JB`
zTnsmA+Nmx%Z@Oh=8*pz_j9(o3(&YcUy(@W!PG*E1u&LQ961idnVmU6-16=9siB1kJ
zi}}f7P=}){SelA!a$zNvtHYF@tINCG&41mH_>piuyT0CE`62mXNzb>q8rfZ?#R2!(
zQvq8M8`RNgp~RY{l(wwfv72Yd74N#Fr+|xo{&K>Oa3KJ
z4Ci-{)spKebiI6bTHOmbZ=h6(GRx*;+4pYtV)7zv+W*G_sK$q_U2@-TDxI$mqQtAn
zmjJCg9>V_M-j<(ApY#}dQV4N=md0A24Qp&nk~Qq|8w-f*@)t;PpTvdq#N94XdHTf>
zN7q4T;_94#JQ*M7pC5{GQ`GjrypM855rK)xEEWRhtg_kL@{j((TA}wFk3qdIgf@j{
zOhNpfU4LGb<)@vfAnz@as9hZL>00*5LZ$fJAwzyo=fT|NofVtttm=-6-DHMXBahQ7
zI|{x)x-=G@?C75(pet=M9+$&LXc^Ic<$Sw``wv60TLr;pWx5qg1oq`arA`StZC(%2
zLuP{UTgf3oKFZnM5qvdlb$C;Ou9_7`nT78vQ`E-?8GSUqm};1w!s2hCm+5{Ju{
z_yt)jS@59u^a(B{j~q{Oxum$~A6bR1mfPeww1Kn<#3VMBKk{NaE5)BM{+qSnT`g&6
z@Rih%80&VuK{(s!>S)ZRK<6;->;uUHc(1(ALj`pSJTGh>*8ux741=D@!qzQHpq<+rZ9tbNbf{#uqEN%cZdml`w&
z^m4{Z$OWiGN_skFWg}uKSo42!UYj`_{Ip^xR|DY0Ex@oP-rvth?rOpbd}R&l<*a&S
z!{*R)&ZM1gqeWHb@x(isM)Lt;$`<8jZIye@q+tgQWRN+>GaTBZf%$PHNEru377|e9
z8_yjCZ4zF{1wL4WWkrt01NG)t#NG(dSw&8CBY7+bV>MaHgu@Ds+mY28M`Gt5$a$1D
z9ZC^^A;!RW=-Ib%&%(hd%!^PYp}DxYg#KC|QZWelwZL@o`X3U>6}6xsxbxSXwD2tv
zj9lP1o`%S~zrjdToO3Hnr-|F}B*ven_oR59@B<;TEu+fB(!!DOu7
zZOlzqX6OT2WmfV}T`%F#PCW$M5+N^NveE6~Lf6R-43gWXj7w*|9M>clDY0EkZbdxk
zxcFbV750<<5dSKBwF@-mT;NluX}9MK@ESO<_ikD~3ZE*`Bol(ZNyU$$#AIm=8A*Tw
zn*vIIV>~n*r2hh@wcYoA1e!c1+|C5gTyHa;?8V#qA?-OP<_n8&q*h9^VYj}t3_p@j
zmxMN;+SZ38EU_0b;$kunqe?xpLltH`GI-GPqQ#x8s%$(S&rHU559piuICgnZZW!k+
zxdc0iV*+ficC&SDST!
zTGVo*orA-AIZmlWR_edKh*ub?L`&o-oOUNYDVbxZA{z^u8oGtPdi}nDzhq2oO{h2}
z*?g`_Kvx@K1^I~h6!7!TaPqlb<#C=0YndAlO-#sqI~6iUDTRgEn+JH6Y_oq4X+muG
zmzdhks*kDKt9D)a3<51PB0YZHm)xJSYo~EImKFm2F7rd~YpM({!i<@%FD4+w8xtv^
zM|INARvmSP+hk4J&NA$uIjLPBB-atCh`I|1c8+lEBrDD1Lo;exrIJl`zTD9{Zd-jw
zgB6J|6xwby;1W?E6<2Vv#oqR`4;HB5c!K%Z3eNVZDj}3{_yr1IW2uf7r*D15r-pVN
z!M!t*aD2vG(r&MX@`Ox<*ic(>_-3r+M%sML%VZTz?+KZE-z&NK2;MyfaO*$$h1$YD
z$m!{E6XPJBb8F6_m$P5J_qN1Eeqi3}{h7h8P5Br4cO0-?>Z|_|*o@r-{SrN1X*p6=
z7hTnL8rlAPChN&XZz60ka^&()WYynw`}RpDX<#auMFO`ahibXht-~<8FOKUw`JxoM3V~m1x;Y+mKujsvr6Z({bOiE=(n+T(bcI=F>2kZkD`SxBKKw
z?`Wy)a81lH(aZD+lhrzXJjI3h)h{iDJ!o*Vlb^pPv4{y*Ovk|Bycx904vTk-6CEET
z;0ru#dP|*Z9Tc!yr%kf#K=BcFD>H!mX(z=MxzV8Xyi3@=BmLpK=!jb4U=W>sor^W6
z$EW+g@>LTG6`2#G;5__-h%|D1IZfGNIHfx+!y(N>QPb)EQ}imfAX+F^@sZ#Jge&`2
zlGUxhlwj4TyY?@XRyB0o{Pod%@+jA(r4O%sP~sLQ5A&~lP3tRp^Uhr^=Xh@#ONpqB
z_Dx5Jb!V?ec&^NTAk}f{Oa2cTtL3)T$UwxO@9Bw`8!alH{}?L%=7<|xl59iFytR&O
zXVINHCwna``n^mC>e0`98TOTk5&b!`dnqSU&
zqZI@o;pNP#*Ne^=QH=+NW$_P^`dyO0bRCZcXZP>ucREjA_C4CVy(oO4j02}JlFE$#
ziHH0+!oQX5CFNN)!&pHKIVm?;w8}ed)!0*PM|3_e@Gpw6jU^!hz4NCC{NC<@o_XJv
zfLMa$Uls@Ud2#pFkw1cCnE89jSJu|>DZ-nLwQ~i0O8^x##&&$Y$%LuYoSPPQOh(Vg
zduDB;_^jRH(lOB!3W!d9?w6N2owWj_HBj)%je(qVT27hTDzpyot$taGYq$EH_kDCJj+M^`Rt6)+25uUCl
zW~%n8sC!?g=#I!NR7@5j
zKvI^6dGdR;hfa^U1GBn^4m2(;+?9`AK(cX6ES>)^3v4rpb;LKp!ry%9o;0H0O=^Dc
z=Zi}abz&&U+dWepctSpG_{Us3+(}Y+aGn)m*PT3WxS86YHju4
z(@S4zfNt0hgrXW!v4?U}mZa@;5qD6SzRi;*Aip+3m-hU2F6<0-q~QZpxY1
zL3=;_;BtvT1P$quVfstBQ#+O3F$+lmGH^&gUK!#tk8tmP+s|g$2XH
zw`!7Lt+U5*N}Bx`7Z>NoBNyTfFfN)HSDZs7P>rsHYa{vn^=!_y&zIRcHrE9kO3f?B
zmmCN8N3_P8pBk-wlE?F2hW!}OVhk7|N(`_vpv7BTH~PJ#dZU$eIx%~ik*2rY?ttmj
zAdEW}o_({mvUUY0amd7vO*>w^)$ximyqMfoow8XcdM>j(b4hopG-II#9|Np0^`KN0
z*dP2fTOOb%px}IxM1Sge76r0Vk#li}DA;>vu*XC&BPyDiFq&n<$glK%B#6=pN?^Q0
zl4hA;K1Bawy41{A_7R>j1xxH@cW8uzmq+|lp0E)Eps@3)-7M|=-t~I*0~k6Sju*Kv
z;;iEi1WOB%4_z_uIzK3fI2HCK9xC58zu_eKXIniv&dm0JMot}n%F8VXTF*jk2R`4O
zQjr91Q`6~oa#!hg@p7Ul5=K_|bJzIWH)&hk>gri`-g>m6&{9mhT)CUs}41a
zM6rk6y;ujNs2?YR)P;mZb4_V%7Gf&1L4ahfaEO+M+B4vF_P9RdaU?ntR#0C$onF(t
zS$n%w;_sUaFh6LT?DMk@$?teGCy=1lJJM0mw@lfugp|}+QVB?fRNQv(|16{@YvR6g
z62IhY^2n)QHnGnk(kunVjVFc2+KKXLA{)Juf>Pq(e^88Z%w)L^F-q3v3YOX}m
zm9|U5*T0;07mM;9`baMgR{PtKgPg6k)5&wRJ&t=DZ*Yj=(8yX3B*YY-RUg`ZU10#N5Z;sb(=O~R6neHq2xH&HLT@9!b
zy`Vk%S~tydungs}NWgintOJiLdes7!8bjR^Zp%gpDbEkew;i|Ksu#rLlj7cDjpF6W4Bn|>hj%NVxF9w@4hsn
zx%+p1Bbs?$D0nkHR2&t`%H4E}l~z6-c!e9jEYyO-Fl@jjP3NEc6%YUgcYLi~5&Cl%
z!Gm}2JpSK2aPTUoR%i}S2oylR9q&H4e6+%Lgl7ve0YfuBEgppcwOmv_`Jq3Ji&iqc
zo}H`Zpy9%wL`jI6fNinVQ2ayKllNfGcBB+bnfDpDhRz2r^rEMvEh|2eHmi{Qs)-Vj
znB3W3$`pWDc#^^Xw;jo9e895Q_ZaAZM*T$1{V6WrzBMw$2`x?v+Evrl1??#(I1uEp
z$Rta@vEuLl;VD(*`Q?v=dw(t#m&g_pm1M|X_~?8Z)gV5owN
z*8h28>WM0GJ!i=UD5Y#|_F7Cwf3{>d8-KUqb5knKHO%T`ako>;
zW_1=(gO;AE8|?7^h9&#JEU=N1HLr1;_-nBe7ohn^m`>+ZIW5hcprIyH{>X^y=@Ixm
zF46hRKT;o1)$w|#P%8!i>*)E)UvfE@HON47|y^(O5v-dMgzf^yiVcEDgZ
z;-QH#XNaWK+?snjjR5+B1{jaC@PN~fw49j-gMKdXtD>mq`{zfUx7C@Cr_n>
zaO!#r>nuMh`ZaKq4?wj8@nxQ>_DA)@e3Vkr7T*hHix9aqO@Q0Cg7fcT2Vy*lmBpF9
zxX-59(A-j46&l3xU-WFX-BTdD$1B>DNmodfE@f^J3E4*!E;hU92cOGa7aad>c=)Oh
zNXR(87*dGDmasvnOwPqD%f%R5a2vkXXtc;#E$5F|h`?CtN(l3(
zs2FS0PFMNiFjsx5ei+UO(tl-$1D&I=qPN9a%82Ha#`N3JtjC0}uaFjJr?FenF&dj_
zi%4uOvU(xfEaVoq^)$!}btTA4P)k6XtH#P$oj|6>N)v1?6oN|90&18;jH$=M&QTD0
z+57SjcSn=fQ-|Th`_X=QhzP8L1Iz@kjq7*v{Fr2|U+GN#dSd2EOy>Hr8h=C`c;a7Q
z+;Bl6QE<|~_Io2PQy6H6kW!(X0y@OnJfTdVwcv8xr^*~`mJgA-Oi*R@0v
zz!YtiSC`=_aOfw41oge|&op+9+^+Ye$VVp%C1RC;ea>>1^VQW|G#s0lZ5Z!YN(@%pFI05K=#FbUUn56aHFsr0luWs3n|X_vsH?c&+jLV
zj%qfwjOC9;7*b#CTm%EwkyPYQ*GQjfN=m!22w|ean@&O!^K_&_l(uhzFb$&a-dv~W
zGGt;c6x=jOcB#>p-vxV%vWyDUq+%LImgAjU^PP3eC68=@*7QPO-a^(*c$CFwo>qOE=T3fT9j{jLL3Ni&b!pC@Jct|WHK2Z{_RZ1W
z+dY(r1H4tSoo(i4+dhr@Pk5`|SMbF~*L(INF9>9B`v6I?w=8_VM8in4(~@xG;jS>k
zM0#ctfozO38ll?;P^O(0cOO2$CI%f98_~{p{>4za4|!k4mh&_7mQ`Jb9TKy0Fks
ziao&2m5^l=zXMy&$K-RhdXr=lP{n=lud=|2r2cp)f1h15Tu)b~Fhm7O2%AcWCb)R@
zbQc@W)xaO>7*YPJ+0M{3{)cJy{@ELi{9?N7!eHGK>p&ZOMA)VOwml
zmhr{rA>R_oHliJzB~=$UR`L}n?rBH6PYz_L)|6a0+8$&EfjdYGCuMI%~4vE>HO%mZ}9TU8~{Ucuu&q&)=PoK|8xi6rbNXYT2-E
zMr{<@kQQx9u`(ZK8`NqLaA7{bI2To>bKEd18*whD!v+I60So;Rfijb~lP`@*zbo*V
z^`QA&&hmyLvFDEGO2>wVf(7@qn=Dml`l9i<%vT#NoF1Cfdpnm*t=Nd
zml@y!UIt8$8trz3DGlYx-SC*9fgM))V)nib~rja~9pSNrQ-d9Vfzk>*nEqiBY
z^v~kAJzvxLTp1ZO>Vp84s(2Zj`?axy-5=wY47`T?Zh9K67F$A{Ry%*P05as9>-&ex
ztx~|*tdz-Vvp00oUICeix9t1CQ)TipKIF%>*LwpD&$Go;`0@Jey8UbJ-BJ;yUCV5h9(zORF@2DR05
z(|*O+C|GxnycWYsJh;-@EF(0(YKPSdb@`#t(yirHbl5Gn8V5cLM#{l8pqvB?mO}K4
z^+?M_Ea%wADKU#A3CMlIL9R)65Ib-;o@Cv2=m(&
z#U+u)a=Q~&s~G%>LLMe!q#BBc#&H!wEE)!`nLr!yN8W{&5chh-3zH{}rQF=pDH+zTV4Y?
z-4;Af=3S}J)f?1B|FxVfrdoRZEcz$%5_$9nK%fvt`~tmA+re_7Kh|885nQC%AFFs7
zbh~m8)hfgoV^HS{q`X%r24GBGUT&()buD-9RbL0HSton*fS37w6i;Pw
zp>7y$vKrR2ffS|rAe7f64z`^1Vj3=u>w4>)n3RvyE*t$Y7{
z4GT+Yao&@9;16LEo@c+>_{0${#X2~jCI%bat&cfG#TY?D_*S`uCH~m=`r)3bCyV|K
zU9Bw^Ld~V115RhBP4X{HkNMZ|D?Wx%jnG3+tRg;wGl(+sa57j(C_AXEBVED6!zdDv
z7dd24pc64E5)^Aoe#ds6uz1inM@Tzk2g&*^htlHlv)ay)KT_dt+C1)?WCne}DRi=?
zNLgDGgW(;0nHXeNpLHLbSh!m*_g^x&EHnAApD*)(#an1Vv03*H!gIkF_ow1Y{4rZ}
z>sh1tDsY8#XFSpeSS&1F*OT|mbHm%;HF5*o-s|4}(6rmVF><_Bv7UeP%Fsm1>OD{R
zL+t$R_+JAyyU7f1g+gcEV^+6U;ysao#NF?d!tV17#6zJe7wDtKP1X1$wnReAz*72c
z`T&fAOCWWp)V9?z)3h-GOYdLDQY!svWz1$Q3r-u|KQ;#*ZuI^0;Zyf>Fc9=$1XInD
zEI_G3&;02H#@xpbTfU`$7HSsoMIj73iyjDgKPOpRTbBt1f?$-)(7*L~KE#N}68*44
z2I5D7P|(m80LjAXVx8ytW?H}k@Ld^(y8uSXygz>wcB0Bu$`eRJ`Q5L{0o6q$9*3o9
zAP5Bm97wBL@duEj&?r-GxbBuKmnwa{zahvW+;-{rK?%`&1Ih`5M=s6-fZVhxL*cCM
zFr7b@#p5WdEJ$>Bv7vP$PT%A8Bn$}&*)LTOBxUX87b{d{vTEeSoJAJ=WTE>(O*@PW
zW2yeo+}Rd%e2AJ=-C|s!Ntm2vE>^aYu32&>^|NrzPqN=JNaDy>lN+(UkktaW>QTls
z@#@dvE-huiA%9q1kdqu6inox=Fu>tGXisxn&KqK!5CpjWprM%H2?7&Q|8XNv!mq+g4*io%1>
zP`Rc@?!stx&XxNl`UdO=FV)-^vsOpbH-r(={x(0K>6e~-HlLbu6cG|32o%kJ9xTG{
zkARrXLkR8WU4vKnS=If+S-%tg$F^$MO1?EC0{f5fxf%#Kl_Zp1=bPeEDY&;|8Y8Sg
z63h?=Zlddkd=!VA?~|UmF%YOADGxQfJpR|d0kHF=#`x&w>;eP(jNIm{AsSjEdDy1T
zDt)ON%P#;df^$iLiqIAa`S_PD7EbYpkC5Ind#Sl55Ts8|ch?a7n)eE2hSDqL@A7CY
zlbWP|JBK}{n6DROqsBQ)c&R99hCTXwEqVT1S>3M75#LGIKEJ@E?YM?!Z
zHy#2XrNgNDw8Mm;j{~yyr6$3nw1=tx?{#>YZn1n;R;j_AhL<@@dM%s5YD%4s
z*U(8^nOeT0a9gYwb8g_>F*gd?g?z3wfjJA25{*w2hMq>
z{I_ixURVt1{&Gv6nBOaJu}VuBu-pbW7y!QJejW&%JnNPE@Td@Abu;o}hkUu+`L`Xj
zDwwLAR;d94t`
z3tMch(3_rAvk=3h=*0B*S$pPQ5XY`L#9H3$Oo1t$))#Aq;!?e$?^z5AiO>U#wG7DX
zAUoOZM27EhVRdhkdgKzpK0M7jZBqy?>guG;V;HP0X2M{lM`?GetQ~8RAyq$99RY8EOc(59IFi)%U0T@YMa`30wO`
zAzyFQ(GWH*+`p--R7q~-6Y?9G*K?8L8h*7`%
zloDVR1C{29Pg0Gm1BzemQjrRf40?URtwRH8XhIm|=n!}`Jc*0!CfbJRs!)sNqywoc
zhmEaX8l)0pstzp_`f2zHm{D`qvVyl!*3qz8{{EvQRp-~?^R0!*xs6AYWUGt(xl*&H
zy>bR;B*AaLo9W)h6cj3`OZ{^QHI?ru8ODkga4R&5NmXmKb0|@XNl^X5-YT=onNMnt
z_jZXTQ)?_pX!jIhh7jGu!oT4G0p66^8#pkmB&CnL9|@aq0%9>4=?q=cOy$cXtYpW~
z^?0Q6`fd$2vvz$5&;t#ysLF)sm>g0ZLZtU7@@Xy5TQw`0;AF2J6
zxD>mBVazV2bUeWj!Gp>4VCdetxx@;!8ad#9^O7lMrA9*t5b*iR3J-w-<5mc$vMoI2
zvw55B81o=d5#aUXNNALb@{Gq45X#6Ed-creHXQ#W{Nn%ZgGM1QQ5jiUN{!R-r$;oF
zmV|=;`Rms&7Jg-;loO->8*^rH7#6rc<|9-1l1%akX$2do5I%!HapUJT0BfzYz5dWm
z9+4Hq&L7jo{FwC1oY_yp9Gn)vY
zXUoQ@HIhmo;w-bnL`*tf%ad*^`l5lRf(SIUnn9ArN)hhH{shn?$>RLo^4Ra@sqhT@4
z%{{_^Nf?DFZ+QecGiC>W&7MUpg}GNVtzTr}sK<>CYBrX&)X_mvVbsgGw^bLyZ$2-2
zlaTuZPzuV#lMx%e?+jP#t%x}+){-6CJCxEf#LDGUUFL1J;j>YdPKu98w<}?k6TqB?
z{Q;xjM-Qj*Mv==mPPu*!o-Eb;AD+%KCa!4V+PDsGgA{it?#12R-JO;q#ht<3-L+8M
z-Q68Zad&t5=H8q4`$a-RILSWyY*}kPYjdSM@A+SW!FUM@e&2{B4@*?E&WV`rm`p}7
zn=5`h$CdwRv;EUEzJYZ*JK!Imv_iw=p9C;!PcIY(U7b`)xLj?DPSs@RC-k-$`Bh@ye?BWjgJOO+9w
zU$ka{Q?V)t#;>nC>kA{^gwP$8gwyWr&fT^mLir?K&b&8j?xVCtbw9@^?~3nmjeG*Q
z%pP>p(i&IJ8kD8OG*a0!tRFBBy^5Cr@f&Vf!0CZa0Dz$R1(0C6V9~${0S_S^-3wNA
z;LJ$?KobRZ&Ml|5TCOR8otvkvPhbu_8!WMFbndRirJ4#PJF(3=t`R
zi-aPN+k+%-X0y6XpmUo`+wAy=YL66MfEo-DcEVQ9K+At3qK-XBVgfSv3;WZ5ARR3n
zF1PY)MSgPiqaP>&+ow~R?wFbEzgmcjD?Fm-ep*H4_&Kg#|&%=ws>(TH}l((nmFrF$F
z*QYUCi1>%zw*usw9TAN%5@nJR=L-!H>nM03@~cPJhnYTFqE-^t-tZ~uxXCMKFFFW4
z2;SYtkh)V|2+FdE0{OCtpCVT`-;&vpgln({HuiDwL_ymmF!PQG9AMuaZmtjZXm3rF
zz(gW&ZX^yu(?*FVR=T~-{;s6r4SjR+)d}>OE}bmAm-v^7Tn!4%3%D9L7$3k+_0jsE
z&O;L7$)|$%OTG+~od780(Bv6cA>m_e0KgiZR?6W7>=?W;WaGE*T`z
ziFtxg*W7qkDn>6i_A6OT2AyOMCycLdm(oFAuCsJ%+`b)fZEV|47@^qm6K6NOXiVS8
z3p{;|JW}h}XvDFoYhhIyz1!WM!wzSTd_L}W`n7IMttwjs5e333AJ;wTKG2XX6cj0*
zp5cfd>px5YR(SG)6DZq*MNKF-cI5XH=ij)NBITBG@Lk85`AZ06HzGe(;s~^oEYLIN
zE>RQke+zbiJ99h@NupZMP)bzjx29
zhRj1kDVE7PSF3FzJ-~F~wx+GoRF`&tKjRL9=i;mb!NL7}`$=;=9OnZpU*
zU6?4bP@#U6;u>2e59=&Pk$G>2sl2w-)nH32!$^QkjzV;+U#MkSRkDX&(_C4)T;tyx
z9Jj+t7pH!G4~+Z|!2?5WRu5pYVtK398*d?00#)IKPr=Vh;;qhyDG>vh;`8XjTqfYS
z89BHES|TPU#@u$osf?;2Q^2c<)O@I|tH%S1%}NgD{d}WS&ryTa3mR&LGOW{QfJhy?
z%;+~GwheYkq(c4_KLepu=ZEpfe
zgf(?5EZq$hrN
z(E4%%9$D)p4Di03HafDG2KnfHX|^4~={S{*#hX=k;-XP4iSu?Ws6xv(ghCjWS2`_`
z0f|5p1q7HmFCB+rb~oBb=xWfPSVg*wnr?fMvaZN5j+ncxehvn@%F3x_L(K4~
zWEs~!zKk}@m`A$c4ZVwi&H{Oi&;K~zGYO==hO&&{vspl@#vNL`(_0|6JmYuyzJ_#A
z-W^gT=K7Jc6*c1JO;|Dg6W;c2ZOBR5-rsIBfLCc(rQ?amG({@gI>_4PJ+V7vqIlrry;B$?bN!hYd
zD;)?(Je>WpPSo-_iwL;uuzp8v*{AX5?kiN5+4nCd+B*C-L~Cr~0k*)yII<((>Ksi=
zQ5%6xoP2yp$_HRzVZJmt1Vw2pJ_H7;G(yB%QbC0hn4&?T&QsE$+NhVSQekA!`F*_g
z*#+RBME#qovyiL#zo#jdO~#BseD_uaV-G*v78`8l`{w7dI=vooa{z!l{vR9skiuW-
zBmvRI2ZP^H5%94gztYfDE5LDC25F^;)Xo@b!6gg3^EyhCqlQkx;bM@H1+DEyP8GA8
z-tL#Sh&%z_X(WSVPhAq71cmUvGZVSEpH@JUG+iq$*AgB&Jb=s
zy50t%7mS6;-?(hwVBSOwjU%_X@UJSmm2GtVkh-`sp?|b;&Dedu$B}c8ynif~ZdM$C
z=0n!do5+PEyYZYVHe*gN5Pk$|N^r``RMlV1v3HnS&{=!W;YTz7C<@*hiF+^uTDc>`LA!GQv{34SL1D>z
z^8|b7_)+-{uLp9Ke4h)9u+X8$+jedD)2Hl2dU5+e`5rB=CHLfw2}uV`q#^!!ZH*jaJR@0e#3^Ak
zq=9)OLJ9}*429*I?77|?PQ`G&4r4+J3jJ9CoAe
zmG=ErP+5E-LIrI%y9B^|W||IM-3Dt&I1Nj0$RT;KMp%ASC>#LQN9ZpGQ|q5AMphfx
z2mIOEf%-UGsl$$e8;*l`hS6gZn^yE1J-r$fD3X)Oz1;CiDU%?+Mby&L5*8ZJqgW)N-q%+->Xq?dXf(hl^HN@
zhoQ~MAgUf=KomyUo>Eb@M1=YlOZ*VsjhZ$e>EXIFqs^dqh-HXGbW)0rctSS_ILU2WJmNWEg6|zB9@YJkTL)-M=_?_U4Q1pG=Q^Wi&h18CL+$WJ!f
zy$#Bj+Z@Q)NkQD@f-pr?S%}&mL--Q{SyPhIqs(SPgX1)0o>_`yafxUmcY2HBXytnD
zV(z|C(=@f2Pz7gICC0IENB0f{&azTgR!X)oM&JH5<3vO|W$cRC?Lp?DDhzpeWR_!L7sj6@z`6x^+Ob4^N$PMjV0hYm$Ni?q%twIa_vdb
zbl>k!E^z>1H9uZ{J6oxISgm@~jW_yQ+#DstwQ0H*ddOvK?D-c1my*$Kf2PFJSpF`i
z&Hgi$(Sb*!alh}|#B}7~Zzlsb;P!|Z)$ty@^)kZ)aVAxe-!R?=&ozWy
zvqWuHxNpvGNH`VOitf3?^xE9M0lLX@gY5?qkN&I~l9qZD<2}~gEi9H$;kgm7SQ0C-
z7&Vg8z(S{nm`bQfg4Epw>CIDxJEWEO9azr1mD$2a%XY!S+;Yx!yLjzS#zy8BRC{2I
z;d5Bk+qx@(g<*UL{umMCVBV&c&YKuIsASiM!7_Vw4fzI&YV-UHZ=+f)K6Gb8zM^(!
zTe85O;?5l>cVh}9e17@Y>P%DQC;>Jdl)9lz)}vLwAI%mPL|p^ftd4KTad;)fh=oL?Wf&@h2bVk}qhyn&M>FYjcA`0w2e9?4
zCSPkKqP8)LCCC1_$J&UmMH)}vV=GfEW-H)@6fi1z6i-nqrtnCaOUij!Fd{Mf;hb?W
zSaY(vrIcR`R3n#@{r06FPm|h(b{m1D57AuN&Tc~CxKX!4sH-$7AymgMX6a=4*XCYj
zqiLg2BGFSU3sThI>-017bsX(HW)qSl?xpGGQgn$F^iN?~U1PKNt5{Lgmu=L@1j?|;
zy48NX3yR7vQ&mCaZ#+YJm;*f)@qU~c+|FGwve+mpvZH#}$o|1Wr2F425FHFilhd7D
zr193A{wcSMho5a}w)p3hn}lO8Wub6|S)^y!aZ?h7ai&f$5i^o*{(1+6RpaLpYrCnH
zCFt9n1=l2hEw;U5|y
z(2)R&duZLKtfSxWd(1tP?6aE+Q;mOk6?HKR=z30S^L6Dv$O={iPlTxVYlVd>(S={0
z&peC~S8KP2l;q*TgHW&e{YDcZGHQ~ylI!5$c4hn8o27X(O<$?p3T*ckz9eAH*3Ki*
z3?q5>#{I1)3XW6j{`c=(bHOxkmI`2|b6PYu{Z^~Q0=sBFowzOc*gvH+3}Crhb9wLo
z#$8EXNQlisnO9*fkMT8k(Ws);Mw0-jMpTopY;yA}?0n!cN%f}YXlOjH`QeDm*|m!v
zzHbz=sYt;Dk2&kl!U;WmK@9Zu?y!_?T}-4{pLKEp9=tZB#TaQBh6k1-ZGsvM*Yx%l
z&lwsrO!|sy+8?13BHv(mb2_K5!gb$KNl4y|{2+(l_kh{`BN@8nT}8k=om~1uSu)Yi
zqeEvtI+J~IYMI(i9Y==J7RnX{g}{v~^h?;w@ZzG2ijW|p7g?cTR6UPkS@Vr9mS1AO
zUZZrC{58ynS`W5cR~!NQO>jLUTxoAw(b|5Wum`gH$^S`24h3!cD+}?;*p4$
z3~S47?kn&c5)Tqp${xIBM>q3iehHJ5uU3f0{6u
zePmQMdG4t`X@8)}sdKsB-~1#W2;0Q>>!-B5FgF32FdQZV!L29DYHc!uK8ibYPq_#1
zp-b6l=Ffy<=f#hI_xW)$;qk0NL5AVm=Fh5w7%r!Q+oH%$oAB`cr+qlH;Lr8nqyYc&
z{o{4sq;{>l(r2D4#P6Gn!TwIJQ$a2;0DP?QuaPc;JNa5VjLxd67b{6r4#Ee~39G?ay2NJ!6CBwHL5r55~3r7eYa6XlQ7|hb5w{yDI
zk7Kof>eEDE<2YCL@a~o!Fu_mP4!_!qjp;4%&Rk95!x^P@7jxY
z$np#Uu1Hg6U)cQT7Ff!a&fq9tB@6WS_O4{>%}GA^KNdhESXbRWTdgnnfx4SSCI*)c
z_Jg!-n<-8FdK>S=DMc~lgWnlB9^-O=t;;!43ZQy^=@fD_uj^A7st((kta3UM
zrfFg?jX&REJdd_$K69Gj^44ip?Kc~3wn4NGqDC)z{exQkVoiMM&u@YfA&71?r_2I_
zJM)k&g5umm1})!I^B{w0>uX|q9fp)U?r<48
z!Md=BdVx4Zt8@clSsU@B`3G+dw8u$AYWG*m-nxd&>~
z8dxO1SOfm#P}|wWI9zboVGm-c`<8-0)leyY9lp_xH8;+CU4?;bF0WsT*CUJ{24+=Q
zJAO@k6Z9=xt^87FITiM6GBZzd|Az0_^Y}&W--%fy8X{U0DIBSEwK1uY+!pwuo
z+1}u|)|1%^FzcqxA7&5K``&CfD;^xp_iGG!*Xk|0z3Z`d=hDV{u_2(gio8*N)NVN|
z{MH>srlxKCh8-iei;@m=#V)|+I-N$;(W6WXvSy!Svp&d;;3E0!3x0)g0R-lo>@g$d
zp1Bi+cdAf@wWy5aN|6nWkHk8%05&7%Inxzc||}kCB&zaey+EPV>aZyr&e?V9ye42=BoF+Bv`XFpnTcF2R7J>ZdOVUfwcWh#rF+I4*OF@jWU1c$X(;>}%3E?Ra>b
znp_pW-K6eq=8r!cW=7rT@X=zUqD(Lo>ol@>evcm%3+Mbm9isU{hkk-GN#VttySNL7
zhFd~@4?)8Y&!A!~v6`9&kYyH;pfd(a`zI*cd0nN`{%01CA1oGLCGHBk9O6@}#J}H`
zcbMOP)80oFs|&HlpMKSA5(j%5V3nQKt1hRqzm;kJA;`+gn#mV~XK`%g7w~>^)NlWV
z5|!<^)dQK2eLPzT?|FYxJebMC^_j|Sgrs~b7K$2BOACC;cr~?vGEpDD~`i>QXAUBr^tM2Pg|r?&67IBcfePRwz_1yyA2n)j@?>-En|RCv$pS-bOiWH
zkL|ms!i0nLq*BL&4?-8JxD#h+IqO=gKK4cxFK-#RdyOq7>2n_b^DqNJV=+ChI*YtB
zOAaOmzO<@%veVpe7cz^D!;@c-;{cbeiqD-Gm-UZQsf`f;*>8C$SqL7s6Bs6-$im-&
zWcp_jUy@fx<{Q2koo#_?H69o7gYCBgWOFziZ~2Cm~d$_z%db$19EYV{e0
zZUW))(`|lz#qX7huX0xVm$VtZ2R3dOk~Y$97ROTG$zs{w$H*&RgF`m>8q{$=Vv`t|4D{8*Za1KwSn_b!y>W#18`yco>goQS__7vkEVb}AYLu2=P
zTYl#a%Ct4AWIszfZm79_R_!%15U)&XYlvxiz4?$s5OW)v>b{??;T|LM5QTV4ibDO>ml6ckOI$fLLg095F-40<6xAI@){jJ7Y6Vv^3zlQ^~LjJ
z!0=LOL3%=g5J;a@oDPfbbEL<35)1XIP)!CtDIfUgwDa`Vi12ME7g?l$k)2L%JsfTI
z94odGjq
zxLuK8rJAm6_7yh!lXxj(2_1(%DnvB8_UZ!z67S6ko(JId&KsJm-xN2#2^$r87L4z+
z=leBb)y&JmjL+l%G2)F}ERJiXS(?HMZ8(C`9yUmH=cQ8F?6>3!gW2P&8PW9vvnrAx
z6}mlK+yPg@ji~zD^I{-?kpA$6etIqkkBYf`q)1Tj5@V(L%LmI9q3KBm;xvCUQhd?#
zB--?d?Z2aefOm!xgq=O8|G#gPsn6+RW4xM@Qvu-B<9Ah~`l
zlT6S<@9o{r@T~hD>Rij~G-+@lIIeU^6(i+lE-=kAx$IOm82XQvM`Rocg~fKnifry5
z=fn4mP(}?QV+@-xrG${gezT-JWoWVAmMFy@1|phG!`_Bw9t>9J9eh=t5*Z6g)*9+q
z-OqeFS}}hgA>49&NCVW#U8u+VPOz8P*Ic59_LVVC*IIk-OC?jqa^D)~@wksdm-ns0
zN*A%6U)bJR_ei(%vb4!+wu&TKdV6BFsfa&GPV|Jnd$_-@Vg#Nyxr}Y#R7**
z^P8eTFP!wrQl(sz62!|oU6vF%iBb?j>`DZ@;lWP*2$FGhvykd*Rq
zG@bm^PAsGjQQT(C>W6*@gEB5SvjHLX+nbEXZgLrsoUx2?7F)KOl)!@P#_V9sruVhK
z1uFrJ>&wlqxSF>iC!xv1)9xKKDG6=W1gM%cWbT_zF=+~-cd(_o2ug{woLK{|Bb-s^
z4582In;dx5@uHP)L<(hGL(&SCWm<+~Te5_JoDMLg
zo#wN@Byjgd0Lw1{4M+CqS-r6Na+{HA00#o
zLJ)2Vv;JX#SSge7?xtBoz3qlo(G~U69~+1GVOP!XtC5G*n|a`?6M5i_#{UNS=J`}>
zLcdGp?t3bWsiE4N(aiO9xa0U`)H}_zj!lVeb8R#!`!sT1V)@eHru42F$ZZ*;U?Va^;bc6i*@_t6T1#-703iG{(@_xI~~-a+nyd+q6X-bP~WgvX73+A0rP2KHR=7qg&@
zUgAges|R+cVg>ojx?FME+od9ZWx59^@TqHHqZ7?}|C55fwZ@31-`-MFgs*PQV*7E6
zo}jpwrf$N*aafUxJFZs5s;ybXiNEgPK!7+b2JE7i
z^|u1U+zx~li%R5s*Z2-dKGn@}J-$MEM9I6+v@N<0&1(Bm9$h8(VWAfJ%3h1M_d96G
zcvB14zuv%I52HO9Bbl=IjSB*$(3i+6*;PvtyEzh|U*Hj1{YcWSWwY9E^4n89u10l#
z-{*b2Hr(W6xq#yyDkY{TFfE@mv)K*@;Dz>OFdaFS#cqez-Jm1zq4TBiSEDSD+a*t#
z)P;lvkUW-xS`2hQ`UGlK*_pz~Ik$VC!`-}N_bBBy!vskw31^=q&(_I6Ga5GH7Ta+J
zJ|06$OI~Z-%Z%EAQ(JRstbTNoidg|6uD{K2JV@SNFT2ZB^D94U^5=dbV(r|_3OJ5c
zhW~tf8dV10)FIlqfSAcH<2xAkp-b
z>1Z2|?kSO0Gy1PD3lu4^-6X928%!p?>Ox?!l{uojPY11esajJAshZYsa(eOCSZSCi
zO3^GUi=`A?IT9@kGRDH>98OJDxj&uMLo9G|4|RVCci_psF_pg(^sR~{vm@K14=Fe*
zrfMHjO)PY2QU&n&J$YL4Q*3;?3w@G^NwWBqEObS98C_T!n;#KjpRgFaj4E@knAM8?
zgy%2f{E5Jf3K}3QTxq8jpc~qA;L*nqWi~y7^&Og}d59_#1z)J$IgNck@+cRgsxo^2
z8C86L$XHFNxvGBCXFBszGqS>orAa9E+&yCmdR(1Q7W&`SPuu?f_%@8ur!ZI}3M?{S
zaoZv4s9MK1iJdZN2gQg{7ifYo9X|`PP!pi?myeFiv*CNF?(Kt>_nC6i|q}kaUdNvJ<_`?Au
z`+hB7T>4cf_SKxD_W82*cVI7ey>AM`fA{N_ir-W#T3cHDe`07>8~)K?0UuYTv=2Vt
zA2+XVZbZSk1r@LsE1$+9LW|z^>pWsCnITja!0&z~OiQ+h+~RVKd9y!GQwL5)gn+?+
zNCEtMZ(zngI+biv4;W_|&!`h-)wy6e>U9N%qmfr)R8NBS+)C0d-m9mQFdQuT)%s9`
z2>i;7@LIeTc5<#b)yDw$D$Yj`6#;vSK>sH%UYJk?U-+TM1L~nmscMWU++uhV%-j0X
zAA4%=w;SkrR3U7qBE=%2xOocj=VU@gaT#he(cd6QuLRl&6o}SAIt9Ae;u_=C%RV;adOeZMSHKVR?+cCEwkVd~}
zJ26GA2qooQX2plm&r6BNA*`YD;bF88kkXt?%Avjgasi$C_0
zHl@zvm-n`jI0Y201tVy;=u4HPP+jPVQ!)rS8o6k&oq4l2u%=g|F3-PE(*HyvW3CgM
zN!qB_*mjS1m^gm%@{`ukU1Ym%P-ue_;naA)l_nn>KonB6vQPtiE#--~^kI`mkUH)>
zM*{GpG*tp=a4>~_ZV8cS{&xB6s27}UeC{q!anQI&|NVGPyUd}pg7vK)Oq(IjSf~KG
z>iybc(CHbBB_YxKb%CS#{2#6M^}?sjsGfj}DU%_XBub}d)QtNV2~X6;7u`$&(#J_2
z*md9jxz+-Mb6cEk{W_$kSzuHqDb4?GVQ=>)Voqru#R-WbLxe?TU2bC
z3uxuE0VQD`lpb^o4yzbBMgArkoF+-+P-L^|$z2@F-w{#*sQPJ*8JUUYi-2cy`T#_V
zVm8EX5!vzBx46Q~uT;#;%CGK7f0(qm1Xf
zb9gR8U#6ryo5TV31$i2!fIpb;V6CKpz1KP;43C+KK!Z6Jr5$aSeTv(2C_MDfYYf?s
z2S(a7?otE>^=c=Z%IxxIp&!bPj7{#ugE?kavyBoy^;(lYO8PAXC>6>58Tb?u-!ygi
zsuxj-;T7`~#0+8)WbjR_{nuC7c5kNkLn10u4qXQc%?_)A4{vOw6Qq}sS8aoa{9}?m
zhg5%)=$X?AiGk}eu)#@z#V#xRjSd}JSR@?!U~mOgJ~*bd#gEIv{e2w>uFDtz_hR$_
z;SMmntymNxn>QP560L`1Bf+#o6Yv?^db6ujQL8cv$YSiTn3FV;$<6$IV}Ua*VXTj@
z4iAdq+po(F6hpjS)*5cu*vV6NNdS5_sR&f1QMaMH*;V_>&3?wF|CX^qZ*{G2fmgcf
zBV>0y!+oypIKP8B%hx3XK;O??t+JG5;?T=3%n-R~
zI4Dsq@QjGAk9|SLU)Z<+HL{hR1c$fwF|Z^B23ROd(Bz`jj~DtxB2D|VFo>H%5<=Fd
zLaEu^z?pooMptBtz4}hWHQI}h=iHD;Z*WGs*WSK3pIvGM-|U3F)b2-K(iQKS%tl**
z5AJQ}+LRN^8N>i8nY<~2J)h>I;cRR`JKYo^vM%KxW-}Oj@*cGZC*wSGT;iWRaaWk3
zh_bTGFIcGjGhFZD5db=v0Iozu)s?})v!M_%G)IA@3CZA^Z84tC
z3@$2z1G)?C?dNp_bqA~vOh_oOd+WUY(fIUp^OFD!S)x%STEB0tR*<9NtWXzj<9Gn~
zT&MBtJi)4)l}zCp;uGDJ%RaLeQSwLvkupIvqSG-Bt}omvPq&X3Pz#=k!%TN1g1A(X
zIB*Q6_Q^?ga69+cIiWw69g-P_4hQp1BCKTSeGe06`psbL+EPjt=z!X_HJyc#58De*
z)@{X)bf{(UCNeE5S@TWX5X9=1943wq(ipe6)-huyjzf=GU;}&nW=&8yMg5&6
zadr3=ghb1ITfoCCRXfD6Coh8wI;f9_$Xn#dssz3kYUK(Z4xBUBrnkHGd*Agx{57Oye?)8Z;z4WHuUn`SZTH%Yt2~Pe{G64
zvCZGE
z{2j%8%}f)FY~_aKQ&%j~^He7~}kkUk4ZOO$(2@;SnyfSKBk-R9H7Z9e#U5
z>1T?MeY>o0$_Q1?CG>5u@DZhHE42&(aSH{62wrhG+~HIW^aKR+AHO{)(k2SP;^R}{
z<&L$g7Ot&F@;JU0E3?$>h$xc$=~}uo
z-a6){gzXoFa>{YJ%14M@@Sl^;HIz*rsNU@zJl&R|ilP;Y6f*fdFy3$!QVsCzv^VIP
z2Nls;Cs#TvrF0Cm+NFy0w}iedeS}<3eYcg)q+ixn2jU6_yA?DY_4b^af%+m`GeM~$
zt*V3ZqbfldY+l<==F=Gx#^^Q8y@{pMGiF;XLE{I?FugGA)ELmDeotj8)aK&_&p}KE
z9kfGl#bCIRo8=>2eH?G1C1d#)5^!i91hxQ<);Aos(v{q#`;bBWSc9UVsYY;QHQI>t
z3auznN`?C{ueFXKhKI}lb&JZ=WJdvpO_TrrTRNvr?ECsm{{44s89m-|rYN@wPmbK5
zw)G7?pMf7#K?4`0jL^QuIbuB>Uf;VgyPGB0sCvU}bwC%6=)j<1n_N}+3LG)>!X`_P2BYe12U@T)kKB+Ew)7|IQb-lFZWEo^1qs9tOq2DF#
z4UhfYiU&hHub8#T_#5Y67TYDZa)xX(rc!yHT9X{b98N5z*1CM7;e$1@6@RC4xl!~G>7wne1Jd?8E8P@
zX%7(?)QkSJg{t+YXLHjib1FrlaKsGEiQ63tKX}VwwOydiU!S>*y(ZBv`w_-tKHKU{
zuw9w&B<7)0Z;4})ymIR=fSyJy=nX7M4emhGY^hn6sJEPiL4{2PYew?!ee~rp@v%0M
zhv5FTah&sv*Ew9RvbQWA`8m-(YOtxUD^7&72>qXYrHiqArH=qIvIy_tmuGyau<3p5
z*eJriG3X)KZ^oPFrMOb`~Oa}E=ySCFuMX2MX)Kg
z4TpqRi4Bkv9s`H$=rl|fq{pl8qDIF;O}_aAXH}?vkG|F-4G{Vyoy`U#sxPm4w3K0$+Yy3o!$0$l2ZCf*9Jh-L}h$(uL#hgOKp$
zO$I!NhY^u%xt~O|uvQLpwwzFo>e@oQ(b#4pj%m6*opj-=Y^mon5glt%u!j?$6c3hS
z`I+ec`025w)dO_)?HV<6oxuafOjWhJh*q3~UMLiLj|CDH}brhOmu&gEep;
zl1kx~X)Sj=!10cpJaUf=nSq|U;a6>
zd7aCfa$O$EmAl+UDz8`2me*I~VqtC*asT9B$JlmKH6$Mgi(#OguPWvg3n*oir$6YI
zCuRzwv6v2F-r3FraS(cR6(tOA{e>_E$f(8(?!V16sVID9X;LpCcIC}Wf)7fVwryi?
z$i`*>0teE6!fCrRsel3b;yn4_Du`WZe7gI~O(SfuZ=Y#FU}WhHoF~WJ4(4f?GxB?<
za;1QI1|#SCz&AufjP|o}m}=<~0v1X#D0x^G!;sL+sHV=p
zU*uX!Hx^tN1pYsPn|_~P_?389S62hZy&lf!M8$u8%K;OB{s98j#{Ui#IZu9Sjre(r
zFXMu@1!+P=L!r@QI%P=rF#LW0yc(xIvyw#voJhq_pyq
zFDlH^w&^tqC?WFpx1fJ+13MaceCJEvK5_TOA(^mAaJ3X#MvbYMgL%p4KL(x9mk
zLEWBBhKNhg8g=Mqrh_>k@`HKwd5jP7HOFsBFPFVgrrcU4Ri(kU
zcw78!?dhI4V?Lw;MO1`_#X#G?_E^V^lwOE8pReKsP*8n2znE
zS96T@Y%8}D_E(xd?gqOJ;37jZ&1GOYHd}cbs*$Z$MX)L81}p7WA4=6QjDXU6I{jaK2QQb_1y`^q=Q~pecxf3)M$&2#)Lj%>ykLSpHm(%Z`bO3
za=+gD5GO#j>FS(Ay*B)aSbATNFSV7sj+!qKaaCv74T-N$DUhFiyKBI`c~DLi&eTB{
zve_n(*J~qQ9B(6&NC$KH#CGU|-G#+3@zaw_QwI9_!CRl;Yg+JVW*C8`;j7c?ZgRV&
z%D$k0-Zbz}x+Hiawr~#$j36#&AQ_x<4u?%b*)Zu8=2ycG?Yn;+twR(vGMJv#Ih@G+
zO@cT~`atkjo?g9-l%Ga2hNxU4k7cW2kLvpE+4*d}Rn6-BrndD{`O_!cxjKSmD?6tg7*)h$GMr>&XkWT4#nT!E=7(vtS_<
zQ|h=>_@nzRAxefPOXb|VEvNDzYIu2chG3P&@vi}rP1SWFn~8B4cuRdCZTe$yZye_E?fa1DbN1xTGS@Ej{hv6*w|Fc)-S5X~
zB$Wa&a&-(+bHQxynthOnv_aQclxQ%hbPer#A||!zVZQ4to|=k8%U-JQL`67L0Nf59
zRYtRNArAq&tk~b_C~vH({Cs!_^}nT3XDq~iTaaCyHrKG{8yGuy_oT6zJUIM;@;%nt
z8O%MTd8vC@1hOjX^VQL)_wPs`^&R%H>bHAjic(ofs(NZ$wS8X2_c-wQS(i4U1Tp76
ztcOTJ2HcruBb2L#Xa_kL+b?{q_kL+Pm0S9a-Mqr3rcTDk+zn|oaqMZO;HSXC9~GU6?D^KZZ)7q%1KzgtMSG2}Iig}K
zYe9o*NJazSJE#!S+OH5y0jVzK4cRqF9S(qYUILqF$-d`xQ%f3?h~n=av5OF1|+R$bv+9CK3h7E
zcPhOS1<3V?;rhzGx3|ZCsIIb
z)T4>iz1R27pkXm`Ha6+vLz^w}ZDOS#y8d2h#JeaS9tTvhp(Aki2cyI~>{8#Xj*GJN
zL^!I|U{~2QxdcsrM8$4|7L}aLBETp8_s*6=tuefyLaRC?L+W~G@aN};k7HO7Sb_<<
zhEL;Rc0QN@D=Z03D9D?g9UZ~LVy2#OO3C!eO;JkVO)=O4|MT&}0P<;q7!8m|q%%N_
za|UvJeQ;oP1NVuY&s6$bgDWYV_GW;B*2um&>eYr{-da6wsa1=lAs^1d9KhFf=zl-p
zUo|J7z_Bn3Wxa2Bl=urYH2}I@C4BYagcx1?q`iQi=|wME7mUzDO8~4GTiGazlZkP1
zRB{Bc%wAGc9<<^L8%05urEOuQ8gQe&3J(D=H{|^zp?wwMIRT_=dy-{&~)%l
z+U`S6S>h%ZuKnCc;h=)aUmgA@_Zc^Tu^zV
z1zB%MmbM)cZtYGF@aGd19ZX`aCkEX&sqIWvjH;l!!0vDsN}enj=uPd!55W(h!#`2t
zzCQLPK_na;+$st87go(w&UINGr&V&epxQ=f{OQ)b$a+Q2QpUD14v+lcO;WItytNwK
z`64GX82f3zj--i8DWBImT;zh^=WD#vQa0D>?fCzw`pT#*w=P<`yE~PT?k?%>Zjh30
zkdQ`@lI~DKKuT#4Ub>VN>F(~XyZOFzzI*R4$Ix-cdEV#Qd#$$v7>G{G8eR7=RNff}v8-~~RvM2xy%H}?qQ^fGN!oeLW*l|zZZR`|skz%t
z{r=(pE}j384ewpgWmq}3soi!ebAob;u#_BWbA+!zePfld`=X%^((w>RUqp_lFqF_w
zQw!6Edzz8y-NT8G;?T0j@Kk
zs_52PBctom&^Yz&8E=wGySz8jpsK1IRqfPbuSGy^?@%MYN)a-&Emx8Bq)lwF?t@OI
zMT>$H%X2OCD5a^CX%X^1AUMQj)xl1im)RqJb>8MpS;A`4*Ko9ij^DQ{T$Edryi4Eg
z6yY<%pr>YY3j|FB?R;N%Iwx$&Z>@^Zs!}kPUb{|~X@3ke5dpjn7nUTA0KOx9h*wi=NHz+DW&>-6^2y?ihqNr(xbcxqFn+}CSK>GW%&3Zq(fD|^gXk?hg3>U17FIjvK5
zKT7yL-K@>uHQJGZvrW`QiJzRjKZ&$^uafU)L}XZoyDc1!90$9@CN1{dA+tnie@%lg
zWEox;pe*{|$Nj{Uq2WOGn7%!093eIO;sX1(SR&Hy-9X1&*X|
zYg7B6>UsjK9ytR$SFM|#xociH=~rQbDxKx#T^n0;Gb066=dN+nAsshFl>?WAbXjoOCyL_mvc&HHWEEvD8=t_3&Xn5
ziuzDniv4kYD7>rOhBgogKO(s4l`8vTe9Cdo;7)Wf$;6H96WHG=4TL^`I?Dt2+WIE9&ALFN%u#o7YPJ-rbU6N8dOg5XX;NrDMo653+Q
zna-y(dX!iR9q$c=X2!Ppd4caM)qU^$A8zscRP(xMzyHZnX+*s&)mXhWHYF2&rxB#c
zHC;;`jZ5E6KCaeO(~3}69x7HZr(n>7PsZ>EucxR|zUMO|^{sA5k;H@u0O_&!0
z9}6zJzo#_462U|$W;OVA>Nh}emEt?ul3hvav376A?l2Mny>B0#7UO*Wq4NOH;u%NJ
z3b`l(WlhA&hEryBF=`T`1D_5P3FTOE%_C{tD_Z%gY}rH;by56E*iRO#HPqsS@H5<_
zcWe9^*Tq*~M43A~Uq3MlpJ|@@z7$^Ty=*QS=zN+;W0Mo_lI~NA+9BQ|
zRWWLe2JK#)Hc7TZl?*cPdkT~uSW6`rA6m&b16(R#9le7x<+Z!`GiYRPkQ`yZT(3TU
z1WI9@ni=@e3Yvgn=(<16agy61`~Lm=;3G);9H_}~q7f0&FchK921NRU{@zFJ6wW^s
zzGG+btf)pCe7c|?GcJf3_(l_&smhP&HY-z*^Xuh+c(|3?Rjz%^+w2K?Y)%I7<%pA;
z)`-gZQiBRCaZ4%CYU~;&>w0}KzBOTm)t5V4Ogo8T->Dd|8``Gk1z-zRX4P*e&mMnT
z+p04>h=sxZxh%NCJbS;UT(iafp5pY~5yPYB4lyCS*`0`Q{iXl8iP-s!zigJMv)#GB%Zr_)55U=Du0neOgB_s1h_@=9coET)XB0wwpG
z&Qet!BYd3GTQ3f0_Aikw9gXHh8kWXQc&al*H0W^xG43z7;5>-9DY}Z;Fea~$Gd{|h
zR103_9HVGkzp^}DzuA@xP{dQxlY8j>EpgUH%i(p10yyB(g?nBUXcV$V{TooIM$>oV
zMPIkXf!^+DnKsk?%~qZTC-8ohio_s?BY?5wI=%ttnw4@M3GMIS48M8Gx=qvZo0xpTl_Eg$9`l=Eas0t`
zwe&%tUN11`japSf-d`OpRV@suAlOewl(@n2Wy&-eupG-PBjHf2@mDM~!bC)2HY|+f
z21yo>ts8nd^Vh#tcKH%V{}}q?GegW#$IK^lmK6ie7jQl^B-|3OuY@APsj*=4bR3At
z6>zx4Q>J1itY)cFbLf*2Dlu*LL&YiG+LbPC`)KXrME_=^f%Wt3iCW0&xqP}c#LVV*
z$HUXJdGZ$Ku1@N
zytcXu!0hN$K;CZ}`RkYNdS1g55&vD(S>z$8WkkM1t)a5Vy}Bphw-l6H8l&)1K6zlL
z7jKNTRwdJU7Kg
zyM5E|Ms!*ONpa{xf{UsSO`@7pxK9=h|5UshKUoVGJ6C6(At-0n4B)GN<=3@vv)mEt
zdV6kHa9(QKE(+wKJ^id?xyv2m@F;~dDQjMyb(|jQDC5&FC
zL^dh+C_^~tvrN?7h;-qJb!k?1Kc#70S@~>|C+#)!zvze(o0mgKL5r71sN1~9tu^sA
zmMXn|?c?IMTl<&PhO$*+&jN!;$lw8$*oH@h4K$U2(6SY3YSLnCdZ_rE_lf7xNqeytj2$8+)i=EQGZ?h7hL
z_h1Y>J*-qt*oj4?sOlG1Y{mNPj|*a!LO=bUC`yKnyT@me#+pleH?B6yr?v%b`R^HP
zKe;=tu}0eSI?}q-!4dH}P_6dHempHwFx%NhpWg9>atcT7_PUHPXxErB62>u$aiK)L
zBK}fSQ>6LT_46xBe@A+1DW~?!`t%T3{l+TiJte4yOuvKvpL{-RM4ypj%Y0u69H!Hb
zaCofSVxvID107vW1rYY`i<{VY62t~anrKqf!%nSv2BDF$rl;F>TnQg+AZ`bfbDdXQ}eVg@MGpUQ=M+*}WnAR`)oQd{b<6{CvH5-S!X`*G8
zP%?mcLFk`f`?Sv_S^pQieFVhGGr!cC3{qeL^{;F{Ha7?HJremN{kuvjuD#6R%Vkxp
z@E8Wp4PaB6$Sh$LQ@K-!`!REEkE194i#pLfU3sTE-lOe5wAGTQ7k>hsuR>_59zT~0
zT5<54u>E|2@c#Z=|7nOATxfm?-VRfYcm
z38+v=J-k1mzb_c6t=j*DbF}pyk8S=>8{jJtFDTexQD6=4;B@^pmHAdv6{5>(+BU3DW^
zQ#tPYyMyAp!z}WY*m~qqvB;8f)!@8zyfLJ!*sM19syou2j&%Hv=Sq3^nDs-o5jtHc%dQ*rNG1s7(a*ah}|=Y&aT^sO!u!z3@%L^
z4W?VKr>K9hEOS>?qce1ahP#OhB5ZMNjOFba|04BwU87Df;7zRWE}lI8io_ruqvgYpn4yxQ
z2?`)1hHVvm32FF27ZW1zm+Ai_5V}4S2n|1{Z_7TQtUyXeOcMlC%t!kY3^wy>`abkg
zs1~Z`QIz65Vpe>7%dGY5NFpP1>wbw^`*Y)UotlATq`rn7g>4#}q^~iS0(&no%X4_n8jJ$s!naJjy&r!#Wxl35*%hXg8{p4*)0CPUvV(?Yre9_LF!ZHh&<%KdhCp+iY)6H$Kx?GNjX720e3h0bwupbvR1!3DM^7
ztjG$(hPRTm8M~4!m|}=+iB~#<$(!Lr9Dn`F*5$+4=k&81`4BTSH2g06
zA7*eQHoN~?W9*D47HijSCmx9z-qq{5!iIWZ=c|bNG24C8#p9CWnpl3N-=qEz(&%+f
zq7gzy&PBB>pMYcB^Fj3S{-@A<D}O82zfS-PNziG8->u~&ys-=0%Rc1tDjMRX4~oYv>LEBoVJKU}@PP-_H4!4zt72
zcE~l{V&#h*Dy2xj5wXv3zOLQWF{R3#$=5CU!odUOO64@tt-_Bc3ysuxT5X=I0OWP+1S}fdc?$+iV-Va
zk(R(Fc6mukR#tX@x<)cN^ULVQ&+GFY+&FTf2*3dD92rq%*jrdw(55i_1pCW-kxg+R
zcu`ZUDZa?4_FDiYN+@^ZTOIR1oNf(lm56O6SEZ(-MI)yh;%)jL*qKhU$y=h2->lZ-2^Gp(RNT5%*K8z
zozlzFX59|ql!j6G%>S~o&VqhozSSuWvPjvpm}-+`SbuvT?!U*Xir1e;h*}&=$n2m|
zMKEmI7JXspoo~}uUz1XSFcomwX@$r5*_ED0k2)m7Zlbv7w_$_&7loJCZ5WUQ6pT*<
zA=VT~Q){cwAA92#%TG5ln3Ck8)U-hYP3(WZIc1cnm>Id`HW7SCnDucz3x|$(uG(w`cr5F!8y<
zU9MG3U&lAS{U3;5+f3I=^S`^qp+i5!{^T^t%i9Z$Vdt?j%VNIbwDad$;NWf+RJV_RrnH4!%*9SS%apDRHa(
z^Og6C24g-jDlp~yYSr#cE&16Qg(VN!Re|O^}Bgd&Bte#DJVgQV?Sv
zj)pqJFUpz61aD0+)+DRl@>4WV9hA}15mkm&6!iy-C32=?bxD1}5?>cuVTVn}p!`n-
z_ciT|EPh8>-R)kGPS5Q{Jjmy#&JyhR|`P(w9z8B%Fis4mR`kBx(PzI$-g%2
zd$u~7fm)NcqTcETRJTT}EZvhmnHYqLP_`x*dfWx9Lflo1rFHRCwaFJ@FvtY_hy}I<
zK7@~@Y5;C(*!anrjn(lYbB?Nq^T;A1aFZr5*9QhVLG6#XiayjXM1Nb)?`)<mOC~
z?Ah**9L4u|ydo)mIs0e2rY?KwiT{N*aGhtimM~7d+rLoX5GCN^%qAD6Hj2=}BiTVA
zwpK`fRnxn6$0UNtB7_^1%2c^>P;#WfL0t3pav(UyQ2V6{gUI3dUcKDIx}?Shxb54jJh{
zg$p#lU}4|(02hZ}r)_no+4%lV2>v8f8e!k1$Q`Y0{Wqe&?cNLoDNNdAS;AhHe><(d
z8du<`0K*rUEUxX@N~^EUH3Dnb8SXw;E1@5LIf_&QX!)|JfHqCh3R{5hZ^Y2Kyl83-=fk^G52=wNuT$EQ;2XEA
z8+!gig^-r2wza`7-LE=zC|-w@cKB>~51vC4yKQe9sSmJAlyVd|PH7u{6a{SD@Cym8
zoi}})ZVM2yR?hhVBqsCU40HFV-V@IIo)Qf2;L<*-R#D#k9CX~UU*q)6tvNeR-kI8{
z6_r059oec4%&4htvO0YiF-^Euy>GFUQ$gKc_wF!m
zZw5Y4FU-kJ{;}jZDC*FuEz?gl&_eZ_Z2i+K=Qh#r1v6jE?VSKYc~o^u=R`)K_brr^-<(AJ;R+*_B5oL?4x9o{Ql)i3+TF(GRm1z
z(6qf#g7`zJIC>Zeq}`pr-akDMhRP^oO&Po~vfN*6H$UU`rwZ={?l3NAo64<1egp$Z
zw($$UGJI5o2}MtNAbt!*O-6yNDei3oTfMXH8KMJy<7`RsC8nw-*27J$KRsCMlE(et
zShT&MYimj%eV-)ktJdsc_pGV+{@tfcohxnk9L|ZNfHil!qxCI~CN1lsyFqtgGf{B;
z%yBPK&5MYS#{_zG(x)Nt@KkhFu}A85#@VUEMLdINF*%VXO6{yA7xJQQrJN7V<0Oqa
zHQ%dui3iheS2!Qp)ruB2hWd31V9FB1&lH+VQmWnln%e8_0FJQ5ePhkEcc-E@XIM|6
z+^?tbO5|cw2C^_IM3r1bOP;TIA_6Yk=kV8}fziB<$(mR-P}yC@bxsQgOmOkJ^EI*h
z@nixxun1`4r|X$l20F{n6}OOps;X-1yCQpAdVHPyr9(hf48Aif!qfyvslz-?OCuNd
zXsI31^a)2LqKLNpBOxYjUrg7w9c@jMv6Etp<{aq?l-QYE_m-snn=Xi>x4P4&?9cx&J8l8OFvX$#hf3r}e{4$SM`PIl94AnG9gc!l^Upzf95ms4d5;B;72OReZfGP9z?dEc|KVr|syS$52wiR}oym`A1<6DB!XYI3o6l;aua!dKAz
zo1XggBWkPtnF0~tCtDfTAs?rl$2_mOnN4rvEqY~J$$C}oW@{M(?+*AJW^3gcn7+7V
zj0M`pnLsFceax$r0*;qoDyAZhgjEv^0zN71>gB;)f40C68nZ?7Sp2NBnS=t+Q(hI2
zlK;iu7O%r>@*$&-V(sdb*plmG#doh%3HUfB%CaLF`0;{w6n@t@#W0#)cT#fMRNb?l
z4Cd@;rMqFIsv+j8)Q!Yfnx8fBjK8P{*Y;=5stlr>HGZ)#U?2JeVDJPVCYU2j-eO+d4k^1m!+DY9e?zOP<~Tm$X=3;r&_Kw
z2Gi~B=w6@SpgzvDVTxi>(&hvx_$;-=YZ6wi&0Qs=|GjnmQ|m_NLwFCJg4EdZ6cnt(
zQJ5|o(0ptmm(E_m{u$hd
z{)iY`?tt*%Vp|SY*p|DaaBFr&B!@8-M+DF&!EgZ|anxG+uHM+Mo1&=x4&HJ8bJ~2L
zM%$~IM}ZaZ%5xMcK<-Q8WS~gHN4X*I!cgv6^p6-MmYROLu)*UV=l$l1oG}sF(9Fr8
zFZj*tZJR5GLo;5O^p+_KnfN8z|2}k$6GFS+$tpY<7yt6o(_`fDOIlMT=U#3}?SaHs
z)MWBqd#(op(Hhx;nkW%qg1p^m5xG06KzwByh~w@-E56MwzbYZ-SY(cW7t;
z6&M(3wx?o`C02_74vvl~EsMq?I*`yFF7fz%HAeuT91j{5LZ!_62JfXaS
za`UzN967l`#HrS%m9;DJLVfH{tjfQIe|QO6UZBNhIxLHV)1VVywwA`>Ux{`mLg(xPA+m^*f#
zu^D=@0Z!P(8+La^#bkzH2t?drt{&YDUQ+Kj_y?kzD>Rr@llDLnJ>>P>-Dtn7!zgac
zeQ9fk*6P3b#-}sx)Ut78tHxQ9FZ3mS2JSk{skj8KA()AGwQLk3>!S6wk^0D_9URiH
z9ojF^q~+Tz6%ecj=1~(YI=(1&AU$l}eCmT68Dmj&JB|5(2zJiSqAMKx4qM`V`yQl`
zv=Jkx1CH+F1FN|=Cu%O&yWRcS=l^mCMQ4`=Q%(o?E)}FJG?Gs8FATh{iij;05YJSYDnF<4d3H}-vr$b&+Z9F>4A3U32O^^g&J+ScQ#k@-RQx?k(p#rRwF
zgcw)%I=<5qTH>gUMxI7!-uDJw_rbKlI(C;x*aHp)KU#246c_?2WpZOvPzb3(Ub3AE
zZFOKqmk;Il;>b_s1R8)9b
zD?}}fuVe}C%Ut1?W@ZzY#qUrbFi+HP
zs}je@KiMl<-EN%6a~SV0w#R6dvUrudzuHJJqFMtPl2@*CFW^hy2z!$2Z|9(^EH5K6=x3^pcGiN0PZCTfKwH3~5z{?;SKB498RN>I;8z
zv~6E8sb;ye9vKRz#hv<+?ta|kIJuxtIa43JJ6S^{`Vk`H2A134fDFp>82jPwN;Z{Q
zU+QQ7W3&GaCtaLpBz-u|P~OeczA?N%o-E;eA0#+roOv801y}Eqq9XBNOeS~(hRU9t
z&pnJRYTsDY@{?Z|fF)2v7j$F{Z4Fks5^lO^L=uv!_X{wV0_S^bC8b^c<(05kyB=M&
z>4)kj$`Ln{b_#k6xi~gHo>~;Lb)hH37Vb|wq#{4mA_6#Cb<_BixLWyAbjh8k>0c@v
zr{U;~|FKHuH%e)Wy{Gy$mq56Ztd^c4dZo4>2`~0Jlc#Lxft`iCQnr3`@##lU$IqLs
z+QoLClMrxal|hwP@WWLoW}W2+EO+0Qp|gO;%HM49J+m&;C_=IN
zm9Bn;U2-lzP-28#6*j3GHrd)O@+kfRF%Crps9(&o{LT{-OJUUPwWIv0@cw!bi5(^e
zigwuq$*9>Q3JjJRmcm^)NsfpIx%!SpU`}?gP&O!Nb;Q;tPl`Kt?{}-sjW@^9&Q?tp
za%=p$8K0ew?fA&)Y}r{g-%KM~-qa3;Hcxs!d;7KJSr2FHSLu6F{Dhl~r6wqQVa#6q
zj=8jpT~+BS!m;dV(%eHMpbDdWPAziMHzquVzbM8{T39Y)iJF>voY4WCPb5-iuyvWZ
zQ+aF}#_!?*WcNDnf4XnQ?c@0K1|?s~hH(&B87Z`Kj-8|T`4&u9_nCyBeuAvVt
zn<(!#Po5xUDRQM^%srIrRD1@Fgn^`EjuhSav1&)li}JJX`hAf0U%WrAC_R=iuEOD+
z@^-h2c+glf#L=0m2#ZbOB8a@Fx)7EW#i~sIfyMb|Fj+p(z016w)5eO9{xMPn)8F25
z1d%QoxIbCtjX7Hh?^b`8VGx!g=|r(6ihe$gu7hgIMWt(>rjAWY=kzkoO}hMw95+Zb
z1y4)oGzW=IER(i?4rkcI58xWTQ}(emH+@|Fla4Q1%?)s0EmGqHr}BGy!~xM_`4DL_
z@nY6sd7SK*UB8##Dix8L>~=X7$SSY9=4Ar0I**l6O&NSo*Qu{Rb>yOIdcf&Hu*h-a
z#)zP&(ZbHj%K`vPolwm%BHH!l_EsV5L;asFD`Myhte{wxuOHTS%WB=XRLTIi0Bp3y
zywa`QFI!j-GxvD_Xp#g)N
zGZFNZI8=K)D8NxFBRIIR>&gng=qt|p|e1TwKl@u2UWj2~{wDcHk8fu=-K
z=i|muT6q%xzBE#GP{+sggY5z*phQ(BA8@-JspiB?<}(yqqIR
z&ES|NmZPe1bC~rx6~7De#2fUpyRFNWT!xd{{P!ur!v-_J0g$8sVsRQBdXCkdP#GUb
zB=~AvGTB`%*v}f?&rO1nm~Z7Y<81+cA&kwSr81@6(*IxqM-~#)f=mR3a=!>6bK4ru
zu>&k@jE*Q$R~9&$SnOtFgvAMZqNM8V2gyaSrUSe)xaX(JsKeQg4N%|3{D
z8I7M|L#zWpW}P4WgThNky_^lowk@s?vi(Tu;t+BkqU3^Qatq*d%xwmN$^?#1IKOq-
z#Dn0`OdxM$(NE5ICbYikadgVl@Vi|^Xg)kXJ~ErR{DE8ZvXFx0a1$kWKp1)M-l&Z(
zGyUgwJjYKTI8XvggXc-1QwA@5_Znk*lm?Z^0V@9EJ8l?$xGQj&B0H%$li
z1>lR#S02rq%ZQ`G&Fe#gP9)_TE&DJ+1lEbUtZku@r&bW2PaJysI)-srEDyhXge4s_
zl}q40IvjdLLIRdE(X03|bSxKHXjn7tyEoJIj#;Eszm4Dgd%lEWgMl9D*jC>2BE={*n3(62|$q8WmFEsdIyZR>MhpG?Db&T)y-L=#h{5!1`>eP7;slF
z+_)zy61S+S#}t%c4%`vO&-KsA$kyzTsxc8%*S?1K;ZjhjfNiLSP36LC
zv*TipGd*L%V4IQLrV1`kBw{+%TqR()RSa;e+U5R?F`?}t@&8^NG!oEdiw0Vcil~IF
z9)>@P`TvAnxn_H)fPDc(m+(jk(sUdI3;}I{$YU))&T~xe8t7dS=wF0=Pt{6Pau}Q!
zTUB=9`G%1ADWGjBZ?3Nss;a8iNT(WmfZruf?F&~TV!y~|fs01?%Hb3&Gjs&q)rhPM
zcnxi4ys1j#FTfkqFUJLX-vT-`OW6SP@DOkn7)@GzPDhL*XIg#T4Sv4k!Dlnc4&ZEL
z!fjgmCn7opLqN0mD-eJkANB>)i22VW#e&3GSmM{L<`&>?%fXqEl!DO}9vEI1BsY`Sx+2`jMJADhS}`cXmQr_4Z9_XCD^52
zAo(dYi10Wv$uXN%<4dqs;lmuhr5JFz&+|V;zI($C7k>1=`-+YYS;>F);hwIsECq~2
zlE3@eWy?>o@Myz8fZkIC4%Ds6K)yLG>#TB0Xz$%05K@|b&x&hP&444oZ=^;xn8U;O
z9Ym)<0{07$z?uO1SVdkgrV8C$!Um`me0$MBsBO@a-
z3!~ox2smvKUg~!L8$#sZD1?2e0?0l0r#}gLY%gfzB;<#1_Nr0Y315>PGAe%jhDOWkO@_cbu
z5?X9RLDM;vo;p9I6=c`JWX2>v45>8}CY25D%G7d8xNkC$q~~z*BSAtUr0U;y7UYN|
z9C)8zrc3=*)Cjql|t~z
z
zgWIwfO#$??bYtccCB9=KD5|Xr$B9F5;n5(m;b?K`BJ&9Uy{kd=v`{R4!1G-V%!T+>nzb5EttK(~>ZZCge_AF?d-YqfF~Js3=P)HAv5zJd$r
zW#5&7k4Y|=%r2v+V`c#-OdLGd;Ec;0=_B}3IWrj`6j@+OT|&TOFbe)&B%q%{OceYq
z;g6Pl;>>UHnRE)5m(5C>wE%LN)?QyJ(t|9opm}{+k$?oy@BuMF
z0%DaByOZUz5)u*=j_qWm96)cfjxffEjdqy;N!+g^$)AlV1Y;?3)NFh4XNx(`5*G|b
zI0VGd*4GfI;%8`9G7?WA@);yZhTuc{dDz;o2pM%2f`Gl(7#$Zaau&-N7
z-Nqq4sACquaPF6Upi03MANwyXs)mQohPyuJD3VObE
zuB+WOMS^LpsL)-%v-pggd2ef8AQ;klFMM(%+1Mfj@tCvw}usJnc)N3y3=`arfhAhIf`&mFRrO#3TW^on7
zS{ffy{bmowmtcZrpc#zGNO$!bfRMu;ww7~(xJ0mbM0e17FyAy`zyH$?o(`3$_I2W3
z4Lr2-n3)>81$Af$oaw*eA7II5kDb?qtVa1D*~{kc#w(ocW1$iWXSE9pP9VjQ@I>EU
z9UW<>BIfr5xpp}T;nFV4BLRpWc`SUU6Vkf@K6#OEpP=V1mLZC3RmSsYTus~j*%MDV
zoV~(s*Be7TJ|Djs>*Es+k`yo{8#r!NhCtONk3gF4XV{2pS;h
zA*Yh0&%-{42gE$?qn8c@t6>xQ5}}3U;1p!pFGe(I^}`@IuhiPdPh=^*C>7{dL&ml90mU~xWt-zdg+20U98bis6tn
zMPVNkz{QaXynbFj0?c>cp@fUx3{Z*Kg%^OP^1BoNE~G{-fReF#M39IBZ+GZHVoyHD
zCdI+QN$ICflLDuNb%5Z`^DWT1)B;gqhuaFogmuToor7Rz_BT;wA_*IU6%tUx3#~U-
zzg{EQTcd7S{quJ4;D9x_q}by%?FzXV6%|A?!s>#b^G(0_bEIk}BQ%4X$A`9M10cAs
zT$v0oLUWxKfl=(^T~*8XPj|nOwyA@*h;nU|)RB~Og)_>u%VmHn!e>NG3aOVa{To;x
zj@(36F=o(%f#b>$W`gub0(s>81(MT{XwVW>t@b4*AyEKoepK_$^NpXVw0mHo+gJd6
zkfjM~+TPmQ>ZdcVWRdPyYYz-4E
z-kH(tfH3Yf<9*0YYsLbs4suf-b&idXe&dM+`gd+xf^<|0-g1~9_|!SrL3N)%?b>vP
zwLgAUW7(G&A0J=a{oQj34l7s$~nP0$h#}>vG`%2oi94fKCpzLjSkLGqB`cAVw5aF-uD
zSs?;ZLHR|-<`n24GpK)-(&&duO8B;8PQqkUR2}ughWaSqYvuf_&&ld(Ujx7y1gy4U
zcOS56x51qo3qy6OzBp)Ddiw`Ow+l%x;yGp^(Vf5oXGgsz8&j3ri3
zRrqtt2DkLtLXU!r)esRzOlxIj<$SdV)v$c^G$u%qmi^uNbJ>7P$*keZ=zTc<2YB=C
zoa^KNmXLu|U_Gluue6AI?*4I_be8od$W*v4S1W)q{`qcl?6xBk^+0}?Y^`9pd
z!Jck^-~MOVu+3i}g;{@G1kt?Dq{|3KsgsifCN~<2j6_ng0SI5_3U=$oqdwY#NFh_b
z4CzK{qy;)3W&%^o-bLUQYV^CDtV$1|x~a6#U0ffUdp$%)2eq!ANcUdJozRJxSR2P~On9(x^QSoU~0cV;CZ
z8}?jQ_;ypj3F7(%vL&vPBK!XuPKRL-kR!eVe8=r~=>UREs*vYGL4Xx>sN~9KFIf;2
zGcJP%*g7|8bSdyStn5C~ri4!Ch)C^ge|jv>gLuaTky3%6znAf++0#B`xefq953tOz
z$Fz)-NrNRj{VCo_&SZkFkyDn69Uv8e6>(OyG(jLE8+*|}jdcxfOkpSs|MUd_IC3gK
z-VAh3v%`4Fnhj?QxJ?LamuZc!@S7BiJ>9#P4kR$?vcA$Rj#;={c0z}Y1w7nd6y17&
zc)Ubc9yA?V$B5wZBj?eeZCL{DPx``+{P{g7cpcKuIXFRZ;3JK)Y(YhwX3vAsln3F6
zI$b5iDiX9w`0HusNodWWVe$eli6wkKgqDMb>iOx0Vre*RurKI9Uo0Xre;u~}0r7a#
z7e|iV{{B*Rqty<)t}XK_`62$Md9Vm)=j;K>vSbcR8Rgv#1!hSoQ9{5Fpv^gf@^d%^
zM>kG9AM{h78$!mK$A}^EZ_~HmJ>ECnkE*1x=CJ0#enKL`<@{#clKNcPgLwZ;gyYcq
zUx}lMmCkMsb8)y}=)}AHfOC|>&kps25@0D$g*>C5V}MdFXVcJV!F4iOuG`F7qMVg_
zxX@D4vIROVDoCIe8G(RCM1@&n9*77JkNpkr%~0y^oQRd=w!fvCvQED$4I5M{k*90J
zN16-elLU*z<{4Z{(|gbJ@RustdTGPZ6G&zO_nTSHwg7!-
zwfeYbob-R94GTyDEtHd4I}*X)8q-pV)@rj2H2j`jAK~|77~x8Vf1N`LhDH
zfKmTBA2w7$vBzMiYcLdszbY!MnIu~d}HgHlcieJ!k
zz*NRQHy{dAIeCdr0^{%eZ%)TRdr6f&tM?6AFsLWSLGZQ(d+q1fCzif6WJ^k5vd(7`
zWJwH7PcvF;r`dU?KrP&tyrmC%DBuS`S$5$6C3|k1Vq#G6y97P=bB;fu*wX@`v+wx`
z!xl`>O*^I#42F1VI(Bxow~W`fx8rQ>_cVkgNTkml9#BMhi?dlc!MvKs3KNGx)wW7U
z2~mU7X-sB%tA2AN!9{vBpH5`6adt$I?<@aq
z2zPAhJ21tmE^StxG-qDaBmdjWc$0FAUL}XPgtu7m)DQcb49UN5faI7Q$1*HqKN5-}
zS?xi0&RW_YFcNCt4B&JQkkd^dL9&Vejsj0a`3^KU7$Jw_V@B+WMpu-70$%_Yq?lUi
zez_X3BTtYb^^OChCVZ5Nu_5lH#V>=hGOG)aCTj8mZ%YL;yJCXosDfZ6g}nAZ{5AcC
z&ad~u*_kc??if8>@)@F0RP9;jlx8CAz$*E5N<3R>2REJlATka-5_q?=F_)DNoBVCU8Hf0}*
zlZ1JU#-$k=_6Qq=+lFtx&+cSaEv+e#=
zvGk_kP>7ePpspX}IVKqP1-lqC417{sr7>)N_f0hO2ksZ&5P?1xN6e<53}go>nZ=Md
zQyeW4&4q+ARU!SGEYxebp4Q}M?)w?`8A9m4X4PzMyi*t#V1-I?uk?O(o!=q)wX%jI
zP0wcUw{K7wT5Wry!6@q^iXttZ#coy#hWfBJS8w3FjuvE-iih*R^r?*g%6q@5>@37i
zO79BpW}zV^RYs&QJkrYyC0cfnfu6e)7<`^8{C~r~uw4HDeT08*E?F?-PpKwY$_#8y
z@>0n1+3M?n&K{3UB~NmN0|VC=m_ntP=x=bnc7c}I+g7MPP8g>kC=A>cxXR4L7xJ&O
z_?(MDYozRpz&$iHl0E2Hb*_?|J-DZ%Ka^q
z4RM~T%-KBf(FmUoBNA{c9H5q#l9HQ?rQ>pe`L`I(r-t$IiYMZrnYAX$42oyVZVlI$
zdcK6r^tCqlXE}|XrL-U@Rw6BXS6woOxK_JRTRU06x`I-DT#LE$ZHPSMw+3*5%Bj`B
zBGAt*!deo&Ml2!L4z}GazD|;8>`u{F7`B`hE0z0mj#q=@(Xm8}
z%}E%QGr4DZmViN#+3YqxvROFI@K>g41GrfT4_8vsWgFC{2mZIA-Mde=^r+#r4?ojZ^D)dB2CaA
zjJqpevRa{mUri+01HbAr=mi{5##Fdg!ihg(IK;jd86T_v`<|fKiB36J3e(91f4FFx
zE*n(|zhV0|F!$RTJ#^&$Ka_n{TwFo(Y#^s#}Y#}A}iJDQsc(>pM^Dbgn5TTTsfusyBC*G0GLoche=4N
z2#DTN#%Q^ha5!*YS&AorG8$#SE-~hRODy{Pjlf@0Z07EmNK}6F6LB=b4c1v88??~D
zOEri2`|Zsmtl0%Y5{I?2E%|cHh^O1kZ5u#&>HF;hvtY6SyjgaWJ*NNu+EN4w_iv#F7Oewi$fE=a0S4c?600@9;=34&tNJoHSDCd5;!Bgn&p(M(I%WzZyEx_Ty
zKxcFa)*qmojUxiDyFLMmPfUJsSEz#|#T1GFr7Y!a{s|yEvW*xYmIv^P57bxew~xA}
zOjiH9um6Vp29!!gq)^Z>KA1o47u4^O)A(GAFMnFqd=_djmWqVJC#js+FEhN3aTfdi
zCK47{=;&2ljTazY3a7CD0qCjb8xfm&od2mL08^UD(FQ@V_6ZjZT}Kd6F{jI-f197%K0QCj#>00-!r#zXs3-T)oN3
z4DtQ@B;}m7KYYLu9D&Uw4~vAI^u+r2m(Z~pbK$8B$waN#)82(w?tB44HAf1(&s>oJ
zlU`$>_%Y0(9k25_W^^`dIQORoc|W8TYU_Eu^Po=h0REQ1-~|U48QBD
z9W3q(lp8nvsBYzdBc`+ZRl)}7z2J5+>wt(jEavNdlgC0Zox0Ggl%>d^U9FCjKS`x9
zUU+74xsjB~^-@)`@&q@vKULyR-67YK*GdioD#`A25z|#C7D5mpFs^G6dFEO04{D1u(
zt=ft6h%>KA$QD5R&{0wb=gpK-hm6T*b^33wq`0~3x$D|W_)2;ICcHac*{JlDv;0$B
zT|A3=`P3PL1O~X1=2)Y0O}hOGIiXMPx_APMb(B#)#e)CXvXNZf=1_0{YAWHx`1;@3;!{OkqzI3u1y9~VN>XbmXgxQ+V9GouW`Do=
zZt(_(2y+L4O+17Lw^*V
z=fj{be)Io*Q}&xbo_F2t#*&cV>jJOMk0f@n;U$LOeZL&z9G4&x+Zm`V7N?7h&9dk-cfe%GiUmLu}+!XQ#u*xcEw0)|XJ4e4^DDe%#fUBjW;A-6*rz@D|RSEC1<
zy8YRU?`$9NqR*Pibt#EZ@sP+-Q$6U$WcjqkAI7D0cSPTAuuZ(F`ucth(7APpZ9@w`
z4Y4XM=a>n*XHfkMlOzowvY>J=ZcNmBHRbmI6u|Ri6N+<{>m&x!!^>og&7NNk7NjOo
zXqgKc({DOABjk$6LuBw1(OqGy;(`#fVTb#9&sG1*~B|)}7rqWT1c%6>vyu
zdej1Hx;CX8jg(
z4Qkn~v@rQDU4!(0xbbDTJIQTTNdSXRl#IOnR7$$b*2qIdzI)c^ofGW`U-wT~n0GnT
zW@FQIyCNC58VR_M&!Ef{trzr8-YaG9pKYqr>l9;N?Jb52W4qHjTpi6w}t!V6_~YX!w3RA1dam5yZr$0LmEmXucr~6RNA9ZPWI#o34{^f-k$R8JIqv
zfId)c!d*ySG^hcH4s2FqWI3d2d2*RS%H=Ga8rYBb3gE{2tzj9a4DMnzb05;RcxDjl
z6^8zMiZJBjJfD!Vmm8t|L+unKq=1|GaE+o+QSiHLOf-Ca`kgP{i?yGS-I)_p@O_gq
z5~w4!InH<9{u&n0^s^caZMWFjxA;WYV~QQM-oXh6>qx
z5cG|Vd~vyNX92Y!AYUi;6Zn2?)keQfQIk{%2^(yY`k5=c`5}wHszt8hfCnveWhx|?
zNgcXK3n>oW8j$qyeuj{97h~X8U2m05X*Nmfl?hL8_$%$z%~$r&d6ceBmugFWI#9tR
zIIqfhiS^47PG>f3rM$2nOG(!MX{Y>fO;VU#@m!QVggr&$ZZ~n$89X|?&_4N7TojE~
z@mvvyAmO=FY$#^rP}UZS=8)Fjvd$fpRrs1^6$lW>LhMrxcG@i+3=8=nx<8$>=EN{fu;DD(x_QaI-PaXaYu
zTWBMTN$$|0HH(cUaec96&1+rv$eEJ<7#j7|(Pc@f
z2?0NLnQ39EBe6%5Z&aH{(kOVu4%dEHJFqL?q0EoodNz(5mP+1TY$s26-;m_FB!kA+
zpnQws$in^k?U-Rue?vW18~mbfu%Fc|>)gY-Djp$}1;!Spj3Px_xtLN;TMebqD<^aG
zn0(|rTFPJYP$0rTQ-nIS4Kk~094P1
zzNz`uxNdE0D^v4pSVw`+!a!K&Vy1D0znXJQq|fwd@2=11N)fUM(rjh1W%<}pF}L!N
z5Gu6I)s2ew@WN2SdAlELq(dP0`1B3ptU}qSuOpMmGw1_jSs8
zRDJijty|iU_+Y+)My&SyWOZn-JHPF6GszZHL{7fIA2uS1vy|@q^I9AlCO$s(2r-^j
zKxp4i)l!*lG!1+n|)9bZZ#3wD)
zmU1FN4lvk%pW0M%uNsw{b4dtlQqylZJ
zP{B}ZCVAc`5V!3Eu0@FJ&nO*2!}VhG5Eu3jFP2!4H^jmb=)!yGl6lY8x$;V|HgtJU
z5;8G^9|bEh<+FI%e96wgjuR8knNGEV#)Rdu=}?=U*2#tyY?MS+oE
z$3mTo>Uu`TR8E^aN_Nj$0*ejSHJI|3hf|3VbbUf7DBEO4%@QkQXzpIxW+!!P*~4J%
z&W|0d*_m?+t4N{yKTrM?>C_lcG)9BUd=}V)g_CO3Bc-E=X<(Tof_KD<@ZISQj-h|3
z;IJ5_T#4?PMXc@N>Enwir!!{@p3Gn7a>}?f@uope7NQrypQsK4yUcDs24L(9i$BmA
zwY<^LQum5~M{_nxDci=7L??$Kf>PK^La_Ul<6*bVfrlX>bfd>p1&{qBX(h}4ltCjz
zHv;ZrM!gU`V(kdATQaDAm0y4*1LlmFCcv%C-mhDJ$TQzk+519k2(nnu4BQ!#b8gGm
zIsU^sQ}at4fVZnbXVvrH)lp*XeXKw)w8JGS)>)^cFpF9M9STsslLwaV6N|HaM&eCx
zlA{1Lr#i|RKERXjnYlN+<<2PZcU3>|i}U{EZ*)zMm0Zpm8)v)amImtR&dOLe4xwIp
zm7U!ZoTF6EvXwTMw;1vW8#LnrTqZ3a5da~pi$vR_n*4I_p?=&Q7eMmv?_9pPuGBl>c#lBdbz8a9e{RvH9)mFwG$Se74p!S%}KO%lB=FAb&XVtyYNr1$vnBj2m4OmwgW2Xu7%$wd7j5J%R0=!So&-&mwNq9GLV@Qe0;dEge
z+y@Ql_20%w<`BED=6?@saZD^$DVPTUMcK79u+H8T_7FesDqx*2nwA7+jC~qB+8fGq
zcXPF6d#j1PM?`i&e{b`(=)BLgDxUFXyMVL(TT@?K+cg!Nm$5{a0GzqL#P>4lwXM&&
z#74Hg*69|1hCQaot3z80`RkP@%LFErnsBc>GFn3|qnB3^y_MRYn-!^Oi1P6{Vy9va
z9KVfhl_s8qS~eQ{C{{i5Dv-pANGf4x6~CyAUIPoqYRe;Y*bzXzZ(jj8?Yj6ZdWbyI)&^L~VCb($fRdRSLx=xi^1`$quXW
zn7q{CMHiWyfnLp&c>`BLw5;yt_jVI}i>_S5Hh$hxI|!#m_{|AZNpcumpbepP?DFp7
z_~K{+Iaeu(p5liY&|INh_H}m4L{Jz?5w--tJnuaR!MRY7P=D=J{6?ryiW;D&fHzrE
zaC!>g;(3Uj&F3An2tJ{$2$-@0%8it1EairEfvjJ_Y0rib@szv!%oCeG-9T4ssZtae
zR|VDrPOnGFeUgp4$X)gCcDg3{fv`5cN#SmXgM1n1o4ouq!;D&0S9IG?rMOBt*NLd_
zlSu9V5Ih{tKNEBvl)4Y?aC&X2E|U#1#S0kZ=yVALUw}{+)o&
z7H$zR;5BZgNQR_^j+4&ni9dHurr=|^c@l*;x2AQ%G25k6I8DPVYz2>kv13J4k
zxG7fwySdh*B0Ck<=QBS{8m^kZn6M*q=sc&y9#~$+%|0Xmaqx`)oyPQc=qd|`^*B$@
zQ^D8=fxh5&Um1#7lY(=N*vZrjl_`@x64i^<(a8nxvHgyEC;a+-{$+Syr8#lq2^#TC
zv!`ZMJKhscUruIqjkRLgw`CWLi`&cMtWF!(uJ{Hla5Mc!DoT-s$;F!3B+V);tCFi}
zIX_=M6kSHEi1-mpVg;#>m$obntnxrH34R%yt=2Vy7_s?hJPj!>(|J}fjF$#1g3&t%
zeSg&$$n_1MBB3VLuy)WnEiZCym)#e3R;sVA2yM|laHjvzz*w(rvi>=mM4v38RjoH?
zGdAr`V7150Nyhrqb&F0Cn8HEvDEBcALiP^v1?AxbVUlm)yMPR~F&>a^R
zqke>T)m<#L3a$d|`EJ3Oa23N76X&OAtfH}$#1up_7(yWuq0i#hjk18jtG8(_>V7=u
z2<+I0olmBgL9GlFn~A#O!FBB^BN_BPOpwmX>2LW#!Noo1^y=k;R|f(Cq``YE_wa#b
zCHh?Da%3`)&Yu0tbXQLQQAP0=QS0s}J2qp+wQ>93SmMFhRZII%PmnJNz6*Jf%Z6IB
z3+%xR1`+qYKT5aSvNA|JHXh_gpXE1SW~+@97M?H`dOU?{EN9sm97WLwlY7H}1~Lvt
zVie8kHPshUlL+iV|A)Jwp?mk@q_ytszool
zB;xjOdAsHu{wEDIms?seZSVaRWE_PGYH)IKo)i@u&hMv~$>4Fp*X4hU_-pDuS=6y#
zS*N`m6zC+e8WYkdunoqFn$nX@gn76<1&K-+?VNY%XtGy8G*3lC0YxqCe9o
zP(=agj&P#TmJFZOKt*glbL}Wnm0nG7k_U<%&O}hDAQWXH`N6Umli%}8ZfjW8GjJbF
z9aZ4i{Bpwexq6h-hurdHlLY&T|90A4gysL?NF&tZ8}0S^NO6uOCqL*{V|-nsm&b{`
z*^T4DqB@*Y#`9n^hTE5u4~WeeiuKmY1zZfNQ$-worJCUzmVjJvo)^nt2tC5nlV5XP(+?i&O>%p0`@&27W9_z
zCvH<&rnmY|oXkGd>+&u7F?=b?5tl;3BmSpOa)0ameJCqA^}Xkf1AZOZq;RX*3aQzs
zjrVS+*Mv@R0GZUrZ70NcdWU~FB|zLb_;9-i*}m96NsUA#%A(8D!G}x|#2WgIJl*Hj
zl%)x_nJ)H!FBL^CL3(lgBv4@6DBVej_8b+T>xBEHmLJ$cIsZHsN7i<{lJSNU6VYE&
zwnp~ae&O;M*-@2R<=8NEqQzSRLzETgBU1-BzU|exHmMn3cAt+HD8dA<+u;f3x;$PM
zD6yR<6mH7J*ZPJ|s6XZC>gmbE5f-x#CQnb83{-+sw60|hxs#uKkF(|y^7=v>;-0cz
zA$L{)kY(rmcWR}qOKv_x`M;MmsM2qX4Rga+?#UNxB{eFG^?wd!!12NjdpmDUvG4~!
zF!7~R96fm~UOcr*+!V1MI&^ZOfD@LzME0W_hm_7jnCOh*c#NzmF4M`A&xTZ_Ka~%w
zBree2?nZ3>&6G^&CI5zyRO`%*R={KLIetn*pKrjJcZ;jc%SUhf#SS?@vVNIlWUhe>
z*c&cU15`Wp^xu0NFPFIaiFrLuY=qAJn9oleWg%O}dRDZj+F^)i5Gnv~R8(CA4!x34
zPpE6|a~yL<|1KIohOdoW7mpXi|3w_wJsieXM&D2YeX)E8WCgz3Ln1i57^hP*
z(Xmq&3PzE;2S@y^4K*(Md6!CXMjXkgQkGa}3x*H^QbV(?w;k*JA3h{{UZyOJE}%-L
zh&Qt1cXfc3&F`CXzB5q7LG!Wk4b5V300$Q?aflEV$qc^vl&EHiCzHlkGbvW*&L&Ls
z>){CW-g~~{wid9U9cXHGc^iIt
zBT}(HQGmKWmd6`SA~Z?Pp|a|}udI+#6(>yb{s9deTe@dRn;futao9<`W+%St?PyPJ
zZfG;rddsW}mdDwAHIF_RNEeEW5*;8tMmDV0)2q5=L)?qx{-PW;{4=hU`iMwCgAwrd
zKcfRmA&^?>^#4Z3c>zyp@C_A^VVcVK8yHI%`Xf(Zu<{NW$}#6_1t>k3T>mMa?hIBF
zuk#eSZ3p%4Bq=%fus^7c$)`un#!u>?Af}f{A15UlrYYmM9OEn9GQz=zwH6ok5vwk#
z$NRlu7V(jzn1$0=ngnX2FB?Yi>$CI0>>zaJlN&7Sf(MxRV
zNICS%bw~3`zqH%FtQ*S5lY2%vz`*Y085!^ZM9rP~Fi0*m-lBJE&sk(Ln5}>olrqd|
zeyK}u(vadr*p|Dd(uv7rP?zAB&Jl}FqioMP^poh<p63=
z1>K9Qd=z@aM=qg)!luf(K!6oI>QZaS3Vb_e$!WF4O0P>QqLeP27=gnezv{b`lF?|R
zv^|)TPazXK(aU5&m)tIL^~%oJtL7fB{GvHb=XWN=8W-(n_1nNaI8`E8VD60<^kgsB
zt>S+A^_ht4al^Gh;+NMEzF5GU=t!bp*OEgVhjVvsYXFBkTGDZt;sD84r;3pduvLMUy5x92Gap6H}p;gn(x>z+MAD=*eBw{V_SP*1
z@nmiq%^(XQMPVkp?#*qc>Z&%+At^^xP17Iu0H
zKaGt`^TFuGc>b^YFGS%ab4bf9hdjgoZ${0scJx09)X!g`$kQHb=kMAE?8gXXqof^Mn3X7e
zO}IG308Ra3g(j;ndI=H^>!-__IIIkm0iO+xc*S^Og1_Mgita(VCtR$Fo5P7aY=H0Q
z`Zaxg`*0OcaMPd5VU3DR!qxk60ZQUJQ|_DxIS-~B
z<0dWzn_Us=W#BF+nT;WgCsT!TRgoz0vI;0{iR<2_122N+
zUX4;iJ!6)%grjAH&0$dEs@Gmhk57F5oX5j!b?|_5#P4}S%Io2pNvlF#PRH@8Jb45W
zcQBQSDFlf&jtQi#cdRB0Z#j&*Zy=!)Pp=BlwurKs+E^G^iaCsW(<5Qyz{D_1T5@u6
z7AC*P+azlDq;WC1V7Y<;@l5mGohxZy7Pq6jwqbr3T{ki~PS)qSAIR<}k3`9IS_##w
ztjGO^wnJLgqggufplsi`wT59Y!?S6&Y3Lu2lz#rdS1=Y3{V}wvTRYz&!&jpKVaOX1
z6`~NZ<&jMTK(aG+EvYpFAgbUV1QHN@v6oU^I@bsIxQa%fx(06GF_1`9a%>jEUjAly
z`yIHC@G+^R-=Q*N4wV3Ex~Zw|6-5PHbI!-cX#HF-G3}Xy-Lt$kkSA46kwAA~GV(bM
zOn`4kz)Q#!4*%qkeVk+RoaR6XbMTFTraHP|zT}JUJceFf?)ZDv{`q&1giv7ii^;OX
zpQeHlI87
zd;5TsZ?Br0;dFMj5tW0XOMsV+uQa%IvzaXYib-$3T&J<)|3t0cG~uy^y7hb=hq{ewoE^?!V%uVm4+Qt-vRPvyU!_Zsuz3W>V=;(gxj
zGsITHU*V`fOahP%SQdUVg3za16x|75n_>m%8Ke|f06@DVTu_BR8JS*u^HBkrX%WXjLhIu0W~UX
zL_0mfY%t(K8eZ@SdnG+?azm#2#7`%y-Mq?C(>f0z`OHj;lv1OS9@;-ozCr8NOY;9&
zX|T?9+l%5xA>^<0oC4nA_pcH%xK%tJ)ufBTUx_n2#W}K9Oa{fqwv_(bBUMTY~c-&KNMpQBo18$!k+IJiTGY8eL$2M
zZ+6}0q#quy=nv-^oi?pa^~%OqHkf617Ybd^G^;5atxN>Bx(m~1Gc}w2kT_xa<1gk<
z2vGrz38FvSqRWQe=nf_5)7_5COQqhGBR`A?6;#@aF+lXEsQ-Do`pALgvGL1E8@BXO
zs&V1da%A^lY^F9qHfR|@zBfWDYQ{xMh$k2PfTkgaz+=m`iHN6w8SJ3@+)#>NwY69WHh`^rJH#z9YyTfq#3
zvWAR}%)zKdXoc?z=6yUpO-lA&WA))mviI+dip!huu<6@PPIcDW%X!~>Wq1)%#^2$0
zFHfJfL&Ii9rB#3meg0uJ0?j0gja7jmr}qgFexq<2gefsc-bzXOCC5*)3MV@?Gn~X=
ze@;vuA)D&n9d+npdxU~#k_M3kVH{sJi5AWM5X)9sc8EENPAf3<{T>KBs9E+{&_g4a
z5RC6*3{=jAQp+{VWf$r#u@(`6KAMpEyy+Nl#wH`hqkNPBe_AlCi7DgoOYy^!jwPhu
z97<*?N`-*Nj9=w#YWtRM`%bJvuCp~ilyyFzS02`4Km$h=2L5#(^;v);!0i@YDafi)
zwOwpAToz%RlLx+=?YEneJ)54`(G04Odeq}GOda)6fUg|})_A`a$Y+a{Muz~VF-9sN
zw`6h6m-H6FjKgW=>#xgG&!$t1kSdZxxeE8xJ3FE49r;S^Ro8Z+7956rw1b(5nw|SS
zd0XI|+wGl^VJSO%uE>G;_L-j%Yx;mk!{2e$0Ag?P-dD8K`l?X;f
ziXFSjrbY87g!KpwkLu17!0*g|#Kpq^2V|jLW&!ywqF4V%B
    wCR5i#`QU?Tik8b$
zs&!JZIGgW<5XM820CSj{2brpDDr3HS)Q^%M%5foqYbDslE?N<)8gjQb{U;PJW
zB+la1uDDGdZ;zL1E3O3pv+oR}CLQg0_fz_>G24`R&Q*7+D0M11(W35R&vrY#Eu>kl
zv%Et}?iB+}l)Pjbl1WRpebCNX33d;(O`OxY&+)Uu{+_4H6RA9ta^G|71NZvnCw6Ck
zY!UQUFAB$pA9zvFf+nqn5*VbCt))04e`SKYi*1Y{*|7K71d)*EH;-#G-r~~0>th}E5hc+0o(1_ZaPK_Z#Had%@X??uzjSUP*xa4L
z!CYl^aMceyAo3+l@qc9z=B>nH4an{E46N)+dD%i)kb=cxWdH+7&i(yT%%x?$_SWt$
z4bS!N%QYO;wH#v@6Al>VO=;IR_2c9U{$mL=PrBC%b1oOB?>S)Z-A44Ctqo^PVY7I*
zczPla^czk%0gfLmeVn_>vdQnQtU@sA)!0}r;?$>;+cmQqz@Yc@MBsACQA+a0C~Q6iq+co
ze2aCZ=dR~RaDs8-UPo4Si=Ww;trohQdvt*!&IFI)_bB*{-FN+_>lmU!pvm|kZ*X5MURi}OD24IzT8eh2(pW1)SG>vI*4F{b$co|ZK=8$`&=U7V
z&s5*uZHJKC(+1vPqCjeO3J6De;s8s*{pLV4HRfo4p+Mk(w(WG3>CkrNZ@QQBppvMl
zK40-7R|SI_v#0yEr{#NAgIWG8)ZZ;3?#1$>NtYWV1_$7c?}i2PmKO!$cnOo2**|Sc
z=KsJRq4>$a4T;05fzBSZ&f6vTPlUDJd!FK9H+sr#ZWuMpodG)KhS2m@%M)+w1Nnh`
zIxhs3bhS8WE_eLg0>!38w}#;z-<*Boq=-Vz4r|0d-cGnr^gf;?U3FG0fF^OFG=S(O
zAWA0@9uRx+HITJql~Hkz#oYn935Njnn^mS?vdF^n65GN)hf-rbZ2F|XFH}gR*48Pm
z(Iz7#G7#6>p6)U|CWGtIcMo)SDulzTBf5<8F3n>>^DUrhDIXQ$_NS~Cc-KF(@Hm&C
zt0)13od}v^g}{o2i{gY6>~^j(g6IL@j5R`YWm0?O<+SWmvSVhfH66R2&fc9kaK3^!
z*a@NWI}9?p_Sv1DGL55&C_*pZo`Na4G6`wN*a?_aN6629L*Qh=ZG=N$w!W3S)%J%m
zOu9>ELsJ-@#-(mzwpi24h+>(+@*jna{FoOHNBf%v>WNwTHfS_5@yYsFI_J1g=)>fg
zLxw@=4$vLFXaytfL%!71cPQBun0+Qcc=wAtr2S7@)UI50btZlXZ2Am~q2QZh5Pk2-S{xbZ%Ro&3Hqxdies5kaHjixO41nJ*B7s^xJ)Y8=pOVf&lEWaBeoDu
z>4%{pX&kz;leBG{*nQKhpIxWIrpXKambx_NFT{k_t?=zfir$+tW+Y70hH)!1NN8f!
zYd{rQWWYiJx2c#$R>&(=9sDW?tNEq>L;3Y4FGn`%`vf6FSw=3+k@BgL7D2V9J89Uc
zC_3U3AQ=-`+BFe&cEZ4aKmIWw04dWRG#V^ByRnwQRXFnqXpBbZ)==PdoKN4+iv1pH
zj}b{~ye#4Bw?#HETD;W%cL-*IZ?$gA7U@Nj
z(Y5g^$xbysZRuQc_A
zylXy%@VQxRieRVEa^oYzUu9!7J8s@|Bn~+{^1Lts=Zs`
zfRN*+--O!RCI2eVfAZmeb8Zq3!9HyRvsQc@CfJs;^j1^=dJz>a18X)Bjuz8t(R&z55j`%b$jHAs8Y7;4;Qtg)O&ky+}I(3zoC!Lxmh
z${|IY$t*kB#poEZ{Vum=5AYbudg{84hwmS!#jy~c+$}TM&9Y7%S>wuQs%b`0lIhj5
z{4xcGl79{Ry0X9GH}~vX_bLVIStNKI(X!w;w8Bbv-9*!JqeQ+bc4b%B{
zJ%OEQC_|(*uom#JOA^tM%E`?vRMI;
zY^A39Z+j&?3r|1>J!ZtkJ$tT22RgtjDl~8b5L^nl$Mf~t0NcJOpY3Zsw`IzcsBj8zMc2#J2hyo1$l$>EoW`951J%jciB7DVBIw`-H=+SPE!w3
zSKJO76~!@Dl}(W17%eC_?3fjr(&0&g2-U
zuX;J8J5XhQJJV0J-!9*dx!QvbmE(WkZzh52&B?omz%j)aOd@N;#t4-At-KHJnHFwh
zfUj{pKv{L(-yyJH3BWwm0v2G986~Msg{cS@qHJ!A
z8qZ&zOCpy_(vF)5RQ~2eLr7@xsTqmSt1755SzI-&C#5bSlagY;0*t36GHlisLo@&(
zm>gosD&6&%afc{>bL?8$cq3iM=-0{Ycd7ABUKrB*;dPBKP^`+x3~_ZS7>ZCKIHDuL
zMDEta`ccT%={5I0Baaf4d`c08c~yvvZ{%`P|Uf
z7lG!J*gT;E3wI7cxX@&M6~_9YlqhT6KMO9{@;yIbLN@Y`UPQ=+9SUjV+usyTgHqGL
zKbzFn&U);Vz`{qm%_Qf<)Xw&jy$b1lGl_%HH$G2#NYA}cEBDS-UW!9T#P|8jZ(6X#A{rS=GqLJ6MSeTUgn
zz`PA0fo5()y6D%uh$V5H(wKahb@S2zr7o3CCoOb^`IDrCo;x
z?UIBnR7iG`tQa_-e%?7Ya3P^{6ipJHLZ=4x=qN1y;8xi(6hmR6k@2tibr1Qd*spg!
zax+!El%(>vsFR2CgH3MxOTxRkD0JTL>Cl7^ei5<0+uXA{|Db_6G9*5*cNSF$O2Xbg
zTMO+OogU#rX89Sy_+okt7n4&F*>h6Gn;oT8>X>RAx7@MySD&na6Aor+SqYeDl_Wfb
zaJw!&Dn%t`I0c{k{%}kvg}?wIL}J^kbDa+Gd`)Y7_GuHfqbdBf%G6Ffwf52T%hBH?
znM)`b0(T1@lhopDClrB=)s%sNJmQXzXwQsPHUy%z(5`B?8XB4E`!>2&_MH>zc65>f
zq)d3Imsn#iKj@5HX~IkB&_3$lNGp89;2DrBI2@B9}<
z6A<90f4W^&X&=si<#qusGeWS(_Rk(#5|yh}`iNmJ@?qVM-^J|F<@J$){==9H&@Sg3
z)}m)nOFVNuifiVJ)JlKDXiNU$k6E#0bQYl<_&{ZM?8?-mo<)RcY8yU)S2coHNF3M7
ze@0Xf?$`mcDs}b-VS?)(Aaq%%+A2O4Q^_1_9GNH&9;}_LmxMOMf5@vM`}X}Gc9kxx
zqrWv6n)?pWBFA_|uaNr{abWMho?N@600Vh+Se@ZISZ~x^4tT^suY`NUb_G={N2<(g
z{@;~6`Q>>*_=P>9D{`LVPRSmz(-7y{W-YS{C=zSe>%UTt4cRrQ^pBK2S)Wm{Ir8G6
zJI%lg(XJXkbv#zcwmmwaMKGu1e+8-mWJbu@WD?iqz#}pZatEj-m03dc+K(An4nM^3
zFl40Yr^;Ga+SNM)yc5LC6S>Y8S&c(J9c54}+e&VJ9v@wXRqkepNbn14+dP+vY%>Lq
zI8&MC0ETVnCYNv4gvzd?X!bjTq%Y%NiG`7H(C2tN<$mP&NDO>BUsT{l5%M^Jt@@-R
zDnCZrrt~bfJ*G>RrXHC1(T09avdYZcsT5H|T2@2R8@Hm5*x3mT16NgBpQFX=#N|U8
z+zO9l2X=%X)iPI@`uqno$1Y=XkIlluAcXNhi-E;nU}i#~PgXkgg?geia;feFIr!tg
z%c92`>GCf>xlgsb42FjMzModBk%hdO2`6LlB{pSZ3z>?-m7E0Zy=u9BL3>jz4MlG(W@|3?2Bu2fuII(fRt~cVglSZWpF&iwE~4J$
z>Z68r%fyP)Zk<^S;$hPAfIN;@B(&`O#IyN*Bob-nx#7NZRf|df`jB1b){Hpa=wN_s
zS6>~0=_hBSQt}9lNCn9Ym{$#5bU__Z3q$KC!=rxIFd0aEdkSY70i3e{|D>5c+bxm!
zUIiVV^K=8m)S+mUkI=W*Qz46G0vcktORqshZ+yK^^(F%W99zgV+Vv{Y6a&N~yKA4y
zwW?+TxeQ%!HbxDWN%oZYp~3c@wUAY@yyUka`I2sL{^82r*@Y%;z`kV{p)lJQRD(kT
z$%E-EG5c)Rxb(;4d)Xpz_Lki{gzx&w8I0Fg+bgd6sPq`qUu&|BCFnK#S6oU~_{tee
zU`iTo+Sw!5eF|G0<~7Yl^6%|u6>b<8uMMfKE4Lv}tAO*v59|(2r3)N1z0;LFQ<||>
z5`@9u3C4ehAOxRwaA#HS#|rEiwcAFz{e1-L@|x9(ws!ZHZ}C|_Her4Mbi5gWOC}W`
zxhKMpG3=*;eN&C$RShO9?dEdR4#{994No$YMuf2AbP~VcQi>lgTT%|<*WU+RFB8uB
zuuxEM;YSNo;WZ@E{B!3j6vXZjr&?E^^FjeL&i*aRhg_NHbWvo9t6ZyqANlW6
z*kln>oW-N>3ZedKAHc$QghZyo)7F;ozUa9@%cz7ZpZD%UH>XuZF?i?69+5yJ9&
zk6Er;orTG~CTuWU3Bw_m%8XEgN8jRdg%9*2n0R_XDFu8kElIC7D!`;+YuYYs^-#$`
z3hiz;n->7Szcxs|+!{!r>#)%gX)nB7YFz-E*tiZG+=g+U05-R0n>*f4ceDVj#8_2J
zt8+$}w3?hA6=GySaNZ;g-Fn<29BV!$u>-PK-DKA%16~aSrBMa(P6gCemlfw5s10J;m^%B4nqD!qLef<0sEz3Sx&3lNIa4
ztFM@JHk(AC&J`VpuyQ7DmOnMhw3EJ>Z}ubt>iuM5nfb0a#+Oe49d1{D-39ekBLAd{jHv*{V_4{;jn$
z7lu&%mjIiIT-x!4Af)A?b?cb}Iw$Hsw1CGyu+RQi47F%swC*dnq;}Q)x^V%XgOJ`-
z*X0u8#_o3hu+cbXLjc-q-9YcK!X#wu`YJaSVog_niZ*e(37&3Kr#7?xONDIU5goJa
zrF;_X>|Ba1{u}L9?Dma-d4;mj_g~^R>Hxb@p52!#%!28ONc{O?_wO%>$dHSvhtAdR
zzVG(G0ErZ!EsXSvKA-x$^1Z=YeD7=so<4(*4!Jr$dJ-WQqh6}byE=5Dtz_NJQdMq$
zsPntL5GzI&l?f6=8SuKM_|HqS03zp+oSy*+@aqz?{TBRSYXM)wTME8=_d}k=;Lf`X
z9QW<~GN+4KromK}8s-mQ2^W40Jh*gjq2I#>E)<$k$Y{(mrVkkAK3pCDAHLo)Dz2^R
z7LGy^EI@D#9^55(aCg_>PH=4qkl^l4aHnzC;M#cOO>ieP-c7^p^PKzM@qJ_5=Z>*|
z@nh3_ty)#JYR;u;pm54|OncH*h9KAptWlK3$r>
z1*PwvD{1@v0YmmU0ceY!lxzQ+VWG5sJ2UUyf;~yLN%?u*VznU27baz)U&7#A)lufr
z1GGOsx@*2|I=sy=sJw}Ul{oasrTv^7DmQuccDOIWZ6?vzh$l|)!sB>wnER>M=(7un
zrx)Nbi}SEte(E3=siR6ph4}&cgR_;p-A8lRN^Ac{v(fpO{A_YHFf8DMC=^)^pZ)=Q
zF(~no)#Krqs#@Suwgj}{7ZzHBRwDmoeJ-?F$k5II$nf#UG>-2<*DaOQr{M2PVrq4&
zQiGw9c!G@hA5y5O@CRN5M+R!f7X&kq9mcxtT`DWcD`R)KIFfu!KCvC88d7%V8~>F8>!8GZ=Zx6dyh*$%7^}%
zB*~i-CDpS?3;d@@t;$c={*0EEdfMN;$9Grt8f-U!m$t{vr`)ddTx;b7G%j;RY47mH
zG+J15mU-rwGZEW{E7&A5{p8;1Ljd-mf$D@x&g093)7aD1nMyX1^S_rP;ixA`bJYl^
z5^5Qo!<<`D3w%`DDHYMVFkDfqercWx6VPSa8YxlO=5qxTf?zy1b=qp@Ic0k|PWFBG
zxfkR1(T@NR*{y9McEOGKZh7=~<1)=VT4rZ(O1ZBCcFYG0Q1sQzbou_CxYGKBuUFW|
zTIkynRzsh*8y8Uw0a$rmNsE6Od;AP^e(n2d7A`n2w>!c-<#+Kof~G4g>$gbXd>1o)
z`wKD|;FaET{i}{ixHchsOwUgC*I6)hnh}ze~}=mgx-gv1?+naWp4J!K!vR<Pto5yi~8mc^tuqlDIAG@6?)F{emn6%V&W-l
z&gylDQ20(BQ(GE^yobOyXr)k=32t4c?A67m7BF?)kn;9(&${HaKYcMb3ktI~+|7Or#nPs#fQ1P_!KaMWiwvS(dR2ZC!NHU30lw0>MmPGqCp=5
zIMR8_5_ozm(9e3wT;K#b0?K
z@sYo^r)48MYetD%j)C^*Vd`+Vh=w+4iMS)v?E7Z3sN|g`TtxSpa^OJebWDtJmL4-W
zLgZd_8C|Ze8|*P3Wki9|_4M%?J5A1OeI}2ltbSrC%^S^YLr`*|OK*F+@jDcFTb9B;Kw8vNT7Q)^hsYZXR
z1;3?=?Ccvn;iqcRe|n(05&{_~P^4db^ZY2KQlX%}IyyweW4;ojJVzdYCU($BdzCTF
z67-lej9X23vzq5@^^f`MH|~vJYR+}k`k$1p&qqR@r$IZMooIT5Rc0Rjode;CbGC@S
z2D^ft+M5*3P;8-3>U1~}Gx^p4(6DB-UEp-36cFn7!0lJAdn1b$x#L2f06iI6oMg}5
zHG70OfY9vw(AR1k@oasjXhowHYD_`2oV)_g-ne%C=C7z$eNM!OFpM`j7T*{nw&V>F
zycat;Y(ZV^BmBc%Iz!e*>*EUoGo{%77(R2d(cwZ0d86CoiG*QXtsU(&K*y9%7*eI8
z-f+14jToyPFbUTxj8w-5jkV@c=#$)ij
zVc$uwB5eI>eM+C()>}hnHqZDh!?wf1I`R?cxY^EO?Hd`u0N1BgEisKe=IcxO5GeK{
z6VBa77ODmpORYY?7F})Z@IPP2dSvDWA|fbVBPj?M3hPl(`+c4Pb#*_kc&u&38Pd8!
z^9b#xOj?W3x%3k=Eh8(GI2vDlvDoZ1eGbfughVk0B_W6R@Sd)8M?C6Zk&uZtrgbBZ
z+${OypI;Y8?N_?J6>OOu;(3}q26jt^-JTV3u=R+U8~Y^O3^~@N)*-YU54Y|$w7ukA
zw{{W$S7T@(y}7kM7mP&QTc#F;u$zIIa5xLfE24ok`=)2JO?Hc_w$09l3QVc2`4WC+
zhY8rXmHvh)%f5w%PT{#${uiGhY_X5`_Sl0?#wzNeo{k3loH9?wCFuNgMqjA|;n04=
z=r@@UqoIS>s35hu^ex8NzL7kyhA7VQ&CV)y-&hQsur`wh0x!V&Jf6&f6tD5nHwdG&
zCdcf?-LiIKSU0uH{IKmXX|b1wB8)jgit7yck|y8-#IW9Evg%lUcs8xAeK8u4QVAVv
zdnf=43sE02@}3<_tvw&|`HP+qR3}Q|q|C?N-07WbQ3%$tPb&Ko!OqhrmPg+#{_GV}
znI`CA_FTSq7((GtA`-)Y7S9*FUw-WMTkKI8ip*V_g5vIlFT~_Ocg#
zBH2#Q-P;F+onj77INrV$W2|2d;do1-m0US~8D)&S1gL7Qge9$K#KIs>0a$C5+DZ0j
z%|?A_L;F7~VRx>qNTZfFNVpi*K5;clS`p$!Haf9nvDogj+oGnc_kCG
zLxjpXjr#`sCXIC!MHW+1AbVW{7JN$=PP`s}O66-w&ts<>bH(yRm?Sz&K&KE+3u+
zYp+F(8i13{hibEW=094}D`V6uN3=dFo8GLhh(^KVN(J9wp4>oV|uvu-G3+Pl)*$Ir5GZ|xO)Ocb`Q>aHCzc@)r#)-v>XHWpb`
zpOHYJp-OERJuJiJ17jdJs+HRsR`OwUOnUX+2XF&d*8P-^lTJdNd0aRa-)cR
z03Ku{ywCn%p@2h=_Y>2*YxVHFk28a$Eo6Yth?|kAl$5^1G)H=YnrRWc{vmpfS|7$O
z?x{CjzP_SGFQ303@q2OL1+b3YFLw(Kdd%5DMp^Af(Fci6{T#+5KGJGe4`a7guKQD;
z$bi9uK$8`dh=QMoZv&i(hR+#sObzK2BKnkzDp1XFkc#k+d=&^}s+r3fqN^PbB3sn0
z=Eo-}!g0RrM~=B8CY-qIoNmD@Y-Vwk($94p#oA7sRX!Phy_=&vuzLG>pC%%^>+;%B
zKh$VJCtoCn*=facw&@ih^&&W*RB%888vs?vFCEP(k)YW65Hn2wupYKVNSK_3!8ZjN
z@byIFYZP#h^bV=k;_5NA9QetiZJ@4%uOBq@as&qKT-ee(>Up|-ygVs
zjb=lC7uiMQ8t2YqT+FH*Z6rY{qZZ?Sh0|8(=oYfm!jhHJ`UQiu@e{oHBh$Gjxk91x!a~8xw@}!{
z=AWNTp-*+4I9g_S-9u;I!YL0oLQ*|uwnW{HqN9IYc-b4P)D4^XT*^P1e`*PA3ZV~e
zr%0M^Q7T-axbl?(3h61*#o3)HTFz!ZET%4bpAAHV(7Y_CCVdnPW~5Qb_T!O$rx;6i
z&x@akKAuAb+ov??hPrYNfUUA|Pk!u>VOYMt-)yxjODA<^<#n?WelT~Ok&fx~U1Z?-
z2R5wx4y;iD{d4i>aF>3VyCqc@d|!N*yxe)N{)q9dH>2nq0%#(UV`|}^ai566ePO)r
z%gm7~Ohf=r?ALE6hID{$wDI*tSq&HIP`(nQLclnM40trFX`K{udG0LlAw;+aW)0?stm*ASS&G6IADl0w>DvVBUb{jApR?YkqMVA!762=CGa9Qws9DPf9
zH*umM(rw(q6~K4hT~4u3W}Fj>-X2zS?vpe5I9`}noY4^YYc$Vae${J0ZJE=u3^K_#
zO9JeY+{zu*7#$lc*mD$|tz?TS5p+5(Wi_|~^>E+h+%g1!o$xBfUwlDTpV+7=@}C7G
zdp~9TvBsF+-alu65~0?oEy!(E8pB>?L_pVov9|3;$Bnuyr{3>xeK6-~s2ZY%F2{ly
zF@uE^YWuueAs#gKAAFy~18<}Up4mCKCgYjl@csTb+F1$YQCu+&l}~oQQ;nXR71{M;
zL3O$r0Hh5Fx9?}^!XN;p`J>lgOYTo58x8P2?o(1edqKGwD;99zwa)R)gB+x>fBxfN
zSJwCub;$qjh_LcVgNt>RF4jvc%SeCa>876Ucj)jds6@|}0$gFdQgEx^J`=W)9c?GB{OCUw
z;$=vpkg{;7^;|BXIs5L9?RD<&BDdd$CHE`rPJj-DJ%L*2%r_zI8?9%N9Lw}|^A$(uuZRzYSAMK>v$nxBjplnYfDa*eU5Jeh-%d*8s6
zOwTO7DZKb;2vPOQ6T4;abuU^*d=n{me`N4^v?}#e7N1(wj4Dsr^r7?^q`sm?t4(MC
z5s!LKluph)rO0x_2{$6T|WR6%Q6CEeoh*d(5_|iPrlFt9P%f#z$>^
z@&VMdZITVTVE1{E*y@k?9=M8OXAm?1NaQoqgohZj%%AVxGW)dn2kG=+ZC@1Pbr#57
zU7E^7iE}RH5^OZms?6!HL^cY6agatVEWm62$uoqx#tza`jDM;!rJ8He=pv)zILISF
zg*0ZfH>)Gk7uJEx3J@$O$>BSq8w$6VG_
zTKeV?^Qj8T$I-#kyjJLRkxmH1F6&O806rgcjCbvc3-24*-aRnEC8Z!%v%;no_FD@^
z(RK@VrQoQs-k-V}+-;>EuW=OC!%p;NMX1cey|0e4QBd^0{_Rb{)a%$tJI6Sol9W8M
zW)p1y3SIEgxe%PeJ}>&}zvvcm8`Mf>PiV!cf2HPMGFY3HBGqI
z3FxEq#jn^`Pehzt$~j4Wx#y@n>Rf7WG@dSESGEJ{Gz2Xp4I0Sthb3VY6jE%$Pt5Dz
z0BjzR`FYJF4|CBb9Mik$zN__Hw}aIT$Q*KXRhd*niTnO(%gnU+?HU%3~E-2JKc<57Ciz3UL4`@`MaqiqD5yXi?96I`w0}gEc$UJzuSU|Kzjti
z@Gl*-D&q&CMC)_OfE(rAh0uCNC&ly3#UM-_-xLZbHe;O_oukuY))SbDq}iLU#g2B@
zG>f(n>nDX(&M&~+b&PVcXRls}KSR`5&l_ELB)&x6O(1Vy-V2VV?Uw__;~`O0D=n6w
zhNsKb%MXaiVxzm2N9np_q?Mc&!knwMh$0QZX@1(bRWpa`L<^6|pEDoJqcG#PURlw%
zat?g8an`?W2wgRzLE3k!J`ud9xLS7HjV)h@m8dT%&U$U`J4yhFjBtsA8oTdbntkg#
zXM+JF1GR|khy77O2z?9K9LR%Ob^e3%>~oMg#4e~L$0BFD_FyjDMegdO!1QOK%^IV-
zOTN;WAaUXht7qSI)z`%@4(TbnMjRgUyq_6svxi$&Lh%SF`fDl%#(Ul|*thKsSJD_v
z?Tbz)12n8uImkQ1>(N9$IgxF2mE#LfK2Ngzrec_6bANb#80UBqaX7_R(khKj{k>Hx
zQ43||2;);+mHwTlsRZ6K;thNwnR3aOXuYugT>uLg8fnA?;~jFW?AzhS2?b14VlpG^
zKt;W~cd%x6d1zxJ8sxH$EsqbbM9f#P?sqfm)nH|SfsTgHlCNdISif3$YlAV&L+B3)
zaW~J~xLQ00CZf;Bd0)Lf+Nf5*?C*UVlFDM8h6V2-O#CC
zY&AQMWcE50LRk0`{9{SwiYh#aS8{m3PU22k8&Hl4l&yvt4J3^%@c3r7LjJA(r6kiz
zp91Rc^U5|g!QSV2#X$W54qa{+Fj#BR
zj|g}cJ8;GuwTYwu9PI!1^3)up2m;guL32fW_S1QR$f{sJKIiT?O~Bh6i|aioQ4+x
zUQ{UWFY@(PrM^U*oH4X_v>GPtMw|`g78aR&NK>5J(&rxoEo`)H|ETq1Bn
zchK@W>fjG*n4h8RsNad;`J5xdDPI6GsS1vmDG+a1@BWist`(ey}vjcT-26`hD7rM!@^vy;8^*v4u1pEu(pUMit9aZbLhS%9dJRHg!ff6ipvCYqXkI}xV7>m-<*i!t@adxiDnNrXn!Bo70*
ztK*5dpj0Y>O2t}IXQN_%GD}gjo?$u-rW^YJt!tc<#e)xc`r
z8`_e-N?@E8*CW_@sVvsd-gG&uHcmNDLC;~UU|{t_Y>@w7ZV4#vsS`I$L{2y?>IglsHN-+aBG;Q`6v*Bh{{#1vsru1GJaePHWugpy7f?^gIn=T>c#W-
ztzVvjG(a0e
z^KK@h(j0Gcsyj!jUBu5W>EMdDn$h}^ACPMYYxg{wHYUQ}L%>Lii;K%a>s~iHaseGS#yxxrTNl6P1uw;$WsM6Fcr^+3YNT
z7)xjXGXpL#UQn8fAYGb71LWsFf7&}1Crx{E%ErA)aYdZz>+zs-vZ@xX;5($^rvo}|
zEwc7<$!7c)(u0KU5VrA|{_fbuKNy@je-xC?11cBUk@kPxa`Y(oQ7n4*q03Z;v$)7o
z?YLs_FeR5q=!TjhplFzRO(qMow7xQLoN>B%K=(a?3{_GpMS)3kjwTcFfXTu&DP;p2
zoQwdQDI?i(G1T1s#IPi?IUq1he!k7qR=?>SU1{^#yx)=+OcK+RssPfXMDWZY2fa5?
zJ-~!Eo@?^jG7U9`@Iy)9B^i*DzrT{OrldlP*+q;_@$rWpxI!)qT&-D&f$R*Km53A0
z?^$an!`DuqD%OADGv4ZGy?pV2v)u=hCslcoIiZhO83`
z*-
zhC}3F9b+^R;*i0aG_M66kR(FDdeuAkGqrimw$-UlVjG*c`Ah&TP59kNuQ6&a>D6zK
zg>2>dy3$M})U0=XbbM(xJfoTUaQ-#7)kEK>pKziMFBD48!WGyVxZL7acr!U|d@@XH
z32l(YS7AOAvE&jSyfAZiG9%(k-agy7FZm{z0W#0@O{i1Kcy?WImMSb9v25-zc%7y!vF0eG@@v)&s;uxf7*!Cq^Fs
zo4d32OHX>MaoAB-
zPCyD6?UTgouqKHmk|-Q+lK;i$Xt9`U^3Ex>J#2^PBq+>w4Hb7JVZ8AF$6B~*xwe6W
z?|U>fHEkoyH6yl_9!wDr=V_4nYh?XqmIZ-Yn^o2$JO-MGK;DfN=ND$Eb#%0cUeBbk
z0oFYqUHl-og5``PR`geu5?kibpyi7SVC0rJ`>f8j)pmKf0Gv^aJMWRJ(l`vXQNE?N
zw%Yze7*a@FZS`{6ajr$fF4PgSQwBA_)v&$0z0gd3=^8lTi`Rb_pW)KT`Ga;mi@JgoS!{J#@`9{Ty$2zvZXEF*>+1>c!!8$R3BG6oPbjJFJO8)
z-`nY^4zF&!x7KCoemJdB!#quK(OkY%HL&+^FF!ay1$yg6jNI2qfNU9JuTig$5=)tD
zEEx~yLmL&wSA|+xxf`a9^YPOw8UM)4h~tV0&TX!rCg@fec9Kd=QWZrw&k?V4Cp%(L
zooGy5b&O-!yMdOkzpUEP9>8OUwZ5Tc+{39+1Mn4gnY!%C>;=J$QfLCxQ1@J=m09t>
ziaM5W2L!0y8FPZL$%3)3ooCapM`|KHPd6lRM)Z>CWklqWl!702JT^l$+V3h%C%XJ@
zex(@$jM@}u)1HQo{nS=@t8Gp?&gVL|MGKDojAT5HEy9j-c`d1~d+aIKg>BNuL^wGU
z2ecB_R*nBdAmEVKM$#jqDMCV*iRlOaeiW~Y%j540sw9}LZg1~sZ|UBsnSH%wNV%FK
z*YO!z41`C(D+)-feB9%XXo7yxn07rp)Bli$Ljy=Tubz(f70$|)I>yW{#=0k(>yZ-B
z8vPJJhlC^CF$#pxuo~h-XEUcFXXBj-N}^whZ!JQzcqDYCY_U>Lo|l^qEE87
zDsUTF_28;!otsaQNhXdAU7i?x_ilkOAD0^XmjGGB@c9?>?@v7w*}ZW5D|Ah$)r*^1
zhd5PGE=5AT6{C^;D>~}YRNiIZp|TdidK*HSpxEi5elD!w;h%I*j~JHM1WncaT)slZrP=L!4=L!ZnU
zp6cLXPiEE+AF35BLfK_Gi{JVE>M@9tiV4d?N^&424V6Pk@g5BQ;(|efBdL7xvJe9`
zXptrx^>5@)ye+Zu)FreP7p?#W#qg-Z)OMJ<$l>PX{HU{j2bGLRjU6}?iLU3f7_@xS
zk)xH@bx*Bz&Wl|A3#r&$v_~qNOI4;lNugqq6^xiAd>|vU6s9LZ0_(r7nVw4Sj5-eI
z%jf3Hz5*BzC3>y0Ce}|#mzw%5BWrSi`F{LM9yu#bh9yxE<&p#-zsK2KPA<_Lbc>G9
zej<~a4K}apfPkDUy}TO5B9}liRi}hKhtET-f5O}6Ii4sR=b!xBY0qLE_O&OBjrDFZilKqdsx{);%#Z#d#Oon
z^XSag+%tEWs#VHUz|yn-qhNp=cX@LAh468uQt4AZ0y{$u*e$|_v!vDBI`$7#&T1Y~
z8qKu6QbpQvHAOD!y27CN?RWV8Ptl66LoRaVrTFH54>Hi^eIEjOwY+Yp=sB>lF5ZRn
zkP=+5Y>w_^zwnNEpvr?Uj)RKDKlT#VZS?YXn=+
zhR7DaJE@4UrlrXp{UuM^e~1;}oxKT9jjYuzt0v0&7IKQFSEjgniE9!KpZy7)
zRj+Njpl$c3YNhogF;C+pDMbGvl$8GN`_96z0AZO@bb*^|FNg+sN=IYnTU4FmYEQd4
zYT7+euOs({?xe7}+MrEw|6r&2RPZ1VuW~6R8Qw!811145$8!*lB{Sw5R;4s<^0nKw
zje)@3HE6&AyM7VU-9ZBpD&$xzW|Rg6ViK=H1wDU#n7TIC5l^%^t4-wx6sW7d9uGZL
zP@Rz8*W~zY=}9I7ZHTRdU0nj;V(NnBsi1BbRC_B6?|H`S-D>sHSpqDV=$lHBpz~8l_l7tK(XLbNdzUQ;=rSA^=Yw?B
z;e#GxtSYzkVBfbn!(}(|KS^|8@lNyLA``bLvz#jJ6-nwOSmZ%-(kz-P?al5vl0SW^
zkT%&=KM4N(`Ur+Y8#%|5^LdsckhS=U*MSNS@Y=PTYo|qAf~gqQf20}ve%jKuyTKAo
zAmm_7zMW7%+hysLCm$h1Pg&ZXS2HHFq`PPWCW+W1`rO>Tj-lMwvA(2|t=X=VTp<`0
zRs=IrgHZfi@$Te>Xx{~}wZ;C}mEmp0eC-#G*wY4;`?G~SuSS+}I@3c=U*#CZ=UW)p!)4{XwJR
z4WRYJ)buG_sI@~4gofp`>w&`+tKh5gR_5IQkHatjog>1gFVLutqq!BeYFVN8vQ?$#
zNh*ckzf{QR;*=W{}GdKhDS#^IK*2`vYpbSM|uLf+8Z54U+4=x1)lhpKPTh&4&N5+lz=EdDV~mZ(S|p1ebOw8mubi)
zGlHC_X3cl)K7u{Y_ZZ}I$F3Op^e>VPJ8wy^Z;%1v#gDdHs1C5R&+qG
z$kVZl==EhP?wJ$W-%Ao*0>_o+%QjN0`O2z*r2Xht3WxDZxqFY8g5KOx21*I9@Geso84Ynip=OL)T9o79#?2!
zII&T2eys|08-LZ&Mc$b9e_Zi^Q3vOxj|?C@CMaL8!7ACM?Y*4C%Z70D8->kTGMR?)
z#!N(N1oz+G-fvrmbn7C34tY2#ysx_^pn06~zA#`fO?b??P1}S>ksvT~i(b)!7}X~!
z<6WGwhTdxryD1b@4y0CnPKK1fc%+8hu-XtL<|m__#h>g
zPY;zhevtXN)Q6^dqh|hd>Yyy=QX~A
zR1EH9U^;vLf-+w4h1H1{0Fb*E`$M3Hv*MpvOQIcJ6y|edjq2&P8fqlQ8&}tgRb|Bc
zyHL*BJ2$g1##j9vGVV4`L|t7WC1&q9Wf^d#Z%i_t>7}jyLV9Cr&O&1Vz*v)oOnG=2
z(A&pd9Vx_YL9G@q1Nl#9T{H5Gk0@TgM;mab>3zOV-fB3&*zI%n(SrbOIyqsN)G)*jw>vlEjBtPU2K6+$idG@}4T-`J!v&n}Hc
zUcc5lqMY5`D!c#L0@H0yNdpnVp`CF6O@mb6hU*$kZZAwcFVxb8mBOHV|3d^;m)tmu
zlk@_6q5={Wf_iv)$aN6fjrinbEmZ7>&FwdnC~CVo4cDSP1eGe6e3d|%4NDR!
zQ4xWEd7T=&-kA&Aa1Ut*b$PAJ)#s6bW%??@#ZWNM=Ss2Q922S{T$8^+N!?C6Yetjnu~PA>V|j|N;v`zBf2l@wX3w=jV#&=v3U5`1~S
zUaQi)e-^~_YGUae1{1>2-X&f>&r}1l-1)$$uukkKGL%?3k$3O*C{Zf!Tt;;9r#zxn
z6G`2Vk^65c*rrr|xd}B>2F#0YLOerOv{iGlrytk&1i6kD!23U|tP*biz8HqC1#oLA
zDI4{jn_E#ZaYB9m7+-FbvTwH*b?CSH?I$g8)>j~kU%wp-i5&QzQ^y;juk}fJ!AC1Z
zk%0tiM2Z-{Fd>tx?Zzb*U?qwKgi{-#_H$SRWIcCwn3}xZceXVHe}`3sI81O!O;9?o
zt7>3Wp?{?g4%fZwrkIyyVM+wtn5q^~Pj(F|9Ov*nyc__|cqI
z?5`E^A69E4+eT%#XcXaQpHZc!nz=jL27G--DFW51GkI)AgMAjw{|DbA=7Fjm*FCg<
z6eWE@`}OPBtFdF}*<*^>d;u>ox+aI~8raCBCHdv6yo!@nmhL|BkU{i2@>>re%Nh`x
zYI*~z_v7OlCsBx-ivq^x|NZ+dFUnTBOI|SWqTQ}j--*p5qfu@_ZR@V{9~>@&2G#IT
z%kWZos*Z&2(Jj5CMp2UqHgr`Z$2o0g#i0(gs-Yi4d2yffod|9;6TyZ|z!u6A5H)(&
z6<`cDNu2_~IA(4sK3(5qkKQVz%t`C;S_)qFXkd@G^$&A@T)5J}r4XT~#=pj1X$Tiv=^W3T$>O83sFMI+ZghH^6G(vnvg%~4I(sQLHU~H>fC1`Z?8rU
z=xzF@+iA4d#i9lj?yet8Xo50n)tkD`4`+lcDtjk58+wh6kByKLD0WK}&C5?oM)adc
zuU%->Ss1eis>A7K9Bdzu_?k%CXzCE!d{Lr6G^_soNss0IUjHTLtl)W|teV<#CHu*o
zPeWcthTI`h>aM|tA;CLaTBo4@fw6qGil@Nt1(kvxwZ?82TT&|xd_j=}d|o8=_)8#?
zLn0!F*3qj=qipLy2%z5<)~I5;)gxoOvw)XFKhe*125Y}{E?44Y8cG>9SGE)k3uM2J2fXvwCT6DjMWKZRR
ziN?{A)Sm~-@ZQHTN+uFAOfWxAS7sbX)dz_M(3=kZr!-@ncNxmJaGvS8O7Tx}U7_>6
zia}I}4(`>Q-G8a6!qD1+EXTAjeo`_zZd7F=l84?&N(r{m*#w@gX8)UztyyB&>Zl=<
z(+JUFbu_nsW#*p#`7+`?^4$880qL32wISg)@tEp
zvcH}!AOi=Y&I#s=4$=OULgr$bu@ZmytTA7PXd<FMI&HUyu@cNLh4?tyU)B}G&P{Ss=MhW!n@s#O-+G??q4TScf>;;6c6|o<;7FD*l`tBzkzk@Cn7JB{0
z2%DM>N_LNB+qna%b(9u1TRk~B_W+HubRdt!Y01zqJBlq?zP`Jxx%?n;M!v_ww;y`A
zVo@g6WFeSk8Kq1_6_<>`V<(X|8_l?k;b~T;-(DHT3Xh*f?_lu&X+H`mrPPt35+{j?
z+`xVG$Zh>SHVhR&!uNz-!U4hgEwK?1`Tas66p4Z_E(3Bv&$IK2calHAEM{e{dI1uy
zX=>p{Dc;->j9L{RruDVTRD%)|6U*#U#@r*;zhl@X36JCvytbbdA&ATE5<>!9NgKVJ
zCmU!K4$?o-d>~*;3nOiSmiM0*cCU3Q;f&?0OSGsqV?ANENSXNn+n3XoyBRxGWcSCd
zJ=aG-S85aFxB^Z+d1ilzbU9d-jJU)S6h@VzSo(vKJc|5`kwl$n3Krz$3LvqVp
z@gx==*Ea$nX|6tg)uKqluK$I_;Y^|a-MU0DW`n6T*XKwMW_bhVzTc}T@xm(NX??iq
z3Q5xyhK+wbT6ca17!dyvlged1rMf$oa(I{D^^xL7t9FRh)riFHoJO3eRfw8f0ym;T
z#b#RAYdPDk>9TKN^lIXgZ3Y8;TktOcG+j!;ee(Qse?BE!Mm$eK
z@1KOyFhxH(KEC%ku<)hv0#eve^_(26t`keZ-)>Jk*YWEuf2|=tWscEW3>?mYaM`3{
zF=-1O@IF)tGDgVp37&T@+TY~Vy+%SoyhteMQkvRl{rn3szZ~3Io^tLb4!_LFrx*BT
za`8j`Z}F&o7cAuc5G_J{;Z6u4%xY_5CWIWZeP<2H5*a8yr+%vmeMF_0VLDeX>Jud|
zQtKh3ionO~Uo?1dGW>3gfXu!q)_DS>dp$$4n^~RFvDjyIi0l#Fs5Ni@W;NiYLiqrpUI7^`!GtLDZqCHiJ78%G5uY(xr$%*&rTY|o3Y*8e3+NaYmbSY+#X)1>}QximLTyr(x7}T-SENVT_
zz4Mg~=DM7h1b|;kpB-PiHdd1*Id^66insr*{hgew{hgOgxj1_K=Rr^ydAUQWc;z@?e_38
zvj-DCwm87KTdf!bWFqu64CKPtr5t~WiaF88GL8C(l&MD&jj|e%qKtONESbc~EQ&Ho
zAeZda0?of;6vbC=upSLWce06KT9%^@ctpT&27+v5g1f^IYfkpy!=7UF-_(!Z#9EQ8
z$`_-0jkbZD*PS(}!cVi6^X(q?aI?Kpkt@5IsNTh^T?Yd};8SWN@I~H-7C$+3gVG;7BMaC}`bG)u!LyrG`U|YE8uxO>3xN
zj;EcZVCU6RuC300wxWLN->pbh%WDZR=m>i>=mEg4Y#B|uJ7JB;S$C7q(k^{I60p^)
zmHW66UqWq;H@%J5_<-ksrh5M_>}%L=uD`4rtWd4yZL*)s49F22_ElUh({8~g6Y-Z^
zZ3@sSQ_YQ~8K^`&*qEk2!Kr-re(Ko3Z%#GKR6jBiG&mNOl*}SSi477zZ}h#kt3Z;l
z-Z9ebCW|P+R~H_on*ahp$l}zvi%PSR5!?QWR3IDN4F+E>oy)tlFbM2wt~M$?9d8N$
zi|s@v=vjn#jjsCH0yPA!D
zAKU52FAYcnN`AS{e&A@EXo!m(n&2ryiMqUN*cM(hbqqReX%f1Ez)rAf
zY60P36cp>ofs1MST4cjw`*y6L4Ixo
z3gtr3(dFcZ%RGizuTEPavrS;*<3tSQ;Ad~X|Cc8{>-#t%`X1|N`84n=q@cYbu8}0C+VJfor?g
z>cS0J8J_Pl(1CDCVt!N*GnB<3e|#T9%p(u@(p*qR_}@Q;>jgXe4%bf-z4gH;f;oxK
zf_#_4(5tJYxP*!BueEFjkcHCFpDG2qxFs!&v%f2$kYGj`6%`bCG@?j
z8D(XBLV~zSz&$PC;bNFZ@d)FI2f*a0rg%7?0UiY)K
zP1R%?kk@D{3aiIC@JH6-b~~Orv1a_kqE=Y$c)4y<&PgA
zTgVo03oe;pVEBK$!3MYYUFh^8o1L}H=I@_q^JQw(u&X(ha=+U%PwUC94y2bL=C*l%
zy4HrvfiY3dn2c=eMK(PC-yZ1e=T)X71a{K2JTH(_*-An&NKkTI#K+j2a{j5x;?gNj
zSA;nK*EK<|)f+$OUB1^ggOPZeJpqW$?yirwNt?^3t1YEVOS+6Y
zHQ671sBr&ZZyUjQPDtn~fdPI+8$_SypNLGdY-Wq(#PtHIrwgTH)Gr7{9`DRg-*^Ah
z!TH}FO3>Th-uV3ymwvJL;c;S76+;pmy~Fx{_4SrfRdrF@Fd`}_CEeZK
z-Q5ibY3WAf0CH#)q(wTUn*$tL>JTaoQqn2iC?O5s#_xT9ywBsg84Si?{n&f2wdTBL
zt!pk7l}MxDYUddqe9bRa(I1~9UTGxMH_}MlM6E%g1)ZwPX?9t;n_0fGAq$OEB;k>f
zd4%SEAF~23x0m%%RWK2%FBSF)iiHI#1V0b?0F}+OIkCZYL4{T^7M#Lo|6TI>*Vj6y
zDGs9=Yvqwd;zy7EL;l$B
z`7W@wTH{6P4y)ZUdcYInD%}FgoRnB`k`O9)FcmGZC~?f_U8m6Disi|8b`UJ{$?I24*NeORTGu!
zv22z_xB5|hM~Qa+*6Eaa@ipE_qhYnBV|_P(C}}Ar?wtSePCqrkIQW%cYn5NN`W$Zo
z)(gDke->(aYBK5YYq>KqJzYie_B?+8EQE$wLXnUau!ExLXqjlRbg_E}2MZKpNJbNx
zb+?*#D|zM9j+(K7fLy;jsj=~1-jvLiD9#YBCE)BUqEv-YQ!E%hPUx^
zga@>E?kWsr2o?jE_JlS!YfJvetdQ;z4v7`%vtXE%iptXon0aY{vWq@4x7_r`Pi35>yg
za2=u+jAiEu;swH{0(1`rRm|0iEO?>CTNiLCC!%ivDIrdA$KFi*^r?JazQV9treY6u
zr7UL)ktGMl8R&zE&>8g6BiBE5r19E0Ye*aq@}BI~4&y{Wn=#DG&)*IuG|MT@4iLGb
za+|5G6^^*iMxJ9oYV5FeLb_}NRI-LLdsh1=5
zfW@GkNM2rkbk40^{qusC@$yrf716t+0nU>jx|*YB9Z19UMd*i4XIQSzU
z|Hs;*--`=T-8&v5H0S9m=?as$&Ldn#jggPEidz#P^FqL;Y8x0N^P2`N-OJ_#6f_>i
z*`^c`5fPchrK+k*`|>{#?}0gB4VGD@O|pwotVc|2e>%kjN-_7@t=M&;d6e9HK9~Ia
z*%A<{6;ztGj5WB-DG_bk45jm5nmPlNN<{8*kW?iqU}RXG8=16FEUm1_7n?0QLw;0S
zDZbx4TJ6Ln7hw|AiQS`F8vBQoukb8x2j!J9?D=k^(;f*4}cSymHCHnQr!ERsLo9
zQ6}`M_V!qz17PTX0QoyHc15m*asahirai#mI!gwd6nykAI6(M`z@-Jo+pvRzI|(6;
z0-9(_uU6{n_yA-b9s8-3^ayR(10VLOdq;I!>_T87Va)RPs8stsm5*6;#G0W_m`-T
z$dBl7LJozuArq!XN;x{R+*}>Df<8PYsGE-AFmF?P^c0WkBH0SC$TbJ5
z^mQIJ6`5yX<1yhDv+pxe5*0VXH}L)O?KAJaX=zKnEuf~Pp2{UXp?LJ#3+`3nWj2@M
z_WUOfA)z9<2sj6@nL2>CQUbKK+cC+xJ5Dk4J>4lpw(05NL;9MBgv4yFEGMnkGY*Bd
zm6etL9^EN-kN3awFe2GE&NdJ&^52EkYZfm}6?-&PXu@2a9P0fq2)Jv__yx`p;$yHD
zl~&M`RgU*E@<+zx|4*VqTReRyh50^Fp`BUtT(2a@Cg(Jsh~UkLew)pnK3~QDGtQd^
zFSA(Rn*AM!upQ^6?@adi(y&D4?fh*bxXeJrgjnFD=XCCJw;o6XG;l7_e6iWA<|{KP
z66(qv%u~NgDBbiF-?~UKD+DU?bA-H_e~wjPUuqzF+WL$!9}J{p687e@$H(lKF6(Bw
z_l*N904BY#X2a=})7-4*Y@1&ckcB6~^+TwoQcTZ-f~kA!e0F{?lU)2lHEMk&AwF3x
zwy@aBhL6RGXI;-_=YG#f{g9zKQlQ_nJ-Z7TFVK`IEGa1wV`e+qksJ*E0ljk!{e;ll
zr22L5+qGhfaY*av;CMPNB{%CcOXas1{SI#XV^k#xM#(6g0%^(Tmn0AK!=nJW0H4KxkhBU^{`0R;IuoFsmMK;DBE~^_QsO6XMDcD~}l@(IE>9R2L|D$t-~hu4MG
zMOCdn`Rh!WphM<*u4KSD__4US8i&jg~I3uluzo>^?gk#*b}{f)LN7rI1*O)XED62i3~73vtxj
z>v{M~@zU(*)d5!X+=r~&9Jt^u+>CP;hj)>W{N17cliMkN8wROQzaT9w|$l7fFma&AUOVqur4?pDUB;`WtvOo`Qyj_}ykYbpWeq{p|f
z!f`Mu?#qtmH8g+or@HwaU5GdR9bjoYzo{5kSv;7C#ypF1r#@bxhggkDG~2EyG_OMU
zPk=fJ_)xH#?%{UNu+5k#P=tS5%wuo9HsxPs-v2TXz``~+60_=fAub-B&d&My2f)dM62frneAB7qY_h-c;IS3TY
zVXFNfC#N1_u*&JcU(0V#nFpTovxeNSw1gnKHDa?NtpGqgF6&-OljImiy?+M#VL60z
zP%~#3L|IrV#${gl<~ihgbD<7=#;hWoEng~@+U~eSERv8VUnBE!^`{4)^Q11#!qU+G@(zmp#5DH#f~Y={WgtqPb!J{${F8S
z4M%{R%OJ`()B?SDg+j5EKe4uTGceF4@3uT9M_9F)b|yB%g3%GjrM;+lm^DWm(3RWI
zt{4l8!oF1tCH<6=-oZfmRY4P6DwE%Apg%LY5`2x7O#T`Y#VEW#J_YoBUe>dWe#NK@
zBFhydIP^{}*O)=Qys9P;*!)uyoC}tF$zrH=VDXW}nG0K4BHKYY=gNFVk`Ly*E26SP
zT^WB(T~UTtOdME#D(tkC#DsUcP#yx>zPMj4sh2r
zRpp00`nOj6)~0nBNbK9>h4t+Mf>-#ec)9sKv6~w@8BRH%JgpzlnE2|42-WK%2=H(ONE2ViOtW
zNRPy?CZhUgqx({Tq_-c%_u~l;qYP(!oAjgM!_xN21uwaeA&GvMpIWS4hh<>}Chno>
zW(zD7Xo&al8)_qkK4x|QV`EPhyr|{#N8KsSD$J@T@9VFI<54&XX288?!5&xfD%cy|
zw_^5zhWk-iw~w#{YK5nMMZUDS8a@8eV1(xBdV0&N*W#@Glr_->eUu~qB;&)U)d`}M
ziRd>MX1>d8_XS>6%RRx}JVWgzD@!SuTy6=-ij?#<;=t<|7x1XQi5!6ogFfq*^gy3l
zkKB$YDVDIcaIt)Odj`918=iBXFLKkZM-iW>kNgVO?s_c%2TIr_cr6+bm!A@lyJVv}
zN#VD^L1K-sb08g`J>icEZfvufUq$uupiW77x9t6+A=B0n;M5AqbOX(hp>N_n)!ul)
zc}f%0?cl34E!a|yBDM?(E|Rx0!yUN&t+s_0AEB#?!WjH1zqRwlhd8!+Ndr#Qk}si8
zM*s$JzP99H(G%Mq-QT}tz_R)FnAd*u-Si1iX^=0ca_pEu46_E%fRARHeJt9SWTeAw
zwWKGJeW!+ktmkK%Yx43~@p!GhvftB%>BeZ0DX27`_+k@}`t>H4x@nq{{4>G=Zn%?9
zRDage_;`)OcJxTZb07Ql(Lh-kgk;g>N|kXj3Aq%pfonv?YaGNek^e4P7{j6e$v|vA
zTt%M932@$fov=RrJV$-5$5xKrBg#}_@@(O9-D~zjM%3mNDxD4UQfz@K+7JU>!k%2xN
zH|ce$;%MmBH&C#dI4zMg@8WbZWEP9q&@#naiM@;bOG$XujSKNq=stB1
z-1&Bz=}v&~1tBR&23)Cz-nXv{^I@-1dMvP0qfQX578Um1
zUqefBQp`$@3!3|tSFsQnHc2U_IzB~Fh+CT0&#&j|YMIDNzQr$i;sWbyeG}n&>qjb2
zz%2K{R5wipk6jM;RzmK>@Pc}~_{a}|>J0JxlbsC#A%HcAV27v`@9iS=*Xit|=X{ZN^yM05HaS&AUC5%T*fP;f98NigFnpFO0rFM_
zTBB=8kWU%|c|dc8uI{&AoDEA@$kleFR;^|>>2>y?Y)`p2zw8VZjE!LTaVm{
z3NVC|*?v|4jx~(01!9fYd**KeG{KXtdD9abtl0*bc1U^`UUA#BI#bPQ)ORY|~-E&U2AN9jw_M_pk6E{YCd#o-}B}z}sMOaFrpIBBsK+1J(0
zu27rP$RB4yu$`El`=|cI!UOQkgN8q^bxh#PAVwzH~46V*lOT@LPd#IC-RRp
z5Jot-QdgzRrjafmD)O8*1dfO0y{MKKHIp@9EvQEPq9HBIQcx*9e2!lIQjk*-gB8!?
z#fE(>ZZ6?f+U})CZYu_SC6cU%Blvbtc`~kh
z)1K*>ZxWhoL>e=E^qNp9|=k8-+7^t^bCLxjE7
zpL$l0U0Tx*S;s_QlK9sd6Len5b_yf+H}a5X#+9^d9cONizc|}%@=w(
z#0pxr9Ca}SFT`p?>IVo~5J5Z@_R(L@JknzodS?`yGLp1FDgOJI;xq!4wFFt*9K;#P
zVb)f@!Y(GnXg1OO9P32#-toEIQEN%~ozD
zeUINMsVtvk(mXoP?)4R%D*C^DYW_swGb_Q&%@as|oq7()3$fR#lwZLd)ZF7WUFBG#
z{_ML@C|vBkH}h!bIWmBeVrJ^@ke-3QPq0a1IjQZpa&g_;$vNCH#0i=zT&9kLHd?LD$P@Wd@qT0xH@|DT#Jw3;
zneLlEO{QYTes}75TJt*>la%071n{aBhIVrAE>y^nTg8J@gw?#@4xFmYS&HeJ_34NFcVCaM$+GiMnaApizRVImuL#ywQN6|Mr>Nkp>ggj=e-FroJ`K2FlC
zJNkq3etX!xY7sbhSmvRV97z^r)NWo}eyop$$$PT0{BfLkZ?O-DQzk@HRcgnA@NYOs
zvbT@3j;M5xDt=#e%+e6A9Hk|D)#Or-Lzdc<+%t7f`;&rV8Q=yQmi2Q}1G%FtI<5ME
zyS_?&dmy25>A0^w`1sN3t2JY+CJT6@7W&US9I8{KE^X;i>`DdQXlq$sjPSnWGbB#
zp=q`Kzby_*Ecu9+Clw4xPU)E8o~;DjXKgpB6X3l}O7XXRb=(o$(zq^x
zPL?bfcM@6s3Spy)H#=G`W}{rzwqzA*LKXC$7yjpZkB6=~BMxo^3@QwIdz-D*IUk3v
z(;Dl1lsQmVeL5NeLxvljv^t+LLr#hp^rGL!eK<6Jf8eN}RaBu;0Gp!sd3XMoHt$SE
z(lHeS&t(8weC4#6JLjJas`hIoPC-sfMwAdInW*Ldrap9S?ajTkC5Yd&yU!Y2(n@rt
z?*NVqcwp0bBB+_{Q_F?W!p$kBqktLaOI$_I&>*+;(@Hsh#BYE(5&)O#W2NQ
z5+A(CHMe2)F8{g5it^)wW0@i#Q^OQ{M)h{YbSKf^Z6D9v!g%l&d&K4zX=P8}54G$_
z)@JPc{yospzm4bxr~i9=qj3CW|QCc`KaUiWvn(8k7w6&hr%Y;j9a+>7*Y4*~tRntyN
zX*3Q&T<6)iyRKa_>8&zkV1aChxv)tOFYCP27v6tjIO-mlxFWyTuDH%uOfr1Mp0F-ErB&)eYQn
zPkbV{S7N%ZPT!Z$;Z$85_81wSiO(o0Y2u5_+-iPnN#r$3<|Wdf~?&E^I8
zKT0hP-!{L0%WIzF?OI++?sb{j!ObhzFK6rh5+_HrMoBswR3-#x*7%Q!O=I^`WY!8t
zlk;N#`UbkI80u#^lVla52p&WwJmb==ff7GYYcg-r6M~;jN%Qmi4ZIw9`p6A_O{UbB
z!@EtmjnYB{xYi-~7;}P3$b56EXOO~DfPQxM)v;Yp`>fM5pFL8wDCLc3G>T)uS&+Qd5->9<(O{Y
znBWi(%>u-DtVIRXxFYJezJuEF-~h-u%Q4A3Th(F81d;R>bCQzFZ5#?IV-3a@H@D_V
zoAgC*;mPI)&OyKX9A%@Q_y5{w&LpLa_?@F1h#5BzkThMUE=@n1aJSigU!6oTb4^@*
zVD-BKmXdh<;zvg({L~?Ok$@l%*ZdJ=h~a#@pZ4B6A(PGJg9rGlmwm58A^7^&#G7}NTK^1TcnCw+JHUW4~Y64
ze+BJlV+k^*{pJkV&Ays&54G9&d|o#thb#nKun`U_Fs+Y2@%!NTx9y6#S9n=s!zbM*
zuxm!?w55yVoZC3h7J(K4qmJL$t*>KjihI)pW;2iR3OKq@QpiyvuL
zqf)%inUhg6!<|R!hPz+r?>o1{gSYuihe)w^M-)e2R8f6GM~KKQ6GPsmkPxxv5mcWW
zbZ9w90T?!%Xsn{E?>>%cyf%sw@PDI|V8&ijG**QB1XufBqwg`OU%5BzjP{UE0lE
zJ@UiH3l_%S*@aVu!Q&rvsR6ioRN~r$*253+FQ=f?pDF`7yS>Kc2A$KLg|v!K%5S|(>(GCuXW_^
zX7s^4%f7I|qzOly^%-){?u1>|jdjwS2^TWmCTj4CqyN_#m5A4*O-Zbm!_`jA
zO=qv5+@+pp8Sm#qa*n;X&>ow+oBGJM`OW-bj+#m$owCFFkdi`(PwdvzsJ~^^Q`vZF
zANKWMlSP&LgoKNZiRn<6j{hnX<=5oIkRvhOFpsBm+1};49}jPjc4Y7}3Lan~J>Y*LI`+Ktyet916lBgj3s4CPczhBsaLE8o(GVal5f
z3u3k@B0WICmiixmsB|8Emv~H3N{hbg{=!s+G5WndSo{V21-FM^nRA8n#~a1jZyrli
z=O^F3ETxz*{I+9Y@XI`ew3e01u2%|`;TRT2nx*#`|NlEqY3DY1Y1r5`G4)0$_J2t%
zu(47FXmX_)nUfR?v2`EccfJ`@DCqsu5^4RdYNX9xr!4E&;of$_d8++r6DhL3MyY^T
zSp!&ESCy}@^xX(Vna&g(i?`AzW(z3z0vU^n>f7K<{;g_#_zjf)9)Jau~T+7r5@iv
zw(4%oXM6@*x(E#g|I@Dg0`>ae9hWSnqCP6N(AK9*aB!U%e7#DqtQ8CDmPxggA=YiT
z5G>Idn>wxio_zHr8k*&F{Z^&m#`Q%4pZAH5&WKsJg*7U#Edm
zZS{&7@lFE#qCKeoii_dtAI~f&lk8a?R?Skh%tOtgI7PQ!P+vmJ+J>aRqZ;;k+2&VS
z_M<34Ok%Vz*{wby*HRO|uLF11I4Sx4Tvm1l!dL&Tx9Lc3TtY&58a`tS3HE`e(DH=^
zs!NQpXY8B@Wx9f&tmK#8O|H-3D)e90Sarv2;hau}j7O1kSLeq{302IPu|^fy^{o)U
zP&oan?;34vcCIEs=`66w@OyHDfWMO~Krf`V=6o!8yushyWU>q`%!Xe0;gYcD*HUix
zGQt~~{~_#u|CM^i)w#U9)Y*xyNN
zSt^gC?a}OuPrGrGC3-@3Ib7K&0)7L2_79~;3JyYRT-wGkHN>nsb*6IQslhCozfcaEV`-je!u%wI2yxl0;|u*Ndg722oNiUcKsYZS~A8hl%-q
zo$55LpRZT_As1pcZobqyDIb_T`6*E*O0b$;5HARKsbVkj(PuWG9~Wb{KBsOH5h=UP
zsh!8my>f|{(Gx1pW_JbJ6pM!JXXY7j|nu0Y3m(bZj&;~)(46-eAeKO4hLJw
z!h|2lPW+1fE_`G&$ei_Xf%dfL3-pZL{L1uJS>}lhPTo@kMGuF%>Eg1un3Oc}cxP%q
zH@11_C@Ww_C9&0inlip3eR3<{x48BhM8#UgTN%8^==o2`T$X$AnoEeA#qJW1kdP5e
z;yPTc)iJd7@~dCZA3l}5`O81j@=85pSm7br*cYpbHde_ZTbAboU%<+
z9`4(5Iek66$0q4mb2rB3`LNlm3x8AZx+UYOweQ4lk2b;fMmN{2Yp_8qnaC{ly|E|S
z+#__jHJoLtK-;bXZ}WJd1_T0{UK&w-^gT17zzF_gY;?g}!wlWceF5G&?UrBx!VYhI
z?2BIa{$CpmPzq8lPz&2a8?3CX*7v;ym4FkG1@sD`Z0@cOLuksXkx6qL!@p*B8E;RM
z%A1=yc74814DFR#=`MWV-a1GENxCL-#(JuRL5|?9xu#yFVn_>iNBoVMWCfXmfB5@7
zDES8tG#erVjyoz>)*7=o14EJ;F*lAqBX=!UH%0$#I`o6k!nnenZEcJA`1pDotDDA(
zlWC3khR)OjMBnR+3CQngX2l1Vot{*vaXB0QQICjmHy7wr32cUF49sOyB)ywvTHU@V
zE5fu2R4^;sD)b65rC89Tb!wPUk(Up(wzHdC6myRmY@EH7&6WA*Cy=pVdm7=t{}bct
zgQ{m>(BJA*nGEfYyuK`R?%`x#8@;KT+1DF*DBf0RCNOo4*IoIFkjGoIrsde;3^+MJ@eGUx9-V?MRqg(@w8S0+%*4#55ZJJEBUe5uxDV&
z+=gdmb~D!Ox`owy>*B)R&LWD7f@bB)mOmsH6A!VJ-QDfu32>)w8@J9=e~TS!6&l`Ch0jUFJg2NQJDq

diff --git a/doc/pub/DecisionTrees/ipynb/DecisionTrees.ipynb b/doc/pub/DecisionTrees/ipynb/DecisionTrees.ipynb
index 6353b4ab0..1161b80b6 100644
--- a/doc/pub/DecisionTrees/ipynb/DecisionTrees.ipynb
+++ b/doc/pub/DecisionTrees/ipynb/DecisionTrees.ipynb
@@ -1935,7 +1935,7 @@
     "\n",
     "4. Output then the ensemble of trees $\\{T_m\\}_1^{M}$ and make predictions for either a regression type of problem or a classification type of problem. \n",
     "\n",
-    "## Random Forests Compared with other Methods on the Cancer Data"
+    "## Bootstrap with Random Forests Instead of a Single Tree, own Bagging"
    ]
   },
   {
@@ -1945,6 +1945,75 @@
     "collapsed": false
    },
    "outputs": [],
+   "source": [
+    "\n",
+    "import matplotlib.pyplot as plt\n",
+    "import numpy as np\n",
+    "from sklearn.model_selection import train_test_split\n",
+    "from sklearn.pipeline import make_pipeline\n",
+    "from sklearn.utils import resample\n",
+    "from sklearn.ensemble import RandomForestRegressor\n",
+    "\n",
+    "np.random.seed(2018)\n",
+    "\n",
+    "n = 100\n",
+    "n_boostraps = 100\n",
+    "maxdegree = 14\n",
+    "\n",
+    "# Make data set.\n",
+    "x = np.linspace(-3, 3, n).reshape(-1, 1)\n",
+    "y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)\n",
+    "error = np.zeros(maxdegree)\n",
+    "bias = np.zeros(maxdegree)\n",
+    "variance = np.zeros(maxdegree)\n",
+    "polydegree = np.zeros(maxdegree)\n",
+    "X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)\n",
+    "\n",
+    "from sklearn.preprocessing import StandardScaler\n",
+    "scaler = StandardScaler()\n",
+    "scaler.fit(X_train)\n",
+    "X_train_scaled = scaler.transform(X_train)\n",
+    "X_test_scaled = scaler.transform(X_test)\n",
+    "\n",
+    "for degree in range(maxdegree):\n",
+    "    model = RandomForestRegressor()\n",
+    "    y_pred = np.empty((y_test.shape[0], n_boostraps))\n",
+    "    for i in range(n_boostraps):\n",
+    "        x_, y_ = resample(X_train_scaled, y_train)\n",
+    "        model.fit(x_, y_.ravel())\n",
+    "        y_pred[:, i] = model.predict(X_test_scaled).ravel()\n",
+    "\n",
+    "    polydegree[degree] = degree\n",
+    "    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )\n",
+    "    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )\n",
+    "    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )\n",
+    "    print('Polynomial degree:', degree)\n",
+    "    print('Error:', error[degree])\n",
+    "    print('Bias^2:', bias[degree])\n",
+    "    print('Var:', variance[degree])\n",
+    "    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))\n",
+    "\n",
+    "plt.plot(polydegree, error, label='Error')\n",
+    "plt.plot(polydegree, bias, label='bias')\n",
+    "plt.plot(polydegree, variance, label='Variance')\n",
+    "plt.legend()\n",
+    "plt.show()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Random Forests Compared with other Methods on the Cancer Data"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 27,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
    "source": [
     "import matplotlib.pyplot as plt\n",
     "import numpy as np\n",
@@ -2022,7 +2091,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 27,
+   "execution_count": 28,
    "metadata": {
     "collapsed": false
    },
@@ -2035,7 +2104,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 28,
+   "execution_count": 29,
    "metadata": {
     "collapsed": false
    },
@@ -2050,75 +2119,6 @@
     "np.sum(y_pred == y_pred_rf) / len(y_pred)"
    ]
   },
-  {
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "## Bootstrap with Random Forests Instead of a Single Tree, own Bagging"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 29,
-   "metadata": {
-    "collapsed": false
-   },
-   "outputs": [],
-   "source": [
-    "\n",
-    "import matplotlib.pyplot as plt\n",
-    "import numpy as np\n",
-    "from sklearn.model_selection import train_test_split\n",
-    "from sklearn.pipeline import make_pipeline\n",
-    "from sklearn.utils import resample\n",
-    "from sklearn.ensemble import RandomForestRegressor\n",
-    "\n",
-    "np.random.seed(2018)\n",
-    "\n",
-    "n = 100\n",
-    "n_boostraps = 100\n",
-    "maxdegree = 14\n",
-    "\n",
-    "# Make data set.\n",
-    "x = np.linspace(-3, 3, n).reshape(-1, 1)\n",
-    "y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)\n",
-    "error = np.zeros(maxdegree)\n",
-    "bias = np.zeros(maxdegree)\n",
-    "variance = np.zeros(maxdegree)\n",
-    "polydegree = np.zeros(maxdegree)\n",
-    "X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)\n",
-    "\n",
-    "from sklearn.preprocessing import StandardScaler\n",
-    "scaler = StandardScaler()\n",
-    "scaler.fit(X_train)\n",
-    "X_train_scaled = scaler.transform(X_train)\n",
-    "X_test_scaled = scaler.transform(X_test)\n",
-    "\n",
-    "for degree in range(maxdegree):\n",
-    "    model = RandomForestRegressor()\n",
-    "    y_pred = np.empty((y_test.shape[0], n_boostraps))\n",
-    "    for i in range(n_boostraps):\n",
-    "        x_, y_ = resample(X_train_scaled, y_train)\n",
-    "        model.fit(x_, y_.ravel())\n",
-    "        y_pred[:, i] = model.predict(X_test_scaled).ravel()\n",
-    "\n",
-    "    polydegree[degree] = degree\n",
-    "    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )\n",
-    "    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )\n",
-    "    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )\n",
-    "    print('Polynomial degree:', degree)\n",
-    "    print('Error:', error[degree])\n",
-    "    print('Bias^2:', bias[degree])\n",
-    "    print('Var:', variance[degree])\n",
-    "    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))\n",
-    "\n",
-    "plt.plot(polydegree, error, label='Error')\n",
-    "plt.plot(polydegree, bias, label='bias')\n",
-    "plt.plot(polydegree, variance, label='Variance')\n",
-    "plt.legend()\n",
-    "plt.show()"
-   ]
-  },
   {
    "cell_type": "markdown",
    "metadata": {},
diff --git a/doc/pub/DecisionTrees/ipynb/ipynb-DecisionTrees-src.tar.gz b/doc/pub/DecisionTrees/ipynb/ipynb-DecisionTrees-src.tar.gz
index e1351e4b1c57dcdeece9a9ea51ebfb66af4722a7..0ff487d36292b6af24323c88adb0fa4b5eb3d383 100644
GIT binary patch
delta 29
lcmZ4cQ*iB1K{okr4u-4Zjcl!KjIC@;t!&I&*;tm>005ib2_pai

delta 29
kcmZ4cQ*iB1K{okr4hBonMz&Tq##T0_RyO9XY%I%b0FcB8g#Z8m

diff --git a/doc/pub/DecisionTrees/pdf/DecisionTrees-minted.pdf b/doc/pub/DecisionTrees/pdf/DecisionTrees-minted.pdf
index 0f6cc40c4fc831e195c0e9eeae3cf7c969d280d1..6fc29fb0bfc45c85ccd51f1d6c1e80a460be0b40 100644
GIT binary patch
delta 6137
zcmai2RZyIZ(u5!zEVyfMcUjymxF)zH$l~ts;RJU!XbA2Q+$AK~LU4y5i#vp1S>T>?
z@8kdQKXg_1RL|7ZRQ1cOrqOMt(S=4LN4{uA!MQ^gf?D~%U~Ln=L-~LTh;D_L6E)G
z9rmhY$|im$%<0O^A572^jXAZn(R4-z5b&a&v`L+obkn$7oZiCKNy1^y_}Fx|8NQCgq%-wICoekrj7}
z@Da0bYg7X_(}*rn=ICA(c(?rpoj=0~DO!z_Q@){UE%Z0@>0(;yBs{jE@{(XMm&k3@
z_!^AX78Iby3NN=t%J$z@M2u+19_oo-yl)V1Z@h(9!AJFVL(PCrYGlZ}NexxmbmI$f
z^ZQy971%0GxuRlWv$mOrLs+9R%0_C~bNbhE8y-xtS6dZmMz`|w3EO|ErcR2`_W3Un
ze*@?{NKi7oz^<|5clmr(A3UU9A)fI^9VVKE!l|d&J@1t`I1+Ye>0{)H?31)ik&D=c
znAt~e=g5zKIFJDXi*6SD3L4F8+Fg>gzSl?EI96ic$mC7k45{4E8>$};UxwGnF!QT~
zQpERdz7=vLE9lC~8``=O+{TB|Q*?oiJU<32VBl*=
zSqLLaWD|G7AxOLIdyX$D!qVrXS{czRB!W1hnuV`VRNVk-#dY1v0``;eutGw^Y~Twy
zNiIn({oK=|M?#H~=2xx_*940qB&r7T=R@>E&+oA!C28!s-j>JwO{=Esvt4~#&0Rw0
zX~ev8YG&cB4GPs5LfBdJl#CNeoa)~SaGWZEwg^A4R5_Hrfit00jF=Dhi>%w8sC{^-
z&c@o}iKqggs&j#nEoS$j+?&Whi`S{Gwz0|kI&O%PW3&qInfnE-CO)h^^I(Q$`Y%j
zj(vD$)0k~>92~VpXTK3NSyDGApf**I^HU`34Ynaf;UxdKyuw_~`0E|o?uuO1;BFuz
z`~rw}@PC*%3+FB#L;|TYbp;0-kwQ}HIrq8;Uka)QKzNQL6}u+(d#ptcd6k;{ewGfV
zzGDAaDp&9P;!i7^sf4oG3S-kZ+hvb*%EQJT${=vShpkI4Z($BlogOmrxEJnqP8a)O
ztG{yEj9BTxW
zK2^*-IdY87uWYniJ6Eq%j?+3LOyiC2sk8o8Y!yl-6nK&TZhYnh^*?lX*!*^VX@2<
zA+0IBx=-AWyn=>r3y1iHD3Xk!HsF)l!6ZnZaON*Zcz!;zS7wTWb8K8g*?|Y>5;7|N
z3#EZJ1Tv~BGBO`j2?f;*wL8*a02-F|f-t5N1@#}G{R8y2P85*YF)*gjejoCUNY8+=
z=$J_zTi42>LwWJ6-4iwuOW4!7x8i||-4bBgkj!1YG&P&t_>m>X!3WzC%_mD8i
zK076#&PpqFHfkg&m|3ZvveX!$X`V5iFhI5fsw-vHRcIK$LU#EkcTg=yz{qHJvrDu4v)^@QBIfM0g7XA{a@VMx07kGT&eVm
z2PHiN;%1yb_n#SdNA2AME?X!xCwDHF{aAUldNs>`|JYFN`m8s_h>36PK9o4$1r6Hkq(_f+j)d6Ru$rY7&_IsTYk%@GLlTxUF?c8{KsjKbN)
z8M-E7&Mek?1R~86G7UP)-RLy2gQJ}5^3EZ2KDVDtrLA(+clEXMoC~`LuJt2cX%lN&
zJVvs)q&)%q9l~ubcQq40PkSxk(8C>>vd^;d(~iaB$9|^OmFi=#U=8lAZ=YqwWVXbr
zHg9`e(!;d&VSA~WgOu18x9K#6Lbu}LG93d|aDE{i@KhYuNiO#~s}@vDU?4A<5-jpC
z7!_taVa>a5SK|kB>F%rWKI~M8vPOmF48WWQFSseC#-+#y-=R``rzJ_f8yS#OUFkPwNiFP#J!-_YmzFriuD8
zp*xy2>ddetK{Gxbcdxn5KDS7K6Iwc7GpvFQNi|X<#mE&}5dZCGBWtV`k8E0Q^E>M?
zro@s+;(O{AiO3tjJpQ$x1bRj~M_{jY{c}ABKun{)(|NUHacW|CkL9$ZOH0hG%Q?}`
z-CvEx@#ECnU^&~+7V8s%dKdTQMxWd_6G@?WjZ~XqP)+43cr08{c=imRkM5E$6PeHi
zH~9u!lNv_G&dLlup@4{?yk~rvGk`Ft#_ZlKL%~Q4eiAr4%w(b$dd-Z3jPj3PZBm*6
zUpn@vz|Su36UzYq9^4_)d&1jN;dD9pDdutk~vphd~rcUCa{gWE3^|
zXtJ128hu|nQ3?pvtu(}c7dMn{Em$+>-%%?++Nd2aG+kTk%sYQ0VM@%gs=($75D9_3
z8yL@fO4U!aN)^m!QpBK&;Tl|j(;=scHq!8AYt8O>&<%^5@&|7FR`iL#fD%RO*ur-k
zR``^a@MdI~<9k#q>)}Kd48;*`hOhA~phgUvNskc^8BT5FvjV;BI1Kg-uGQLvwRf#_
zj`kPE9SVKaA!Y>ZkBfqX2{1-L;}_OXp%3V1p9h>u%^Sa1u))Bk>dd^ZHEQ*EtEDte
zb}sVVBy5Qc*^Lh9YXtn(RP2A1X88C0hILf%2$j(!Ch%8vQXxuv_&EVH2T@C|wvfGD
z-Rc@~oG^?g93?%TLp+#Q;dU5IbFt3Ee798{#Rq~LH-=t4UD2^E;Yh
zMl$a!GYQ?7f4yC-c3~^ltX@{N1s
z&D2X4+np7LOdKdC0cr^Bk6(;6!a%NrQ*$uuMs-lh$ItEl>6;V;iq&Sv5EPyjs#h7i
zfyR|BS)7yk{JT$?S!uH?^Xb+ym7cM4z7scK<6i-YejE|^Xb^1)l9UPsp3U(mZbBZ^
z@{r1AQW5z8m2!vFN)|5NMr<>Doo*ELpKu5c-|Gy`utAkeKouQ}Sam?JTDGA&+h
zp5+$bGL5BG*GU15DY=3GcbxmwlF?UuWlOnZB2nA-O?J#u8&C$`qziix;nc{R!IX+S
zqWj-ki{R`f|_z}nkG~9IDUn$edM=r3(V%i_|3bng9_}8mpV%byGXiUrw%PO
zVdcMokE^ApUXoVXoQ=h48I0R5y+1^ukA=RBAzq&&)J;V%LdDp7;QzY71mZMyBPSM)
zY?;ugbpmX-aL4u;>i*s@~%JB~;lHdU45T{NZ582|pSGMgKPq_%zzELcHHUQCM
z|5c=@S84c$r`I?8&(4p5f$IHdKYur{B&B+DK!Z3n?S@mnTG;p?ObGFm(Aw&3akdl#
z)aRb+IUcVm9sBGG+~)?d;qH8qNORql3CWcA=X7j*SXC()(erIm=pAkmN`4gX4JKZ
z9)nWHC|?IC&`D-#?dWx|+}=YwUo3F|_r|>t%RrWm^kSlm=-Epzx!vXE(6Abw82G?4
zj$i9E;&)NO?{go(T%9zM^Ke?G3!Fdijl%2y1>p<&P~3*>fBs`@;2i1RvH=_%&{4tS
znUjvVY>?_8!?hKGJ5xyzxCU&k*O5(7xRkHhSCcHrR^q&i5uP(_`%++mXCDe|`3XwX
z;)c{vlKj~%a%^88x)`S2pzQ%x$jqBNVGd&ySQD(!vqJLKbldP1dr?!T&t}JpaMPF_
z!yR4t?^#*6`o>al4!CCD)2Cn*&Dx#UBmo@`pKf!7wdcX^{dlFBhVPFl{K@-pe9k`u
zV8ml#M1CK!*gTS*Sue|G8nF73Y7=~H+(s<)-1~fVaax_^ed#kimFVwRc@r`oZ(Hnt
z5rc9QwT)u)=j+A(RiopLq1WG|`{Rw9TVwH=T4uP0MG;a6m)O;Dne))8%bUuLGA}Hw
zYo;|^oZXOooM>P!-u=8=^h!F=-pQ<18ThG6TsiJ@pRk@AGVtoI?EU?=xgt=XkCbz#a6;a5$
zy3N(TyW4z>C+(|m?DL9gkrG_kwz&;78Y6SDDjv>bfJ*vjZ*wh2L5k%m7TJpPS)oXM
zNehNMkC5}_T#*>H<|}IZj#Af)+0;i9$wWjVU`ePLv|{-U)SDUlT{P(f5<`NKv2lI(tBi
zY3(e^;rIEYK_s|f=ptBhIBPxwhA@5D7Va9{L<9-y)f0SE4>hYk
zx3nk;Uxw?lZ=H#6#LVXo+JbX%hX1$)JiT9#1fDx6$$lSx087$nXxbZ#VWeVdTvT1!
z5sGe5y64}!qVUTJOfX1u(7&g
zz^?06Q5C0b7>EbzW|Z&|8lt?;x{lVwqh115sqgo%%DnYWE%?1$9KnN#8LWYlP8Ro6
zV5X%hpf9zVTiJN!>fnGq$oPe^%-(yh_Z;Dlw2q1|WfT;+ZWWtP)A{9Pl@;zu9D={|
z|%ttrN4IqyUDF7jgkN2PK8?3&CVUw`V1s~N9&64pWp#m^iiEn<@A4kd-WOfB+>Ra
z8fzrNO1YOx(4zUvv{XV8X}Raq`)NwZRqMUT`nI{wU`4gVCk+R(bp-w|kFdK*9-z(8
zX?Fe4;Kw+1Ty~cAmK0IfZq`3G(p@@j~ODKs5iOZFNr|Es(A-SU^EYQ2-2<
z6BdydR^pSF7gUrN6%`QV6Bdyblvk9ZllXrvW|%zcHZFEv_H-h={B-~Qdjd=uYOWeO
zWYM>6lSX!nrKH}j25Qp_r4=T}x*5OD-dn%T6Uz
z6&sYnE=ePz)awJh!f8e=qAE;%Kq?i+!w$Nmk-r_QHsj5
z%L+zMiXiO%yfLLXWqmlvMC(S7_o94&d9yx6V6yPW(0wef+narn8Lp$QATsgh200i+|g<;x82M=!5Wk
zUvAC0+{al1m#<)L-Pd!F`>5N3PR@+rn>pt%jmvt;QLPkOJtWa25`T=^TMjBWz9nxI
zpw~29ot4OPto58iMlfAz{3u#$&LYb3VKnoB6fGyGo?7p-f~)=rx-45~mSe!OOjJ`t
z7#@`#5`XbHR4MdQrW!KTDm}YQIh4$+RR2%#|4-n{(r-kl^!;OxQo>9mqF@u$wXbPA
sncVmu_pND%RxY7U=l{eA)6>h!!^_XZ#uihEUr0oN7n6xeQA-K)Kgy5QQUCw|

delta 6093
zcmaiY)mIb_v^CwKAX3uZ9Ydp}NDUwz0wawKEj36=OAaukbV)Z3&5+UqNH<8QLpk^N
z-TMc84|}b>&)Vm#{jy)qMhf$C3Ug=#W&~CfHYo~I1p47K7I&*83fmqBLcIqVr7Kv(
zO;PM=&69zD(j^B)YraCuhf$T!Q)F_xY(}t;hPH)bVvU6Df3HJb!~an8!RT2T737Hf
zx!P{PX-hl!?62gMFxz%h@akTOA9P-NFdK#|t}&l?srLTtf9PiCZVB^U#5?hxa)%S|
zKb>hjCacqzeP(fujEpycq-J&fmiPI$f^o;=Cr+n}ozL;}feL2`HSmM^4@|k
z&U;m7EGYvgfcs&3JG&Y%~`lGvg=sfpkRCOZ!!FBeB>sFQvw~mpNktS7%2t
zJ7y$$Gn-u;qT7CI4k;sd(&uP(_9}^4B!RC|)%Y%bO|@`_Zd@3)egQxMjQh26F0-6@
z-N2!oy6VHYMSn144~mO76f+MXYO%Gd}Ru(F1kF`fDU
zg;~#6fABvw^Q4O&+i&X9Dlq)-Uhw9kQ2{hxf*ZpF1xi~=9^W-7v-f5cJitZVXWb{0
zJp0ulT*{=BMd7k$c^}3yYK8n(%zz__a5zUMTsn_M>1#
zH%X-xr~Vn_Ins9*yCW1|I?2~BB1X*qaqqW*9HYXQJs3t^gxgfm*^&To*+;U_uG+AN
zmndD;gqBX
zK9z}UFZ>|og3*>Q&bodO`>g5h@tkmzN*i^Xu1!*B&-~E^R-NK+N0sJ6TtNmsOOEfS
zuiQO^NOcl-#FJsaVoyK5y#ZMkE5jQyugKV@*bhs7R#Uz_k<9-ayIR~&GqDALlBvs-
zGiMO}Y$2RmF7>#0<>Qn(P6PvuX}cGXrF=TyJV=g2w9YM6ziQLLH58tepK>{^ET-h0
zFc!E>vE$|@_bl52aNt!&He8Ok;hwBTH|>6d3_)|``5nmLS&ByfPUlhrSuoLMzcKEf
zI|DzzC(9hgH1gM2jsKfE$xK{{oU5OGO(rLjh~q(fTJfq^ay_z(K`6QWfjA{%)S9=(
z0+b}>t%|8ObrIEFOA{dPE&6@GvY5u_eVA%)g3MLLK7F|yZ
z7LFz+rVR8gHjX8Z*qTu<^h*jBWi%Wc=Re^72fS7|HooN%M8;4B_bGy>v{FN5K`WD{
zxDW1`^kWsgp8=!f3hI9wE`zPbMLSIXk_-~uiwk98s^WEi4O&UA@$Iqe~mt+SIHQ~M^mS_uD4tya`HyD8!z;%@`7@8_T#!!eHF_JG+(ebZKW|8r(wzFa(8hD7wtKY
z-x861gchC6d;7BDTkofYHm8eDcVTW9K(A*O(TL9K)mC6WCckT(7?Rm=rwp_yt2Tv>T7z%iE;P-?EE9_6`
zUtVtRE*GXSbO6{KJr`g}z%r=|&z;nP$4>I~@1)*8IGQk>o`v-4<%@l8DYWsRVz?
    vf-HT`o3l4WJ(Vj&^kT^F~58fWA>B4Tj|!+NF^-)4N#AS
zBe*f9TqA8}R$myn!tA5sG{WL)fM|rHwO!(Y8Tt%gK4v#GsbKL|sx&@B7fP5YLG;D1
z@hDsccKC5CK0foBm8Ds!h}io)I}{9?<{>X*Tq`PxCkIqNo8IvZNRlUd11m`pNL^N|
z`z)7PM0J(hVN_nrL5HDLWtx;sGsD`LAxI(VQth{QF&fsC_Y%8JXKYX4fn2kf+y5(D
zzkqw%xiqLPm!q`d`yBY7qhhUW4&t3{<+I%4Ra*#4n$Z32<(Kc>;k!&XxpkSLgM3z4
zdmk~5;WD`BV70o{06LLO&t|~900DZig>n-O5tIlg*V0usTcs^KUTn@5M#dAc2^+SE
zq24>eC5ahtydjP2imDm46tc0CMjBW_-)Ap#CgC
zjO{#W^?fOjC^fKDUxU$#8tlir~NQgXTIA!Nx8kP-3PFq?%&4yz9j~@2H|%0_+z8F7O0iA
znEy7WtC+<{Fb)T1cFQ&Un|>^0sBtjIpPKwwcO!AwV#<{?`!70PoHe~BW=wCWu-~a0<0DJGEpi@LvHgHbK3TLS7qeXMzd#3
zi0oqryWxGUQxq*!6O!FUo#Qf0vZ5Hi^!BIyn{HqR_-)oZn>7=t6-$ej{Kv_aC<}fd
zow_J!o}`V#en8srHlY~w>msq}8#8Ajk$*N@OA+gA{5YHVP4P5w{rdRK_*lt%Q$IQ^
z%8&H%rB*X68HY-P*wEY`maSc*ZdrK8GmJ0NNi_&dUu3leB}2d}s2yw>FDGbY=n#;P
zYpdaB?GNKspjxh3O$k!d1f7(+bTocxoj3l2)KMnjW<6tEa#u7^lu-Wz!y|l-SmS>K
z(|YXXvuXmo;vW0k1auKF
z&65Rjz4Tb{uDJssgPryxGH>fZ4?pq|}3L2O}-^7_mks_9yRs=dvKqn#h
zsOxV!^Q;ECX2tfj9jxE|@kw$JvvjtvxpT3ileZwL9^S9IoP5$2cx4zBnau#beRd7<
zJ$=4Q^73t>Uq4#B*eSfJcry%C<9?#
z&GK7_00_f{ka@kG{fo*&)-Gwm+(q>nx>O1Z>he9jlbdd;^U>t;74Uz!{!=W+hc2q!_;e+16;0nKW&P^12eZz+h%e
z|3=Nk1+-SQeAni{pWY68kR93SuD>e;ZzEcc=5`LHT$Q#`+)-a+2t6`832I
z!%eqK^&+MzEl*p};-qcOUFc?@ryEk@s4GaouZ03@b}Wcz!h9x`@XrwsnQBMUnhQu;D02u7J#Q9#lb!j~
zxuYmM*hB&!tCicqchOFm8h~NkgYr@8PCq1-`V#mL6*z7CNhLTT
z@x(glO!R0ZE3Cz~C!+vgl7V7nMMGUsPqc1s%kY(Ya)`Cmav2iY
zo`5_Vb-0we12PW3`OLmP`sI-~`qBhxf76;Qmt1yvjkt4?PiDL8z9a`^IC!#_U0UQV
zG)`Ze{MxY)Bq%t$tgU*s>&E;Wa=){BdP>CVMX&Pw+2&97kW}sE#V~G5Fy;a|Yr53y
z>#s@y{Q1LWp1P-3LO!^4#*oYApxK^V9XOXd&jxAP_F=?mV&$dN{PGsCrJK0FhaLLeq
zS@=*Ey#7HPY%;ljESE|=uw~OiLpp>g`m~+_20*VaY(m~9UM)ymiiXf1uPLHEP!G!Z
zvJgJ+O3Ys!LApI0kk;qYJz!+BJD>BvVD$XvvN6Na)=6R^{na_}dY&>b
z&Ly%ERv^n{J-(YyKN4h>pAZ{xbuT+1!el9_-%)%wXgw=AKL%N7baF^Pc^`!KMPDEm
zK0&L30_POYzqA-tou~NkEd1&h5)$3*qBo{>Olh|~YGJ+6tzCnlP+%v^wtg8M`U_Ss
zkK?|7&>D!Auk3b-s4ibUPm%KdPPV_a|$
zhd|}B=COZEXmRbdm-tF%PwfTxide<}LMZMOD{O`h*3|6DNlUGSaPk4yaCmwBHKO(9FA^*Ijg!g7gJObphbEcZSDOjnqH;lF?IX&T0
zTjFQS8Yau^VSRhPvAzZ$xpG`Gs|r!O7qP2ka`4|+T6hS0EUxsNRdAj$P;ZSQN?aUn
zLfD)v6PucLr>BijO*l=V;|Q~%_c#j*$k=D0U%#52#yKu?fuh0^zg`N=-G8-~UcTLS
z$0rEg-UV2-rg^;#G;^U34I@>eB9tf
zEP8B2v?Zj8by>0sbz78G%$uKt(Z>+`;$LN3lfgp(D}@qY6l-@BsTS!c_K5We?)nph
z0FH3SH$}lT3_?2lkBIM5GdpfqpE?GAevi^Oo)?DA33uGFb?stZMsS=~}woeMr_xf=$nO8B|SgB82pYa^PSdDo+zVOI}Tm08P1
zM{U_~@_EVqMPnl`W1}%}s`DSKoEpUL?R;oksU1*&p_h@5>O_KS0yv1q(8B8gL7yY>
zDWT{m{5V`O34pjLH0d8cve);&W-pGRtekYY+VO*XO~M{>
z*HdBanhj~3F=Sz!9*(pu^(#rWgd_Z>SHBJL!ZpWh&Lfw{HppWiT&fk
z0+J^_p@zYSs=;}|Kca9ar-OrsP301(Rz`
zw+$L;GM&)WjTh3Szd{=~UeRDs^XCTcVE*q2nL*RarP9H=(?jk
z2}r`93lk0XNGqiaI~_g-LQt%2ID+APFI3e(jCxxC>)azN)K#hwCO!Zj#|lRJj6zIr
zPwLT8Bt#kJkGopgFD*wAow;qC7L-tKamwrwiGH1Q&q268=_r4G;O>?1I~q=s)Uim7
z*X%5%yclGJ!4V35!kcmDs%h19W6*afz3b=q1pf_RGkwUV!`LJ6iU9QSNnjN}5BsY1
zB7;C{;DpF%fQooQoX`xj^6O=U5TuX#|E;4~gPy|)tRsY8OIy|jsn&Pm_cuT9zGmujVK$WNbmdyyF;eQU
zJgMpXzrSYdK;$m+2_`M)E3@Swrb_t))7$oy**1`tTd_uWatPa#+8MQx66<<`Mpg3r
zB8Zfu#Ef(C{Qn8s{}W%Clh2fk*Z;G>W{XrKN_auXBaSq`-S@Lq-88)->Z%d_4+Smj2mk;8

diff --git a/doc/src/DecisionTrees/DecisionTrees-bs.html b/doc/src/DecisionTrees/DecisionTrees-bs.html
new file mode 100644
index 000000000..263295b72
--- /dev/null
+++ b/doc/src/DecisionTrees/DecisionTrees-bs.html
@@ -0,0 +1,311 @@
+
+
+
+
+
+
+
+
+Data Analysis and Machine Learning: From Decision Trees to Forests and all that
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+    
+
+
    
+  
    
+    
+    Data Analysis and Machine Learning: From Decision Trees to Forests and all that
+  
    
+
+  
    
+    
+  
    
+
    
+
     
+
+
    
+
+
     
     
     
     
+
+
+
+
+
+
+
    
+
    Data Analysis and Machine Learning: From Decision Trees to Forests and all that
      
+
+
    
+
+
+
    
+Morten Hjorth-Jensen [1, 2]
+
    
+
+
    
+
+
+
    [1] Department of Physics, University of Oslo
    
+
    [2] Department of Physics and Astronomy and National Superconducting Cyclotron Laboratory, Michigan State University
    
+
    
+
    
+
    Nov 7, 2019
     
+
    
+
    
+
+
+
    Read »
    
+
+
+
     
+
+
    
+
+
    
+
+
+
      
+
+
+
+
+
+
+
+
    
+ © 1999-2019, Morten Hjorth-Jensen. Released under CC Attribution-NonCommercial 4.0 license
+
    
+
+
+
+
+    
+
diff --git a/doc/src/DecisionTrees/DecisionTrees-minted.pdf b/doc/src/DecisionTrees/DecisionTrees-minted.pdf
new file mode 100644
index 0000000000000000000000000000000000000000..6fc29fb0bfc45c85ccd51f1d6c1e80a460be0b40
GIT binary patch
literal 526688
zcma&Nby$^K*f(e+sfd7pfP!>M$0igcq`N~}x;G6XCDJV|odTO~q`SMNyBju}SswA6
z_nmL9nc*)leRyK6`~KCvB$Iw8M8`u+%k^B`1-}`7Zy6kDBRpA4C4CchPEGH
zFmf`ze(}!_3WJc5sja^C3kD%mZCm|!`nnc+`Y61-C^oj%`r2kF4wFBW#cUQUFzfmX
zeud@~6*ha`i-{*Mo=7W$l4mz_>RW%Onp~h4q<*zTb$#g)_~4^#ls$drL-axqw_lq`
z5oA*Dqu|Q|Ckgvfr^Ie&?>pY9ChM7pUADc8$ZD9IJ)3Zz)M
zhM84W5+Ehc>{2I4nh`WtgbLT|$4^dZbQN6AqLOTvc6z?*3Y&&OkB$S+3pL0hvrrU5
z)}3{|XlYwTKgJxc^erx_%H|HQYq_^9Ra{pR8qm(adoKOTJ1dXAb~o+G1E{e~^EOkV
z0u39wDF5xgy*7eX%4B~MCQ5^L|5r$!`VwUFtU^taRC!DUDvnVRh1q*!$o|$bTcwo6
z)YJOT`4sk_Z|8^-qA_z35<9{#LvgI?dtBORuJNl3RZTK0Z4islwz(Czn(PlRy=ndR
zv{0ryM#HZC-D-Z`dQE-BQEl}66S+dsS8BJR_2|jlDt3Kc*ER9>&K5Gv7`|eQpJWjS
z>*?#lTt4MjvV5Ol_kJ0(OEOe5N$q~yw;2nG*FBYgG%u}GeUi~8lAt#3>@R4fq%1#p
ztNkG!`j~Q0{yEd2V_!(Bgz0-A`zzidg
z>@B__DaeQ%+G^%wtr;VOmjra_?N3Yko{0FQlu=?emuHE7zb=V1_{>7SnXRvLQm*~-
zurhAyF2l)7$9uB~qj)ZRZtvP??_I0-Zw`Bg0XrgPODdCFdXr;^)2=-wM(|@BrwYZg
z5|W`jnkGTHSx5e_$i})m8|URJdx*%|jHikuqh|C>1Y5OyVJelPQg-aGM+hflsU9t;
zXCiUjzqo%}`9*#Iojw=y`_)oT&pPfE;Qt_O;}qd2v0=6|6#n3O>97y`F{5wXggh7I
zB^O7Ox^vK@kFEM)U^D3B3tP_7c%421&-5XV>Q9AsRRO*7w-WBGjQZN$-B~H&K+u`u
zC`2fv@RO_UwxSOYdE9+6VhZ5iheQRmb4JI4>L`5&0#j7heK
z=6tpEM3S$X=<3I5C#c1Ep@Q_due#=^ufILh3PLeP8-q4i?Wbv-WziSY(y920q7A!E
zlFquX)I8G`DauNeb9wF@k<#2+7a>!xYtek68(LA0g*GedvQm3m1Zy#zy>A~8DvEc8
za*?nQT`1BbYA_x~XbRRgo4i@0Q}1G-%_mGp&hK9sQPI1L(>t_w$SP|$-fD!W@#ae(
zr6=KelY`Z&*x88)#S4Vq2K{#&l?eCzW(`F8X@&F8*2=Xw^8w@$54)l@RjnWTO>Erl-fsjFj$!X;@J#6vfYP2L4)
zJ;hQ%_Mg@5v->R8C@RbWbEVLfu$A*}e5IelOjm+DR)f+ShV`tqiQ}HUwly9lR%QO`
zdij?-Xm&#Z6kZ!HXort$&AF4JpWoW6fNq?zT_Lxzp}<@5_7p#6iK1vIikW&v?6~6Z
zDrC}Sx`oR!rg+ou^^gXmJIL6YpJ~Xs%e82tdvAuSwRk|C#^?4ZF1iL=dBpTXTNGYV
zsUfF&`tv==tW%U1HR0vYhwro#GOXq!KN!s4iBY%T^A1-&XKP*CJH36s;JaVNpuWzc
zkiGsUq~8y^F6&oBU1m6A_T3NGuyK?Cy?EO@=>UsY}{NfK^{CZ8ddU3Uls=19f
zl1uH^iM9!t;{|DQBm!ntkCuZQxxA`2CqK1w32m5|36(RJ3};k+=|FlkH6;^J{xP3F
zi7oijzG*^eTbvv=(q&PJF~;jkT5yvn?4ZFXnaK5Eos@5T@+axMQFGx)?}
z;1_3mQd4m!L<2WDC^<_jqww|H-LvDT5}U4-vajbT@Lax<)~9Dt{M1OHd&u*Me5Su-
z;5MwfFqhQv8is*%p6D*Kd1mgYNU1GIl&?E8l0|)LJ>$imz#vq@uMi>Rj-2?(;}?Du
z57PDA0~y{TzgwaNiZKQfJknvE;ulY=qCI$@z1V-Myf<7C--|HVomJp2ZftGysn@N$
zq^-X50B50VsGRW|vlr8QI*g~zDLy_-^@}582dJ~@CQIvMw0-zas2(9TOAu7kYH8O*
zHR7;uca(-nYWHa0_)q@~P8ty_jqq!c;|q^_ZYXWM|Cri}Ph~BGAgFpU@9b3cmsQvn
zGv;;qS!@UQYjcd&;uKW`MW}+p_WF;_B!`t_cf9_%YyF2gis{!MJt*C%K6`%RmyC0`
zSjP$`w#Vla*UtCM4?mlCl+tV5~O8f
zmeM5Ch3i<*=Z%|J?=hc}N_A^Hqgu&h(dn=yu;Cn!)xWQTz2ep4Q%^igJyO5>#Pcy$?V;;>B
z(>lHL5bn?@tTNTdGbT^EFs}|5R8FFi!&>JaKk;}(c-Bn}&HDLhEjNlX(wt^!sA@Dd
zW8{(}(Bv1-BhTCBQ4X(py7tFDV+ekXT|wq_)xU2b7}g>3X`(lvVN(-UR>PWfyg&R5
z$7vP8EaBV8@cru>r-la&f*~H(E8M%rF-;0Z>YC@y`h4S7r?~^JVutnW4)>}3In)xb
zJtA`o{KRB#l{@IZxTZW6CpNQxPu(;Yp4l$Y%cHX$%M$di5tH|$9{k<3CDUyis<2L9
zDQ|&tPf_DfZcV5mnkbzN83FEvKWeoLp2}!C5KWD)Fu!m8WMezbeka6Bx6L>nACfPB
zSt2nceOd;2M4RT@&3_w-`CZ=b9d;;xM(AUs!f?k3@~^f|`l>@%s%i6L%}A+jHjgUz
z$G>Sj&n~(3Rxs^rSnqBvKj9@a(pf)`QRDq&e%H^R2j1xka1~)LpoJ#zCscga4X1T4
z<+Gj~A)K`|BKEgqld3pBN`G?rAwBdn+7$8&tTl3X{bHwTGG9H
zZt>+AOHOu8kuzcf;71vTKAMzxgs&jmvBgJaui-*&!hg?<}s@5vO
z{O-b=4}qYPYYp;Nnx1VB3LD)_Rw2xBHMUb;Xzq~cys6QL3Y0uHDnMFE6=*l3%Ttd_>g4#8AxSmNcW1Q&0eOn~5Nj*jy
zZSMOq@>6ebm5))z-#V`_weQ-XeE4K43h$40)*@HOa!48Lzt%vxr7&d3*2sa=q+_e*
z?3htD!jaJAZ;np?L;%eb2OA?`!Kc=B*ZKHPDsQ61wLWtFrVrN{B|AFn@O;~S^80s$
zinz7NBb9X$;+7Q!NpwG!q`Z)r^Dqd}d5HC^O#Hi+9!@oEtmUy02N9|7De++cW4+-L
zJG50&oZC{OS7Eywu4$ip9z!h3s_))^Zd*X}CGSdfaXR%5E*YBbr%}mMWYz+W6v08&
zKu5u*XVS+fER*N*?_csaKX*aOj-IjmO2KqD6?J`Z=yjFgc}{S9+W=)(>HEEm(7m!`
z_MWm&jUzNJ?{UJ*VrL;3Z;al3ghzDV3OGa;qI)D;NE7N;{N0^?J9
zR;5;e|4{wTJzkrbk$kny%^zb8`&oSAE^@{3thI5EYtU2@g{+Vt(sX4}qU5jN@AJP}
zv8ag)4*Ng3js1U6o4li?{tE_WDIH^dU0W0eIXfNOfBxpZnYN)m3WJi7-fttz%EW@g
zAfj(%_|f(S8w(Q(1HX~2jkLbCfQ6Z*g}J`D?F$YR1_29G3u`$`ZC!A~cluBxU42<$
zesD_(Z5xvp%)j|m_CIUP!p_9R@|#Tmzpe4s43fOO@D$Qsdefk@Myzd!=2NQAqNN@@
z)Gwe^Bs#*6P2~k)-(_?gX4Pn6Ewj!RFtK+n^aI~coo&M|&7ima=-wdenSGh@b{@;Fn)^8mwccJ3acXxJv4GtzgjxH%-nMb>lkdi9cv(WdQ_G=`2
zzLQQZS|_XReRVE6meWgsOH3bXm1`M4p*2tN%6c)2{~B?%UWS8c6soO9ucYB*L{RV^^JI;D2;y3*jq164qID%b2mJ&>+dg|`s;^XPz;)F55rRZ{QOEIN4$`+H0xaaqJ4F`
z+m(|XRybQI=H=-YVsYC-#EzG6ygJi$s;u=+!(^$+{yEa0YqLg*ii_{xX~^H6dc1#D
zkj_&L#U~&LCWPS3-MM*jRpaRjD{i9QhO6p_ws7{LCaqmHHMO?3w#Di{SKN}`Tv}dM
zQBmP>l>hp;&7G5n#<_WMzo@8aFYxAyEWt9<37wB0q;n0orYq08{8EJ<$DhOHRjvEl
z+dnS02A*l&Tse$lFcxJ0|SXTCx;3fqnTKRh1H$$vXSoLlA4<5a~L-tA8T>4HPO)U
zc8=*Y%AYseu2sZ
zUEfC?U0wG6ezWjJp9(kE^t3dn>6GX)!RU`|1qIKbf7i?be^iV*qE4!i3tp1wxgcS
z=mp##B)w%OE`;U`j$vaD)`$PW#`e(h@6m+fo*FR-caVm`1nT?6F&N@
z2D#?aiNI@A3r|B!FRfBJ?5=-mPybS@Kvw1eS7wm
znwnaEwoHN`$HKw_yxHzd-TURr2^IO-6vspO^m7^-8XlfnieMmont8$n@u8=grQM{Q!+R`vrF_k#xy!cocVoaH1PH>3_ey)q7&HL#Ue
z;*wcP!4?wC*Qn!gkp3DKg)<^f5F;w#m)S`|jZVP!IB46!nSE;*cGrGwaj1QCG)gk_
zWdf7q;J0qHDq&D;{pe4OGC%Ow<>k?yKF6sg-RO=9r^xEFwzeLw4f*wJ2R0K#%&D#-
zwWdyJwhgi0`31Rw3+ZsP$&zf9;tXs=)wT8sEm5$N_326Hve@$SGN^yWAIc|rTbTMd
z473^6ooGZ6jnu{8g&VwFqKvTcI_{t*zkJ!++6r1B6@HRbb|fX9wGcy#jg9^2kv+~u
zh1+tNmsj)nK+G~jQf^L8Q4IgI_e%GL<`O>%EWsvIO4eDJv0Vy3_LUg!>
zif3b=-Vd$nSd?g5)%J~mj_&U^#5|Okm06ZfY6#<=9PF134Z~)1bU&7Ezc(cbkB+90
z>TVy#1oZ~CnS#98#&dlS{ZAZT@Ufl|PGwCH0eZ?;;N
z#56cWscU97lj3N@$;nyeSzTt4OIw45gHvTb!{s7;)2qfs6UBzb?>QFMcYY9X6Im=D
zq~8s@V1M+;qW>pNw);Xb!RUJt5pMV~qSuUXAlJ5iWNdhNYbul7->qLI&k~@bZB}Y3
z))bvVOA#4Z1|`C&_jupB31j2pTI8h`%$)Y;V3m=mwS7w=u3gx09qoNlwo7RjlQ3)lXf7qm1EM^7d|*8(tDd5i-Frw8JTVWzesFM*Htn$det1)povm$-vWioq7xG!qUtHD$DDCQWnzJ?L
z+WAut0#T!)$`-ghHJgE_rlqC*_(A8KxONuW1j+zlXx_>)>gw8SK6c?(7onWg31-_z
zI}(m3XI4i~6tYOln(oKuYgWCn0~kGBZzIMLA%i$L?|
zl4cvm_VUuw&P*NEldHAK^y2XF#Xd+GAB}UwW8%z=j1zdrBG4kT366Z-j>qt95mR4!
znA7rsTwAcWkC%w3Xpv#`9ej4S*RPdp3eUE(u@bnT-w%c;g50f5Gm9Rmun4HBT`p8I
zvlnPZMn_wWmaGp5AdMAhn35xXYKr3vI2k)D|)aP;?J26AW?e6RA!F-Q&vXd=o
zY6w7KxcBGMZ6*3d;*jB_`x99;7D=-%y!Do{4f~|NkP?fwL|olq9sRUS86R^T^ip46
z2>P!a?5?LzpNao8e3JOZY+qYYV
zGNcx)tgJgRsia7ce(yz#2S7U|TD3-rU-OD#1sK0a1yZEj9$9LKP8lb>ww#7`?R
zT66-qgolUM$)HgqqM{Nr0xK;1AeH2&Uh);a9~YZp@9=PYq_}OBJG_No;mMOHru%yN
zN$F&#pJQ2dT>+QV#dmiCS{(uv*KtCq8{7!vnckhP!o+{lsAHRE7`)RccK2o3-9N=3
zy(xbsJ}Nw9QJ@*m4Lus47ciyBLm8@OUj~vCXTm&Avc`Qis@+@;a3Y{Uh9sWk2ZD+}
zl4Do<;!~iB+|*Y~vx=DwMZS+JcrrBsg;!YY0@c`T-;!gHg2P*%JDh&k`SeS-Rf_H6
zBsM;!xeVd{(aLcbSjK)1|A-+qUY8GitZK)0KtrLRqV5ykFn(4?hD6+GwNf9$6S{MF
zvkrtcjag*h>WW+<)@%3SpLc&F!RC5lJFii8J)>uC-b>dTa8y8%myUMxWrrhKj*x2>
zfN~3?3}e*J$0LWCzK^yF7DCBhb^lcsS=%|9VwFA=G&H$hXBC~AJa+JB!EIBuHfL4=cYmFHsEkaj1evXXjccgfqS8~epR(>Ff?2F|TP*N)YI4MsL
z3(S3OapMtg@d^j3T4a8msdzyCsIa62?c=lq;06BAo>GK-Ld!B5)rH57ljSCA9N91a
zQDyxDFhrOjYWCjbvOmJjEmP|GwHFQ8XYykyz4IA}{hmXOz8H?5QmR;1PR;|Q`&O`o
zZkcsaQBn0bMl$=Ky-`=>4BCIaomMGYg1`iHVACoQJnX@`NAnNK`sJ$+tWmMdz{_`KrIQhe%Mf{l>pcpvunkfhHh-0Res?G~$zit2V}e
zUneIuD=W*VpxoxVmkxlsj1lRC_C@RJW}boKd$ODRx-2{{p#@7<
zSC>@X5P*$*h2fZF{=4_@-@8EN04-GQbQwwkmYSVK2O4SJ78loWgAJF|+OjayCH&_b
z1_r54WmVNtxL~4@DeC{KWMf1PA&$|+k?j}^&Oi8Kx#m2_A*ANvI*4tx#Dy*q}6
z6rI6@36(k%PZSL&BlvUVII?dY{zb4~%p>q)=xDxX$Kw9@9Y_Ry0-u_;DTU*kC%J%N
z(y4VIBkCzJlJU%&&d{?mc;uNA4tmHnKQjMyAf_Hzur>O-hXfp+){R>AK^+kS)TeptKOkC=RVT?n3$LZ
zUaJ~d{Wb6jm6Za+zs_K~|5f{j%RjU^Iy$EqM7+6)aM;`){pM;f-$
z8$)M3&^u=r79O731#qCTr6q=IuMCbI*#`9oJc^aA{`zH-Gr2#P5bib(mg~YArm+fe
z5Nja~eVs)6=Pm*QO{j)aa`&~gh(SHm;P#RJ#m!quMu0?FUFh^hDT0Wspit1Sot>SG
z12Peuh4zmkf%PsP$I!1B?`2S8kn@%x7%LtR+6!wRk&+^cED{X%OHO{tZY2ETv6r`x
zBIZ;w=n2msDdHrgB;&Cio1=L-T5}%QgVoCNHQ!fu!v%HydHVmtNd@RIL*Ob-cc+~c
zPu6DQ4wCFmSN9#Hi22|TqU)U28`rVg_~x4dCSz>P*nqAIJur!8NaV3w07B2$?E})k
zCGvRcTo!EjP+y<@X)mEOblVmRwYIa%#)gGXCO-a)k#9^_{nLOR0uHbp{B(7(pPbg)
z9OB;7{P&VuvhwoshhtOP>&wLy7Mk_&ZlT@;UdPR7lP^;r9@*XqkE&$F?y2+7z)Zrl
z=6L6}tku}BDM?A2w%y#;;S+m>uL%hWfx_<(3GrX;wY;pc-|E+tlN$mUOJM@fU6uN;
zufdDk>q5W^$jZqr?mKLb*HpEZdhxiTs)|;n
zj`&Qe!;oZMH{~g)_9nDi6;O17>(3t7zu^VbVp7x5$%GW?cN7qn~zoZ~9M5Y50
z=dQU&(M
zdjYg(sgRPULd~Y8E1=%kn%wpoQoEYK4%(6=Aj;q)nY9T(vjT6=e>JZI0i>G&hc)>s
zduMxxukQiUh)Ep;-ZwFlM`zjpD=w}2YhoDTpXI!jB?wDsORuM8Vq
zuOV)XVRIzcP+y;wk22cb?!sx0;C;h~+CRYoUB
z0z}`x?@3@*kiZFt~`RqH8aX?>C(^cXQ=S|Gty;LFWyu3UI`~v`7
zB~4IMb9Q<EpMT-N+a@n4Iy$qs7z=lb%KcnjL7`Ma?^QV(1}g_B
z2GAT45fQ(WhEAK};;*ZHSUVK{5B&;ZFic0tCG?t>hNk*-Rf9cW1MEU$HsDrH&dQ&k
zZgo04v!|868Bte{2ez29*tmu1ztX{Jfda`hG&U9;8jAb~H4Yf_+9`GwN288fpeKR(
zc5!ue;7&|RV!)l6s5%!`Y-?}-;@th!cye;GD8SDTc?>%K8FZ$7b@
ztgNiSgGVVSV0{P0dV3|Ce9?J;wm4%`J6uJ)2m^>trM^1g`2NBql@C5?TuZ^pTRqHtK!?A#{S1boF(pazxn2P
zdnk_n=|I&n2=7F9W))>*jJ9^KkM;%TFMjinkS5?C3{TI`-QC^);@cm${nE931p$ZKDvan$Qbs}G|*rT1f8JHPOBuFlKt&Sd+l9Q4$+qQrBU+Xy=
zRgdxrRaXq_@Ei^v2uDn90-OXELf<fZy77C-
zftzF21iF47x`cx}x4
zJ`bxdpSWlumH=qv=;T!6Ea!I7ybnN5QF#>v!AX2J#@C56Bxby>@I4T(?aWr~)6F?g
zG$8>?c;1TpgNv-W4I2k%;lG$`B)&AF-c$#cP<2}OpLjr^!V=ODejro&()Z#or+*sn
zEf)t_4KZ)wwaP?lfon1l!KM=>RzMAsoZ}~cbjiw%0C`%#_)YA5OFW6Ub|tS|U>#(P
z0KP!FRCnSnE-qeDQi9L2_oSoVZJ(C$Mz0!mdLZHU&f&n*NxN2!vVYnBF#C6f
zbRZGHfFQj(Ud5oAOLm2H0-eZ`V}dInAb^Xzits}=b*2T~jOeBUA#QOdV*0a-bwe{C
zph7rE0^;sH*jQgbzdZk;SqblIg=XOh+3~L=ppY+
z|N5tI>qD8R+e4SRJ`Z)OTlUUfaH<@^oq%8fH2J%$l?oCgBS)IyH3N95Y3b>xU6DLixhZV*KL4y7v{b^_w#;{yXFBuCFc3{3Xw)#wiuGV2z=@-1L~2N|Z>=!*$=
ze-Lp3Qu&JzUZzO;w=uvYqls+Hr61ZC8}y!Cm5<3cjE{e7$sP~d;Ns!}ph8R-0-{Y!
zzwLekqyC?{AdmYf(*9+>sgsEJ@kH*-&`{G{Uog!%8;o$v<7Vxp4_!SSAye*?jf?w$
zGv1#AGbe)p8O)!o00NA6u7D#>_1Q;L3a}%*NqsnKZl@X36BFs#cWMs_D%w}e0oxcm^HxuV)XFFcD+I0>Qq*jHO6pc{r8Se$t
z5g!aZT2fM7kRX+%;&W4abA>-9J_S}D2*_&ay0|aUa(qr}-@$kyE$!8DpeRU#a*Fq>
z983+j*1EB~q6*<9RNhC%^J#Z>6hKiH_Dj#>{D&OhEQ+J7L4YqI(F0Ho*k#G~uJkZ~
zP{uC(Fo8?P7`#AyN09dV`332~2=Cs#>+$A|ulh#Px4rvgQ^}HS$Y>FfkrKo)or$EW
z5z*li37i1BA#kHEa<1tLH@f8l&8pYt-D@BkTrbpadeBINs=!x98V>|uSXkKC&9P|0
zorzL8hiu~iR#AKILoJ~wpjhBR{*1MD=PCus?qeNfdv&VZ2Pzb#VN}Itn+A#vO-%_c
zP{7S#$Qf*M?mJPOfg_1iH@ZzQ7D;1`=+53~hQ|8($}NYt1)(Tu!c39v3_#2_U8ZR3
zaMa%bpN_ZE6|zBP=<4WjlYK2NBamu4Xc$QUF;-4=;L`LECMIBQEvk~nBwTbf%`3ot4tshG`n#U
z{shb{x1b;xkbI>7>O?LST88;U&A;T6BjiRv{RDu2Wl~VyG`JQZ0|0;i^ZC{uFja84
z%LH}X|0GoR^F~hUiBeaB5r9+=b3a&EjDdjzz-($4HAIe5AJH^Q$M2hW8=c1+7LB(m
zlVm9!4T3jBLqk*ci;2Ng^IZj#TtHAjH_1~zeyChdwLM-IDeCr(C->`kZw>K^CE1Nn
zWX;dZtI2Y>O_O<&{R%jx`X^KGzXUhc-1vk~_&aeuQ`xUIYHjCROKE7ez?y?O9tn2V
z5?Iv*irIP_6Ck{&E43_Z6Q~D)eE@k86x2fba)DQunAi?XI#}dPzb`-GI!o+G63F}g
zsa#y*AAbKWi%Z-w)c6AvBcsts5m}rnY5X4f|BNlk(~qus~3-m$SvFb^d;X9xpw#pyHyZen{0^z8<9`h1g*&c9p&CatdEL3JWk
zF7wk-5(HF{JTB6Q&dz5)-A1PXym}v&>X-Z;LM4+sH^>Yvdab-n_(dm?ZWpx?5ysWcXVqs@aa!BPn!C1A#rb`YPp
z|2y9m+j%E!$I8T%!rxiwI$OUvETg5B)V5gbwlB^R{I5cAX&9HJi!BdldOa~TSP$G=J2_cBL510HZDRumjEvJk1U?H!?!W0&1_F$l-l6UM?=U#$@w3N;w)7
z>By_tTA>xgh1z%L$0grwj-ul`48)rDq7J2@AKLU@1s;zkfb>^sy%c10ImK+3P)`GV2(=*hSw5K$E3
z-#-11RM@aLel08Cj;d_&*N48v{n&dCK#V2ZIyiU$*kU>tBrw!8G;)xUngxI(5a>s#
zZSf>7Z;aYnf==K}lKuUQBO;#agBhrh8OU_XOr}5VZL&i*K(!=)zZEL7{pg79aJ_s(4i698+S#o$EE~G>6@%wBHu8`CQ1kuoD-~sBQ}Xjc
zgrJb%xBV~^%=o~pYT@E|S#c8c;l1st(jUF02i8T7+?pUW1vOfty*(DE28gfWq4c2TX6M#7Hj1Lbw69Ph;=SIxdumyZ
zAVC7@a$bL5AK2>3s8%RAO=4_dtg5LQ^f#D(^P^C|3rFCtC(`NeXzRG7_d0-LL#c$b
z?GgXrQT1YDK^=ZL--gg1XH)6C=_A6au@8=Vc!?S7nr4%M-MM2i3nth8a~qRzLG1L6z~ehw{N-_!Ym
zQ%COEESc0jjlZvv{_i;@;})lS5v|wT6Qx7r;|^Qnurr*Si-wIi#|qe48NpEk7>MxS
zaU)RILBhQ~S;mO-a5P`TzmME;NxbnHymfW+2bt?_Ub~
zQG>$@-=`*F0!Ua`STtf?epnQBF6U2@T>d-~oc|h37xTxV1C1{#F5Z3C1wPM<`G}pF
zS&n1!NeQzgAUM`QtLQetew(u@4FiY~i^1_c!R}CS&!{d7Is(~hRs
z02v5$Yuw}VScb>xXl*!K4jbEiu8z*FI=isw;&gY8&+*{ZgPVQ*Zh+70!cjnyn3%|#
z?A#SaH#9Uv@cP3oXn%gjSb-LVi%SyN}gRieJmic01FYC$iew+LtHW
zDu{8Iu0$N$$k>?0bj3aZ@H!F0mF}3-^z_KYM8uRu{rsrQ=jEFniO;3-3fiXHTfQ*8YqV^M|a$M5Uk#ku?T^-=ITDiIF>CQ~S=}mLp`s6OkUADNj
zMf=YYp{K;e?n~{MPRISp0$b^TY=AKARIjpq50cH%Q8spVaQqUafCtR%uqvxXp`i>3
z4Gj&c1TOEm1I{N{SZbxlWSWWb@w9Ywt}+j2oT&L6_Fl}?MbWACCh}AZ_+o&g9PAde
z@FZzd!_EjQ7hd;;W}qo!!2ZwA&ucB-D1lqIk}?{{gT8+4Tx<)b(o!*{;Nk*7MJUhD
z<2GXVii!#^A0IbWBv2FVjt6?^=z(Ab=iILe?U$H;qvXttD>y|#vb?aly|%e2cb=n=
zOF~MzPA-e+2&VvZX}L_vEAR#sw6xt|#9dPpe*l4QfyC>gi1?;-Z=F5I1`>fVn%W<@
zMGkNm3U~`RS2R5h$I5rxUl0W6uZW40x}q5tHC?FK&8PbZ1}5s=>n*jdt&0Kzy4hZW
z9n@tY=#51#QVTz@`tda^xbwb*L(@uPSGSEK~K2gAp!i{m>Uzvi2K)!*xH
zZs#e(`1$!8dwcVxl6aX7xM?_8&!EC*Pfv@PDsQ#DNHjt&`7_S!
zP94xGytXtGLzbYPu#(D|K_3E`KbxM;RRg?{@SA|$JRV|C|LT>OInB&Ru4w=5cY=Zg!aCrU`jYsh#Ki&iy6-Qv
zP`Lly*A`^@G-6(Le*VuJV4F)!Ck7`cV_VRjPIr2umeQ3#d*>FL
zoFlF-*N>pvlVIHOVX5Au+I|;gKQI;cyGR&$ugh%qe!T^OcU3Uxi2X(nHT;H?Zrzf0zB(F`
z%E-tV7#J|svO8-P3dCRD-M!wKfuP;L`F=R5zP>&T1_Lz}$8OR35v~T_bu8_D`0%j&
zzZc2R0k{EvZ((5}k50B9HTHeAZ@go8zOS
zo#{%?@$@8EMqJ5enSY;U4kn{-a9&cRYgettff
zo?l-U3S5M|6aYsuy|r(C=S+$d@R+Z!@S8Wc3_cADfj!v07W8YXU`<^NqjxK#df*dDEFHJ4Z)HjVjxDp#TD&
zdCV1`@{IynZ>^9{UhLbqZ`;n**4EPG%FP=})fC2R{WMNgPB31uqjs)W2ngHz`4z2b
z?-z3dmtD$K3BaAUVB$p^)PFYJ?v114c|;?>8Y=T$x>7~WsQsR;s!J6#jAG)a27G4%
zqR6ELNAn^2YYb$H^l&8(=BxuWrOZYpL27voWx@0BqMf-8*k7Q{8`o>-iab7q
zwc=H5OvKN0)OrYH=6^hY1)ZA^{wynXeM%fQsu?B3wZVpC5KbWh)ZU0YT&x6?f8j3!
zYSFK6(G#9bg8zEIz?)*Sw@2U5RCN!1`}sS(dG0HH$N4Ao^$Wgv
zM+H^A!l*4$c?Kp)(+J_X98+a4q~aoQ^
zhKZ2-=iT1~9s?MhQjNtuJvle1KXqT5=c!BRa(`6;vr&Ew#)nGhjl;f!{8+IQ>g~qo
z=v>7W1s2TCu7uc*d?pEkuGF=Q_NbG}8`OPUqn~IdxU#bS{d_Dh$oM}0h@bA~S4&{n
zG|l6!@fOg>EoEVLJ6w?4^i`>P>npd)*CpOC{z{B3^h`d*!T&ap&QSypMv8HE-&IN1Xaz5z~3l1v-
zi?Ih|MnugA<vNX4y9kY
z=w8vLy&_X2`$`|{N=wRJA0AgtFexO3Aurl78rF
z>Ny;BOqKLpM{Pd+?sJLN?ie|x`nUCni`>|+2c<1|b4*9c8EGLngw(PX2Br*ERwM%f
z8)@G{9t5pCV5KN5wzidWB1K5z*!bE>1fnFvzY)yX2BFNr(SZ2t!>04LUMTcLI$K0#sXR
z)M?Z7O-IZfgV^(gY0B%NUm3&SgT8XJN=82>#&a55oT@CbWlRyPjxjbB!O&aA#%P~E
zdTnetR`?y|vc^tY4xjSE&d4yUik|A%s8_po58?BG;Vk~F)P`)v9zJ~71WO^ZMlKgD
zAxJZWrEEI6tcbhzekhlXlLG{KIGY*#7XmK>HkLDz$n0GCK!S6{{5yi2Mpar1(VByO
zV_iL!`EH)pHr6_guFZGvsMM-CLca0M?zuJPP1j+XkG9g97g}t*=_I~%6Irj?LYNW6
ztr=%gQL$JR^KvitW%0L>n=d+*ZV~O6(eK_pW!vBq&
z-N)4zJJ!|YWGc^$hX}1+56wjooYv`C%jz16Y1ru}T^!>C1tf>1r|PT^G9$URX?#(Y
z^v4*Hmx%}uHP8-OK_{%>{3
zf_b48rdx(n>~wO0E|1M!x=zZ@t`n2(Hk5Uq!MFrSP41o_OQXEVm7cG2CVB80KKxff
z!W!z6N&L+$HDue0ag18j9CzoIZ^6)PXVf_>-L_xXH{JOrn3zv-g=YH4+NpzA+)s5i
zKc>~z(j`|*H~8hMkGcd4ms8Aad+@N5Tvlc(cXIe;h^4cNQ3u;j>(*)AtE;VuU*~F@
zJH&W{{ruH+q4*=~3ufeGR5If&x(X@vaoIkJIgJ%%DZ%s#G?pO{h9Fdd*viY)NKZHU(LkGP~|Wo=GpK6R-Td$GH}4zPR-vE#n2L
zoq?%8G17w~6s&q1i)HGNs0fZIOt;MMfm~W;rMaK{9}f6UJz!7g?RqaZ=slPNpD*;`kXTS3w;J
z%hR2V*aZAlCbE|DxUqmM^XX1m+aOYJmNJC&f$|eREHoq++2!o
zFHauJVdSB4>Gyk}*B7COlkgqxRUy(DygtD1&2atERkFO$hiOnYTju3`WV0XlxKwjQ
z*`oq8P>y@hvx-t`O#5&Gv);Xx%n)OK#5Fe65Wg{}&KtYY)ggl9IF6!q3WtZnw9(g_
zmoPji3?D(bY#xpz8!%V1Tg+=D)rE@<>aM5sMhO{4-KlhUAEdiWI^Oa5Jowa=bCv*KR&nz^SakHCQ7*9Gzjo&@A2THfG%UQbsukpPak_zPx_0S>`>#c3BK
z`G_co0E#NaVMX~&tN{0^LQ8{U7c>qdd3APN?${h2JhK8dct=L89C>)n%QgS4LjH)P
zqIs_fD)M&W3#aFfv{L3n`WDe=vUuHbQ9r-_oSKmM-T~cg8pcc8XLYIxdU-ad^)9vC
zxy;%<1#Z9iBJSU$JPfd_2fh^TjW=)I=V!MPoRH#F)tX#$R~SiNyE
z$zx3MDPN~mM6YJ4nWkrbj`$VGC;CAay}1zMgI};rnkU@sR!|pI
zQ;J2^S5(AbToK?1AM;SrC+z9Km@%=_l4(B6V0|D=;|J=39R`nUOr!
zJ$$V`-nDU;3AQWaa=o3aQXPFCxwCbDj|duX`FZ3j1Jhwech~V;Ac*e@uynHsDb%~Y
zPriHa-#1fK8MC$t&YSsLu|Zj2jjj$!H)zz<$2LkVN4_t!$SEAIMWT!QV&h_z-+!@K
zIgnZ-NruwnO|+E1PSTb($NNrGu_&&5piEbtQAq4;U`7egBggSA(81wybAM^x(rrjI
zrDuD*J&BR(gUOm|wNz+E9lwGz9-GA{9!`G3>T!Y6oFcF_3>-L7XSFv>i|ob4mI&q
z{^~k67k0{>Q&E*Y_a9^?HwKF48Z%R+rSPLMBG1L4t)-t--)Xw}7k(;MIDcGQ
zZf-_HYy7NMr0>2%-Pbpla=NklMJeb`G5!w8Vn3CbB`)K0&CGS6P{??sk7Zr0sDD|~
zL{MnndTR^gVc*NGbj_fEbULks4sm75n>tM?^S�H8ZT{%|#{t++YT7;%T;4Y2EDG
zyA48(9#^gR4=ip(Jjr#GgK^UY@H{XHFfaE#vFOw`X}+1#1b0ENSj*Otq$@EXZt==&
z#H3{eM?7HX$#f%`l~LURZonn
zxSDNe9_EY2*7r_?>>=1J@!jfO?jp<6;kE%#koaUfI)@}ODmL?4=F+0Bh}N<~kZcDC
zm8I=__=5T`yIUpKvRqW~Lj~L2Ag{F8m&j=Rd5UR1guQ$_dB{3X7=IOpmbv9F?_PXJ#nfJ$T$li_Do@^qZ8@?
z4P8G`<-jY*seyugA;`PB@=d)V!Sp$A8AT)U;dXmxTLR?y5L{gZQT*7Cs=Owi>!qRA
zr-0gAKRxv1<<-+gk07YZwe1=<)(pbEzXvA()*`$*#4{-3~X5*08?c5C!>AQc~^+q8!$gp?*crv#n+Tv-ACUHMKZfkw
zdiDo8Wwnei1)Ld2n=MLDBpuR^gw;{fKKin8=ajVMTMz
zJzo-3Rh9hm_RB)zsG(Y7N~WJ*O!CVQqf}+_h_{-Y?8Ox9B`de#z5TXJ(}NN6dUn!9
zY8cCI+CqZxR#>BXZ>lRi#F)+XEHzA^U#CU}M0AY$0!>9Gb?~x@q~NpKLZ0XJE|UX~
z)b^quai=4j6HsxLCnAz{pr>g$*Aqef7Sg`=UFJYQL#?^0AT=V3t>z`75|-y1`i79K?nS&-Pd|p*CYjA-xDpb`K7OE(r$TcQc$*2cJ
z2I5v~URL*=Nq7lby?4yL7`g>h7j05CVTy(GiLc
zV_|@OEy&K)waK@k7+xa9biKXPth%@;OgrGc=hpbt4#moHGC($$=Tu}7Q&tc5cjUo^
znVH%cXW!rVp~_0`f4SGIN@-(;`;A%Z)#Vi;JO`J{2smjeDu`@e(URyDtcaee-TV^`FGwH*Cw{6u(%#o~r(oh;*D
zNAPpoOI2(}Q-hj$)Xc3?MZ_;k?8;6z^-%4^jWzws<8cuv+s|H<>XSfda_~@BQ(H*q
zTgs=Aw89u^g`Ev`%)9f2?%{q_ntsWPFu+*?nRNbP>}hPEA51~J7BvDx3!L3ZFsA|0
z1Ta1$a!X^n3x6UPXq0q+`!Cx*I_Y95C`!)r`7l
ztB2sA5Wf!!XW|VHsx8j7-gL*MKC8P@C&iKPVIbVge)?>N2^OGqn!C0Tj|c0z#m@6NbQ&k05C^k!y}*?A
z{+?OQVD=qtuXqI~z=gcX$odd{Pz3dgi3e+1Z6@n|_SdsnT@hATJUE%gwM)reC`=;=rT5`9SAMs;9O?
zNj-gHS_5VSdz%;&-H95@to=Q5h9h`bFgrHoxR_W&ixop>+ai5FRh6mF;MgwDLLy4f
z{375MKw-vuUrLIS#LzY)QVPwni4)?dur
zjKyfwI#T5P3V8T1aICG}d2W3a6fb=%x1gF~uwV8sj9re&OL|NuSiE{;#XS#@92CK-
z+R9d;+Iw7dtj>;J7`oido40pHGEi&exuwJ)prpsiT-{Nh`4M;}W^+CrW7PcY5dfY}
z;+aX?2%lWi*~L^0cixNz;YB^ptC`?!N@#2Sb=c+Y&3@lQt(ksa^}*#0EFeS@u#%QC
zW&#=Vzv@OptxWpA%2OFYl(?dPjVRVLd^i|iO@V)Qqi$rmy1J`}SAJcd^|8(olnd7d
zU7y)qT=h6>j@Z}O*E*BpAW?&$86g!^OnLn&;-bdp17GTc^^J{`KK}f9WsEOX0BDXo;7q#Hd?KipY;ozmKw@
zCY#9e`*(Ar9P=&q`rZu}*QSuMi&4pZB1)Z+Toe$g5NT~p3eU$pfg(xCa$nTr{yAR!Q9W(DDtbpGKPO|ZZ6Uo}rOu;}Cv7dV$10H3*%l)gp5+47+DE}l
zN8wY}O8gSYzw%E%2v$YQHC%fY4OJMJGII(G%1^2)yltvl0l8%y9UC1nVtk;6aD8vw
zkVs3ij*=W+aBrU8gXV+N!npT=%H?3_wO?!>+j>~YOQ-5LOq239D+lg+Kg5nWK?#qx
zYX+=3GcSF*|7XBnr?8o@F*41Rl8W`NWgS-rg!abCN;yTD2T;Ai$K`-(@O(9!wT?hG
z`Tl>XDPgPSjg86m7h}u5{Fh_@BXU#NY)ovwHYHXXDnsIy=Xn??aXS-&Ni8rI5>U3t
z7RZAoYvoXJKgG%|`3PMEp9p!rGj}(0+k}nsJTq^YOUaF{$
zcLly=Cx2Gn2I@L@wLG0;vH(9Ah}#Yiv(jIcY>jA>ev<_iM&LMmuH??1Zc1TElEjn|
zmC)4`^Z2nt`PYciyu5~i=DdURC{Sy2m*_k(BE3npzZP+{Ue`;_+t$q
zcnxCi7f7W&Z$JW#PHP2NpDaTFV12-192Dbf%rL7N;W>0f+4_I(Fmu$mGe@YQ;3Po0
zERMHL5)}6BJU3UW-yhl=zpgsV-~{w*H#6YTJ-c
zqZ5gf633&r3FCPtYx!%Z)s**L^J*VOogD8|Vhky&
zd#ocv8^9l0R5Uae0lz1tC=^1Uh+mANsn5Y@nM!_+9*up389bgU*C^*bdtK0&b$_o$
zIp^HaApTgGbmJt^BqhGi!DM}~DeC9-KtIk>Qd1Jj{6s7=4w7?R?>IDps!Ia-p#wMK
z5!W^6OJQHJT_Is43;PgfFZx{uaigBMkMX5Q=;Y+-#4b|cQ+MC6H<$aktk=X1z8OJ%
zUAi`e5YMXi&gL0E%9G!Jzohkiimwrjj%`}=kbr+;Xb!1s_L%Jp>U`Z@S~o14U3~(z
zY9Bq?N2_KS66G42H1iNZj7e{ei27Q?JlOZCq&q#G5%V(26>{bN^!aWAp;#jQPOiO`
zldEDihd2@)muXhG5yFx3aFcIs+1Zezox+A{U2u@(G8Y0Dy`Dztg+jwJSwG{B7lX6opZJrgM
z9arjdP__xko^1MdX#V$ED4Dz>{f$XZt~`h97weB}I1VO6!Q8#{9S{#V-w;Ah-9s1^
zW=@=rS`$fU|L{KRc25eq#!slXMCRa$EM*e33DVIaHz1~U!%V~@-1gUJkn
zi#-+wMbu_+uid=}Y>Q4zak-Sdzeh@zYbcbU$kNqV&xa^zKUw9$THSf#U-UU1}uw$Hah@$-!LHp&5`4
zMTH)|)KdHMsVjFPrQ(~G5h#+PKN|&y*mR^Ndys^}&r>xiPQ`O&>DnX)WL(>mX66(-
zZ4sq&?tAwtVsU@64Z~Eps>0;ufe3@>fFNUW{<5nR(`tOj*G>ZNVicjtEA@u#Da1^=xUxbTtu*o`*TggESJyCc|5S4U#qo
z=5Yp00-AK~%tTw%f>wWA3dQh&a=o)
z>PM0UMoqeA(HMLALVS^`toyp8wDf(3Kg~S$$Sm$zJ_+dMQL|x8(z7WpK?BMb9>khPasSF+bQ#}#7=xhC?I#+shkqc6s~T75gYULY3*b)0ZrvQ{4N
zR7FD}$Rj0-wp{OsHGyJM%HtV`gQnW$OoH4;i}L?P4U{K1!6BKh5QxO%!Rk(}gcL^$
zI~uha$1>U@YDJIqx7*#1d3kTE_CkB*Ouq?T`46_Mn82l&1rfMPAPvG?DNmqhmDq_3
z(=NAblMUDP+Ixv`f_(ZEBSC~VFdNmOy8l@(j
zz%2@CoZ>pr)Otjji@Y+XM8!kJ1XlQH90Ff
zSoZp^Yg9Yiazj73hf)7^2%M|mvll*M>F9!fE0yd>uYye}i};3lLUFm5Ijs{Domn%{2O0V}*JgrrHzLITMSnVF8
zSjZ(cV{O9nkDFW5?TSb3qre)x@TwvXq8*ZQV&=0l!7zt>!=)6d*Eqr82s5r{WS6A~
zI-vU&fwP=HpknuCgj(v53?iGP&e7K);HBp3_9dz=Il$br@!)6;^C439Hdlw7N+0|;
zToT!n)#=7}DZwFf23Fn?SVUl;gEwa4a)n`h6eB}Ry9w%%7e9ix#YeQ&-HqO%%G`U(
zZjfpAUDa#WzSJP2x3k)~#Z*Gk(ufk&Qs29H{sJ+Fgk3^9{yLiA<^Rvz@8<`d$!%9(
z#U}?4Yz?+s9RqK~`$-+}NEjV4YaOFYt*ly#vxuStm8pdTfuuJ7i11MK8OcwtDQVDRK~M^{n(Ev>+`rt^clEp|Yet
zF0UR0ZzhuucyBi=DQ2wEq?@YVDb13zi3Gy6D2>&)Ud`kDkdqgV{#7MRJv_+b?EC6j
z(mN`xd_yfI>+}l37hXT3@{8N;Un%2xr&XV*?jB1H;Kyo0zrtG64VT$2L0mXEb9`lX
zB<(c5u|_ezCd7nudUo@q`)0p9KD{eh1`m51ch5|)HLW;y&FsgqRHvVqYaL1m=-H_BH6uzDD75zq^9KVo?$uRL{Yoq?lJQqO%4@bqW3zQlPS_iA
zqPS4bs+iL%lV)a?gv|8%m=XQmBdJDz&UOf-m_k@^_cE|eBDg?KIqRH_uKw-im`|ZT
z;oKZMZP@u~-;LI`rmAa(w<;Uo5k<~(ah&=0+Wl|5>u$KdBhd9V3O>b%HYSYx7m64p
ziCC+2$-E=jv#&3p&d<#M#x3GD8Q7&ce`Ge3Mi@VG#fUl4j$<`zSf3AVqF%rkcZ(|fzfmo7!QddLs>BCNmk2O8Lh4BZ{AnrqhZrgd@tjn
z>K&Vv(}ZzU=0=Y~Khcq;?Y&Fbdhd@-amkG1C;@Xd;D0*;EpMHBO{U=3>}fQgz8r1K
zPKzWd(~5rFmf%-C#zcXX4lojbIs|r1QeR}o6of^7K~uVBvv2x;0o*(D-}&ehBLNe;
zNV0fIP=!B6gmYMaa}P6e!QN|!)n?Dd7c(+(CK)F!tZF72E3fZ6b*Hehq65_Sso~wc
zWE|ylJ2v)S2&H&$cS|tz&c08=n2~znFdpX^j&5D>bSu!rUpx`dkb-L=f-gO0gp{3{
z3O6Gm65o7f_jOkv9c5B2oPd>yqS^0MgQml2XdS{8HSz`TDfQu(#;VN3Oy0ukjP&o>
z__F}CP6rkBaJH18N_>qVlnV2#C}`rMckxmg*_vP-Q+V7m%ZZIMU}!waddC}w+>Zkt73pGk+9E{b(7jW{Z&)Qe!}LHQx@|H&Ye8`p0RAPmhcyk>nY
zyd%b(t~`t~$373-lP$njwF>`I-{kGk{b+u3$QKjyk2BW}=>_RlrYsTc26j5V&Fi=V
zn4qp2%3ww9VYqU{-Rs^xq=k=CdS^c~FG9p3oRWZNkewI3r1yG$@Uc>wv^WwP<=dK#
z?N&n$IPPwV3(ewZMmCB8#a~|^Q%7%SBU3mV)#lQc0UVgiJq&PQlLj`9CH=UX_g$a$A#d0O3C|4_oPCN(@#QDX#0G6|tXc|&Jvc#Pm!C_N~PMt>4A@AKmzQqt=J`Pp!WiMZp
zd}PXoPlHs2^158?<@lYp2RXSXCnL|adUxg+vAk%Eq_SDsKQnf4?kG~L@r(Nl-Q_r6
z4&(!*<>${OhlPQ#?q|^v%EBVAlVoVc4r3;vy<=C#iy-=_Zl9Y$aPuXuaP!XtACS!~
zMC`Iz-}AvPUDhli23GHDA89lfl-fAwgUl=}v!JkTo_0Tff_ZN*J$W2|ui&Y@Lo>Qe
z8`X;FZ!StCC9W^YzmOL?%}%D8-v3!E2o`)hRN2lGWZ@@)tmvPd
zfJYW_r}lx%u8wP$j(s@4rp9?S6ts)3Wm5BxqlE)N?J*CNcd^*u3Tm&gFc-gQ%@!$F
zxa3ur@oV_(fA`Zvt1J1N>t&X4yzBn%9@mV9ldfKnP&~+KyOw?TVCU0DmxbXx>iG3-
z@}Nojs@S*K`q?QRuh0eDKbRYVH6V#`t;<3PhEo*>K>QuI>_Xh}3Tyfv!ou0gfA#Vg
z`Qp9RXYq(7N$v75eyfe6*-D{%ITguqD>>L7c=4nvi9{d79wgov3dx2(pK7ejgG54?
zGFsF}x(AcOJW-OKO-z0)i<8TW<4}>D>JyNC%)1V$%uAdwn?@sRN;2d~xg2)~_P2&<
zNVDlUy4htR@bK#Ilp4NfYF*#slD5iuJCZE*
z%^O>Mx%1M^IXJz+&LIP}!j5io70V0M4z8VYbNxXvm
zb1J|NEglw@vN%7cC+w1mi^Db*sUAqLOgbx*iGa?Is{y!p!mKHF!Kai9vN%{sn4-sk
z;z3vL-flxn(Ixoq$Q-f(wMuaG=GFLf@6TG18qP*8uLQ31a_i9`^WrrvxZ&(V{3Mz0
zOJky%i0c8;tV%-7+w-ee$D+9WJ%Ve9z+#dSGn1fZ)g=uQC8z%jPs2zAKvnD(*_RFT
zHaFTpjt70(g?OK#@s!un4Ub4EDD_MY8XDaCbrtMMg&jdz#{89oEFi-)N4LFEHv||Q
zg*ML4rz#hqPI}!wDJH27t7|Bs1-0pw-zXG`j?ViqpuQPC$gzfP66?Qc)4gdpk*_4^
zEYx0>XyRyu
zRAMv;^`3eg{EDEX{3)f7_jZJ5mOoGgPmn$wEro>43IN;r;?R$0ng=pz_^fSdAPa)3
zAZ>J6?z66iUThp{RZ!ixx4bjLDKukkLb?o!`z1tYDxYZ7S+it+hdq?^Uq{e3VOabd
z`zb^2AnxI=zy=4kW1)J0E<@WQX=UN>tYrfuuoe_WJ+!q2v}DMm_-%${QKq2tyU7fZ
z=-4rl0uqsrr5uHVyPnbJtJzCAaEI1K8LTw!fusjWNi^2^pHmc8zT?nORcjreEg*6b
zjpWOL$Du~{A?*0KE)%=nj5ig(et|-I#m`ibV)OmIPA!aUzrfRXcO+x4yRv2+VUwhZUjc
zC3!f=ODlb}3|3|@p=@3iA2@p+>jB64VG4;SDb|d=`4y)T6kuReNMFW22~xcw&)GOe
zs^~D(jh2bzWeQ3&fTNtUsObj`-!4mM2l3#8*WyE_Hn4$fN}54*Pma-Yq2;SziJIQQ
zZ~RUAOq@7ZK>TZ895dgRi5rn5ak0=Or!NB^en{ND`Y0Kl$Z=`b3yFTE3GowKq#;nk{N>jxQORPu3;~|+#S_?jbCezEB&JTjV0zf*T
zK~SF1^Ft^wQ)hN`oQZ;gs`V#q@|AUcc+rt=B?0Z@wCw!%zbg-bX9Uv{T%6dUmA{#K
z*jiiyks9n0>KvF6&ydNdMaMxINa3arq|1L;K;mi!C&uLH@$iK3kL6|uH6(j+9P11G
z?~dbf0IO<@bEn3J`&K~Rt&@<_9%C{n4gDB>LhvGA)3uZwyJ@DZdcV`h
zyNcjl8J%Y4-gBjQUy_&f#MHD?XI74D1n0;EAA@u>D%tcJz6R4O^XWq}rw)`dHNm$?
zoOZX@Fp}!3?iTWXGzJIJJX6lG0%l*k`N4{{p3C+{ni`QKx0`aV-t=qSe2
z>QX-p1Kd^eWWZw!h?<_Y5Nu?bb|?RYSVB$DXTjjKhD-tW5bOeLY9Mz~Prq+XUk#Xh
zP4I*Fvo6}~;PRFM$+mvlfsw(gW$`mTNe}7Qwt(3;ookYlE|1u(t8)6UMe#+x8ytYH
zxDC7_+&$&fi~FKO0GJIbk;UM7fs!yubsiH&1YzUJMj*Q122Sf+RnjFr@kg5IVX
zGW+$%DccfsLs$MZNFwE@8k?^z@x$;Rj1jEu&
zDQVXehxN?oi%voTwHu4R2KszI2wxSBe@52IsW
z$AC1shO>CS>C)+_
zkb>4wm7nDcJU%G-LOLrWTF*Qac4KJZ=n$L%s?SSjjx~K1eT7KhN{>mx`|#Pp;Gkrb
zWK4_e-?c@`;OB5Bj3O!T!=r%L%et;Ik)~%Ov)2hydZm{^0l7ZJ4sc1;Eu?-dOs&e_
zj(eoy3rH4JwNTY-jq_{R!lQOzHOau&xdf9p2{!=*u{EJyc-MY}b!xQAiCPYK@=}Z=EV68cWtSkA@S@WV6>jA7zKDVnx8c`J^^hYD74l^7y!yjnb4?e
zNsRSg_yZY;{)Tn5{6Dnp6+v7?{cgGs^G7f{A!hO--K$q>$y6mIX-TD0H|PLdf^c
zu|FZTQ+YtB2K
zPOF7yDX!%$)Z_a``JNe&r@-(%lJCGfOndIa03H}gW8Ldz*PyQDnEgBtLptD?3?H9u^1Fh?n+uz;)x5FJOLVj;g-Wg+^nE!6$o7@)Vd*0^JXGBXk^dxZL+fYH0)FGI5{
zW84T;uJ|6!kDlDg8@QJ6F!hT_Y6W$u82bONS41OEA4p15J!b6%49Az#0e{agI)d?g
zHfd#f-?87vhB*i3cZJ|Qet2~Fdh6yzE#(f5nwdl@ew1E*hG%U{Ru`Dt)Vz9A-
z6`4ulE;yXAHQ`cGuWtZ@YL`|l_sKgO_erp1kPw~~xta;Ysra2o{;f?#mnqJ0znRSxb35P0XZ>#};N7hN}>^%$7{5gBC
z5s8I+n75*n$p{G%PoTCbg(g{hwW^y65UZm6L56>#J!k@s4#v6$>hhk
zY~#|Kpf$W&nwtBNCQ|JJ)WU?50kiHgVLt6^%*>VuRl6Zo)Kk63QX)3O>*!7&L#_=;
z{;hhMmT|03pr0rm8cAMillO?I=+aVs;CIwYmi40A|3)&aa;2;EItZG{A$Th1H0_-)
z_gr%~QgbP0NiVl!MNZy{0@@h0|9T2)J3lBl3@x^@v11u~s-W8^X$o;ZcY4Yh?_aO}
zes<8?p8{oGj@{ev(?8fAK`x@dmDkM<3xnzWN0>V3MG@E6*)7|SgWvgU=Yr@iKM8k?
z*Z^mAiafxXbT7q&$?tB13|+=Jm2wyRI?+;G4G68|v)eZn>a3fBXm+_-uf%-@N{b1^t7X4}5P=9V;5V?a3
zp)%~DuS1$tGO4JbQqaYTntCkTZe6y8qJE*V#RziQEc0Q)6hiCPw~(7ob%l
z&W6TaQ@yWOMNq1ue%#v$H?bkKXU2$9SFEne1jhUG|@yO%AQ79nYwPpOSK|2G@Uq21SJYnq3C^WnMx~8gn@%-tZKDuBOkhHj-
z-3+gGg%k(48#r|UvJNzsL5l(GK9bDk98Ds#tSY(pARPlZ?QofccJi!F6S`Xr>8IG3rcN3b6=3d}#jGCKX69I)H$H;rmZon$Wo<3oGM(-=JolJov>sTeh*SsR2Ci@fF3@=ZHc#HZ&*jKN^Xsk>CHx3H8W%9{B
z#oN&v@<^VMp(pRbf?)ARY_g&4GQe|=q2cM%M;i=__qy1uSbV+`N;`f|`yXDyWPMqc
z1hSB^$Q4i`>3U8m#B36cJd+1O29(Ic^mF9B{{vpQ<^@>SJP7>BlS&GSPaNa@TjsbX
zdilg&9+j~+mt{DC_G8q_jYM$HQrv$t9}iSv;4f)R)b&MUS_n!38Ig1EIROZ*qADfA
z{AorAx=sl~9dT7A{6BccXlcPPeI$D?(c<$MS}y)OTV2EO)ZY?7JLVhC1Q`^2U#|xh{pIo41%<1H8Gqr>O>hww>3l&~G8
zsE$A>>=i#~;RL2uXOyJXr@p@YjEERsAcZGbME1GwPWlBPYF|jFkC=i?i?>51_30Qt
zVRo=NOrSZx7iegOETH}!34A`SvFdhSip6KVy!V18^%Xfm4|Z*xRJ~lAn-3mbicjZL
z&u5+$3EwR*Ma;rdDmy;oqfAVNC8IiCotklQtao0$5R^z#RZLvAvOqXH8^sAP=m-7j
zn?AJ3skKODz_i8z1oV(dW5r8nccluOrt1f@`|kAH8z|quT{y#6WxAem)gNn`-s!Tu
zyAb|heIig9Hb1>F&{{fSW(qjY
zo;PLM1IY;FpszTlfq_r7n%&iI2DR%Qso%YGS!nj~y8VUisqhzewqNemEX2}lnH!mB
zcBqISzM*|tk+5o`4%&M&{oO53jfGUR>bDf+w@>sQ-+z5=&4f-wa=VpyocS5ood}7C
z_(WNP>p^?=ldr9UkHUkC1M;)-MwXM4@2^|#@z#7d7k~JUpz8Goztl@c>Y)B%#y>c*
z()!)l$R89*5pQ0-XrmI(36pS_5EYl{&2{W0XF^~aB9?Vu#U-lKyVW{)&_q68hL%J(
zf7$;2and52zw2zlYibccxbuzM4{ZU6GBPqi+Y9K_-2a-=IqG}#H$4r->3G%|^~HvT
zg#oP`AVb6kZjWjj0-99``wBom_UqT~jT3MSig^7+j>`k4bL`aPZZ4V#y~Xy2qaE=0
zt7&(mOOfbkXoQmTkMZ4d;cvQl@Kf)zZ2bB)#LFYDWge|YJdlu-Tq2pNG#k4-*{~b`
z0dre(k{OPMLrK3`p^RuaYCY!DYvi_$ItY5i&x#ZVTK+O@+ci)wM
zwS#puxXVa)=WMzExHFsxxyB6q3J?Xo?1H(ic1K)XxOnqD@OtoDt89qlPqb~r7@12p
z>N`C>1?p%>L_A#i#RT(DcDA<BxveZO0}U>3#_H!jg1Y0!x91Br)z#n*Qvbj}py;tP{Ri&(AlU8CH^JuS`ZrF-
z#>T)T`u_9GgNJw53-`rp6`g~FgR`?&9oRZIz?e#9|2Ct|=YY(3eZ9Jpk`s`m+64Fe
zSfb)TdiTVT^f--xpeQeIW_DXE@M59#1?J%-C{c4+Qvt6k&Ne+PM#EPF-DHTH3?whd)R@B76s_
z8U_S~)6!7U(H%CyRTh5+$e&CJ)UNsl2iH$753kO^3>XB8eT>3F`8`Ou9-EfiYWp<_
zpcx4CtEw!f?tp67G}WKia>?^tEc&*$3jx7CU}hzz04F{BEFke|HOX4+YXn
zfruo~bNX*CxxlSS+_*r-JtL#S)5Zt##lf#ZwlN=eJzY?(n|MMXtu
z=m5~n-T&DT=;8Cv$}Q#{?Ub|>tCqFS&Eel6{T5_wxxZyRd-2D@d{>8f?=@RIn_^;O
z3JP|>7u;X%Fz!YLK4O0%VEcR##YOYG&LRjva8I9t>v4g))y-iW@_%!K@cp3WejWz1
z02jZapd5DM+2{sx;l4Bby&i&#Wez}Q69}FaQqYYZZ`O9OAh~`19nO-wK^2jmN<$
zDJjXx^FB}EJ+HQ?cG3TA_v4-e$NM`&aJUE=85xmp*%m+3`z)fonsh(>@eBw+9#S@bAJim0xZblz^M&7BN!`3wME}>@>-R{%7KMPekCRinR&~3&EwN(vp%8pkDab
zvO~Cd?r%;K6Y;?SfmR%_N31^s>|Oz1vcJC%ZhSmEI|IsL*I+gNx}M<^TQJ2y0}qIN
z0X60}WM19Ce}_0T*Whde^KmqLjW`p10$n~k`e(B
zmJB5F{6z4_$^;WZymWDOjsNzoM5AUOm|Q#34{VJmz(^ka)&sn`JOLLsK7RBFdN6+r
z`mTC^7w*gNp?xKYmq+bDiWVH62prllF*F1Ul1LxSHV-FM
zbjoHdkFYg9Y1i8RJ2ZGOt>K|~?BKEsa5wGQ*%=7*C1^N6YZly!=>7T6CmMbM&ad?v
zgDGALDez<^oHvqaLP<#}KR-W;Qo402D=i4zTYd4lsIHC!n2z8B9BAX^`Zqf}J8&y4
zn32N?!7LaI=1`~k?i8RiUB-V?N%waYy}%#3-}_Jl+komGH3I{<6v%rB)T#RS2x@L_
zZi^j(jT5b*AeFW9fK?!Fi)L)TNKZ}u2yVPfVA+b2<@v4dSyyLk0xo>WGWqo??BvV
zJ^S-w6p&tN6{aKN;#8EC1HS+8eO4A1zymyhtnuJYyi{%4zdtcI_V)GxJ8pBG?a}`%
zs{6iycf$2#ePC~I51bzdYvQopZ`LRa6zUVN`92EeeFW&{qhT!0|BG%O%D7Ie6Q-R7
z)_?zTzNBmI*Ir~;riuXf*pmuf1rFSV3G&I$pWLLRr1bRkyu2-$nNJ8Kf!P50Oiw>V
z@89SS;&QkM#0ZygdroAz&z23mGe3nt4?>EowRkL=Mf73|W2;N|qe+O5Oir8hO_!yw
z*>~P-F(D)EjCMgW4+%%Qi3vrZd|R87vjLPIZG8`se~m2xr>2rk)-2QS0LqcSJ@}dS
z%?Exi`DKNhWF&iH1bl*3gUjcFi>PpmWC*rFftt(SaW2Q-5)z2zl$4&PG&MC*$s}HY
z3hpTWiRay=$%Ti$jZH~f+K(k$Qx@DgeW2qy>CrtE+uzcI=H;H*LYn2($)&zbcJsz1
zgx4T>Fg*Rojpk+C&x`PfK`n2VeLr1|38jM+2xeHL0lG((;_gHC9Aqk-qyGf}x-u5l
z(e(oIT3pPh@lwZ=1qjS}&zbWYZI}gxmx7-?ym&&kwVbl-9enp#kuYAAf(aq<9iN*p
z@77~{+{Ng?qBT@=uYa&*q03i15oqM#(u9pFE}5IeUv^L2@E73nGjP+
zdKNwafHmr2rw8CMj%L(odO?c@=_yW`f&!WO$!O4olypJg|3}qZhDF(hjoL~GQYzhu
zlr+*QA|Tz}-CaY2NJ;0=-Q5hKba&^_T|;;7+vk1vw~ufAgqgW#R$OteE2L=N7|`Wk
zpD5hMM`99jYd+i*4Fgpwwqh(?{QEhR-aHyf!ai11U!Syyc+X0R9zL(Y+>Izu*)|@R
zNWft1+S=Z99;Y`iwt_r>NC0XfmltOn*cowHEX^Op7ORLV5t`z1sbbH8}h;lo4-RMf%)rCq`a_u;6us89UVVN@K&6MPEr-38+JqdyE-RkoPs6_RWh
zwUv&;4kPW^>w;?prk0G-&@Np9fGk#$))Q33J9jvanc2y-!{FIxM@PQJXpTl9pe#(v
z!DEvXa7C0x)6p4-dS_`#rR4b4bGLrAGF_PH({)evoK)dXNpW%H;>c2i-D*WSt@s&)
zUmfdd
zUK?;B0HWg1+%7Q{|N1KRP=SYl3G`5o9v4e-WL;LN^C!r+mkbZ^Aczu^_ty^sN=D^X
z-Tp4?uuV<^x~xIr%ARw21|Q|FQbjC>*YElS7O^nVyZ9?bW9HV3?BBlr;yhwS0@P#K
zZ+;a|J^tL1ZiUP~shrHyVvLXO2Nc-td@qi1^`qXHThnR4jWadS){zXg?dACo(EkxF
z_~nlmAys^~b(<~^7TyO2Y?PDbKU|WeqYGfRwSm8lpGidD*7cbrHnUlybFfV*o<#`)
zf|nIeeQDW-Vat0KKOhHTakGYb2cd)cH1Hz0FV+tzjngqbUsTS3CmOqM#YnYDfY$qu
zL(1K^tpnz*q>M~R81gghKm%JYC@@51e~Yrwf#P``n2K7?@y5U0Slh^TeO79T;rpods;Aa#dlvQ4;;Cb2yx@q1fYpsTd
z3)a)UvA)7>{EYtkr4N-X*4#`OT$)=R&~o`b>Xufa<9X7(_2m(0T3?!~8y#nKtaV?+!1Ls2_F4}Db4l>81_^tyTAv&$i_A^$&H&gsz@URAuy*sat`7=qKTd%w$%v$iXH4*%`&u!h&{`
z>KUJ6b`jHj;PhBpp#Xvl?~&;r@(H2<%3j~pPrIi$ib4KVltX|j6(FYifTQG{=j|n*
zU6^(H23#fiiOSvA{e-^?2Lq;)Y})l)a-fX8^1iGo-BBlS|HU&y+a2xfv#I5HUIx1s
z-SU*Uid3KqiEZ~cC
z#|?9)m8Ernk}GKhtrGRz$Y+lfLS#8nciXE=Adu|Q{@K3)Wcxl*B%_AJ$6ZHj1Wmw(
zybg&_SVBN2LM0yY&dr`D?dIABNARpEl1DjcG>ER1Zfs)qU^lF=f1;F(#IN-itDZ)!
zv}sEOm{Emcb6KxRMjmCXu7B5Jsn(3>
z$;9Tb+thVw-^KMSk?gh`PD0t1xNs492+ptWLIFXL8`;V|LzXfaV-&r&q)-YH*N
zkLXu?{6qfs`QUT13XP|RvkPd=_4Pm5M0^Hv(xtJQlntoAAC@=5jD*bmuAce0UZ5VP
z&L@MMmbQ&&Nh`lSXi-{KaDSj@jp5gw!4Bo-8xbw$q=q{_{(j?1=jABi5;({}@wI_r
z{`UTFRr6=I)?dV4B%XxDlGASa9PDFr-=qTb3kQ?;%5T=KeZ~JS;_@=KzgJY|9{%X`
zJdym>xFRerWAK1jjM+%VFUK4gg*^zn2G|yoMg5yB*&SNcH#WER>|IRbX=7^9?^rg4
zp})spQ_0tYnK+UD5Ec!~RR^6O=4Z%f1;z4XW2lJVplQ2k{W3{wp;m^KRo0lL@JS>t
zkkgcSXQy!uwHQO(eM4xRu^(5eW)+t-)<8`h&Z+^$@N0zE_9k*cxCN{;A@TAj1r{1SXB~@#mI|#*eO!!5v6vOP9SvMu9{GO~
zwwm1CmIJ!wg>BI3jamjn1_X(rl9Q`YZeMNbk7;}Q2FmdtE|#IG9~}mc9xuh)0#%gC
zBCeb{5D_gDFyjJK!fFxuO%nbR=evC~{rs9v2IK14`sN~9F^%BRz|6tNjyXVZ`RZ?B
z3t}41#y;R1WHqN0TadL*Z(15%KPQOh1>E7ph4}9Qhev{3cWh3MZEEf-;JXiZB0-q-
zytBUwZnmIb0C^hL2Sv^`tkoQBK&al=#aZ|XbQq2@vyh+b$})qG;XtE01&IvZ!HJZlKEYv%9bQKiNVmx4o0>N-5&6gyh=Hvq-W8
zdDB_Ax`e)-k)Y}LnVQ{RLQ4Pv($Y9TF)OLyXJupo0XZP?q<}G3XVk%&aRw%)XcI8{
z#BV+-88x*nBN}Q0`8T`og*2g;d8@nnQg0B^4FYAO&U0OAZ
z+HOMcMkt}oiLzi>0y_i4t^JMpWo@at^s&|Dlau}H2|u-u!rzUN^KNu{vUTKlXvFlu
z3$cE%b0j-XRqVEx`)eTd&b~#SK|8(?LSYMtb^j654jgp({-uL`FBMV_zk`NSvuyfL
zaNbDcAlpp7Hw`Di9VKDRyTbh9v(^&vWHyO3bVLCjT|ZULW=j;MUkt0Zzb|Uvm(W-H
z+tdl}OXq|BTsL4_es=mh!m$p^;C*Te`6@K42}i}5wvxUniixF-wqw$0;*o5A*2vjh
z!Yq9jr@@?7)CqRwz+|W9TlM)$gehn(gHk)Cpb|Z&y1j#_h1vEEwT545DPoariLu_D
zw`7Ud54Ry>sNfRGii9+Zp6oDBZiF(%b}|hckIU$)(t}kzNJg~7e&D3xHv;bGSy?&s
zk}%gVw?&K^IWb74VqI5Ese79SC9g?hk4$fSeznqav8JNT2ua;rvBjI_&?@3(%YoEH
z{?5g-bgXLkZ{k~g28%6tKWWdlJC0oCZzjl{!$?m!@NI&fzv9fun|2{eZYk)0xezWB
z$6tg;_kX5T*dsQH)UK-M%!kb%0M&5LCGeSWM@bG1rdJ^vroC@&UH&sJ`2M@~G|{s{
z1^bZ6Ilm`FZG6wIMaI4Jb@#ZE$Z=9h!gZ!~8Z#%qll6l=UY@#IU}`^``9S1%ZeIG%
zcv&Lh+=w$SCC|dnz9I1$S*^F{#!ZJTdv$=EOT5rdzfuVpESq&3Fwvlt)2RBAgc3}k
z-y7Ub{7#4GO*!vU9sRupcvY4C>Bj0wmRNR`19?*`ncJ@QjLKm8#I=uk*yP-25sGr=Y+(|?p$EAm
zxi%bYA{OgXAtM394xkj7YFylu(9@yUmiv{$MI0Ke&_{&`--<#e%7ixrB$ekjQdI$*
z!qs2{YI)7y`B{SXTZT?7&dAynX;u~4_O$}ii>9tFv#6usaDdh2pTV>8jdK=Ei(epw<3t-5UX(}oe_1j<-2gpoNK$gMY79+v
ze1fDy+NJe~nmfBjb3Ri94Vg~tJ7WnL%;aG^!P}^qi1wdegxJ_ZOb1bxCd;_Rfwc2l
zL->L3ltwz_X)9lh426ZL<2v@zzI>P@e}`REQL?0+2D3F|x@TNerqkUl9UJq5tDm0P
zQTv@>QlGH%EyG-J&fTO*K7~5d_2RTSE*0;)sDs~eH2D%#LNvzHU^TR-_Cn*_@x#~A
zKe%%wI#W!OtEyuZCJClJsRvdpv{zeBt0(``^qg)T=_Y^XL3*>WB;OX=fvQ5sx)u69Q0FrUv_%9~T*q$qUu
z6!)YXL9sr2q7$t?97&*%CP=Ij*i22U)}9AJSoDc$Z(F@T!ky$B6cH>{jo$TM&<3gYAPCB?SAcYaxzpk
zSlL=?Q^pPawSRnNoSH;SC?6x_!($N=t>Gb);8y|C&}E{%SQbbnfAZ(jO<2H8wj8ii
z8#`XDVddcs@6fMWC^PbG_!e3>z~4|>EeTZAI0UoVMuMI*Y~hjl%vKlanVH4jn(L3$
z4u+#&%91Q{{Di>c5@0q0U6^8SjCKFIDQM}a?4;ht+^bLIe$O>=OBUhXi?y$cxV{wE
za;NtGTS>6xUP%kgGLy2cG$_(x+W%w~ed<>~95J%k{2?gnM$-Pv&rBo{QgW1_;Nl-L
zzbN!#5Zk0b>jYvVv8!+g!Sy6mR}9z&MY}6NcELT#n*2;@sOuVM81kN>e?DmJArK%)
z9^F&Ys8PHQ$V}aX-)9B~edAt|#RF_yrNC88Gd@KVP7A4!&$}g$uM|E1*80UXV@`N|
zBw-zsaq;uui08qC2HUjgA=VlHsM8S*7f}9ST#mR8&kimdh+Wv3x=^IP4QqvD*~xLz
zcqCiQ?-vWbU1%*B7nO2sdBCzd^tOSxzHkXZT*22!Mz1s)RFs!LbS{v22xqW!J@ZpB~7UBko
z9;EIcE0lv%zh7=H{?z!8(|(Z*igq)oFmLUnJZU{AG8|4cnZIhFK9tsZ70=Ju9xZa1
zI|5!9erNU|Y&hZH?Mm;Rb-KfiniDsSyq~HXznNWHIvDgbDkP=)r{`6xvP;=yBK?@#
z7wGFnGMiGNI*b}8C9NDP9#o$i9Sa`*P%N9CacGd^!s(DcoIg?9j5&dM$alq?ZV;FS%y}-+wC93C
zWVv<%c@9g%V}6x{;p++ME|JLwm2&C{*|IERn2hr)vN|MQf&%wI?KYt`6YuH5d}`5I
z{6CMY`U2Ng!cN!$2=^zN+H;u@BcmKS<*;yZg@(Vi63PEQV+%o(ph!E-ZwOeep_7+Kx6iZ2%kHkVp`W_M*x{48icJ-TFXFcVJ{Jm*tg
z=o-#Yl>8ld4GA0dcjcHN&gofsHJp}jLVed<7rTdCb3idl8RflXFipnuO3SYMHGDo(7rd#BZIC(41`V8mIXfl!mE5S1;r|PeP~dWPq%QlWGaf=o8r>
zSB}2>?J!Nz*_jwNnVoix=9&Xjx4h^-N=Y0NDUH_6ylS}OOW))dd3l$z+~74ZqAoDA
zTy*HO;fEAbyGbG{n8U01TNVyRSzk(971~)weq=??J}ci`EzQ2XeY=E!i0h={^K+OQ
zuvdUwvD9I$86ZbL
z4=PNxE-YL{y0@!stZx$L91
zAEu{e)%HulnOo|&w;KeY5wzgiYh{t&a+}HcA3emgitUlIbGz$F-0;`vWXG`XwloA~
zE}5m2WVWB{(kboeofm`tkrqBq>-p#-TgnTQ$4$)%t4%;E-1)pQP(5I}r*m(8Lqx;V
z<$75dA%w+$6+b^rba77|{5NeEC3)RRHGozuQWgP=6$fe3rXE~?M5NFhEN!#9?WOSc
z@a4EFkn(%Jawwr1MhP|Vz%|mbp=IuH<2l+Vy}TxKDLqtW9bD_Qy^ri0PP7(=W`0Dp
zkBU;rUJt_L(dOC3McI^s16ol@#k^{I5ftSx?0X_b1$YkG=28#
z(DP(lUS?B1`DO7V|AlALApU;+{@wO{v+ipS|5Ei>TY3Ze+}r(7hBxerONTKFYdbJ`
zNa;BGaV1zjMxwAq;lRCSv66c{y6zD$SvK+u+m>h73V{HU(%_xXRqR|~TFLh|X5S4z
z*UO9Rs3ZrtG8WS){?N(TYv9_V6%T_8J9o*x%mj|TZE8kp>idmN%E>3M^s*30Zb6PX
z_{7{Ijf7}A6LPUGm9s5n&r3@j-jOxdFB2f_8f8OnmHx?FC+`V{mIbp{KR13o{VGr?
zL69?cB$kFWiZ&hP=vs?N*PJc}@(-})LXA5ekpp$?lLC;|QbC2iX-H2&Z_q8?v%x?rFPNb!77{2^yI?&FI5Yc}
zZ+BJYg8zV==XwFF;k_U;H=VrFh=dfjUOWP=2vkDq_oooov9`+mVkVltaE!Mf2Zvt)
z)t-P3ccNV2^Oe@}TB(a{b|qnN?&)RiKd!>e974a1@ztU3X0Dl#DLtvr^l8y=@k)!@
z)A^aM=2sY%@B4yX%a4<3445l5fuxzp)Lb3qs+U)tqj-I~g;V>?OxF3z)ZQfJgBEe*
zk3b5;`Dj517IpYvnp&VGl;m$15abSE4$6}9!F23#8gApeq}Tz{ghhUIlTknw8lEcJ
zoql`aVc~1$A7mOB85b93e+SYHBANhE-`+?_(4>m`)^H|1jwCWGXSW+l4s#~s>sNFq
zxF$igOgG7Dp-JKBmMt2iD)Q9H8n465-Ked(&1QDO|M$v6Hqn3ZnN
z{L#NH|1^-+XD;#Dz4NYw#%Mu=%;HbgsU2G+_JX5G5jp3|+^}>JgSQ2rS%4DlxFF7#
zYkQYdtiapJ4|5=lS&=SAAb+O!#uH~Uq?Kk0rPCUZLH*|{GQe>APjea?xOzsa+L~}r
zaWhZM){07(-W%x>8gFIgYGdo~f9Z6aGUg|Y)iU$x`
z>(jxTBpH(=QFR!cW8`l
zZ2>2xQZs&Ut?_n+JuRW7@l!HC~7Tu%FvH
z{jMwxJvgmgdd*z%_ne?5pFM6q4B2gP!z#rXS2j6Ao?XexOZzkgi9VaOgR(9ivfq@0
z^z7G{As|I05+&mr(XBzLl~*~mpq3C~B`IVJI(Hlf*h#mqg75WR#&!um&LiqeRbF9)
z&dS2{Djj)DZr%f%E)%h!zGcxW@1G@cy@r+wxJcII;f4G=<(+{|MOvG}k6iweFOSPl
zGd)V_M`hXXnSz}*FQGv4*7UPRs;8)vPT6pHp=lwbm0jM{{_;qDYKGpckdx#6)K?b^
ze3j^Bup`YI{t3{%)mxK$Ud`x?)C`fdYyOmhY0%12wd|$i@*n%3SYb3K^!7Z!ahDt5
zun$emO^^ChTbjE!Rt9d8*pEMB9n)zl;-I3n*b+zvTB0fmTfZ3xeQs-9EYX%+&8j0I
zkIKy*9Jyr8W;l6Hiw3*rF!S}t9hz|p4#Dd_0>6Q!ZY*X9iBSSbv5Za4u1NW{W&Dmy
z*a)5KZ9^BZ+Y`kWgkZ|eUGV7$KGo8H1POjD-CJR5XNeCmpayC#h0l7#@>q))*%+arg0<3J~x%Chd0_7D=o>dMHpD@zgF&UA1<+t3G6Pd
zcP_aEJY21DiZE?9>U%Sx+xNEnUqCF%T`(=eeiFdlW^1A
zEbm#A6l=D4v-mv_qS0ZXQYWgHCX&~2_P#iVDs#B55lYR|6gHjp5`6rd5IbB%RX@0N
zT+uMfW&Dg&fu?r+C!33rrcoN00V?4D3&CjCmSlKa(Ok*3RX#6kkwjafb5%m?EELak
zR-fM;%zG7*tgDgnhM&`aBrC#c`n6ddhqU3KnxeiqL0PEBF}{|cU&<#BvNb&!Dv(6c
z{_?;-mB^aKU~6hB_5xZ96tLIGTkVg(n}daOh;1I@3izS
ztwj)#?Rr4CIC6e(2);RH*P84ER10PtSnsd6BLo-u+FS1q#Gp?*ceLq#{H_{VN#po&
z4oCW1vE%Rg+8sk$2>hX+PtAE^SbZ`Bbl=Q=vXu`~+{!YgEFS;h&`_)1^6_Pt4Vv{b
zv1HgLC|UH{!%
z;1TcH1`OuOHnQ;68}8`;Ha04W_VUTY$V$R^rSWdYhnpSeZSh;RU~#?G)WtRV|KZ9y`OE7VAYz6I+0$p%oTAaeD{gdaZ)
z8QhT*HPW~UG;FD>3D}rE8QWONRPNep%%;U8@SYzJi-zIKylWt0hx*ywd=^5lG0|8V
zDehjOf2YK_C${x^+m7HPls2>qw$eQvSL7a^%fdH~>GGWK!hcYPFZH-PUEX`!%KcUv
z9aT7;i*RG4PNmrt;i{f;D|5%FS4kg|=N?U9s;|hZ^rf!YGhQ4b%;fXTO;blxnwgIg
z4Qq$8QVhQ_G4+|v*f3$L!g;18pDYFMr#e2tmmy07ir??`2gBy&q^!LD*m_iz>glQ6
zbTsr^{>!gfj20$eFqDy!lw*35?mbp#TaU*n@beg_N26mdzE~9BPW;2w}R|
z@VGAM2&8A(mty08C_^MOSsfq$Dfj`1D5|R3k^-V&pQ_TAc@RKr)dYgZ!Q)zGP;3r;?+@O7y;xpEYAdd=KK6k_~<
zZ+SYu8}=f?sLL6odGMK_h#n+u|I^(?1&6Z0D+zBDIMO-{F{%^)$blJJCge|Qg$Jn5#ry{*YZ#W
zwbkF15YEoE#8Z7Cfmrv67;~FLF4o2gMG-;K!CiOdjG0Wd4eXp%oF9G(n{5&sMj33T
z6^ikdoKI!Jj7{#LJv#6Pe0Kv1ng?`THw^xew$kznTwkXrh9r7g?eB$CmClS!)rC5K
zBfo_4$49?Yu9f<@T4QlrSrtXRi)#<}75GEKzrLhIL^OU>V_jI;;VF~nWiD!>9ZeJ0
zA;05L&>g(I@+Zf^G@&fp!COpkGp8&7Py{w{?5rU!&}>3pVlD^x||NFQBhQW!;(GmVLSRORB1
zT6g7AHzDPOT2P}tynG1t^>)Mg@sYKXQ%#~e-8)s4C3!e9|H{|b{z`tA(2)-u`HjuO
zdh3R3?6M1lnbU=S?i_A>M~i
zIPW>ADcAN!F=9DB`0}dOU8+=+z6fY&C2$lL7KH!hs|7;4bmlQ6qu9*Cm*-p@TsCzU
zqO;P1{$x%q2#C8
z35CtFp=-olH?9xqKY+Sh(t{sbKmYT)~^nJ79r`mMk=}Y%IN)7PP
zXmnDCq8&$X>S0mdHKYwqbSOF`9=^j~f{FrL*{M$RzMLNaT->)Pp6gkG<;RfrKg`(F?U|F8Z
zKj*j3%X)hY{I9Ael(61)_eKsDeOEXBE%^8ohYH3;@Eom)rD6A}`mg!l^%avdta{j$
zaI0qnA1sZ!`8nb5+VNhdYX2&TTHkU5@!giQSyHMF6MpS%m(KfYh4Xg37P%KA5B!IAYlQQV}{vhGKf
z`1gm(V)3K(Z}})d;EMCO1e)T}mFJi9P?;bWX)1zt$eMK0b8{(~Gl;pUOY!?|W#dzM
ze<*p+@+D-dIacn`5y#Ek@*bON++R4^`U7TLzBL-Ny1fnDj;0!wfC$4hk`04@RSd`&
z6!|u{iTn5ccmgDfqQVb~pD=a&rzjY9L0^vKk}?DyR*qf^sLxC4pYWk{cYQtSm^4(ygYugg}XFjU>OPaX`o9G?8q*&vOS1-P};n!^N2yjoS0P1!Qrg}|In0fSxL9I
zqcMEln2kkHj8g8Sf{im_L;~?TDr_S};Et|bCP1OJVkgLBTmzg*y;d5Fb;au@c)fLc
zKWI_>on%Cqo^BN+H+x?KSF=0ctf|BJHAO9t=9_|CAtz_Ftp{AIa*B*!5aFT-HIa;{~2m;+e|*eBd-$!avYnO%p(~Ejr+xZ}R60@Zc?S#FY+}^vhgqxa!PiA`y~{
zyeZpqU>VIGC56COc81V=UCt>JIT~6kEtZv_*dSh;X;qy}^mcu6R*~?_Zat2vK!;zA
zt1l`+>wS4TbE(nJpI$YnZ)U2
z&oKgD490gl&T0<%lRKh{c1jAby!pt~5Zv=>taG`M6F)(3Minja)L4ohgB;u1#dvSU
zJ*8R_cEp-tnfvwFK(q;JRu`g2GU^)Amv$SC+D|Z#40Uz1^U1YcIDsO=hq(L!re1-z
z=VnXo;(jt(h?DPxn;Y8%RTrO^zO1^hk`cr|5ExXMKsj_f$i)v&bsRUAu_wu75y%Fo
zMU>Sq&~D>W$}Y3Z1{e5s%_z%gAZBK;imxQ~MGpB^w{)dZV8+;?@SB>FKd5l}5{g1X
zePR4$bC<7(-N2N}EsN=Q&ku!7-U4$Cb829mWUvuK#w=o2Y+f%xKhY;Edp{P6!7el!
zpNtcXQj*(!crUGf9G@@CPa3X57Y#LbR=-*qE)f8ID}K-tIC{9e5PDzD5VF5mZxy_2
zCnV@u^Tr%dR8N-$y;4wI3GhXb!3o(B?bhX`xtJbKuL*2>fYBHYE!^GP^r0Hv
zd@j4_M>W0TDjUC}((BgeOn2_^r@(q+;_J^21T|kNOFkA#$JaQ_r
zUY?F-)2D*pTX}3s-s^pP@hlGVkLVQ8fN$_*3_%L(QUE(m`e=e33|H`312zwv8~td4
z-K+UwkDcnYFMz!F)y2fv?4P|lo_U_+B~3$Wp|o$6$MI}r|6X#>1BSIzt=h;oYO|>x
z%rxlG)g;nSt*47HpyJ-`55LFvS}UYrVEg8eLrNR=^QMU`v$WnpaYFKG-d|MWbfuMm
z9{7do|Au1`&~QA@BN;6P7A`Anjnwwe&LsJ3!M}Xpj5fsHADyj^v_^4Cp7!PutrZy0
zs4tMgwCHLdhh>&Ny!rd;5tt0n+=t?1)Y`&g;GL%$gbgLanD4gSjXssYLQhRC`2VvF
zPw*Sqct?JZkx%g%5oMw80y9k7;BV7N5D6tnyI+k9w7DgUgm(O>f2!@eC=nUEft%L<>Esc&;UJEwetLRleKi}T>jJf8ZsQ!Bsa#bz(O9SEN!vY+rjQ>Ex
z6UCI7+ZGA+-;h#$+EpYaFfAEK-!Ph=D>Wu56$*a@d^Fp}j!UnP@zb50?>E!SjKAW6c;;pn66wjz^es0Pls180quN~wwgCty@Hz($VF&o$x9ciN|*=AG+FzSvDRpL4B*D_1f-i`a{rYn4z
zomh!xo?eS^!dkXCpo-l!74~Y5BH*4UEJ_rC=7#g?rxxDzaTA@pqbNjnMi%=NU07&I
zy5xk_iYv*br;>amb?YT+Vo8%C_;2b-IS4mEk$CPou_`T#Q_X9_cgKo%en{q2*##)l
z9>QrRyCO7V*`f036yV%jckkc+t)cmmlL-ucR3V(&obU<-_v%u({s|i;4YX4qo~Z`m
zUFKfyJt|Ie5Ud29zSbk~L_+Vz2_X)AczjxI3vDHpYA+zrLkRknVEQp}cmG8pHGP$a
zlw_@-j^eSQjcFdu?e-EzNtYng@wB`O5c6u8Az(4`Yrpq?v4V4FyzMW4gT<_faRgg!
zd!LZ?<6&UMM#ve^D2tx1z{VyC9XLqXJ^i-PMP_WmG4X@tv4%L+>!_0Y2X
z8f9Uf5g^JfiiT?rGHY&SQD&_vo??A%O5DI^9ueD^j&;$SyTpG=ZL*RM^`~r
z+|5@}^fI)^&_X2M5Oy?<_&iK(-!}e{j%d65!r5f9an6pg!gi7PePh@9%SZV`k|}@r
z@-NDU3qFv=KV0ryvFl$I+$bGSfu76U@kf}5nKl}RG!uE8y
zmgz~ZgA?E=A7-R3mM7MY?Z|_B02D*aFpE|{`8|DA88gs-H{iWsVB}j2jC;B&NK7B-
z2PnPFwa!<2Nh68W%CzY4#PUyT^T+9-{qxA4@k~x?eF+OrcFga8@e-v6eq&5yo!X2<
zK&Hojje2I5U%EgDv;)wha=9ruQp&_JZnOXws-$ZGBj?jy;3%$hV;wsZhC;RDB;Zc5
zD}xn=qguaF&}#p>U&m~}n}mo8-zI#$C+>rx{Wwvlu^PXwEaWoVmSyMSL^%bmPOspT
zer)owCc8AF?1jvY`2R4J7SV$J9nRzSCb8k8Kf6{5lP_ivlxraB6C~;m=VvNp@r1L?
zT>67^JWpbkAWs4Si`=RGtsrLgBydy6^hXX8QNjbu&pU;bzLM%c-;Fk0__24M!~SX~
zXOLOie3DQ2?V>7s{WS{XQ)yCq|K+-62>)}m!|
zvoS9`4we9q!ZAqdE^C
z@2w61tTiW(gc~sH+uK}g!Mo>A#fVF3yTO~FuLVK;wq}Of$Z;M0QjwAWTuhUl|RZ%Y#IcAjc(eB60AXoJ?hf(yN#t{dnu=Oi@Lp3Q{UN68NCQdMmD!kW*
z&dlbH^kk9Nv
z{gR}3ANI6#lOE>mnAW}cP828@Y-xlI-$c|@w`kRzfwR$uyQmSc=GZN0=L5jyKfCHKQV4pb>aOqO;b&>vm=?RFzeZ>iW$98
zGYayK627fd8NX#sTauf0GRYf^;*|n^#GL}|7w_8dBoXX&@i?m`(Y%0u{%^fKP8J&-
zHv9V3A6J-bG_zV8PQVB9?szPBlwpZyT;)vK*M@#5yS@he%i?A)M=a|8fXl#B6I_g|IkSCGZB3ni
zR7r9YdiX!)lL31959uY0PmD+x9w0%XTs27Cw#83krbNkhg9qbX-G$NQg_WT=n|y1j
z1d5qLtHI1~=t*?Jv^WiM2L+WKMlr#EYINk6FnIs!9?uclk=PV6;+*;e=dmV;(-tWv
zp_2158rS@C5LAQJ8Qrx!ri!VDu($)PP6f7W@EZR
z1$hWmC~y1HEPd}iAuv6~Sg|o(PZwgp+pe|dfX4%5r>eL`Ver~kTlVVC3#ticcf2BU
zT~u5m-UIvia8s)VRfXV6Cuadp<8s{9=|-Y`oDz>ZfQCk0q!c{eyWsHK%0Z-&PABf;-bj=D#uV`I9cUcsu4S^Fptadg+LVyOn={V6)8JnD
z1uwopzus{PTy?!T)3Lm%j2Xm}q$BAB6W>BctwSyX-{)5a
zY7nI)Qej81&-~~sC7>gb74Btp@+vbdcBBCwlwB*TB-0}bmh!-yc-Q`Z32m0g
zk-E}Igh@qEr|g*-6c@fh_e1Ghl}vQLsGa_T`KoT#H-SaxxxS5RntL;Ll=xw~#d&
z>h04c)81G(ddh10ote}V1O8t}O-q1WTZ~27h(KG{@?~7u1}1#?XNpU@14Tjg@jfk1
z79P&8+}|d_^Z|bIPskW58DUWv%q>zqaFzpPsudO5e{xybg4T@`-(uH=++1z@U}F*@C6K0zM5-GljdYdsJK$GQOPRg@tX_jgUR
zccTd5AJVd%6md0j)ArjR{SVSs{UlS%&n;I6?thBjB+w2AVY
z$gt}Bqgz1$cE-u{l^j1h4X>R6?HL+@K2?m
zn6(5<+}h<8^(p^gx&*c+$s!CVSzJ|-m_J6rhNDxQ!w4vu8Vm47vyoMFsD^i
zKKg?$XR!SlITVO~sHz?fd?7L{zVjAFcGjf>t0TJ^V|I!Yb!?`5q=nQnxaZA$BZGL>
zNQ(psYr#JO*7!AU_qBitDGdd0bpLsC_3L}aaF2Nmc5%o>B7Esv!iBG7)g+dZTYgYL
z%`97wOu!0h*5ZWt-Yll9e;qY8V4x+B3^>T{Vl9Fv)mGTw)(@d0{(^sbSE(fYZasrR
z8{Yf<;9wdHCC7N%^SKfdQBHVP4JI>NNI#R~@fyY+UMe941RgyW#^3B5E{+CW=mHhd
z#vs}HJ`9<}U*hLg=E}T*AzwLv@_&?fFRQ!7Ki}cPBFGS&5A;fy!6GQtCzyVHW20$c
zwGbCTR`E;H=j(Y!+2wS6jH?+00*TBc+@slzsKzP5ilPr?i0|GJCs6QOI^uwS$*oF1
zyh4C3XuJ=fAI)x9jRX3=Sk#17n~rUU5+8|p&JDEs}bJHK?jya4i>^q^N0;)`*CPP9h{hME5E#|=AoQItV|W$@-hgOi)f
zOdJJ~5AX6_A6P)If=ZG8rWxy?1q#>)y8D{vU#)ydbbYOLV6J_f=#29n4t)Nz@;xwO
zXvg=pLMpCr?|D5@jl`tl$jvCuZIfXAAE2+DTx)KwoCTU(4~xCENURxv
z2$bN^i2gqY5DL5E?yW?#%h
zrJEZ^9B;o*iA7Zt6Y
zY(F0z#wAVaL)d;Tj3zo!2#g>wywQp2IjtTim6+-J^1rq`A)Sk}b6!-u7hgYsWpmRw
zPYlgWJ>xhAt6M6)_^jqCNB|B9oU-(~w)Ev^;7*aua7&NbIemHQa%z8cq6=h?e%WCm
zZkZK#4b8QJU%F@i_Q&PCJ+4$Qf)3(ul{FOO-N1meUFVL+*T!iba-epu_LQ)Ph(D?nNNgYIM?;v09VRK!T6%3P)zPh^$v6f^Q
zv=t9qGXJd?yS{IikIKbU-|i7zulu|FImh{(u@2j>tc&fzOP7UKC4Yu}c0!E+)c?2J
z@~oW_jEcYsgu(k+l{H2+YzO>brsxMm<Ze
z8%Ay(kcRH~%HnPeOsf^0Z_0%w#718hw7A*&yVd>aC)e7u8K4%wYXc0flyMA?>f167
z!+6K=9*o@y_liEZhuGi0i^%qG#8tN!%KBots`LI;lyu;7aFEw%tOyb9u6{MOHn*j(
zk_7kI+b=!VW(Za3I$?vKsF)7uBxO|#9-zB7%A%{SpKYXnX8b()$eEGBx|8U{m*wQ{
zZfpHh>@aO$miqhNs=8L4?K&Efl5lq9PFk76UN>^c&_*on8TkraYe{PP9RU7ntjz#U
zlm<}LR}BwHYB#mv4K+lbiv|gR!LUWe&@Xd$^&RiuTJ$d#k_`0)AQY7rb3>uT2ipF;
zl_b#A82r-8T$s6)3X2I`sGr7+HV)p0n!M&EP3t#LT8RzL4CM{g>|E1xjOiH>vf(J#
zcBScw?@5@(r|n&91Z#a$fmy0dMJgoPJNOCR9O`?}=!I(9nxdp(3>DO*^A*19_-96J
z^oad`Aw#1M#O>x2RMM|$Co7>#GjUWo%N;IUXJfjOa2XcxT5
zZ5CtdS)le4T=6y~iB>lcuCl{-I*;!sM&nI4{@!lMuK&(sX@>^(^TILrexjkR5yW{RQ5Ul8+#l%Lh
z(+yNLRVHv*0>33c6{QrF9-EZe78at70tD*+aP`(Abg1gT9)*9H`%<6C$?&-4C{?>hWPg)7#YHF3^4Ghkeo(4e|?
zQAO$-Gr3|m0jBd4Fr~r$)iT-7)WhAkdxgP$rAq!kJNSh#coY{`B}WCXj0
z%)zxFHjD_Do<2He361nsa1O>DbJ*e)Yci9AqE4f{3Z@hFXH2};21hG7)E53WL2wKp__H-Vk
zTmfEeh+%_Wr$N_n+Kg3{f9qftQ>sLE%9MdgojE_N7)uDl+i9@&T@qnE1Zw79_QuXw
z9PtTi;`*8GH2sxd-Tg_Ex$|`$R()B^_Jp~Zp@f0nkRt_9p#q&ej93;ln&-8he$%Fr
zl0In=ih(dQ(y&kkHm+&jf<@<)U{5myBdwSqnpD
z8=dR!GBiGa_q+Ic3zwWZXZbEYlkOW1qVH_
zMw{Cv5TC71IeCQVbR>Po9^_k9ZjErC(Ild_WF0xmyN5rBwy_#`3#FX{=pQnq)5*kV
zFAwV)(ya$Mh4f&NYZRKI&W)9DnwHH(G*2donn6TLrGi{&y;8*>&-w&N2c!4Rb3V-B
zfZl%KGKaLzr*e74SqYY6q;Jy;Li)GThW!7*jG{&%A3&Ser`>%Cr^n$I&lCS7&AaLp
zs{oW@ydpsr83`gjf(CDW4O(bjQXpFm1EQpgpEBV(zg>z+Bg|$9cTel*e(gL@7vJiI
za!M)|oM+0i^2bHbfWRZg5>LymeDsfyhU_<_V~KfmXI1C*?y5)N`X1M9pA$fy3*JsaEax
zgtCV&&!$|cs9w_CBW9BT0Ci??AQ1ONVs;_0ED$?4st=lrF168_!)WLqY}LuW0qEm~
z`IiM@m!luBv(8(0;iHS*FI*3b}Q9`$j)tG)Yb{UEDW4KK5jb~hL
zd4rSquV-jfKbb64ggvdUZpd!=A}JNyA*rEdO>>+)|APtB_k|Oos4#yc*0sATnBE-A
zfTV(2z=AnS2H|2J3JL0N`1(0YK%Xv85Mcalol}N$%;|iLd|fbtjvMf{8S>=Fq@+L>
z-_*OdL#6)P^lwugdU)K%?0W&T&-mr&(@tMM5(6NJRo0;>p+gR^>@HKD8ej}Cw*D8F
zTLi6*0_UQ?*1*~`7Rz&sMvXbBU$TIdR_1g4n9Fv+)y#gy_#RpbJdEp78u{suh(ZR3
zPu+vjsJlFj1mq2ZUV_&6W<7ChqkY%*=2|wbYl`
z__zclAoK;3cR3k3aZtx(qA%BmGC&GeD(ZJ=cebYq*uPQe7DY_k^p9!ds?VV12Eyf7
zF%o+gK;LOZ39|3xV%>30;;4z`SqSr03BOw06q~v;R8n@yk!g-eDM;~m*JUS)
zg<2>mq5qyx*@}agc1*s}Am9!r5p=u;I8z2vyeWjJnH@7Sg^W@Hpzt4l@~p(Zc!A^#
z{jkePZ4=Yp_YS-aBi)6RmW6Kz;N{jeJFxWvGi-U+$UtVc_M!s+{|w}11EX17m#B%C
z4;8~d-zjeZJ^W1C+f2^$)M}F
zFuj7wA5#sX?}{HsL5_6OLA^66^*JHze3?B35Dea`y>2u2<_Y4V>%reZkE=5qtnZ;V
zh#rm}95vNC)sV}kP1UWc$1?}>lUzFveCn&+jlbsRKf2Cp3daB#2gu0~wiV;ep0f>t
zj5I5nOdNvP^1brhD9DRx)vIW*1w7X^$ohoO@?vfj_)i}Zra(%jl~ktoIF?S|kc3yS
zEC4P>b~j)T{j@jGLHXBr3c{2eHectP(|USF8hZTuv?=pYZ@r`_kR~dsl!f`k)ro2@
zT2Ubs0LIVfsE1ui7|i2?fKT+YU;Gn@d2VgMV`zIn)thNKRSXlK5CHz)CJK%H`Jd#>
zoNj78(v_o*=^wy3!mCFm*@C4=T}Vf%aY?+!?gu50uwSJ**dVa
zmRc_dU2&5>8AEVFysaNGxa)$iAv7v@E;t@-hLAXgs4=)$dlPH-Mkk7Q$2^ld~-zcba
zwzH*b8i7~C9bczkWe8~5~p2%J2jsB0L;PCPkybpm9t7e
zBp98*ualzeQ_uU9RMYU4bECB{`SXnB`M|%pQdoi#3mHCXw|R~APm+^JyHHYK|BcLOgqD{_iR3gSIEG(BW)JM`VRPX!|aliQb9p
zJ25&uz$vKo4>_x}L@2NjN4q}`*Z?f}Ulej@xnADc*p-gu7SMV7O)cERXVW(tSzuCK
z$ahP6D+c89J>-z>67OYP4$*pK;W^`w>8TrkN?9}$yJg*G+689LJex9`{FC>KCH1zFKV$%(RhgCme*9H9xHk($+p<))
ziW*@Vur+^>{^pT;!7|sg){Fu%OG3m^GYt4`zb+ZeL3H^u0ec
zwAc7xu5$mf%I8Tyo4v-{pi~+D4!E9zWTlPS9zbd0)^a+2_Wc_3)I)WFB|kp2En-UK
zIHEv@Uu-q-0$|hV0o(6e=8~Q-eTMe@e@}D`vq0YOS5tD;p*Qx<+U)Y5IhNOnt8CHUbum>j?S%NyW%xIs8TMM7Fld@dy9WRb
z7o-U_{Bt3U@xNn>Y`-T3Zm8&RV0C5J%ZgU~c3{;a<
zuedk2w0&@W44+UoCM{$Pq)U5Ex|emV=YxI%-~Cew8_vfg)idMbOCGT#ejg@L81>|M
z7HjGdTKF!hg`_bX+dw(ctXpRCMw~>DXE$JuHQ_uf?tl=9eCL@JwY%<9;4=bS57IZ`
zu_t-e>kJGpKgisAc$@*sN&=r&uDp&@?)Sk*M`~Om{Ke3BIoajv6;MgJ6ka=P4r9ou
z`WGnSQ+HH3J(1M@o=7^F(3ZshahSipTwNm~rtitlBNo^C
zRo$Wcn8Yt*IU3{iuxXhueZ8^p?&@H)J!9$QkFz+;Rekk$mu@4%LUy$9)etww_+BCc
zld}k;?SA9%PCb+c6(BA)342u^T
zNHy%e*inj^>lmSh%e|V%P_u84gF$WfiO_v|^qL7(_5F)l4#jp|A;SNd@X~B$NlD4(
zhN9;$ka~1ZH=Ezl6b6h|Y+dG@E~4BPDqoN?kH{@c7hVF?bc%`9ku)Kp6^bPUB93Ek
zw<30mnTLo`hbHc^ySKX3vX^GrbSZ|RVmez9J~*3Aep%8Yg=W&Spqnp5&<|5KD(b+u
z-m&o2!&PRhqC0eSq+dzE^-?lI
z-}Uys0MLuy@weq$M-R%kMXwmSM_6wW6LpM|r`KB-2z{UX?0AC$nV^?p%hcyqmNlQb
z_RPg*KzR8EpDA8yvH4$3%=Gm1vnQvHq39rSipV58HTFnV-HHlvpSt}*^F)ooFE{_{
zGk*k^pFYgN{54k-_Pe2bN@O;}sl8S3)Mo{_vM}tKV*~%szpJDs4r*lFJxuw0=JsEO
zUS4`$hxz|yt{TL(=nfsu;Qk$*y5<@i!jt^_io3^EqOWeH5f)PslO%I-Lt+eZY=8oe
zl>MLy%f{LJkBG=kw}1W%^#X@JMlO*UC5)Thy%TEDQ%MF=Ml|_z_t+|D@oaS1r^;4a
za#RoiL>dlb*JD&qbWZ`;?Oz_0QkSdk*x6n2z#g@1Er5zg=~=EE^}rs#M&8%08jIEa
z{R!!Dtv~R*z^@C|JK1Q!yWhSSIbi}tSAvz^4`!o;5L2gD^W+~;@98P`4ii>STZ?aM
zDf#Ws{p2f+w%InE{n4GbQ>SRKlhrNAgcyj+iE3aI-&_xP=rYoIH>qWvEBUR_8
zf<`R+QxNAcq1wP-(pmOHb}bnhJ4iCzECR?!FI=ZfwK(s^7-VbG>Eb(-@lsR;S76R$NsP!j?;j!K9Mi-w8{7B=3gAolEE?8gY=?VZ*3
zBfFB)Z;vRLm*qv&^|T8IHSWf1bg#YXKAq?k8t_}7*2xJxUv8DoDtax~TVtkG(-Pl`
z7lWB0&YxncR(gv#ivGhycnn;{artMw5*j*~o0q`q2rIV1{+Dl&e@vkSEif+44T{5ROW!IYt>bl_8!Fw=rUhwM$*3J-G=rO@R
z@#qgA3Ov7sf{X>0Uwx1=uL}IGvC+xtrgVqiWZT)ZG3J=GUyEBrAuGGSR4|#V8dV^z
z_xkhSpO^1`X}e`oO_tD{GA}_>4jgXGKhA|_BTBFnQHmI{5+aR6e9;z<*Mp=M9u8}@
zsl^z4e)S$SyK^lL7ojE)z*ndt1K~j8N+9@>zPDECXJdfDefTaFl#C6Vle5O`zK;bN
z`5Ch-dF*P?qMldUi;=(aDhG06Tkr<7V}a0vgTz9
z$#hHR7E#|0^CnpZ}9F#t;xo_n$X?
zTI+e($jUkwZi&sbDp#=S-a&
zO{Df(c#e>2{I%CzlED&Pl2(Enf&S-g;`>VIxkrVLDr2wf{N847FiPB7WE8so4V^iMnO@K
z^RYC9yyo$o3@j^r!QxO(WtpBz0o%9ADe0?w`nv#C>)-W!=MZDh2F4yZ;Heilv2buzuGv_jblU2Gxk~s$d|ocCNG2RVPyg
z)4%#}I_c8>o*Krs9G|%RWz2{jx7kCP^qbKHCQ?+p9r_b1ds*%#ZuZq`HcncYIw>{4
z9R)=`wWY4!DMqHQexpP(g0Y?|8Gau@`HG@M2w77R#gkPn}7%X+93bVEtjfm;p12U`$javRQg*8
zBBL}$3;BkE$nCdUBHmk{@t@Nnb4N8T<8zMICdg3|pd*DIw+QQtp0q(j%H}uKu$)Lu
zWE(lh=h@Nb0PMA#A~Nqnw>Mt<#^m%?$OfBx&U(jBcaHi!03Gr?X2a0BT#b(QLS5J$S
z68?I+xiVqMRT!b!=H&yizT3L6xu=Ut-x29U+<;&lZH6IJL?Kp4w%=M>37hNkJk3r6
z1t>q?w!i8+{W0Wh!fd8ulL*RGHmahJ8kIBG-v9>b+Dd6=
zzFo)^=|;ZxD!4z<<|06u^Y*L2+v%ln#C0u0w^rVcO}0?SduPsxIAh51YPwkJJK$;O
z&B|BY{O)o!7oX$h;#yI7_}5Zid_RhTUpN;t3O<4kQXxU7^WoImzvawC{Xg`53d|sr
z9Y5tz0-=Dk<3vz@87Th1?FN0Ml8WO&jrL*uQZTOE9W>OM`PcN3DlPj4^-HrCE7{xB
z4Dd&F-&dWqcp0vxgiXHmBDTu-5)Mbr$AK@^?F>@FLc7$5gQ<8FU%stfduC(KgY=I&
z?HkK|OO`O~Bhi5kP$Y7MEWS~GX#rvTSd7QZG5|O$mkH;ijpm)syUyshQqoM{?{K00
zeN(ZswX5b^=I`Yf#zLBbmH>4I>rxc93hMF_an*Q-fRmjK!zHsdGs1|{+nZmddV}kD
zvB#UdlJWY#w-0ypR(-Ym)$J6MDs3_|Icg>59sphXfKZHQql=0PPFCkqKXn?kIJGmW
z(E>^CU?Y&gguG)}bv+%ypb<_z4*Mq!o4==~r)lUjsLDesbBSqPI(TUFP|P!!kH_@4j#t{v?jz;`IR-wY%yO2RSbM$$Q2}33KtS9|l1A
zdx$rVOEVz$0{WmC;3_-gb2fH9?gMaPizeY2QK7=`dnaSOywD*@560p^mmsG@kVM_aok
z@aZ(D)vV25(CefdVz2Q%Qv0bZ`QlT`tKP@`2G(EmMqy5
zmXC4D0!D4*#6Ab)S9K9&-l7e>`Vp2H7mzO11yz
zhy@kr=2H92&WMVrO9|4gOJRXJ^Mh|^p&9eU?A-6GnBeqSTA+Dn;_L;f?l;z=nKIwfl?bG@w)WSVB2=K4*mRX@FGI%!YbhiD@i21tiLq72rbUrtr8h4gWfK<=~MfU5YNkC%HZbV2@CkGl&2&ARv9inaZ#6#TpO
z0aVKlnZmmnOg*;Wzy7#PHG!@YN}CRO-{yu+T5Dj8dNM1Hr*&$vzuvInJWD+J1jKpF
z0Z>M^yL%D-iSlF9)8C0HNOul5qCLMZrpt3>4832M+Bt3VAN!IqpvQd$ZUU7vVx
zsZH5EwP-R&JYqcu^)f7+{is{~Pag0_-zfOtRaHE{{1gOMKD<(Gt@^(!&lvzR;{3m>
znGQEq-F9^wNS`L)YgM~uGr0Og=^&z|LuPlCR5qYf24HEU81W3yUOaUX>F2|M)4Kop
z5%XWBqj2GRK5BE5KiU}bdJ;{7JKAk)OGmrl0@ggJVOi!42I&UzaI#Y!_C-;=r$bF?&F`uM4yk8@bs4%?aI(`%_&#g>>_S6?_*qM#NwBvr-LY&D7Wu6c`B
z$6>Y-)}YGzOu-4Fv5%aLYLX)+S~KVx%a4CLYG?E4)L9gYc6p;`<5G8?wr9~9
zCjApn0EQE!vW0X9D8S0~wm%~Q1y!RZN$j1jutAd|gJNh&VHVxW^6aFqyGtTSFvWLL
zE^9t-peDM^;vTnqeuuSq3LYQ4ZU5G1Q8GDtQP6>ssUy@sz)T!e1cNkQ8a%9}>MPsX
z@1Khx)A$h~>gV6kk`ZkPA5!OI=V!sN^&(CeAUbcjuf^~GehHrbODTKTnxQg}c{ue`
z_S6t>`@G2Cs{H6Smgh!{>2_OsFD1FhM_fL3Xmjx5yt72U^PP;-WKTMW9~ch!)p5W~
zAtkg=cuz~ET7dHljHz_}WSoSc-nVBaXCRNT=&r#NuTO%YRBO2h2B*mZL5vdqWUm<1
z11rNoseAC;vwhezjw9Wn0T77~Op7KXK%w8ioJb0J&=JF`Ti6z@lqvEbz7l1`_|J$C
z+=>G(2<3ad{FK+UAY%+rWso*VeFwF7*ChDJ!@|B(3%2g59E&jo0F3fSY?LZE5_1@B
zr1*F{iHNK&P33u!kuuhLrwR^AsnSP`VrliK8_hYlVftd4qmu{Mg7+1-fD_0g-S!*t
z83*nSp{42ug@MTC@6GcHJ80i)!{%oVv*Z%$L+y%Es$OU7%a04UwT&)y5m&*I=^7d}
zpSGjOqEZjd{P?CwR7IxQ1p9!#2o;Q`N?G0vIv5oA+Kk~ntuCKTVllSbWUpt*J|sB$
zr8(uTCLJb-LV4@KgRHU(3q_}E1>EdI6&27HPeCcTQrVHpRsD1V!6Pni#Xz&#?65&o
ziBQzfq@*wEo$jumc)fQRvl-jSvdlbdqkFHU_RJa7YYGU6(M%M5pQa-8fSyUL4tp*C
zl7BV64Z?Q{x@CD!>!E=bP8}Vc4Jxq0f=TCM`kY#UI0ogiXBY-cuGF0ia+|-44zWl}UAO3T?*X4iXHB6T1u_VTZ*&eVQNrpufxt1*)
zvSK28nNfq9=Rv{-&oggGP^BFMGAL<$*NdaI^9tJ=)B_$zU!FSvG6JA?w9U?4f_Tx{
z{jes{G9pQpZzy6d%~atl=$d0H83g@v4Ekf3GShS^SRWHvRsODUW4Qs%0H);ciwfl{
zP5}d>$@63qq~2eLK4y6?u2Wlhibz}&S!$y6T1BX!jgil?=-gcURKWstb{j2=6NlBaImR!1Y?|~_E>Ixlh;9S7^
zmHwfZ6QI8DY}yfgfNpZ+>>#$WxNPag!!#^Uaj(tF)@hbU*|@S%S--z-WEv)|f;Cm)
zpj}Dug8KXg*;-U0$`c7V2lGGTRG@C^KWFicBUuAyFwgYs{z*-t1?@f1a)qk{beTA;
z+P~_6(RpDSxw4x4IV1(O!Ia#xtF9sAw)<9R4%JBSGlatC4~Fl~WLj<;t1bM44lP+;
zBa>^k0qhXV$>;js;{!A;X+hl%Kw(gV+d)PFE`EC$(SqzenwfxFbMD!Edo{=W7kTtr
z7{%32FBzDW#rX*s>PmvtzTRt^SSit)SzBU;%
z;d{nkO~J3D?CgXcjd!+6|em
z(Uj#daJX~zKBEozf|~O3*^4J`bY3*%8ZW0mM^BlLh8-0EMH3)^B)2387=bb4^HvUDP-W3U$brH>GQi=H{mp{P+B_M{=Ifqm5}
zuP>~MZnr>+fH!mVeJ%YYd#ALpnA$3N91ED
z6%`dL?nRzGJ)Tr#CneaTib$i_)kO{s5GIGHvz9Z^`@0?o;+^27jZk^u)Cr}};eb1E
z%!2UbX`LNc59{F3NdwhTE^Bg;p-TmgQsQt@nMkyCg_CCFEuD7yffi(mKuazMGqV72
zwsO58pe${jf9mv=5rl8~FDol)!K$_wp_uZmkTl9W#=f^qfp-^xU
z&`JsX#TQPUc>jN$Rj}xM$3EBlO8CkOqTFKn*_>
z6=9fhHRI&7Jm6Y5jb~TZktZX2@rt(pO~>#b?m(VI1Zax&tErTfz;}EpZUcQS&B{R%
z0{_KzKNa`rnz^!;?aFgY8BjENzJ&C&;qgLvb#ewuQ?d_0nQxSKy_3@z2j|l#_urkG
z8ej9(DByV&jPt6YYOZuL(?z(=+j|@uS_`bCGzsY=)-ub|1eN-lFZrI|d&2I!SJ>H&
z;nep+0sExwRg@v;PG)lOvgOBuYBIhVRh^9%vm($g&lnaJi891WPf#ks2~Q!kZgKx*
z{68%y;IJCpqs=!zJ9_gW0M$C26X~OWen8EAZr9HV+g{%JyN@c_L!lzuOG2cjnr1`#
z`&uWuPgK_M?9P*As6y=cE2Tjh;~k<$z!lI_x~Un0i<>yAFDL%
z8eLj+`scLgK7R*>DH4cJu75yWJD&w+x7kYH;#;;u;Bz%bDFNj;$=zCWwXypNkouZR
z_r9;;;h?7EV6KL4ei5mL3>2(rr-NQp1R?OOV8NI09R_sc?^cq-AF;lH?stG&F=16}
z=7)npG6JO=9j-1Z!Stj0tHuVC!E99xv+~um3OCXyy!slxP7?CqEW`_CCw#nTe9W~T
zDl`%|>0YJdDnb2t#ZL8YUqnG?r4=zBZgJv=SQ%Lz4@dwCkBeUXk!|dZ$R=%l-8wJa
z*h_*Lev!?tDi?Ou(0*eB79jzDEG09I>JWf_pg)5t+ji3$HTa>Tj0`1J&5w6o?PWOR
zg2bz=pNQ^vljIh?g@68{5z-%d(OIV^O-#ia}^M1LbxXr{Xa;I9q(2d^3o?Rj@i+l&&WZ(et^~!BTUj$
zr!!70txf)GONW8hryZM|+eX@t``+qAvN%XzN{T6`)dTwgiL2N$A{_mAW()Nd+fKyO
zQb|tu9Q)i{_ngt}Ee4owIxP>Soq=w>o52Icf<86`g?FFwPBMX^kO?l1|)d58#1Fh|=LI1zGmwq{1Tf~w^@P`%kvaHq18l`UwCbC`=DjyC%WC$P3
zpjC$DCLx|P#&GbmcTu4IU5ej@qoeX+iMQLO0XVMjGelez)F=5$*ci)*{BTUQ)65?W
zn;Dris_aX(Xgm}BVbcWW=~jISXkKlo$R6J|g*@Ov?dF-o*}l7Jzq7D1
z^0gGtzXqCE-p!GK;N;;oAn8A@DF5wHZOlRZ>a{Z{BEBxV%WU2Ir(|_|{=p6mzTJAZ
zl%|Z2i}~@pssOrj9wV7}?^15n8Oz63aV|^a!`8D@G4WERyE`V6Q}_by$U~N6;jN~z
z{r$PQ6waC!&O;dg)9L(1nBK9(-rlbF?e(+vC7pntp5WEesnI$#fdSZ4mnm+oPcMAl
zIekwZM1ZdM^b7;cFr_-^=EIkuIm~D;B4H!HUrZ9ydN^I!M%#ey?KgxyLuMwMztdyq
z*s7rYL2j+~gwq|X>^7@l)~$`uq06ofk=_;)TQup95fgb~Pmdb@0Ru{rb2(Qh3lwNC
z1sIzdx6&dbl!lk9i0gWR;mzeAby`$x$!WJGZY|O;kz6x(^NMEqG|8&bqQ9_E
z_<2qxDx1g(#dJ7xf0Ssfxq0o`Z;g@(M7ds_--54NfVljl?Th%`v>p3e1?%W^J+W*3
zH1U38NFF#-tIKiHJ6&
ztBBib$-5Dj$|N$LG8jdYiu)72Kt39~Zvkgu6v_K*
zP=e_QfjausHiT8nwk#YNTb;G0j6y%@MZx(ER`cH^!@*!;G{1+!ezyygV{}fI_7-Oi
zPTA(8zv?FSubY!1wRoE-xtS_2tXy4f3n-0ZW8B;XtqPclIX!DbBi7&$3&@j(Zdy{J
zk#`ZK3#VQ#gtll5C7~@cO<9NYy7!|j#ah^zOTkOnBr8h7%9@xB`y4r$;CE^yUFLO+L+z=>l(~AXN}z-CVeATz;=_hEimx6lagXnP4D0xt&TIcOo8hA
zp~S%_vSbuK+6#)jsBQg{%eT!z@hut}{kt=cvXR?aX|tzUsXNf-*sF$Wq%JXBBC03z
zAFTW9f#_@7`cEV`-eDc1ODVX})zVz_YSzgQ$)nxBe3y3)B0@B?TPmU&azjwu`sCfV
z1Co=Szbv|wl90Tlp&3%i6~IhaM?eWp@SH$`CMNLwYd8WQFnf>W#N85t~c~>9L)0N!g&C?W~C9}or
z9ebi!M9HQ~H`xfu;Xwv9;U{|fR=>Np2(Ho2Cq_J;47@0wr=xjNbQV5@1GhO#h`5zC
zirfs4uan)N8PwlC*yHprGGg-j$gHI*`vo3K5g4sM&ucg^`)~DHA9NF~=
z<-9KDUgBOD&epw8jt(hquQRs^M+U9yViL_#-(~KgzX4+HgG+}J6vwh2BX
zgE!bgu2jIV@eeP8a@S4G77)y#AJAw*P8npMb1@iEUPnhh{_qbTdOW+S|8;(bc&OAl
zMUKpQf2CfRT4;&A{s8^RUr>SgmDu{}lWNIoF9nN+l-co;eCkBJ1z3O_C
z)7FvK)zwe1A1Ej&@0gPEQ>39{mP<#$3ki*@r3t_|=5`Nr+m0%$|Bx<
z+G);Puf68?QP$efFg7u7i=_H
z=lI9P#kIBhIy!Px@M8UWL1iGfjjM04)*lm!>h!Y<<$C|G!85GIEZ1#X1=r@LD=^rj
ziB8+3q8b!4lznb*fL)Aq(kR!x$Xp<6Zb?l|Z6sv7R$7XLgmkdn9s`jfCWdO(LsP=S9$~rPbqlU<
zZYCuqtu}X@FP)$4sDO8pm6a_Ru<`Zv)z&5{=f+yl+fh8fIa{u&uEr-I7#<$3+Is%i
zwnX}&*)6w?3=f-dq=0^feJ~m^;-#Z_MBBSAK@9caC4HfYMBX5$}
z*!X@OEJ83bBTz62uebOwt}i&j)Gwbao^t%i!_ew#4hW=Ww9d&#TU*=LSGXK<|JtAV
z{BFG!6=RcrN=xpVKkY=)ejgER7xS)`bL#v_p
zW9VYDnzZx|@X%oF$B!O8`V8;ySWJwL0Roct`Lnx+$IU6+u94X9_vDg5fQnPF>l_BIIN1%jog8W@7vA-
zu5;n{>rRm=;-}y2Toe)_VQ+u$p*VQmThERTOBD4|%?GV#*VorzYkrR6p`ucFU!PBm
zj@kmXK*ns6ot+&)#M(y?JQ3aK`G+B*cfQHVqJTx*)YKG&!R2!I%Z~3Ky`3i=PmCqu
z?R^bg0{fEx+?@7cv1RM<(9X)ru|@Ic;@y_2r+vnFpNpjvGTDDqO|rvSbzX
zz1Gyx@fwVyfwfW+vFd|?_CV5lUG85VhQeyl_MF=~At%RJUwl7E4RZ%KG~9(o+1$
z)rV+kW1rcPH5gG3%AHvRJx^Zo^QXuc09T#dZ}HzvZCo8nWGpW)XJKK%#Xav@^{Hl?
zM>#uOw%|x1n;XfksIYT!;V!3-7rQ4yt
zn%WPGu?&~%(s?)>cpwxS&m{KPoHk{tsdGhc&)-tYY;A1R8TVr74&NW5qF>;bx!+HN
z<9=~OdK~vWcKfLSe>Uel~8vEGo-Na*$fX4auO_n4|
zU0wb3^3QuH?%yqZaP>`GOl*61*EQ+GOR$3z4}ks|Gp+)F)l^e8EG;dp=`=h;o_^;L^qee=QQgCx$_Pl?(bm>Sfv}nLNbH
ztv*+TI>YAEz9!j9sD;NM8)XB-I
zTP+5x2vQgjYcRB4M@HuT#!%vH6X9QdyVZMhbq2PSnTbh0oiCs)04#SIsxx}F)91|0
z8C@0LFC(v?>t&-aj8Z(SX{+
z0e^k`_%P7VhE=P&r@tSq;lcf_p;J>+pZ_XcG9gcVoqPF1&Q)0fU5*b94s300lUWV=
zraNFX;C#0So9oNxqQGs|<
z%>t`>PkgAYtp)KR@FxQQRaX0qJ_jn8yn>XJ`}xs}
zf5lKjvKu(aWo2bRoy--{5qpc2YSox)KK1tY_B`1R=wj3;-`Bm;GPxcmW-}D^^lbhI
z-*i0dwM;TgFF|?m1b=2MnLp{Ll^k73ZoCj=js&b
z^Ue_9!7wl~uAzMO2W}nE(%{g%c=5u);a7hQg-M#>%0YO?+C;7-_-AJUpYM6nre3XfF|q?)lCn-M@!2@-RxL!B$#EMn+01At|ZL=+9$HWX`#k?Hq){uDp+Gx4WXO@Y%EAkoV0FEHB6WFg=THzha4e}JbhUIH<~@>=V|UijxBDToLHSwDh@dcRw#
zEG@PBvpf!G5$2|&h&SQSX3`dm0Y&qK-qjF5)@R}G4v|K5mbX@R6B
zGXk%ejt*^&0~FwT&LN%s8tHPGgr
zjOKE3*k2{GBB4&h6oT&laDPJ+5KyrR2G46~Xn^4D`;Q+!r~A)%F#Tdra9ri@Lnb{!
zAfSgc6`8=%Yp`9+y255v-VRl9V}aLLJgx_xsYF2^P!fB4dw$oy`hmv^TB8DQ@;5Cw
zhB<84;7?p=9@2#j5{#_m>=*Jj(6ahQY(~A*lK52fvf8c|jT3cc3(Y7wm`{vSW07k-(a*FKpMlsIw_kg}1I!a`zhYt?5cP*$OG
z3j9%4oC;xy5|fo6Oo`#4q18KaOLJ4sAF0PaR`OUsgmpzG)0f|u-{9hlIIc^fi<;dB
z4tMj)!i^I@yF<4|VLGa+s=B(eH@iw#FVbmNT?XcER-Sg31WjD?jEB3IyRFab{mftZ
z-lL;RWsL!q*d$q5zFC8YLd)>$L<>FLMLM$9#TrHw(6+*AV+1IbDyKD}%
z6C9Q2-de%zI!X%K1<}aP2TUUSh6lM3E2A(-cPP!ox*7D3T1c}sU30OSr`5Xqd#b^M
z?3Al9w(B>av0^`MCnu!)k#E+o$0D@lQObYI!4~Xj(kA1m?A6RUWpFld=VDc*996KZ
zjK4IO^tKd=w^+k>=^2Lx^Ip;{G`sAzsGos&0|^;veccrK`fH8K(@-o7de#AkVG5Zc1YrU7A#pN`
zp7HSyT4v^(Vb=ERP%U7XmGo_KA_gltIHNK5(&BMHB@=7K5eDVCyK_k??JA;ovj!{q
zUq-AB8gRtPg(j%?V;voi8Wdc78c}^qW9_wBbec^PPa<7MUM;AGrbUkOJ3gm+wL7c2
zbVTkmmZ$pk$3p>&b6}oRQc|1+7CSp1ppF6zjqR3X@geaCHLeFpQ^Z{QZy)G~N;Rdh
zPf0NpV*Gz3on=6jUDt(GLb_ABTRMj>0Ridm?q+C^Zlpm%8l}5ax=WBo=}zhV&hvi1
zMPla6zV}}1TKhaw>0(Y^#kF7R7Oto~Ij$v)59z^}QI40;HZDmvps*`loyCS({N37u
zRiI>HL4~LKQ+3SE#ibTY^t~Xjv_c-)eU|j3D0`DJZ+MLxrI+i=MdD%->ebbw&(&k4K6zju
zU~HBJyulm?gsDL0BhZpBHZDt$7P7sJG&-`z)fO11Bj?ti!lb#eizaafgq^o!tmq=Hb9Hd(bNiw$
z4uwfwMLHfHY1jIpexNhJ#oZpsZ@>9cY5cA5-$?l2{9paGW;t0|LIQ$Ex&SOx6q3IuMy}?kHe>M!uX;E4
z?eMWyUVu7yJSQ&f|F{6`oq&LVBN>cfbW#z2VgiDNI?KuHYz<{T4ns|CXGU#oN4I4C?Mo?A9Ow6o#=PH~C2CIWu;vDJq+^17R?2tKH;(#b)QFB&hZ3
ze}jpc06NWeO+r@rEWQNdM>#n;8*6LC_7;OK|BC#4aV+g-R0KBAU42;X!0#Ct7yxDZ
z@y7K6JZIj;(o$2;dit}{O2bOzGQNeqj896k#iwnrslm+n0UGFUvJvMBqExbSazXIF
z3|!}3Ie(Ei*cLgYPj!o@92T9VW?aU1iv9QNqWmX&j1LPhX~U?}lW6xGt7jkE1AAa<
z-~haxg4v48ZPql!lCZMZ>QzCcqhP{3M~-Cz85=kmoJ}ueFgn3*fY?<
zw9d*})%en)&ByUaJQ!NG0;5W(#kZ{(`nG-h?1%*Ar>KdZyt83+V-mV060t2d3xyNM
zr17UxZVS^TefS027$ViR)K>rVjnLv^>8sitHbP3zT8C+3pj5PzV9|k2kd~5?l7XQl
zKmTIcvCF-rb~1zeqqumG;`dDp@*jM@w`}$H;Piv90FpHDXP&!8WSx~0nlh=r)lmJY
z^YHc-_(H5bm5P`6?c0A-zm&z1UbEBdHY^*~8BZ`wVAIN5eWU~f;x7%AW}TUBi=!jC
z2BFl$kdjP`iM44}WxWO)T>m{b0q!^wNU?ZQo5s-$bjMgl#E*g97{Scjy4p+ghnWc;as<3|Ul@^2qrTRQ$nE&>rV!c71Alpt7
zS8*$B&tmr4JKR1oGL;?K!9~6uRHfm!aFX`+ZtJ3XM?~atyu@w4R1daZBqm}VkT$e8
z7TWEsmFQjR*2^+oF<0r{953I1xvILMK}wg+)gfDU)r$1ZRBT9(c!FV<|I=xEcvo5I
zzMH_3&dq~%`ebDlNng)r;8B2~RUY)3fBul^ylD|*tQuFp!#U23ZDH}5a8U@Ete74F
z<1(Z^oOQm!XnLSZ=`0eP&S9w@QGWVlr^EkAAiLPJ$zSyYIpyz(@^qc#wJ79|GbT{h
z?4@bXl(0IV7&mTBc?JqoD9hK4bt9>yG}nfPA44$#m%1#=G|%4@uki-#E{+y>RMyWE
zbo3SG(q9MeolV`h*o}1MEKH>CbV$sh2pnH*k5<@HSWaXC-3I8P>iGNmM1+LqW@cx~
zpXaBK7VF@O+aVq&C`d>_^9fZ|r{E|wt8keP#)a~xG&VvdC1Eq0ihvwU4=h5Tn>gDI
z&_aP`olY^!>Iz>YMd}aF*p8toBA(RO*yDm1LuZ`Xwq642MTHJd4wkZf7T;JaW^HzK
zq0khGaeiD<4R1WK#)1d6&*+tVX*XfNA-xNUSsCmm1F94vLKew((1v4bTQI|ACcJqt
ziJJd7bW1Ljfz!%!mpMi;)Qr%WGsfWxePJ~yHe)kLYfak%Y#|3NiG6^GrJ*Js&DmO_|^L(Q6-!A
zwK>zQ6o(Rc`m{DwR)SW&G?)D4cIG@F>0T*5jYu_p6tBhh$%XzQrQks~L#HR+$=3B(
zoWzlxVH+&cOmDCIc#t+ORUL(3P5aqq5>X`@)P`~7*Ffs~uR@DGvqK|eu+G9+Z}53o
za1uqs1J(qQ!~6O7MHEXVUDnSGgwO3RE=OeZqsT-Ys%~K1ir`WASm{{ytxHq)^5zC(
zzHJwMTb6*KZfg^`AK%YY3@dq!Gbz%oo}(ZGR~kPuTi5OY5DT86jGx7mU%hJeIu!P0
zNp+OW=sm4ogey)yK}nEZ=T6D-!d^?cyEmkj6|~nEcBM#1eNWBFGYR|t&J1NcwgO?C
z{gjtV*T&w4S~K`xNhm!;Bd88Z)2|kZJbu3f9`4iC>>H|~j~_kxmK|4`Ye+1WB+&`=
z*eUaZ<0+hBd~^v9n>NJZ_`gPt-J)%ecq&RYHbs-Pzu?C(i6+cnS774eIyW?Esd2mL
zIKzj$z4fqcbAozZhXmnJtKTuW!SUN-2d91aNI;_KXeS^4;V(43zQW!jhtjWYu3gBL
zwc7hN$GN4KvbR@fUi|*%R0JI6%3p1u-?#*i#5;Oml>@5~LE89Jwg3$S10gIW^L|^
znOx4|)!92;!yi$gO6-~mcU4-L2dfYUut2`KA|%48Is4*(AX>|E&Y?%dE~d%k*YLmT
zu)u|8;rr3>bltL18+y4F#v4zq{rz0onC}Cy`;#c#Jj#hYstKLiCusM^jdENB$#>SQ
z5o|Q@=1UiPjzUw)wtD6?J8#(p7{G7vAt|X-^<-q2B1ETd1ja3&Bn34+(W%*^d0f;M
zPmpzyX`!gw@mQ4Y|)C>mehRx|sNl5e~W>@w>VU&i|CwxA1SS3IEo^#>>su#y11|
zJ^}C0qMVu=G8=Czb~{w!J19stZc#}aoG9+4+)Nwjw0*9c&aS?06viScemzA&fPvMe
ze?8}beSc8K6miH&0~Q|_4QXrbg%iSb?zEIvk?QHDIgEP!@xgUK@%jq%nueW29&44J
zDH%m^BOCL`@VZ{mEl~)X`nD8Q;BKHQ&K-X~EBbTY;par-;hET2+#1xbH%azr%9D#iR
zKag`xWF*rQ^X1JQSIa`<(3$jFbM=_qi05@RLurX{dCq_uVfSaAwejqxk?rqBe>B)N
z#3N7+XB_77F}tlefgeu3hJFQMY{SUZ>-q7narq8N`e%$jzJ;ktsOkFORww8EE`D&S
z`q_gIS`l_MUJMf(=hU>4@iwh=yv?sD&cEe=W++UXpX=gX5YQ14DK-4WO#`1LN5t`W
zQWc^cWGp6}i;h@YR;`kMUsv&_Y)sLNek3|NvUO4WOis{m?-iZxufvCV@!OHLxd|Q%
zrh&k3f%Eq1%K;NhhD)ZN@=EO%066b>NFN
z0vXWgwU5t(ejC&|Bf|DNl|`4w>2EhM(7?l+MA+98h=@Qm{HUM+MBC6hkCPQdBqVTm
z2;dl~q+|Y5yd;Cnv$C>y99H&rchfR6N57AB7XsPSpx#Q^^#DI5PEbulgOtk(BQp+Y
zTCldaG5LRE4RatRksyAULglEKMkeIYwZFV>XfX&ZKZKzFeWUa?*w1;72|)
zU9m*k*iofdK#Zj_D@jK3kLCyrI-l+mlmtJP^aAe#$pwa9^3lokXV+Gey>60mbu$eq
z-CX3B9auBxDqHV+by(o;A&Fy;0Pwf+Q25v|}pj=(T{0c}#wxdODe8shy;@S^uK``=o@K
zn{--0L&?k~poH&MnbzFhCa$8!_gl7KQ37vYb>6I~x$LKIguJSP0kNW3E{BY!xLUKI
zDBZj+HmL!bgj8c&+T~2xtoOTLBaiNXJL8v9C@mZH<}9&Z%n*^EUU@w@$4@U2xc1EF
z5Sk9WlHj7Ta4tk#rfJvEtcd15wU<7neGCqXIEypZ!02VLW5%eNR0~1ew?aUx^Lgwg
zyqcj0)++ZfXIxj8fTsssL+m^Larl@uC
zqLpTq-&|SNQMr0bO7Rz_V7Gd%*;Na{V8)nyOn5k&K5>Hp@QYiR);hb6ruwV>`G^dC}p`N
zp!t3!&YT`xJXS1bp;8d!Rl^=H%dR|a-A`yXc0P_=L{0No99%8{9-arUOwmxU-`ai-
zZJBYCxxc;)q6qk?I|SUl{QP`Rq3=qqpeK~i5nctJl$_kb`R3rm^>MHZDlwna-Nkk=
zJnEZJAPp-4h9HbBx<7->I4C+AIRrpNi06@!kzf|@S5Q(@10xZg00vgD7wPfSXh6dz>omN
zSwdT<*@>d>-w0z1E%F~6z8L^wnz_wf#?sjV9yRoQ
z78g3!t$u%h4}`339V(#LGt$$4E)TNM)?UPSFly@?A7|m?e`
z*R9PIFng8!c9Bu+aY6V#zUD%-Ly*kgWY$7*6L(nsZyfdmQAXKNHwkLGiW5HgO!tYY
zzF$a`A{DS6?hRE7{RP)&Wv-k|C?E)l`XNTY8HYo(u^96>vTi3=9n_m|ATSz+5!`3r
zYnQk(`&O&27?wzwU~`zMcWI;^i1H?>uL?LRDJ+IUI`7+?NeIs!@R1Rvw;$>0Sp7mt
zddsI65ZbIpC5(K@OuUD~_*`)uD=Amy9R`<`FngLZXlu&9aDQ@AE?T>Xel5$AY+AE$
zbBo|3-V3VTM(PiKw;{jczTJRBgF*JDE1uAk6EK^!7?I6X8$K1xLOe33E(x6ZDGRwT
zll&r?8@hF8Re#mq$r3@`Xz;3QPLA-6*^KDSu65=t^5>nBSPhieFe7xV@
zoLrstDD!2erN$&3$5)?E(Qa`
zFn-g7*e#fmfA17`w0BIEQ|!HgyS-LYi$}ja{MkYqjoyTfwe_W#;#@9^@sv!z%-|Zk
zPvU!o*mtoE`dy#}WwGS((Pc9gT407QEQn#4dc@WE
z(M1It^~&wtiC&=##Qcs*bWz5T_ea{7=4Jw|=iPDJ+dkKkQQh6Vw1w0nys4^SADWiZ
z^#oRR+UsBVxR`-xHp_o(arL^phhCDu;G
z91LIPZ!l*M;TBp~82v31$4^skWQ-Ou&3GbANW^>wvq7XpGnWuvPlDgmk9x%ay%;d(qqMl^pRyRy
z16E_w$xIqcYu4&FVg^6hR6Vs{zjYwDL^@xG4w)HiePZ5E!RTtXg=1x;Elf-MR=iZk
z**3j?O4mqqXkGLnTs(c`IGqqfD3x@gF)p>wkVi?&#Q-A{6n@b5*4Kv?78g%sa7Q09
zI66843%wqB1{mde1+~_{XvxUPgoK1HDS{1&h=}^eaPaYa)9!(7P5DpnX3TanHCgn
zJ5EGUFhP?~5l+1rq@i{`!kDtm&ddAZU$p=8*Yw{$z`;AnC~vTh98fYe{8HQZ)%iTn$o1U!rYJbT1x$EyOCVH_1Sf54
zy$thkS@&y0AS6flbMERytcx~*H6sNj%H*$n`Z^$5W26&VaOCB@&RR{mcklTW$jUCj
zr~g;)?*~|a6wbgQAaD1Gi}dlAQD)D)(&zFDI|Og_2pbB!MxykibOdg6`~H}cDFWG~
z>qZK;hdvxj25c*a|9V`Uj!TwpEsC~E#GuxnKDcA~8mSdUjW1Oegd0EdE6LFUw-sQA
z+MSZBcrSk8VdAx_9Ua)qsH$Rgt#XZ;OSIEuu=H{#DjaYenP#aO@Lrz&$*Lt?iXqvF
zNvXtwm=jB|_HV*fOnSBwE7j1iELH*Sg3Lu>QIVIzHMo`p9r;AH%g+=%Mn^6_{3xLh@buMJ
z)PSCB)mSYJH*!bInhPj6YsuqmKrxVEFAlEZaB1)^dEY#FkCm4;2RzGgK7kJ#5o29?
z?`Ak9`^@~@^o*zt&?9nbS?PRkj;bD8vU?$AB`awBmSLYZWPY=LE2xz7qy)-@zEHqU
z=0n=Bs?t_b-S_I@l2($e-^=xGq|}r&G?U$-y7CHsR`&~6AyO|xCRj2%4=L+pBYt8+1+0ZPaK&4Ddb;uC8Ix$Rwpv{c2+8Ccs*W#&7DS
zJJ+;?7BncnHkinIhTb|s60tUhoKbW7WC`&W4aN23rRA_{B`?&-C%-uAs&OZ(@^w)D
z+kn2CIMEhv;j%FVU*aQ#)B0wv#v)AP2=JLDlIC+n4rgYSp
z4g#AG5UGJ%P|$D=*#XEY306^GpXB^hSD$193b3z5ht~}SfRLu9-a0&Nb66GN7=N*K
zNo)^-i7=pgy;Fq6t;B_mja}LUKl$0t+Z}p3s+;1k2-6c`bHkUn9H1$W?Xau$->RfRAv&imoc^t(_C&PMxqGv>)
z)-NPTpWjakRcRdm$R_QGX{S%92z_logvu{Lp~771yK{2?j*@s#q_3_Xs{JNr0s-IW
zaBkY7y*Zm;uqrzJJ+1(gT-4{wIwPE08-A-13Wx7ag!=mH0kaBuIXnRlZzn7~OA5|1
ztm>R}Y0S0RJ)3=FpKEB#c&4$OA;vt+D)BGGHkBAn5?IWXJ>n63;%};G6PGDi^P)1^
z`Sj4dZu!THi{g?!zUw5ZM#iHPD+WM{p8L9t3U5g3&FuUM&dFXc^eKNtgY86%mqnPh
zT&jF$BVB7o|3vJlQ0ZYvX6xW+{(Gj?|M{^cBm|a~C)dQRv9)yMcIM4nqERqK(mYex
zTwuBWGZWIOad&xvs%R^VecOVnn7mb;=ZYPZuW6u`W2B*p%;E`10yj!%%qXeGqM_5!
zP%+?@sN-!1zRNC_xewUsKep5Fh!{DJhq#{0W@QP&+~XBXbC2$J1-MnNPQo_$}-y^FNeoVg_g)FGva<&aX;{i?6H>lfoj2qv&QG@%vp|p6rjC
z%G?VDn|fZe=3x%C&?`*}#OFpEF7!3>_^Y
zHIdw--P^?t*(&Ht&z?oJkeVi!8N(a_0-9sQ
za^P)N7BPJA^uhu?6%{9->mmTeM)~_Ys5&a$jgnjx6t5jVkXl$;O3>b!rHmYRlT0rH
zxY5-GV#Q%;Y@C-(GF_39ni_^iboINvHXP3@!3K{!
z&hjFF&@*+Td9lF_dos}ZzuXeNlClEB3afGoja;IoT`1k0c~bfT_usQsJFthTS2SJ!hp#+$xwR}!qg-|
z8|OP6zcO*%#f!#taFwzw>qgUB5NwyyL+?UA`r4d!1(5V;u}fBFX#7+Oxedxz)4%&E
z=Tq-he&>DkCVFQ5Dat85CGvE9M5u^ktP_DL&@M87eYZ@(nYvT*nmg398jNrwZxCn$
z9G5iV5m_uZTc7Y`TGOPJl`$<4*#xM3SF+~^(!ocFkVHlhbBhZraaq=x&fv}We-W~6LIeRH=0VbA%%-H>f=CS)A%bI#*`&Xkhnz1>B781}F!PBY>t_V#rr*#D
zuf*{vys5yU`setaVie2_A|oY*X^m+5K?_0}5>UD7pI2Bts|w%}Z!k34nH14Vbl`E(
zNQMj-nH`!U@@yIni0F~ABJk4sJ%5%zsrn=jqgXG0r9>wsr8pKr0~Z+UEQyqf-Sz%L
z;M<|-Qj}`xCW~O!k&e!M*UK{`cp?e)*4_iv#UU#x;}&RHQgB|#s#K;hLVTe=zI|24
zmBHpm`lhDbH*l=R`hdtX`-We$OBZMv$^w73pexx&Ko}cRjkZwCn%g?weasUA{)qpo
z5P6B%Q!rl(Ni(DK-u%-FU2^Qk$nlw6Rie)_cy2;AV%yAo!d4@M471>yOszxxJ0=9DZJww%40cJ5d3d9O1FG
zjFx&RNAEdohh$MJUF@@_pr*32h7DL{BHu%-rQ`u=B
zAZ=KTs4T9|ou)aTr?t6M*Hq=bX-^h(+tx(kaWHjmmshix{5IFB`Tl>7ijk5`7cp7(
z>4#v1^afdmvtNk&%Omzyls)iy=5iH(p$=r!zrjs-()mMRtOJ(GW4p;fp>l7@NX6!V
zHD`__i;9(|}i*!vwyxmSAf!l1eo)lxU@yiYAPbpY!b&Fkk96du5eXP0|l+
zCH_vek_nMi*zUwi*nDqf4uc&ubN+LMV#iJ13}i5w)3uJ4lW5eb`T3i(GtqT`CIj{%
zm{t;cx|Thn1L)T?(5^cE7X-oK2f-{?*MFdV^oFPC{opnoKuKKYcR#?XkvHv&07Bz`
zL!F7mIHeA4VQ{k8>-T2MZ}}-G#Kgn`vk*;fZIN5V?{+ihc3dHVmM_Ha2h@6Gbab74
zL^qQUNV6*!L?})Ji(`s{8X+8nbawlXAHl<5Zf;J<=NbnKOGRBB4w97Vdv|*~q_4<1!ux?gJcK+o!($|u)$(3uKVNL?Z
z86Sv8*$vy6B|Rz2>qUkwR(@xs&Q!^&3eXWRay0xN&EM%_p>VJtX1Jz6B;WOB0
zarzRftZ&U_qId4B>h8`00|{C%>?8!w7lrtG)-76m3OhK|Xrgkk>=4d1Kd#_OXHW)6
z&U(5W;A8^MQe92W=szu~XL+@Re`Vs6@qm@V$l{y9giak;#9T(#rWJclB=!#K8O43K
z&dv0-R?v%pr&qbs#$uqv)a8sO`CXKfEncA^iKBDH0?mZBlz}+xIb&h-{t(k>)Dz#(
zkN2yMT{PGRjaE=pGKaS+c*nH$^gXbN8F=E%3Q|5Upg7BX7%>V264a1jyKWdCeN`k6
z)4zG+qc7IE`dAI)rYtYEn*>HgfE#a<3|0if
zq3~7zNkvR`=}yDF$RA3YAk`^*B>=#^E=#lG0IbL=&n
z%F!2s3*X`>yX+%~XumxaMtFO5p=`qOn=OYVn3VmIoJP-zmf@Qz-tyw=My~BRbWV=EDbyr-h+Fpg!^4
zFi_2#i3P>KmU&yA{gpVHM|~H2Qaw3E3ltZC^ZgjfKP9zTJTYA->W^zc+?n@ugu7CJ
zGjJ*E8&*+YN~dX{p0RQf5%PTLrl`Q8PDKMI2evFopUayMny;V
zP*7W0mk;JMt%>NNj*Orzb$63G4gUFL_g10#x;tZW?<5O6*85rnRMtAJRGdD2!c1_e
zol!vyzR&v8(#|Lns;b0u>k6^zoZ7=g)V!%QU@w9#&nqbn5j>C#hfOZ*S7LZ)aP?-x
z&NM7N$8xYyR#xr@1A8}fTdKTsUEOD2gN5PM@X1nZbMPX9_??$CWP5uXNJ?iuzQ}2^1x^M*zX!oS%DvymooH
zoxJ=QS-@kB>#m}?xe3b{00yT+X=y`r^FVsQr8cLq&VA_N@{fRcfJ0{1ZA7_fx1E14^V-X_ctUC`Lr8dK
z0`y*pU`195$c%vOvLFF!4JFWyv3MQVB;%6TU79z42EPTStxyWghdAOq131DkGw=fe
zgp-h+joKl$rD}D)(GS8rQef{1Iw><~5<@$j)`LLq2n1H#gOmWA0I03T#O3{oq-C*x
zyoLR(t7MyA1!E)wn0+tR!@sx2;;rcBm}ndxIFIJG`h~7=67b(8F1G;PMFd-#=jgp<-F9?=w7VouZ=HAo!F-GO4lQHhPyF!-%LX)Vm;dKY
z|3^b-&C=rwiXC`)1}eN7EXG;sVR@-mf$0P95})?ycgb%~hPV@d^*BLIL|~BBdlG8j
z>(>e@IZlvc-x7h`iA*!b?9)
zsIY1>UQ#bUJB)Z`Pp4#ggPv2}>pU-x(MprmM5&k(wcn6^mN;2#>b+4`Vr&Zuir5x(
zO#S&oK
zyEJN0ky>IJ=Y1RC(r<;`fI^G34tW&P#gLOR5G7I!>yD
zr?R9n>VlSqrosQ57{}4xLCysNq;urU!}ErbLh4?LdrHTFufsL*c72~aXMH+4eGqAy
zt@j0+*_^FlwVj_}d^%bJyhaF2SW)WV3J_rT59k2It2^$98M&i{n}ivAw;JBE?=k^}qYj
zUkRCn`F|t~Uk~he#a25{r#|q{9XJ)o#%7FdAW-p9H?Gd6h-8=}YbGGEbgV}Gggr*t
zv~nXjEJQ({jt!^`MXWRK!ImQhH8Nv*S)is*Ek)ofv2Js)7C}6_Aou)-whpVT_a^o#
z()as?v~-QEO`I+ZMhfaQl`C;US@P#+RtS144Lvhjr>CD{U@tB&4#pfjd_Lk>8yD)v
z6T=nkuLn!~K)l$1+60u^+$%+#$*n97$OB*W`X{PYBz7V3;yn6sEE-vPH946}1@`pu
z^!&+cZ@;K6$z750H105wvqILF6@ltJW-1Md&`=%(+v$w*(9I8t^2(BAh8G0|-%F$`
zNmMnB?|cOE&MvVw921z@o6PB2NaAB%8OKrGm4iST^(DxI?m~g;js5Ojl!Lv4L!2NC
zxMzU8iUR>4^H%ex0Cg?>MbvjJ`WHy`t*w&JdB1(G4nhS>4w=p9&
zHMLF$`xswjRo+LPE`OnmVF|l!Q=@KK7r=4@dcAj5OiWKt5HK$MMe8Z*rlzLgdw$*m
z*|ZjkF+8iKZDN0bt$-mFSg$xZIIyVbEJIN-Fl7_7<8p)HF86;hx=5GYG!?GUnE?qsx0K?mh=e3CwMdPK@aapVU`i9rs_2^L{^UP!Z
zzh3c=6A}GT2U3$`F|=aATr-Lr0rvKX-xqG}KgmEV&RXeuTIghny?KAG+~!^HU^sp-
z@9~oN@8OCQ)#JSj?d{)3<-V8CqRkGYt(kO(U0<q5raeuxH~=U(;wtd2?Ry;FZInT&fHN&rM__EBQf}ZR&%U
zL<~cVBpe(<*XuaB;Nz4Cd)`=_yK-YMrIRrRwb1F;eHOV}ttlg-`STqCw;}_x*Y9fM
z2`{FUReyhGSR61>+$(@^aiHin7ibgYR#d9+wy)~ZdtlloU3_e9XyvH$n>XWDeK&au
z2sGK%18qB>2*Z_zje|c&ec6UQ7fZ%js-$EkG`rnkXq^uY$XXE-_|a1DV}9r|@q)Dl
zAa0j`-S(`II@z6>k;ar;?Q@W_K0I$PJ#tBk6Eba+HbKl#(gIS9dgcCq8
z*lL0WU;!?GIB#6-1(w2p8HAnicEF>87?`gL0JFxycq5jRq$9Qxmz0FPvJHYQtOo5?
zA2nn}9r7Hd71Q~h{=On}2@bgSEJAO2)7$8MC@(q6^#TPA!20B*DErUP6{BN7%9%d7
zX6d;--@J2z{7D86@CMo6A6$s}Z}y8FJ~zCdxNr;PTn@!=zWcgfx{{hjJn97+AC!jTJo?l^s?tj{>FDtv5IFq@An;GaJ9q?6fh|!@bgUow=$CN#ZC4!SYCO
zpod1NfzM-j*7DJmfN1lFJo8K8Ed|AI>w#y}NF;q7>b1K7WZfS7^pLcxM%2KYVy%pK
zdq&6}T*$qjKPt-QcSOWhn%bArkI2;AIoqT4z=yrh)|xdyfk};H2HkDg<>~wk9u@{h
zcgq*`-8UQ1v+i|~K9I!l_4Cv9o&yQQU(4OcM*BgDa}vlE&3)Y%M_&K?!=jqb+@70u
z!9PjPw0c39GamQ#+8Se!c>A#q3wQ@JlOU%~B*;}S%R(}FRkZ7Q(SusIq8W2z7=Aq3
zSk{trh)w3oZyRmHGMpw{$)PiDN|fjYM_P6;2O2|nzCIc4D2+}Z>duNjrlF`U+Xt%5
z*YK#za}n6J5xqeEL95;HtXJE!ySqiV%!>a}g1t7ElfmaXx$d3G%Y2S+KjPyEI=+RO
zT()jI`c%jN?XdgsXk2BUr1zHPq30hI)^u&+YcRwyN4W3s=&&sfd@?2OD}T;({)4u<
zV|NOU6u+65eIIvRc2+Z2Grb
zl6`gPZ?=5fCWYj=rih=_)Qb!R3yA-tCwQck;j*2zXYa`9hdo95eq(BJNbxz6uh~8r
zCqVwz6g=(TJ(zaU#q5nT%xC$H^vew@ov-B_wN^dlHZb{+d>fe6%x7`?9<Wy(V*JJ;ER3Egy?hO1aeE!gtMn#g!mQj?p@DB``WGqh*Cw4TmE>auiCr
z`dHPcc+tBgCu00>xD$5WiD|(r>(Kh&L0{No?B%O3O5NT3*a92I
z;Ad+B4{KA(sQvB`w+PK%uq#dVJ0<5@k4#z~J=x4?^hhMjSEW5GYf37blA_`)QWCk=
zr+uq`(53k_7t2AgZX>q=EPQCv2}B1zQ(Am{ZbgN?y*+BMg77z}L8os?RTcV(nY@C6
zf|V70@J|FZv^B65?N$1{&#lUdPqvW+v4ELPJ0D+p76;g#kuIHs8+}#
zYqk*FEV*E{p<_gjnzCf`tW4yedN%_zD-lG*b71zKetUAi*rQYdWQth${K-MK`C44Q
zp6=K1wxS7d6hs!O^E>0Pk5sX^G*U|(cxx$k%D+EGD-^#ytYLq3fwlUVJp>0Ylt?xL
z>ATktinMC~u&BfQ$bOD!s)$8=&^M!5E>X8&pkG{5M>le!&m?Xv^pK4A9FHEmk=WKb
zLqOEwze)z4a>`9MJ9VbZIu+d?4YJm(Z7r-Ua_fw*(_4TSUqlV~yG6HRoMJ%yp`czH
z^4fAeVlDrcXOi&o5o>Ns8XqYs0^}e@$wiLj9gz|
zgSi^;9@_O*3BVa$@{DGJM|~%0mqbFM(1|i#yk&F^0>21NCc>oPwazn}bb{0i;b{VY
zpp$0t-rL%iJ{cfe1oW|o1zo-oEe);ZVfnC`4;#z4q^UM+N{I#n3xLL=@ZHB4b@Qkn
zTwzO+)Z{kKBlC^1Diq#hGnb^h<|GfDci8k?LxY!Yn_Sk$f2eZ#4_XaxDdDzfU=nft
zY>W=m@7EjY#R^lQ4T3}DQ}bEMgoQ*PF_eJKsh;``GK4@GFxu!bo$9!XLg)!c(^{IC
znb`aGJjeT(2fofEZ8Qhi_ZCRs{S50R5GW2is+QM%f
z4F5so
z;QAo3I)1J<&@+=gcey^4!q3lej?}JiVokncklVKvT+P
z>+X7(&A+7h*)f|;fG6F`RGgf9oG%lKCp+Hm+_yEO&N*!?rxw5<8~I{+D)={}b!pJB
zhc>YKjadyZtT=+~LKHa%j+A+61am^so7(Sf4$f+aYh@>lm5e7HP0jYUe*zvF8`>O~
z|GcyiH?PsYrg{V44i>22X8E|B@b7*LS$XjN7vsq4uXu!gz>0Lb9eXhO$v*tq+*I0V
ze?BQp%lWHk(<{tB16ix48^`VJpER221D@|Jf!5v8;isfD1(wkC_ViR5c0Gf|00tf2
zc`Kg)PJoQZETgEn@zvS7MYMlWCKdsJqTAGBr
zQ&sy+oy1ra!@l>~*DQ5-{YQei)-G8dJH
z@K41~XQ?y}6TsgL2O)7qF<(Xnv7ormsxgoW{Wx*9{)oz?j_?ryV>(%+>X$!mC+0jjt#L@*Mi!GQMp&BL=#{DQUl77#(
z_koDLL%*^=2s-Pi*Ay2I6pYff`m|@G)i5p!%-q@J?LMhYN81cvD{Ij&`4r{H%?WK!
zw21`2{nt3Xs~Cw-KZHKoPr|Y3Q@>2%(QMz?TL)G(XQP`n`BbkeUc@_bbI-9y2Afa@
zO?4FeZ!_nQrexEkK97CqE}I}Sw-eP=FM8}k_R;+}e80V8tW5z;S57R-tBiXv9%md6
zbAErK^FGD$(YaRxXU=lvhn8rwLj%*rui4qYr)sc>#wZIM
zOSxEAuz`dzWoc@P94pePu?}pd+|$$3^96Vw*lh9x?AM8@U*5+{r00T>V(8Bl8XAXX
zkXu$+0q=#3GM4UCv^X=W`>dFJ=qT0
zP}zwp_dHe8N|}(3d_W|YwW|DJ1~Hx(_@>vv_oH
zzkA7dDFWVR`A197uBRO3r~McaRk5L5rK4GxpEYXD2j5{zMDXV_VzW#6!yMAimop
zxcJD@Z{KgiEZXOvRVaX+D}OkTprpKZ>gq+wmNa?2wd_<|wIr6hA07bn76=8mJAey#p9>+ro(0b5~SxXIK+TI>er=5VLe!44Er$RqEtIRtELN`B%4
zn?W%j%q%TAzsf&itiNNB{0A0g@IfH|ZRR_WY=p5rJ3s&V;rib{TPlrDla0&!241z5
zDW1XkYEy%6c*zq+p`yy-h32F5WAbi>#o>C5s)QCSpOj}Pt{1-@#4*V2SAFWkp<$0+
zpwsT*n4arKpQkf!owF=Csi-0Opvnl(Z+{7=r~gBX$l9aI6H>T}rY+M((jWZF)R3HI
z?UNGt{?eY@b!opgL#B*Z5%62=cD3S+o{fAMw7>329@SBl5YGN(abw|TXd@wBA`e6r
zgF>`Qwco(W;bc%a!3N|7$VT)Zi9t{7-brm!wVZ#IZOi@*n~)mTp*uyPwpE)QQ)qGR
za=M~Cc5b|h1it%yBup9JSYjfHZ7SBo(Q-RdrpX|DyHrjkoq1C*M61G(gOyiE`;#XA
zomn=dKDx9(B2Ve4oCWhAA}UhfLx7$>0RGR%S3e4of6@r|MDhDJfvvk&T2^XRqU_6j
z^?uvuTIT6S=-w;=vXWxA3$u9aaWkx<*p?k!wrk|^&wZR!<-fmq#^j*@t?8iV4<^)~
zL;d}7OS@11g>}D5t@}?&lbpgmHd7b@|7MTH!CEL}oFeUuosN=_ztW#qdrC`NNug%0
z`Yg*ac?A{fn%r^!$JJX$MfrYTph_w!U4jbIAzecwpmaCV>5#(EC|wfLNOyOqbax31
zAl*YZqWAIhy}xzWy8h={D$cym`QT*gW;TtgOc?rNHwcOR{jl90GWZ!fqfN
z$@2dD=fP=V46NAx9Ew|RaV5CV&CO*e#iCw_u7mrEjE#-`c(_U-gckwD?U=Q-HE8E;3^Md4r1sF}73+TwDI}?|JK1aGg|Roj5*yrlfM5>WB`!vzTR+5`_iG#)
zPg3@UeM_2~5|<>U%Wns@YCSj%()&tt82@`)b3S&56NIr-G{
z!;lS5$|ve^H8%L*iFpo=Wq(62j1X%Qi++sCEEp$a_GYLLKKQglzWNm9Al&46Wo^x@
zh^Dlz_!XL4gM($xJCcnY0i<1hLZ~nKq_Ku*3AmAnUheh5AOm-(sPm7Tytga$>$%vt
zP+i;{va+!2bBsbP_#GEo_jXUND~^Wy-~yfwJITn9lkW9a`Q1+sJ&#cm>Eelq&GL>H
zDwp&aNFg4CIDU;#VvMFGQ0P~dO+%78$a|k!HwcwKJC@s7KKA;Xc(fHOb>slgGxe@s
z@ouOfbfX$^KKKBYF`E9g4n$h)cwDLTr;LKF?e=y0^|25qhJA;f)jQt#NNUxjgh%x2
zgVR4pShUpC9!~ckw{|Q2#68mGMEv=t&={LohHAkdfKxye88qFEMBA|clgak={#7r0
z?=cJl@T$IajB-Pnf9!nXFzoF7TqPKoIW=^`2pPeWivxAKkT=+10c8ZVLi8(Z{VzTT
z-~grgAB8~KJWf)q!VX5TMdKy%;`nLs%F0Uaa^%@?vGKgp6hjBjz@Xy?aD4~8L3iQT
zdTHm6U?9yVQ%og5e8I_3!itV)h@}X*)mg3Q3pY&@Q}NI}_(cwXHRLZVsxc%ph_87)
z6sw`9C#Q;39Swun$4M%LSjc{+aMtGQ`C~9Ymiy$R0HgT=w@C&%(Tc83+zj#0Vpavk
z57H2Y?6wtwc1kS}M(+Fr1O;#TJUg
zzpG?xiX_8S^p!$|RB#E&m=8%km#b_%TLc}g#GfA$>%UKSgMg8`oyRAZEy;{Be@DAi
zp?8PK(^Vt#r`kA2h?C4n#*s
zhc7Jr_nrr*JR6;P4BM@=E8J+QOnd*W{)~{;z$BxONmM!RM0-e2jh&~!xD@=m>)Q0$
zJd;oG2ncCS(5d0K)LOfyc|Ys|nC&0|I`mg{)c?KV>HV;D&HFCCm_XmTDgJ(q(6LcQ
zbNDc9pX*lRp)}Xbt~PBZck&2JfrlvRC>)BbMMu;p(@3Xiq&q=)TTs=NQl1x-mo|F#
zuT9NMvo+yi#;>fXo@_r<+Yp-e*48dwz`FL_!WU-MAVu+@8pZ2zOIl47tENQVYIt)j
zFb@K+lI%aUJM1hfVPjwX3PhwrYClRVB*&E&ro7^@?uuprRWs?xrrQTM9g}4h+y#=J
zPF)klPffF&ypAu!qHM`{V_NrUdTi1j8Fv!xd=8vefKNo-I-p%2BI+S*=
zZbpT>t1^Dz?qSpD>ZzVs+{@`Z-$3C))4tc6=4d=UmOpqAsWeye5gk$1sHwkwC!hf*
zR^-e*Xu|}1DlVtBPm@*8_Ts2bGQc=gULiwp_S`r?;rwH59XICGx}$4Z*x@&2aA@Y|
zzqEt$!FOFzHYn#C`fQ(UhOgsGd5eke;Lu*)yOW8b*Kn1fvmbvS#wRBy`}!<^;VM|I
z=HTS?AgTYfHx00H#3I_8th%Usf64r0dQUzMaC6m2CB!n4Kh?*g>&YPd&T|IK`WP`E(I|Yg{K}2e>d3L
ziK7}T-IDSp8)ceul;}*GTf5^t-$>*%-6GJN1@CVsA#EhvNVU+UlwMj4!GLW9La6Y<
zA4gs|V#Bn}Xv;*d8Pxj=H@v}r1%-`4+-F_E0B{?tr0vYHi!rLi$TXf)%X+4eoimR5
zn)X1%Gf8S@HrL_TNRez9FVPHn8YgkW%NGiIw=krA(qMC)tJW@`R!UmMjtxt5WNn?q^R`@j-`#+mmC^2ry5iA2)*>ae3=i&$
zF$N=lL)9-~W3L-JkHhi0di~5`L!*lRMO!a4pUked+q|aB&ok#jJ5&;_J(9=L+}-9c
z%y#3Q*cWxPOWHG+-D%1j
zS@I28UTD9=J4N{mlLyeYH~!2SD+UpCvrx*{W_hm}L6o^D6`bdtBcPvo1v%zFd{%AL
z_)J0r=C&`A@c4ZuIVw6C?anqg_DM?U*sBuA;$$DR2XPRQ&FWBCi!`0}GLHL
z{Gs|^CPu&2ooD*D5s_FzTybqXd{RDeM$E6-qK-}Fs7!jlW>G%dUld=`2YDc*OyE^;
zsjccb=0U!@@&4q4TU(3o>G6`3J-4*4SqZR`40gF
zMM(+#_XQl76#xBm{bSktFK~i>%Pp;1X(TYwqLE%IxsYBiuUnJ-n!VR&={e#tqMjAn
zpU?pt`xpJ%p+)d49xL{OZ?XlIR~HHp>NX_{(8HWVdl5H
z9_@l!xww0Bc$W;?Yr8ZNtnj#@XrHwKD%u`?gX372mU%h3H2j^;By#J(u8|eSQzWU`
zDdpk5cP055TLCZ2$))mmNALxAP%AmGrL=^n6U}XN^RmzIH2gVL&B(XSLIQ3O
zbVHmuijOSUjIzth6q{66-@nEs^5zgOtne4tC`6SvmQX|I(K+JGO}NX}rDk?|gORUE
zRX-gIyKzL_t`m$Kf8xp+fXaYFCYXYd$7t{?@yBM|9SxMx33z*z+th{DB{K50nrzhu
zCFLtS$>Gz(sAxI4cy{CK)v!{mI-?!iuXPu79L_by{1%%
z(C{yk{(RKk!B6SCEc#B!%2;!}GmM{UXU#lGJQHpSLNJsqD?SYAL{ZfL+4;RQUpS3=cGb+b`;9gBp|PoV9)+Ue1x3Blk;Vd{90!XM)~IMXJ0fCmpz
zTh>ojrwX6Dgyy!$xmrSAnke~vY1rL88tuP=Znp?xF6p;m^6qs)WyGVhq$Dfj&_Ssa
zql|la{bo7{`X_?=6+_MalzSoWOdIQ^!?^FmZ=TrXa1
z>r%^pk1tLJOTpo>!Tm%g!mP7TEG4t)Cl7sjof5_M>DP&-ibNyoAOQ&HVkJ$b#kHX4
zP;Sy6897vbA504lRltGxFK8n6_+%9D51~PqHF8`e(d=4@H&Xptv5glj-2^$88hzVxfc3idQ
z1=|>vVl^eLFK)qs>y2D1upFtY=(
z<-F4-v%!1Aj;Cjo(1sswarVeVc)t=d8^4Q^TT&zlPN)W_x6aKCG1~VkDy&o?ZZlIN
zQZ5~6(A#*wweYzF-{C6g_3dQ6YBgu4_$>nFF@=*sdGJ^=ZRq^VT@BMyjK9#JamEYTaoiP
ziYxmvrd13-u|M@JD2m&&)poqJ3YI*?e)iJFmaMVkWp#>|+gA$fhMdgjziKCb#uk*9
z=jRpSy%IKL@sQr-h9mG
z_Q-h88f%kmW7Ak`TE_%)SsNFP4HFGMWFXy=Gs#2*-}?snibB@xm2jAi(+beN`^?d9)QJhp$oG8GFS}mZ*zCOU6XwX9S#5U$>%tD5XL)FW+&5{
z77`Ld^~@gqUoTG-R=C)>Bm!U470(&Q!ezJ~&
zkKI%_z
z>T`qQASNwCuI1oa?LCD>gmdyt!X^cUi$UvdRgMfT4jBWH?7-xhMfcRL@Vp&5wN6W-K>0{kp>cn_LLFL3IL)MCN+N=FU>(I|vV}vA4!#A#C
zL~;T}qB`nZmJng<;4Gy#dfrO}tUvhBw*}L7PCT(}3WAi4S4f=V1NkC?`$8MELZZGxBqi!00Sw1~3s|1S;
z|MMUX%A^6y&+D?jl~Iivh-dcujp+a(8U5-YDFCSp=bzmE20<(3$p=75ke{CqMkk6D
zPVYNsyezOO4W=bwRsE!yZzh^Q&2LP0z3{IEp=b4kt^mmQt^Ypn_u;vTOTiM!__@eL
z#A800a$ij;(F2*;KmUNQ0g37~HdDxHdOnzXUw>OTb1P7$mn4dh8jB=ZukgVSmr|Kf
zEy(P?q|slU82^rgAhE?k6Ayq1MI@OgPbu@w
zm1mng+EqHCy`_d<4!IO+ZSQ?v_N4+7BFrtAmLsVX`FEQp1Iv+rQ$}M|p%ODKMhv=(
z1`+Az32DPih>bLh$rjpHO9@4qkS^lT
zEtawRH;#ep&y3Gq={qx-{f?v&I;
zV-(B>a_E~buwfGGL+rWRH*YVP=Z6qErfN~MUt}Mq9O3^
z5f$>HeZnn38thmW=^Yhk*LWBA4|os@=y((kq+@$
zE3O`lN+cI3Mx0-)PFUbpV^&TEf4)+|-=2K#o54e={$&M*_VatSrbd7+U3jv)uRb#k
ztDGb&f1Eiamd5rw!A&!H#<;!3oBnel8MkELF(mq~y>Xkx4OBGDF`U)RyGs+6StG^a
zwd7L?Fy$&gcBqd%>aJ#==o-JQ%I=@_eX(@RoW_d-@A&c))7ud~S&W6&e)WN+0)7N(Tk15BfmdI9VKGYc{
zWN1G@?vuj3**>BLv91OivvI~%n#B5+=q)i-rqSnGABWwTZ;!h_rXLACD~`k&gVmR7
zqp;eIMJ^{I7_2mqitQ@fM8ftk__ln7Hg+^(=;KX71F39Ku=(D6QBb*MkTT)rjUJu+
z87D_27BPCfEKjwhqmHdXMCfK_^V)*Zu{u=f`oVMhJiSVBhE#{i0U*h
zSBaQT@2{3+MO^qAhYt%)F*!N9#kXoXcS9JYQf^LRp(3+#m;A+bgLJ#Paxn2coZt8T
zF%!BetGrMZ?}hw6FZ!0lHUGb!?xNgmg;Xzy
z4*66d9eWYSuXwFs9ZgLydi77R&`EvwZ2K(Wv*M7k%EgI3&J|4-fstMU%mFuufj)2!n*H(lGg$J=Qq~4O3FH+9EeUaPJ!=jqNS)v+
z`i4CMa=L)^;Xgx${e27s1us;)&4VYZzoxPh7<-D-g8(o#?aG&QbbT{3>eSy-MZ9$r
z5G?wQ&lGo@nkEoH)?>@EVLJ>Wu2!A3>j~ji_4Pml;@&W=5vOII<9#NEMp=Pd+fk-YMfN3C_C|17{~?k*|Y%-*ey@uP-$
z(j~)F=8D9o_;0)nvJMP@-2g9_v5xcTpYpHgfhIF`wgD{S}#o9RIX4;e`Jzv%1M!rF6QBl>T22#WFMkvrCk%
z9BD$DE_$$4p5KI+#}v%AYa#u$y>!m5R*?5YRrT)Or;#f?0-k~4!4WhH$;E6}>E@nZyk-CX;FOK2XIkg4Pj
zpj82y`6&nSZ_Jp}(#*xr>Gu0!^*+6w89O{$puIt0CPV$+&B1EW#KOIhb_}NZnx8ew
zRpQc;r~QP;rhkHN0cW7KXcpL8aD5yuru%BnHTn!+py`+kwDkiMB62(O6y!u%`T18T
zQ%|iL8mj!fuZS_6rej^4YZ(c{%HqxU8d=cwJ|cSyI?p$_6D2rmOXfhmU{d02U!PA6
zXAHppcr7qoZ656HM-I#T5&+CzrU%E)mh7!Bfq7nCxF3;h#L30t9@RL^;b*$-;Fu~q
z3c1RdRrJTet`=I1i2RZw-N~QXRr)9%M>U$eo##yNC3!!)7MIKvfB@ngq2VuW6+(DM{CHF>jn@
z0bGk*FT2N3B1U=}sx^ZBn{O-mT{eArTZ%tjDi-SWF@Y2nFAG;cA0>s%9O|7lZ
zbE;Y90-h2aA`oj2AZKB8XecZ)GUy6WET)F{&MaSk17H80k>rj2ec3!!6>gza#7L?D
zfruwE8dj39C$D^`87D|skooU_C3+UN6z*3P3f%!(Q)LYe1zA~gAvbmwD1h5Qezi%H
z+EEF1SWLk}?PC}N^lTa9mu21K(J<8j1(izLdQpnwaqnZ^T3|Q%QP8f7={w!!p*#kV3HPw?O^)fFYbXY_KL(^Ic|
z+VB)?PbLXg609ovass?pdnoN?IZ6JeH*?aug5+_Z6ky`XKF}KvXz2xAHKo4M6$4dJ
zRz7y+s|B2QKSbI_1WGe9mV@rH#^<*3>1p&5eb=&{qzGzto2UIn~T
zneaK9U@NT9XD8T?V#^SNTFUmKaNk@M__^*%fm!?*soEm#HOGdpj3X=Xdr$9P*-X;O=hpb9PeE
zR|n(LEht(hDx~ll8&@c%ez^mun;pbEfphJ3k7xi1bd43%Gyir8YSO?0siZgX%}pY)
zbmW*ZZ1cz~FBbtZDm}2b)Ogu!fF=g*rdqyy=fW6hw6ZEHL_zcw_ZiTM8Ak(K{UHN1
z$F%>L<|lw@X6%9;t+ZhVvma-Q`_DspXF&3fTy1Vc=Z=KRKv3c<{$kdoyl$`!uV5*>AfS#$w6;
zg2POj=Hiq}g2hk2xaO1hhCw#x^=OcAyzv~S$@9>Iw{PD`t4yf`AYL{Z9OcMb*Z&cY
zjba@i*2iUgtn0=??qTho{l5{?+yW9ow36RUuDz{~?mzgASBox0`91>Q(Ok5|O-ao$
z*3!2}6_`fekZXI8H>MC>yp?{V+?k4_+c%E0x@)EEP(K~bY$bb^oycNA-HCs5XKJVR
zQ~^!2T;2wlI{me3hkh@+^+)8HnPf@E}bdW+*U1;=Y7??b{*wtx={}qkS
zz>e$ri^uh5IL@!{`Al)!=&*7&1G4lqS$=QbQqK5Xt&UAO>QwpP!F7!bZC7S&dSI$S
zz3Zo8pwa6QFQ=@Kf$og>QWie^Ha<+YIqoCkhZlWkC)PEJE{(%`iRbPAe^dB>y!Wuo
z^*3n2c~6(vC#L`+FicY(wL>lSs%v~B7kj;85UrMnCy3thWGousHC-t-n0mG((_f!`
zwNp^9#Qa>?$)T}=9ydIc-dPeFTrMX?8EIEY)2y>MS(s-kojEkWN$Tar%=YOfQyHA@
zg&KS&b_+pCOdp#C
zQ!gm?Cm8zN2KJg*TUosYO$}&a#`$j%4ZncB6*E`_eEIUxN(fL=_rO4KlS(i+ePZ{(
zl;!&m#NGu+C{Sl)Vd?2_sH%FesMxmvl2BY%SxC7G@_*DY;*4Qf?`Wv)lh|$3TNunuPd7>KDz@|B8#+_+yr`4
zDTH}Hp_XdZ6#L!TAJ^?PCkO-t#Zd`qy?&>Pu`$TvifJD-z&iVFxtA
zoHvf+b;?)Jds&lu&{^k%owS0nOYypE_u{7`mT9p
zIQ*iQza@}MnPs6P85Xeft3M+tJ`BDkK!uQ|Qqdpgo{TMXC!GBmEs^QNHBL6}U=>1<
zq)thE)8k`Z-JqZ(tIchSA1DFth{h3c1#0Uh0WlCR8+Yba?7c`gb#Un&iY3IIY;b~;
z8s?z7^i1C_Lx2UK9mFw+fq2t}Bl|7yX^|!^rbYg)r@60!l}cR)Rx5qDS7#-ZXw4v-
zIIg^Kcl}o@iDNB>O#6olBFV!t_YMU;;o8T!d#lY8gtm=eT&~?94Y;Mu@^U8*eSU5S
zRv(2SO%K394QtQirDV6K)=tD3apO8PxSRV`b=ovsUjy-vnpZ6CAOuhMymHarBhKIjx+H>5>irzOoZ~kW$fYStBR2jMeU>YZ
zVFHZCW=-iuw|n+Cbn*(-l-1Ev$qQp)DD;GK=nH*D$)EgLGf^?mk{6&t=k-t-CAmv?
zxz}b|bhEvObxL!y)OZ~2ryFz>DQQJ^mi?Gq5A%F-Jt5t7
z_^<4=J$O4?b2IuHzpu{$Nicg~36o*LULkH`
zf3DL74*)U~&p2k-MHlDt8M|23KUcnF^bJN1JkkPqOWBt~e
zg}`MB$CVKNiVp8UeelA(sHC_<(<^<9W^S?sU@7%|i{ZhGcuL>k3|eNt@>8TFQs~Cl
z+i}o@V-V1yln|4>mvB;Sm&@<1gjTPdQUIq#gLWwy0mK{<8T+j&{BZZBN+PB}>!)r3
zseZLLG#_{V`)oC`H7-Shs_T3EFK6j{#t%xkMtwboYGRDow2=YUTz9xwhuJ$NUe5vi
zu0eJ7C8ln0*Q@8UnPHf@Ia!n3m-;O9O+n
zbotY>Gt~e@M-#}gx&L#2JZTJA@dVk`HOUx+v-wN&|DBC6)exduBSRLQQK3!S1sn$1
zH@eiX&b7Ho=!h&KKV@5u1D3=-UF?I>oGNxS=KQA#>6G;IQ5ok8Gg*}B_sJ9TVYTrE
z40}vnFitN43lH*}5*`JuYpHVfu>LRNR6lPv$}P9ibpHpoBqK^?o%=XJ(%%Al;{>>DS6)+9J*#DB-&n=vG`Mvx^6u)iN=f+}g$Sj!9!PmED
z6={@7fh@P0f;g?53NFt}Shez5q5mz4crYd;@xOyzP|udm`$|nV
zH=DL1P2(|rpA_0J(dwddo9D7KQho5f!)nWMDi@c_e<42?(@K|@mrP7dEdU*9Ch~VX
zxlXZfPv$ho;Km5-`ulMF8KasK1nGei50Z_9g@p7$f-=)znUK8agoHg7HGH;<&LBDm
z_;Bqy01fRw82B83C^hD?A635P&5r@6w>5|PN@JkdG^{pzD!V$%>YPB;4b#MRMU}E%
zF$zYH_c@5IgdYqUv&n6wWg&g8NOU6!He!2Z2VJaLe1y7ybMgMtB7^xE$yF7K6r<>X
zAF12|z%b;MOdU(7tAy5G@)5X+-BD}Z6fV9X%zK-)aEo->`xIX<2;AIpjbnEle)|p-
zjMk&|GGU09rdcuVnmd@i8NefA@B4G_(So`^fZ(gL=T)i!gfTlggN}f&{{LYi>9aRz
z5YpDM)Pp{^4;%Nni!vq)F&#g@ekU2|#lP{WF_T|UPY;l9eqHsi=-m7
z(XSyg@CCgp4Ja2ivQ&`n(@cM+OHHG4{1bK4Z`d7+YB>zuYcwx2*9m;~9hXS>`Y5>Hv~sMdC5}
zg#V7yUif%cch3|Of%GSb^oWePy{YeN_}xzm{g=$08#=LIC{$(vC|r{b%^m38e8QZKu`oXQOz9NsP3PY?zkgi2a@ZK6EY1yd
zx2b-k2**G&LdH(`4@b;YBf;j7e}(5lZg?z`G8X-9S4fD*gO?r!PSdM-cN$FJj4
z`l0^pR|w9pHQqxQ)17Wdc~|AGgF-gV;capiRm4^{02=32c@0dXQ#giRM%APRK}DP)
zfa)6iPG}cuc({yTfucn430c822xj+)0$r1`3zpEk&}})tpdK9JxXD{jXWOb^Ta#vI
z+i*U8=?~Dr?cm3@Wl15K8TQZbfYMLY_`EE2$H)L69Jk!Npi_C+$@mQNPP6&CmY3UFTfuw61A?)#v%99I
zRN3Gh?ChX97Kio^56h_$c=~}J+W_Dq`>ShfzVl-M<+%O5{Thdx+_uvX6%gLbBs?H7
zm=t}>Rwl2oaG<9LMFFGc*N@K`|2mm*dNdRW>3Z1IOTJhV0_60$JGt*&huqAvZ6o3;
zvX>Ua)s)YJ02l{U#n7E!VfY*GV~xKu#|F)hC3}6tJg>{U6$hvOiz?xP7F=r?Pheobf+Y)`b;>`D;n)+GVPR?GlGXx7;J_CCiOmn^R_zu7|2H
zTbJ<1n=A(`8CxHP?!eoB{_?pmtNJ$SiGl-E5JIQ7yh}X$`_T&JYdK
zuU#bWfl+-BJOz>nh-Fmm9HPIUPYLnM=QO;HT@F6HYuv@nWhDakIk!F
z5dVA(92B-k$ci|&*m80(y~5p5%nLK$c^1F}0va{wB~Yxajs%l|rB77A
zdM1rCXtbuSNvMs8HY0t~+&?xk`<45eExZQ3jrrNC#?5r+n=
z^MQREMWP;<#}!Y^zf~o~%FfB{=5$FI>ex)Tg5#vA%Qu3@
z*qhh6?UuCE`UpliJ}~qwJRHLKdvr7FRSTn`f1q9UJXR>vRt}43f;tt0@+|;Y{3L^G
z(y$fkpZ8QUIkRd=9l`{KRuuUbjlbtJMg9=O7gos4jE)0j^!gm98co?r$?lySB&WyX
zRTU?QHT(^&5t4X`O^~(
zX~SZ1qZ|+Z-_N{(aBc*vXP<^Xz5h2N!bm76Qa}Q0%sv|E&^4yK=E45gdpgN~#qoGx$$#h6cyd
zx4@`__|;0jxm_gL;}ML7b0CA<-so6vIge;9OsL!=?+(ZxUOxm_t>CY?FP_%1e$f62
zDmL!4m0#bp&(O{?bHoR+jgDnqq&8p8`TGw0b=Ff7VQ-hskNHm8y-i85)QGvpw=eKS
zRK{m%J~KA$PQ#*se{GPAyU(Kr-LOx5Eds>PVD(&
z=<3if6qb9+M|WOd(EO|7A`oNXwCcn0R?M{}#spG(Dj^X6R!8S8(+()L77Z^bs#Ky{
zJYTsbaJ=9#rdBFHYM(qBed^_3iK0h;Cge|v^|fgK#3s6}^HfM_UB`9kbwZFXpygF_
z^6o3qTDS;P@2u-8SS39jzak4h-|SsOBGD2^qOxR8Z3gS=~H*3)RPkwv&iiMrsb#M2fy%OQr-tPU1&ErwxXmwrnh@ygo
zv6m)mlDWPA0=?K)+hJC6(R=hhxOfa=++t!z9Pg@XMv@iW!Y3JLrLunL<=?1r1^z(2
zZFz)xaTENwzC@TTcE}YzI0lkR(|Jf`h15As*;N-(j=`wnLThIjiek2bWLw%qhl#xU
z)?gggJc$zwA*yxansxG)3{br<-Jo=e3F^7V)%5CU
z8x{fa?eNnzTGc6=;5IU3wM{;}Z|VPp5hi9&+Pc`ER|+
z+#Jw~2?J0xHTmyd{Gvry91|-3i8fuOKqyedJ`sblJ^gEzR1kNIk6u3b57=x`ZPU?~
zHNWR;3F>6w%7#AJRxT0ud<7!9G{G>$1+*ot+zi5Cf?M^kuh8aR9+_D_NX&}}>3`05
zWp8i#{BtNtinXSikk52AD&XXsxfel#ZB8OLH}?t&eDCY?-OSvaFR$F{^Z&3(X9}fL
z(DQ%aS{Ww3?;yEe^CJ$84Q{s5vl_uP`e}KBWV;(8B-T<9pPpRam;DOnur8L)@!PU<
z_3scx{C`3ad`9yt+l4U_ougM*XpJ$En7d8iZJ0{%pJ(|q$J&@Kji@$@pnJG3tP!2!
zD8E2-3&rnzI~rBb^SW>stf-@GHwG&5e#Q_o&`ZaWczKKVlPNS7<>1N
znfSs3tlZVHWwA0I2Ok6pa9U{#=57dhr6M{wjjoMEuYrVfJI76&^gYF!3Dq2w`3Yls
zEqXzqx1yBKrJN?lsgGUW6lSFpqa1*twNMPN*Uhcu<1W#cFfZU>EX$Yvb-U~UOBM^=
z=xPjcK95aMWjvQEH4JgXmoFBw2i&g{D=IShQ)M${K;NLOL_E1sxtwMnrWkj+zPYk
z@$fB*s}vsVj>_7H#P7sQi2v|*N5@sc`JlDBjx71*+B
z#eA4~qg@MQo9|NEBZF>q9#X%GfUTL!JNfn>b?r|TmfFw;B2&pGudHe32iCLWKwwV!N+GDvyynL}*#;=k}G
z?|%^qx70^sjhE{8G4pM4PwmKcCGeV0USD67pgM|kJ6%YUMcNnIjmadWIdc#k@9vdH
zAMXv7G2BznC@oXq75IcEs|V@WTII0kK>0R$+{&UTY!Q5pG_TO)F6uYC?%z<~4bE?x
zdD(c@!O~FlKU!PA-Ajjs_!tDIQ~vpVJ}x36
zlHqg5IP8X_xEm5NYrd;W18s`5RNG3dxIY@JK{Mg(8m+Hb$Ci_+iOC7K
z9{3@2AXl@bCsGihZ+y45Dr5?^7|)rMuo=(!btnb;HBj
zsVDPH7ZhWXw)*lriwTV(*Ua8oTu<^D>MGMbiP}$g{R-Y@P?@7kQVw@Jm0F@ZcrK6)
zME
z?H^m2QtS;h!8hT_SN`QLAa!!$$?%uEHA;%b#@dhl2Zq8;sFxLXX|66#M&G}$612(u
zCvHXau+@nlhzGhWy=Tht89KYaoy_BSGB>pcBf
z2Z6MmSLO?6(k1xC^}yV!+f}v|XyZj#NSGVDE6;Ltj+B^s)0(@1{CrF2dn+Qo>k6zj
zN>X5IIWzuJBYxC%x5efJ=EKx`{cVs~TA!0y8shpgzlT~DpokRzmLI8F(yrr?ANots
z0Rbb~<*_mUSfmh4d`{2}q{bg_FRnfaRT(+__y7p+kn&ZiCDaKk*xE0u+x$DB`D+Pt
z8*#L+kLVRsqOC0lIkgUrqY2
zToK??QMH5pc-qge|D2DDMtuDWmXjvdK=Psy9N2SuT^;IR1abmz(xqkvW9Jo!Gx=T18MX1`#s1+BLqv<|knU5_Y*|8=zvwxEK
zsezWh<*YL8RGfGK@r}jo)(8Qse{IUaq(x_f#w%w+iO1TcLXA^9`-^Xp^3u1tfz|-8
zo-58C$0`NVU
zh|b>or=aI}Gkz@QJ@CcB5*<4?a~U~>Bm
z0W5m96uvZmNQJ1b77cqNR4^aACQ~KmAM^?Yje*@BADv7i(
zgn*X|dC#D_43m`B$0=<#ztc6epRr!ms-|}bsxCf_s<@ZC6Jqofn1#?(ElQ%p;OJo>
zPckMpSJ*3Gw1BWbF-R0Fw0~p+Fn%6leQ(uQj)EwkbHn`W+$}59fT-_dVaewzJCRgH
zF!d^s347G0r>_8+#N*4PBvKL*l0xXOc40w5@>hv3d~U?0q)_wPZBh^&FG0$iWDu5k
z1X$hePmk3qf33&|&eB3uEcitY?f>~xOHYr2nC96NkXx5o4T8Mq$`8Q)F8Eyl^Bw3%
z{s6irH}^GZL3Xwu@P6Q2`0nS9T}mEAHqR2_DlN_O{}Wd*cfo=@j6{oQNDAK2qc(C{
ztGZXSDMk-y#6@q(lT=db`yFCW&cxosvx#NYM(R3$NgH;|+MYSj*_EFCx3W!~R}FXn
zu<MpV@|`4T%egDu~4
zbolTPLN(M6x^uyAET;A@we#p}i#fiThgtuGbiRx%5R^9XC7gMJ)dzuQPp*WI78_we
z$G^90rNs8;Ey5D_h=nqd2vYc?ES?re8ax?3MXH{oC-@aW;_yygLxXmm>55<$9K)vl0DGITmuNO1^3HD{D;Q&jU*+I9{S9F|CD~GmR!XslwzA8&3+#F9xYy8*99);}r6xbhEtqllG)b
zf6TsAVkTOK-P~j`MjwAbzcXul(?S{D{c+o*JuE~1;b=;P(zYyg1AI?|tl{CD_`yNf
z^&I`J`xc2?t>69=M@ZhKw~qAqf)BuWTjZ#kT>pQ{2pB*7_UuAYo9Mei(ft1mAM#3n
z!yw9vsr--iu;m_r6?>8Tey#Pi@;yi%93CD9$%cqS_Mb7Rn8e^82WS_%2MEmCtR-}l
zU|?gTQ*;FvR2?U(YdkoA+B(LJ6p=4PaV{;#(>sxXF({}m?_=r2KO-X}b8&IqV7e?d
z{qqz-a_8XakXND^o5tElcEh3y5BevSRu9yne=^#WZ~?BVSM~mkvd!c=Y
zHuo|80LC}fP6bO1;R|D9&LaAev5XoK*-kGpbF}~S14x}LnMRvmo0oB)R}>+WOJ;e|
zx{UKK=nY&<1|g1`5ZoX*f}sn9o>;ZBV%cteQk}#fwmocXC|P{%U!pTvG?PzI9T@w`
z74wzktRpgwL@FPZx6>W%yEI^Jz}briByivAr6EyfA|*n6v>TA@#A~z$Eg7Z#%2EQi97&
z`_T$E4y{Dg9Gbh+<;+fs=AXNDHo;a0Gu*FgI!Qkj5U|BHjv1T?;cf!HV<*k2%
zIkEuzBBIOfaj_xm#o=MeE@uXdQ}D0i+Hv|4N)dgl7hwKLv*8Q0a93e{H3zxh7b+bU
zi$F-jAU~FbM$EtG_Pn)A)xy3`ybtE1{_$!=f4YyFC;zj-EPvyV+EG}cpf14_K~#B0
z)cn|qcvfR+>W4*@Q`kxvf~9ygn4rxqQUobIIyhMpvY@^Bl~f*FPd2xJ17Klc;pF73
zxBrFnA_YAB2SYVy@*67wUh2?^xn?s5{?8C0
zks6Aw`6LbSV6HP|U_%6?>XaGGRQ&h@)Ds>cF3`CV?5z_K5eeP=v%u+e%DmRZsQ>g5
zbAz5CW2O$@-l{0VYG@~B_AnMm_MDJM#Z@FK(%%r8hN%+3@$@FdRl5EkzTPq{>-C8i
zMrlMqP#OdQK?MZqkPZO>kq+r@iHGh~LO?oHKtciO5-I7F?(Xi+GwlC+z1R74_Q!o~
zx6kj6x!0^&vqtg@--zBC65pIw+~ZdZ&!i2ie+O=EP;A~>-jZG>@MzTCg`$B+?!j=F
zX*kY6M39+RhDRRQSAY^{bORw|Niw#h^P#0#*vDyRyY!L5`vTN~NwpG?w}Y6EO#c(z
z&fSF}lRf05w{=uyY^6_}1}n5D*yaN`^i)WHK0Os)YWx6X^3OB%w)QamU9!>cIbue(
z4SxZ`8)buC`iTxbR+r=S%4gJ&4G_h*7(Rxe75l8^h}uh&|KU)?t=T1VmHDRm;qc+(
zO$wyvvW!eO$DMmzz1o)ZC_s>TeIoT8w3fWL?CbcKWMur?vyGpd6$s4$5uf@G_QKaB
z97b=8(yRj?S<)Np((>R-ymuqFFtn}^CE6YsvH{Soc
z6obOIC$T74be0Hv#KtyDBq*U0`;9ld8R(?XG`TADynCbUZ(=iiYw#EUe>sFq(dl-s
zP~%8}mH+ak&;}xykDskQP+0Nn)HG=v=Y`XnWR=+ug*kp=vOvDD`w2hZ3XJjgRl053
zsuoe)S3gq-7zq=1Wbai^m_e$_OBW`g!uGx(R!l7M?kKk2pTN#6A!kek$sWHA$C{M7C(lA?UWw8Zs^TK*)7JNG^NFS5&m*^P6r8f1EEn3~>M
zN~xx2CsRKEpPfi}(fH`VdrK3MhuAZW=7o(rG;rHvS#e_~PDq1ogd2OLV&CsHf2NYy-wqALB>tx$6rg
zV>dg=ET^ShipEUJ4Gc|BYDRD{{*4OD$PoN9bF`Z_(c(PPu4}Ldix6;XK#@54)g8PE
zTJoQ9=!C9vva(DsN)2ZY_apgCFmQ7!#qPC+xMd2P)e4$otdH8^zb;$45Crz-R%$%BT%7#-uT%1zcP=LPu
z?#Hi{3`5pf%#6G9*+%b=c1RG`(pVs5pna}rDmH2yjW3RMDartz#}*;brix9ta{khqsD~VJn+f7WKlG%m3*Rl48$82#bzZ
zEx+cT#T2t#TJQM}p^2f1^9dxR(ZXU9*<@9Xz+41-`~IP!<}M9zgtxG^#n8y@cGA1*=YOos!ZxYQ
zZ{fZ&y?}ul*T8Ra@2i)Z#|&_TMhIMPp32g(u;4jvArO4j)YO}W;|io@iuXeU;nO|%Ts&GKUb
zVvuEU$Qp|?loPd<-G7~s$TuQf@5~1(^Vz_W3;*1yjX<;_*NU9aX8Q)5vWkuhxu5Zu
z+WtGR5X>HDwP5ZFaSeYIW3Xja=q7TWyeO}}8HIo6QTWD9o88LJ)u}
zFe=%NbomN2=%)1|Sm;INahSUO?ipr%C91P_Lr|+aQT+qUp@el{)T@SZbSmwd#yY<}
z8Uvi9Vt_K*EruRC)f}`eqi*J}#8csQJIWoP@B;A`e7kd{o7#SezYQGXTMFj;A8jtjr_kC;@wX|=+@DbO`&delBR1xbo4q&(fy*ICZ7w*t0?VBN}pXqOP{4XKkjdCywr@XzdQ2Qa{U
z9C%PJTXVs@Fk5W6@wkR<(XmzMCmLN_M~P-Z3K@ZaphIHRHcN0irt!l$8(AV-Q`rQG
zQpytRxB6nrGLe6d=%LMAx4U@KJO~E7TD81blP+U>YiKN@NRP(o8&^ih=cH3M-^fwp
zd3iRzb#Wf{zu#mew>QgSf0Nrj9Ip`SbdoTec<^2*Ni!1tIBx3!T+@ZNw|{+|yP=y_
zKABrMw|wXt5Iyk63}NSF9bsf#giIP@sc*o$h{qj9ziuF+Y2{<>u=rUH>|emWxK)?}
z9eaFz8sc1Tl5QAXhhH3U+|(mid>JF78?Cd-%3?3}sP91gza@SX8=>aPp(DFx>jNRw
z@0m}0RD)XgNp+g|=tC0p8ATGOfWhsvmL$2UcwgZ^j{u9Vo1cAsHFyagqEER`h8~3U
zc%O6SZ$YZGC5HfunRgPDoY+oLeZH$2KPQ{4m)~S26{p0&D!zLEa4V~@%i3CzhE?ZE
zYE~sLoxP36?>V%+;e{BoFhsI7(p?-h_3?>6i
zH3AbxA9pSOKsB4ZYe2eJcKZdsem2RfVQ3Tks}6nE%c2ZU+f#QR$xkq`zFm9yO8eXK-ArD8kb1U%@!f#r0?vRz-H~w!BmELf_=%
zlaDaa$rl4H_?9o#$211t5&h-j{n0#PhO>fk6LduC$j6l0Yaf^Gt8LA{bcBfQFWbxR
zC!G{aQ+)Q_$(3x=DQY}4HVRuc5Bk>^&20WHH}nh1S!&qs*YV=lIGLt@v-q&PY&b|1
zPR>&VH$pTBXRxCmPXF(GauE1<9l&6!xaeq%7-&uiH8K-iyY#OEwb>$w6xtd;2(3UWN(fdn-+^Ok&C6LcNVkg)3jiVGmZ9n%glcK$|N27-E>B0
zPrWo6W>1*eyk}=;C6dVWmF>Y_BR)LNK9Qdi%ZM_o@g<9laotX2>>wlBMpw7X;<&v)
zN~0`sa%luUhBbYUgwuR;nES6;WT|ig3F!p}`v|`_BN@dYbTTDJnEuC6Ry5GCmuE2Z
zoc;SksLicDmTz9Q-w$-t(dH3X!7R^pG*13$x3g|(#3mNidfcA2<0I(``;32v!BlSW
zEvM^c7SMn1jkf2bz$2(!7w}$~(D=VgiA*idsD8Mw0WvDx-YiK-W
zU?{N|-hw)-e?fM3wh-fYa|i1nuWq6I=TZm5G{_l~AG&FmW>=8d?pB^E3@IgcGr
z!Db(Ue>u|aigj&S%-?Ed7oSl31FOYlE~fpG{Gl=rCh>5nxe_hTLR-0-&
z0?wImxMLI?><&TS>yEr#sZ;#e!|ZlA@a(anE7TlNw&O
zN~H(AP3Al**kxCW#v^zcyQUQr7$*@|i0l^_R0-ec#S?yA#%+lETar
zbdhC*wTUZ3M=)RE1FC;AX^cj+O;MMlc(}^HD+AWf)W~klA-2G^Ea%DJZ$*UWHV5`a
zYa!zaEZjcx-cKLUp@#3bJZrDI+3mLSzH!FxzI$&%gqfwwNNt~1SS+YrwVWtpFyxYr%9H`V{
zINTrA2zlXb5rJ~~-|8kG-j_OY_`SSvM0hLn=VH+Q)itKd7m42^nMGS?nL>=x_XV>GkWq+UM0g-qnk{8*K_ZCnWDZ2OHOH%jYadC?ZrJ
zWKpy-(LHDS4E-IdX>`Rb0?`=9wA(80@F;G6JT_&boNM)K#{|NdWr4LnBBLS&94mKh
zBj6aIejLBI#G!t_d>=;2uZTZSVhd>em%@(L>9K3;smz7*dV`LJ=CZN^2mWWr)pB+Fd%QjWlC&6Sr>*l}xT~Elcr9ikVlq_)k*62`&ih@Qm
z1J8)aqA-F)u(Hsda_rzJbI_KLjqgG-EL|furO?sD-^AcOpJ)=9lO>S`UkaP?6i&+c
zsk4!nqK&GZr|aQFcX7W9a~UDkp52SkWyDZt>krg&n&f>1M!Eg^5L31YI(5djf?ls$+G*~)D9ae=%
z^?TPlOqF^@`!cwhzlBGHl#R{N?=b5)u#FQ@;=JX2ga6pt1ka8S!zZ9TsXtLovt|(y
zoGjzC95~JsKlkEi%5DsMq;X
z9zxbS-SngpY5A{b_<^kM&)%Iwy)74yjTLM^U9%IfMf!Ud*-vdY&Bj?Ka9aH=^Iu{f
z5%s;}iIxyV`$gCXvy&cSHCCct>*)dZ
z;kBNZP)rcMlxeL@xwvrG(V0%1LiIa95%kEk;}g50ToYeLP=+1}#;+mhTk_rC$Ck&$
z{)tUf+8yyePrqmq6X)fzC@nK8E$sjHZ~uq(1a`Fl$8jmlfq$KyVzEfn!KQLmN7Fi#l*##^d&MrVoKtm
zu8sef9HH11rt@`b4`F}rl&FT%8m~dgT2fto^>BT-)PexA$M)YTR^!cbxYBky=8?wW5!17#LY2;sI1|yIpR~pawIWxyESXW=q
zA+S{cMMXtlSPbT-`Es;w_Mm?3q5Ii=;Qk0yA2Jed(doZCArYMDm
zg$0nt#Kb5MhQX}xqM{-IMBu!OkB^Tx^4sy59iKN`87SpsW;)5H@N!H|zsjzlkWSRd
zj+EQ3C^NQoaTXeNmX?zE6LVKaQwr;^hxKG;Mu1V9THm^xRGM{5CWSbU%b<&@Z5mP2lwx5PuA42STE~No%J@L
zf%ZUzXyxFFLV1!~{|(l_M{TgdeaFtn`7|w8BJ5!sOhn$C?69-5n|SD>aSsM(evXQc
z?-`0FN^fQ1bCeeqH^+aKA)-=hOQXs*F>6M3i98aL&iKmNDwN~7-IoPb&qsZEnvOQk
z3JrgTSaknwy(@Yf#pgOc=q7-h!F}&#sJZ%LW*4rc>&x!lpBPp=MU{ir{c
z{@l>{ra`%_1ZRoyK@-f49Z(pA_-sBo@QHPz`(ZkIJ$lpRRC`O)RM!;N-=WOFC#d*pHVLFyc$`|hp=g~hdQiD2c#aK2?wd3_f*7e(R&g`o--&}fcfRf{O*QOTvS}Ft6mO-u-``?g{e?D)1UdWgH
z5FQnskx7nXZF1*nIe>CiQSnqn)gvLUbk$>1h%=H-Hq3l4+?LVsy|ynDD#$shezd2j
zvvVY(p1qBeMRxBD4I^&bJP2EddA%pUxo}|HRw*adWt8H`rW?dB*LRBiHZO0#&+uE=_-CO*R_0Hf&v7&0bw5|%`$dYR$=R9Flt~VT!(J;(u&~;
zH81(JkR4k-zH{D&#Up2WI~T+&cy+i~iN~9BKUA(A-m9VYv@I@bAd$!36S59sb~~>m
zUR|XU_r}xmXSaA!SzW37o{wNGv8(`$C%!L<$69gld{3B7#<(NRY~p-navI~4_AS3H
zajN%^nYYcR<>kgG_|`0D4R^D6kDsd8CFP}0vezxQ7y35vaRru{Y9O$Lrd&@{hl7u3
zg0N>QSc^Q!R%B&KQd}+&hJ)V85ngoHy+nVA>wqj9~HjZXOcZ{cDZE`VFQ#xW9
z^bJ}>AiICi?ZO9E=`bA$4A#Lgxqoy>;Dp;(>*jHk)
z$vCkqZ?txSfkS&#wa|OZS3!z_Tdv5@DI_V?KCQI4gLF)RDc6-ZFk3r5m$D?OF8VD!
zeQA_h`}5yU^}Q(OGoyb`Hn!2sI1e0ztg|eIW2#IE<
zn;D7>E-poI>c8q#oaVW@p{XNytE^32)_*?h#g)x|f`&^<+oR~Ta=w{+VemUSnUS?5
zH@7poH-Yv`#;>8L+`;qB0D+va?kGpZ(XaYlczhwc?BchkPn9gcdh+COO4|KoM-O}%
zVFKt46Nf+?t=KIrEJW0Thfdnx1Edf60&mP*ao%9Ae8ca*?&u=p*^~Omul%dEwUva;
zcwu>2F{@{fGM>q%bdPbQ`^G&T-@D%teNHhpSZVn}KHpjG;TcJ+g&O9|;$
zLb{E?_ukktT!DHEuEN^8>T>JJx2lRBwuWpC6)A2(oUw;=OP}cD7?q81jwGBbQi3GQ
z>y{58b12QBCSum=85;U!T!ZCTNnyIav9Tf6agUf75Fz(OhSFdylg&^y_~6pxgIuFW?=U;H{hltjhcKhFwcd=lF~7TtJ!r$;p|
zj&`?bq51wW%pAHiqHZa(1X_S@Wk<}
zcMVNpEKy+8lCYNjOS_u1b_qkc1@cz+my^+kv?u=Ky^4{P(?bUrc%OdueMeb|p|W9v%ABWYU~z2!ch3BuxF^z<7UG@Pova7|CBPY
zF>v{*%f}`EtlG;Fv1RIH62wW})%|edt{z0nTZ;%BOU;~LzYw#3Q7wD$FvJY*){m88
z25g7S*zs8m74q@~$mIaojtrI(AGrJvSzc$>#^(5i8TILYljKCndtK=%8SxPcu2`9
zXSlGp$4T1gkXwmQ)o>2QQd&Eds+m9W_L-{j(0smB5gd!Ism_(;d}T
z9W&;A?##?{duIPat^%`su*o7y-mgBS{bO>8<)hfqE|@8L$nAN-&1r;zl6LQwffu1n
zwLu4!aB?+4@6*=d!5Y~Jc@8t_=UekIqNJs~zF#P?(diu@Taf=g;Ud;tC!JR|MhcKN-rn8?1qEy8Q(6L>0N!scwsv+-6dB#Y+#WCgwChm#
z;RB{p{Fz4e_zwd&B7?A`QQwzt^jw`2{jW&AQ;n8tQ=XW<7TO&_VkUULr7sHK-2H`;
z@PLHR2)ziK();0il+SNq3>RZhZNSU(xi`2qynC9hi*Ex>r0=Xh_MAF&t7~lgOb?c7
zEykH%E_RCFJ_X?0ot_;G`EOubS>*vjH=qd=q9UaLbtvZc2z6cmqC`EZu;4aXMz*f3
zcbtVT3-8igW@-9o-DBm
z;iTOo!YufLo)^Rdmf49>7v7o8Ubw=6S&@ry3rdWe&84wK`_YFD%0K>v-ddO}Z)9#+
z;&8NpiUiV9`OaO
zF9R3W5cbNbx~#Gh*hL=-v@8MX2|i4qdw~D*4g2{U+cz9Eu$`T+2}2+UK}G)T>UH(o
zaUFOi8n`@fK6M7N$Oso%(Bq0SIZha$*W1={7wSwqa~PqTC~guzOrH|ih(@X|p}H)g
zLXPl}rTX@DrA6+@5XN&E%u}Mxx3!Pst}Dy+eWe%6&7M9!dH3b6d1AaM
z(={6Qm2%=OSs$awZ)@698)^?}h!m}q=>i&0roeP+{zoO;v+3W;&*R$M3EqN#ixt1eHHz8>`JwX5Nh*Mp?fY4!wmd+ap|
zjsjH^me|iJrD!rQnXG2!TRyp{io(#~_{7AYzkb0Gxi1kBEY3SK_h=+S86^=2&Wou{dJX;hw3I%%$c_#N{FS@Y7Uslo8sONlalQZBA?m?dtGuk^5CcWN!2c!ows8ChA^5IJEK@{eh=?7{i2BrX>h7nogyIv);$Bif(x6oc~eDQV*T
z%>2*rcfYdjtwS|e&HmO__YMtlGBJfEWeskfD!()bpA(Sp=(d;-yzSL_s{G=T(b5-i
zv|j?usn|U`w6M&)xi+v%53}S_G%9SV*>!q2-^!81#FcZ+9xOT+Wn?TyP>Zf*d>Lkb4=fu61x5YJL8!8ls(7{X
z8Pm)g=O#3N?a^)8
z0oikg3{lDLw5P4@?VKu$v@q0K^3|(XK{ynod#aXTVe0ct^%u~*Vq#(#J^nsNo8va4
zMW|Nv$!}CtJolGm4XQmRD&7OK4ijgl#>4Pu@y6r3G4wSs(gjw9MMPMXY6Y+)F}JO*
zK{Z9-CYYU`bUaRb>wbBr{Pt~8Z}BdqU;uN#(u0e+!(Rx4$+#;KbutUaF=MHeRXMJg
z*)2-5f3+0>Bnd2eSiH}~gqEVBB0m<2j}Vo!4_M=|GZ!L56fqsjB^kRde2bS7)MWrU
z%^UeQCaZ>9TQMZNf-A
zB55S|=I;MR6N2Y3`iWvsxJWR>R*(koi^?r^CKy;+cCPPt!<7(m7K+se)RNO{e>3F#
z*|TRQ7{0ziL6X|q+UX|qpeKer6N)gV_fRO@Ur8=6EW{@z2|_(S4{9Yi4+T9zVZ`#{
z1+V8t#Z0X4Mo9S3eGz`ID{@y3_oxJzIZO&Ld#;lU5b6vU`xzyl1Hb2mnyM;Mm@gG{
zOkf`*dtF_$;U=ipxL1O>&-;oyWai${*5vpf7kkf+!GqQQ)UOj2&q(fJpv^<30Y?IO
zONf29-$j5%c{W;Uw$TqA7o6Wt=>?%7bfCY#%Hy0f5e-(0J1+A{@UCMBP~hla7q7hS=!l;Z^AThOm%jf#q+G{*b@70CE+x?s>Ea=TH-pipjq`!wb+6FI
z?%fFv4~OBf9M{YHK-2vG{TtAsLv;e&=osQj;Fo&cz9Ct0`S0u
z8NB#smdUNBYbeS?MXB(D&Fiikaj3Ytox!TYlu@r40@AGC)#18oP1o%C@%E@CXYHnF
zxT$lA`Jg!j?c0%qFv^XTjBJ0f-W~&GdJaM(gj0|(K|JZa8cR$_NUaROd!~}R0<+e?
zn&1oM_mJ_wpF1JCCh(j+Zs(p}^!_#qflk*W8<%u&4m4t4XVhhHNjsfB9V^b4`z
zR@g!iY<<7nO>ysDB-~Fd=|G6fEukcSfdm>OqoWkuwnA_0?tEF?+zeBs@+So9`=Nk4
zUWKLC*})T9S`<=EHyA5#`@n(tXz>#M84sAFi3!ufyyz7X%xvu73q*$TI#;n~Es
zFgM3a4TAHBQcb6%syhAKF$Zaphl`8ox0)&6Y$G(K!>j=mh6bp{zyQA_HWQpNIShEp
zxWG5CzRMkeOCpy|ckAbrdl*~qAz0JXXTkY^|5^hh1h={o=ECT5;BjJ;vYYNLwE4y?
z!}7s$0^O2+3XdqBi1*P*Dp(O-t&EpHPn74rM?%5__0;SIKf60M6khA~?{V146(0(a2Gjj<(yEhp2
z?G;r;BaSWvFre>*X|Sjz!pReEeh<&}*$fK
zKJUMO|AN_QNqWI7#3y(!e9zZs;jwjabOfq4DDyt$1}rW&cRYzvf`G?wAD{X>&GMdH
z4C8tD4bVHm#?J09K~{RpwD0?enIqzRr4njt@w%JMZ9WhPp$FP(q~Mn92O%nAY+-sR
zk#;xp@$scU3AhW!Wp}W!@K+wu($dbTD2Sl0C`*Wmv1D?~#4^Ijwn~Rpd=0{SrDQ>`
zsj;!&GI7imE%yUBt;V4AfqY11X*yc82+#!eHwG{lLw$W35ZE*be1ur=v8fP9fNP2h
z3WO}M$Sy7}P&{0jun7VEf!QxS=J<>nB@6-rU!>^93iSM7=^=sct-uJJg^dI9-?&hX
z8=sbekujOqaSg_4&(4oSJ_%6}pw}ldE*uUT+MVNLL5G!yZ*g;A(Euw3!Ar;wE~{S
z-G}c*gG|F`${A`G$nATp{q%tFOMbke77+04jA!j+kO3YS#3*5$572Yz<9($y`nofTn!
zcH2i6iI~kG$hV?`yW1U)@{h;4in{vgXt7BrLwCC5C+Oh&`0-;;QK+Q8{v4ni67>?`
zL&47i-2MskLh-czcwGzpmyuwvv6=fJLjZ`1i_5a-i;9YZpeELZfeGcr8+Z{vH&NyE
zuddGfhmIj!KYm@xLS*k$Rs@SbI-I-hzKlDcrP)DV0sIRzJfyx)2!jtm+ny9YVvQH{
z5+Eg|4(4{*`~KmB6tKF2?kC=3+4o`GiGy`ZcQ=mXBYt2{J@*zcKTA7oPrWoSFj(2LxW7X4g$xPUw2aKHS28j;^N8WE{a-q8w*l*LrXKmk
z7%Z&sZiWa^q2NG{2r9FM@VKL+BM5t91b2|(G>}xf%mc(%V^oon!c2^>-#I!GKx_~f
zcc^J;SsEDu;}NhVfq?>n3mWVR$jLhkcrgD6nVJ0nN5)wM5ZJ(3e&oGJ2`RS6>0W@#
zqATPx0RiIH*35~-K0+9HION<^PoK6Av@t>L7EY4zGqwi@Uq1qfDmHAxmmstU7!(;9
zX=;tuFgB(^EFpUH2HM-pfc3CZAcjB-Byk+<>@bZp&@o2FVYM${BRluUj}K$I{9gn-
z#(pSf!5zgVW~uAzv-|6h$!6TCtNjrBqnVx_D`G>fg!{G+9tMluCjc(_;gGnU=BgJb
za#(==za09tO032So!oROf!v-cu7c1DkT<*StAK~TAuH^KrM|w%(8eF<-M{tqvk*Q_
zJ%n2TG=crwqeqVbRiz{+!^4Q1Hw0=jsEF0iyp4;EJv%)GB9o;m9;3bqFa@vEmZgP-
z8>Ef{XvfgnT=M77pZIt+fHn$bq38!yhK7dO6tRs>O>K=#OL^LiPoBV3iEdy$;vmLy
zadI}t>;cewxU4tzBIKfkYDZAw-p`}ae4y~1
znb8e@juRR`l&|y2(OWMfBI4U?Mmjpw^m&M7Y<82!;7bX6Q|Be62fw%H%r_hsMUhXb
zhlEmZMDvCzo}HbYx#>p&n&63hOq~}znjcDU_eKCwAjtj`a0ifsmyk(INCbw5%XxYV
ze9}JNTawTHf0}rdoC0aj)%=gd}-cSR#JFQw-q)LfXM69BlIV
z^03+nhIFO4PtYnA=CFs!w;QRa&Y58!qJ_4Tcg*X_z+-j!T|Mc|qMBX03jx;wl
zIc|={=rg{2`La+N(&!%4v^C1mf!482xaK=^@8IIg`uTpt9m;gfsYgC*aR{U2(I3g)
z=-lWGP0pRgL-S!dy~CVg;UlYS=Q5u7&-4UGfg&*|wjL$?BFA#FQYoceANnVj7DC+ee;
zcUs!qPvx8f_+CP0EsU7h7th#xA@&h{nOk07mXVU`y(}*%z_UP2yR`#cm9nz(&n>A}
zuhcX(_w^&2KyiH=`8N5+=1{KIS?R&zKY1GkVj7*PYPDMP%8=sEU-P$t8u0J>08v2f
ztE;)Sb+@j5%fLXi!Cg}K{QbSX0w!@Pyf9xlZ#g+R_Sm?3TKqOs{F#Iaj6&(bxllwl
z;Ls%oKbs%lz6pJ8?<+Krl=$U+1xS%wjZp}CL2rIHFq^No$?gnpd&B#falIs?Yhn@v
zrxOW6`-2o!TAEVG#|O7HL3*2(n#$^ScK^F@Jd;;vB2M5`l~aDf={@&ihGU;_EX9z7
zloZ*og`)L+efZ;V;Wjbcq#hpkU%%%Abig8HY#0fH2gL6t&ZPUHYqQ{*bSEpQsAR72
zfhq|^y{x4LfcF#?6jHaJ8=9I1J9^*h>F-Zh31Hbq8vSW|y!9@!9SQ2!xHvvD6ok*U
zUw?fr*#fV^+By?D4!Jk_7L?NQlv}X^;;$?;%K+MUboBaAuJ3M(QM|Zk+htLoTqH7GTNv{@oBvIq(Rt!O0vV*yErH
z1;Llz76wg1JGVjn2FCi-mOnwu3zewUm};n`+%GrJA;sjDCkNJmX0$I%1zXImX@#8C{`>1bOzV2t33S+
z$bfVT6j8byx|Fb)*eUi#MvH(Uu$1fLm^9H+u!z54109{?^KBl8JWx@+eft(l2?QE+
zKoD5MbG*0c4fa)EreOODtp`wH1AYTmE_ndC(Q8>*3iWUZnaccgm2;TAYWHx>G|?7P78Z)4<65!^%r3?3i=y9k
zzFJ}S>Z1sYsebryI<&3L?mglKJ4^7|{i~Iu!;faxTD(MKPsJ1!{hHp|Jm|YPYbvI_
zy6SpX?b=uu7LIc?nm{rw_y;tT!+BWSc$vR`4b{9$x|^e_bT-wQ>IcDnETa1gQb@q(
zFu4inD4u7}Lb&-_MzIR*je=jyEBcH->yb)O#Se4tb=#+>rM1dmz$$-_3y~j_(6*?6B=3zo
zF^BGWEB~j0qN3aLz*@2VD6h*1JU>{a0Xl+)#;+O`P08_vPC$L&Oq2<+fV;D)X*OUE
z)AV;$h|pz{8J!)cnK~aJq|{)=;jpd;m;e(uS;1<6?ZDpUFvwR)_8n#I_^6NM{&z0!
zR}WEco5>!u>QwmGf`A|kedRh&oIgnEF#q=--;en}xd(KDw^x5m)_)o7P^|L$Cj31p
zwelA)gL!9mGGu52s#dj*>x3aquGf1~ev!m(PS(hAl`f0?UUq!2DeyD^crVV9A1!pL
zu1X4tOr32RrX$bN`jL}ELQ4EPiCwSFxAJkb1m)H*yzA92HdP{eKFuU6vSH>#E=_H1
z0lFu*=B@=?P_SMwaM0Ow1fFQWq;j9@@%gxYxwG~}$s25u*oFFCf#+)7Ig;Khy2ZI6
za204}HHeOOHNP57W4paESS9#!uUzh3@3TsG8cBiGgsk=l=INB%HW=c8zLF(2cce|l
z?jG;sey$*_lv2`TiEJW
z_?mR$cKMM{lV;KtKk+ww^}71e^>%aPmrlgW8~%2(#d15Uw!)AspV;5+aD~@1x=oOt
zCY7*x?1_2lJbZy{I8@RGA3e;R?6VWl-$La+-&gs3Tl+iiYYwnAR8YV)o7z}A?cbS3
z=gR-<#g(glZ=$5_+d#Cu&6hAZ=5`J}0|OWqWM^>9`z3KT;^R`37j$i$Os^@=$eul}
zfgRwtnoo?Otxt*y(x%)BS?c;CXd)3*P{x+IFgB*f$H((aqsDnO%>_}$iFS6Tr2eRZ
zFO=+Mk%^@fr)IhAdThjIjfd-6;pmRLd9A6l?skQm0SyhAMb4}!ng^3dTjWb;OR<)t
zkmY|qtp_Da^0?)y@>`$X%5!z@x>UK%zvXb~n&;yJRCMLBhru-w{^!qycQh@2Ak^xH
z@-^h-ZF38d9{r;f6k}fQjK_*wmLs34bfiHEC)}dhZTqkbsB2qCKx+LfBMprxPEn8y
zD{CRl0*$G06_YpvF|EtluCGd(n}Z3HazJ6~WATQU{{1dCILE{f&C}8|M@sFp02`zS
zqRKR2pp=D;aHa+J()7^0X)whQW<}Yw!UTk8r((w<)
zGEUAI>(oCklTYE5`76VYxf8TAkb?Fu6oi2kzFKAn{*o4B($dB={2iVm|+OTfwQY~cgM`vKnE1V#E7if@Rc#$Q%R
zR4fiXNUcZ==zqO7C)LdY>)U~a{&Vgb6CE8-x_W%XR>Ag{AcZJ8V#dZ)zwTFyf!BXp
z)Z0E*G-YAgdIuN>7y~S%B5$yL_(etcup?e%*o^A#t7M`-EkM}0`3EHxx`t%RuD#ua
ztcvm3v-8c%oR9ODTy(ggx-^kZx`1B@tO6=(K(^d0-gKs8NBHLW%WAJXSA4WqNTs;g
zYWDE6Y4N70L*LxZ6GJuLr_
zzrv-oZ-4#yS3appwF(=eMsk4&`6lnb1oLjn9Bn338zc}%Ws^`uDs-j=Ajci3#mLwn
zo=VZZZ6lAn=2$bOeg9f9G8AZv<=(Vmc&Pu5&wGrl6gwo`H-7wtdKXf^0hVhEVxPLQ
zd*c!VEUqZ7DCWdDzlDfzy95R(axCZ8VqzN!G_r58*2HIgCKVr4~p?FmOOWSmc
z5d+^TWa}bxfx6NAw;H2k_@ys%%F9R&ON=Kozm+@O9JaA$;czP(Stu$JXkS<=ynUON
zmB*S_{lwKmJYj6=e(2kld}8zXVR+1)57L=jng5G2)jc%+T{Jy-cXcrqNJ!&N)C%(>Dx48<1Eo){7
zhiK>4YSp)|rk?$N-5R!^75?@ETyh5Mpl*eQWvj4|)y$k6Zw0-b&c5)5nV>~ozW{!;FNlPoI|uGW{z0Db9LC3#$5R7H=f%6_>K0-
za9NlIxcwFtmd&;P@u?BGAlmcV?c7bY{{DKP%CKI`?)2uGop0i0-r0kue2lSgkAgC?
zgy;H*nilN3+A{$8K|;SV+h8_HjH&*7#(Y^CP7FEiNU{SD2eb;oDMwVl7
zY*WY)hD|(s)&Xq+S>gAi-UYzrzJ`Dc!ztYOG@r@b`4|%a(gRCM-Qm=xBKgI8a`bYx
zr+oWDo|Ni`Z0{!+PKeN5MB=0g8jOy1QlZ{AnZ_q7mu*nB-n|*&VfKyczEzX}%i(6R
zB=qo54XMON5VZbViKOA)yT0h1lp<)3Pe`oWym@(-|53YlK=i-=40+TeF>#y*UbNWM
zNIJ6}JlES1S;scQtBEKlqK{X1EmW8T8YQpBWSGy7R?RF7?%o1%)J&#qwqciVV#EWh
zn2#xWn^0U`J0{@ty5iT7v2s{jqwCRt9Ah+F?`nYMcKw}i%dnzWG-G#OieS9FeYf61
zlAYi71OS0zq9A$QEI0FWE~4x-lWikt2UII(3M
zi0{rKpGFv;&YhB*xT_`-&*gs+ltqM+2tLlDtW1FpV~`uL+YmY|jR-S*D~+K`6lF&8{k>_$$Kg2)-?+}g-labh4jSjm
zt8zH;DfuSWZ2hxVVd+70i|pW*w^ElX9KTMn9~@N|m6ZDUd*Drk
zrA1D6NYO)Sgzr)sp^35bzfEs3M=YWT4?HdpZuy(MZVPy{Jynx%v3}6&XZQeLNL%|$
zvXb26o%>tk;U?5t^2QvC!zf@5LG{Oa&<<8@PvVtOYb9?7d6A9pCh
z;uDNL^-p%w-kU)fI->(?h(BE-H3RP}bn01-mvs$G6Qv~P=rCA1fK6&J=@Qpn8oWP9
z|5s4%fuQ_{^9K3-W7Ddzt^YrSeRWurU)QchC?!frOG$@xgMf5*Bi-FCts*U5BHhi<
zNOw0wNaxUL|+_$A*2Wf|*YB7f9Svhtf-T4a=8Qr{QJ
zSYLgMz}X&c6ByAFiWTjEj2fiydQ^K)8Q!WiHF((FjgM4xp+&Y8^fpDnAopDIw(p%mbj^$
zf9&gfPqw+2os~nP9Wz*=@kS;2tzPFpvpyJj@+7~Ys%~=HZAYZz@P2
zHGRn~$Zn-G3$#HmhbFC9_MtLvpAPktBP!(|qukRA;SpNCr65;{8AN->J-BV!N!(Yb
zHaaF;YdVTq!`4|xl9v6Jj&P^=qqeit#44i}Z}0DSR{%(yHoYqtA%EW#S6w|BpG@oB
z{k^U(F}R*4&Z6*{7bK^fS<^GrN2+GQt@TG2tB65frI2dB|
z;U_NnN*mnQHwHiY*C)1;@x%{J;;eG7_gf+SBw+5d@V<%1=-8Wr9u`*Bbb3wAf~vB$
z38k`@4_9VS5g2V2WIon9p0)cYsTN%AgCrURH3bB}43^0T0#NYYX`2JzZ0;M%LdgGO
z{e~tcEzPcnoqp|16Y_m;Ok7wBLI%64*%TtMHeQNqySd}aF{QMG*iDs>udenA^jI3@
z1+Q7G{b7*Qi4yHVz`?R)q&1)fhxnB+kQ^$0WCrBlc>w%n9=w#;8y;@cxoTL;!3}bI
z;Bs=Rt&5djxpz^~&^QP;d0+h=Z?91ee-n1m$PE#fkl356i%i>qjqiJNQ-3>}VtTlP
zs{kYS?jEs>pi60xxiNfB4SC28Oc_oPiBAw&o|T}0snN-yHcry#FyirP>DJXvw
ziCdF)Uq6eQMjkt5gJL=y`Tc4H4uKS>c}n^Bn%A0-Q%lKu6+{$hXPUoL^CT~i&j~tt
zSgp1u7~~BeE^saDlhiN6+;WbW+L!|L76xly2k&b&t4;XuH@Lf3+Ah{_
zgj*c0{!dDSxQhz*qfoO7A7mvoedvpzhx$wS7Or%-r~Q650G&=h8Y23k^2+%;6jH9q
zxC=B4G@<{?LplhPeL54j`6-7E!AL0rYk#-!v2O#Pvv3yFUHO-TTjC(`2f77B7b=O6
zXo3hoI;G!xFP}36_9YPXizSXfVw;iD6cMzvv@jryX{0}P-nN&)l3nYo$M#oH>>rTw
zfAUm?R(-hq<6p!BU$(``?v~=)mpBW0fRo^@b1|xLA^@cKT`e++jc)zR&3)DSjO^95
zHH1vE$bh7
zb*FnbLD%HEGViZKz$yE2NIvGpX6?_cMu!Cw5|a`72ECkfR|pnIF@wuO%^)c
zwk}w=Gq6M$?$A(uLJK|0HLXF|?-grwk@e$kfF!a^1=so$?|na5HIs>Fx5;tQDhvn^
zmQWCMGz6OU9ps`$@)qT9A4PieEg0CB18{F=cL?yRa#7FK;i2@$C5}0K>$wJNOx78A
z`ifw*qHYein)-JXof~09(i_h7shso_6sk2garH?^DNr1C_sp+lBOnj4TR`PvCAxbR
z3?TnAyVxi0^ycyJ|3+G3JpVyjLMoGp+ZUzo*x0;q3_@8&yjGCY@~W}eimAwrs@6*h
zro=51WC8_62GC5nJap@vI#*9cPtRxH&S;;Ik|0k7F|aVWy=}8V#NJ&%eJ(?ISD;{R
zG8*t66D_IY%BBVB9Ol>EhH#Yk24uSJ7drSjcsxJ(8_wGbavZHUZu=+z6n(0A^&x%`
zhEF#~K~C0+$kMy`8W6ag{0NTwY
zgCprWSc~k#Q$$>*j{+{!pn2zR+^0;ZN@B@%=qBnJAeep7^7Ts%!8MlpJ8gIc(gwtn
zL|Tw+xk{_NhYhoo6Nch(#_zhm8wh;QRnn12+ca-u0`05<)mB4)f(X30LFHap=Bezv
z=X+ppqd?l$um)^bqTn2Q)*$C4)O(OIlA1igtEYvte@XX!td3r7-0=!=b7w2k6$A+^
zNrER1*|NT_XfCRV_4Ty^)19;5W`}~&hI;Dk*IwvK^A@Ejs#2eq^SE?D7gujr#aj}M
zLjN)txEBuwBmN(QA+E6~2l>tqR`sfvnpU-hgc=Rree<+KIy$r2v^bnp?_F0KWY)>N
z4bbpxNXeDepQAnP_-=gFzDTs;x>m_jIsQXA2FRZjm$P*y=2(eFh{!uu&Pc7j#fqn?4yQL}?ui}88_42#0-0vf
z0-&z~5uds{hvdVwrM4#C&e6rZlA*pQgR2H$4FP|M74_G{sBBsl^Pw2N&*s3>Y0fQ8
zGq7PG&?cu6)xPXshu5O-uxeMIe3G*yktD4Z9r&`ERsG3Et?-|Zv~J>i5V+sfWmu#0
z(vspLTGNw;&yyRg&#RA_Xs=^G*gS4B+C*=7<4s_t2EY>F+7^Sn&WrHEWUmRRMX3QvK+
zktX91iGdAgHMDKYMnC?h6}CH0eAxX`=z!HUELKF*+3{0L=y)nru;3Xxhnd(!2|<@94Y)R;y$c_6<4#VTro2Dg^(`*PXh7}Mx@ms$Q2;^in3S#;RVotWHCRMHqmqo&c(=l@D4v^=L%_KAEk~DMkyCpa_drS82hUQ}Ygo9;d
zaJeKDoVkt~8yyj^U1%p}5{ej$8qGfEF=?vj=j1r+@r#-=f&-EtjH$85=HE)IAhmJG
z!hj4$S>pG|L}V?MaZU_;!h!;_iAfcLN?{VT9FC3}3o{ytmQdU4w#cf4dw868qNDy=
zxXbX=(uXEOK^?*$t&Q#!d%Qwp`dJOdW0R*}-0PNQwYL(eHzYwEKt_KzCyBGVA1#`U
z-zO8MsHC(z(n#|xjA#z+Fcxb)0P6dw12)%a_x#l^$Z%}APv|0R;{Zj-%RARC(`Q*t
z&uS*BD_i2!hvl^7+?}V%4Y6L;1kE;+)n#W2>kBg!Rjg)4MQqtIky+Wa?L#Q+!6N+<
zim3Q$_GO=tV#VB`kQXQ3Q>BgXCJTjKt}e%!aKG+8FG<
zkJu^?yc>jmV6P}o76_K4z`@i^$52z2wiP|WI~Q}_|BMm&Vk#H)h77;8SbnU
z(SrTD={Y6^E@FS9p61qZQ#DRj9cnd}qIaceU*gubEtC3rqdt2kJq>H<`j
zm8+X`GRg$&H_xtTQx23dhkj=bmKfEv1yl8^8wlbVlv|Ae!3zMC1hlianKCqRQOxog
zh~IJ5#mFBvBXkMfk*2U{sO3j7n&RbG{-`lkRh4FPfM2F#F?7(t^^+jx$x3;yz3~a|
zO8`_f=uImP3nrNjGR-j`gAh66GqSzqd@B00d)$mwyD-&{OYtOdIClkUq^!?h>-?A$
z-anFKx1~7>7_h4byHQT_&#Qdt>E>o(e%viMR)k&h0E_ab*e0Hcv~!RWxeO8AUtFwz
zYOx-MLQTSP|4e|4r^a=iV=;;mSuW00I?}9P*H~V>Cz`Bg*2?S-@k&?+hZ9jD}>8HZqUX(~-!?ml?V6Z*E#c
zWRPt|!gCo=820tyWCwQ)u=k5z?{z1>GQRp!FVDH6+ke`=Qv;Iwv3X
zmrWCz>`S(xQgOW#`WY4IVzGob}?UD@DhpJ41(u2H54?c|e@5}Id-q-C*zO$m<9
zGwRMY8z7YSCEGH)^L@?Y_|%zyEM{R(q{c8S>Qr=s@?bi4Sx)fiu0*
zPi~0jo5GNhLH0D>J+t&$#0p&;OF+Ityv2VsaWnFfR~(+HFE_V^Tv};W>um%dD0{CO`CApnXZkX
zQ{t`O**1tj^>~z{PBK;;$XpcL%fp;rMv4s(RL^HrDbY>}axA&qKVEJlQ!ll%bq@3^
zi?wOm6=TGmjx7qSOK4$eeV--9rxmmFe=Dc<&4o90n}>G#(cA%GxvZ~VL};e&LE6$x
zekvXcP_(6d%Q>cS_JlPi6LWXZjFAa?{SN>XN0X0S{SW^#Z&vz
z=1`+q@J;h+Nr~6>ej+{*LUu)cmXG@canGAu>gY?odMaR7u3UGdObw6;TwS@{g?Y_6
z>Tfxq5N$2I#+0lxwP=Lr5p*;NaEi@gX5M_4P#bHHS@(^Wfkb1#Pdozc8V!
zQxJ=gnWAig{h36CyrRkeS{B_qYC*5-pHTSyWn(HDSq}@T(T5lZ)S%Ywz`|(VtvbhN
zF6HTE15YP>^F8K|y3WwGw%PSxqk}{kxUW7edFefOZ*f>oXl!P5By798ejMaF}>Z5>4j9d
zw2zc`!0vmQmbJB5Wl~oNL#!OzYqu8+FH9*@PL#jeLnTIwMb9u
zGY`RK6eKnd-eDsh3I|&$Ud*lQnbcL}6Oh~i^_&ORc2#=stlW6Po4fyy^WI_u2^H1W
z#!91pkKfE#jsOQ!!_z&8BDcxL(Q9}@1T^2+^t{>^7fjDKH9P-}QZiSs061-FVLvJL
z?L^RNAE-s<(f%+*hN3!Qx$Mo2_}K5IJG~Fs31TSzzYzl?fob*V96a}r6eF@p0qoCM
zQf#QiWY_FZ(eaERy@RX_GPn#L*Vf)*6_8B4?5-Brh4=&LaQ+aS9(K5LFz0x-Wi*U)
z$yoH7`@`#JdgWBF!DTm&QJQJNKua7l8Y*}RSC&zvqZOlEVG^s#)t^lM#sZ<3PBH{n
z@Z04VsvEd3q6|~>0cwa(;li<l5g$rl5q%HX9we>EI^UuuppK}r`j|%r6|<5$-DE9
z?`r3;3UNiY^xk+`V|I2>PB3(&)eEyW1w4x1Fn3EfvU*j=+CXDQ&Rb1~A?c>*LPoj2u
z*cl;Ik;CF+3^pd+WpG+o46;674gT9VznNqk_2J6I$}xbd?c)|xK}ZFTo!=Oj+g
zkmTqk6S8o4t}Y7DA4AG8JHmgeY(_uybUE1V0YRya~cgmUP
zH-0M#phIG^WA;W{4=K+b-F3HG$2nLrnWKBPPiVe|KjypJ++YF4IneK4(XSgw;S9CR
zxw+c%2i{g%`a50!0O$Z<1og@1&%bU(QwT1D-l%l@?M>4h(5=D_6lpM(<>bU9Hs^^8
zt!DJ0+R(S5Q3?qV5l2gYR+omN(5mEQt;2Z|E(o((>l;J}iWRHTGKXs*AX-6lN{q2T
zX#qc9Qy%5Xvmv7SKyXxqst}hzD5H
z>{4-|f78nIut!KB{-0h!+PXo43B}Zhb_13sJQv%eg~C#`f%lvlXPDGHUGGhN+c8nt
zYRJM9HtB}q?QUk<8NDE$-7rWqUw`8sk5H(iPPm}s!oj|knBFT$h4GpKIGZGNeJ*fu
zI)?D|u(3Fm{}xpDuv(;L38=`--r{fWi`sFk-toaLftICMI#L3bN3Aw#yipMm@jQ04
zOU`riT8KOL8&@CYx=BWLvUkyA?zZ{;Qg4UMj2gl@d
zVddpa;-D4hb5p;BxHy%^h1{z*fKEE~1bHywe7(Zl;KXqb@uMLrneP-NfTeOA{+wU_
zq36H29nCUfP?9KCuVrE7M7Q^|D0ig>ge(AEWso3T5o-kBIn;;lj#cMjom@TQ4d;L^
zz%3!tLma$IcbbHQ;VHW(Q>!fu3fa46uu#rFKYGYXMdY{kTs@+EyEt`POBV-W7T7^q
zIkg;dMes4v(H>_952iXXuJqv-S73Ng&(Z1GuNpA1otF#Z04fdNK)P#}ksG{IHZu~5
zjsbC~v?W?{ekgnQr1fDbCx#|st%AbWxEAS4S6tRU#99LD)lcOsIP_uYB^kp64GSp*
zN3w*jAPA#*fJJ{M_)@fdvo=@Hxbklr`tZMK=+ms@^nsI!RWjV4nV>9vK%&YEN6CcT
zbE=%#a|ac@;)c`AEnOvFYL*^QM!mN{q^t>z#TTwFQ%=LPma#JVF*Z7^pr}&C^L`Ts
zxegS340Bptf&nr`DyylHu@#CF&0_G#q~fLcR#y2NO%?bc)i4tEuH08w2Fm
zqV-jY%CX@!V$F6rQ9=0a@QDowu_h#J901P9ct?!^%;9P+|d*noaZxi<-A1
zk2TOX!$0(Y!3vtNc`_DZh01p4sgKxsfSfa?^-G&&srl1RI@y)~0O9TcrB-RrFj)Y#4qe8R4AK*#fCF3++LPJY8sZ6+5d^%)5}?&T2NleKSzr
z22kt*X{$#p0^cI_PsY`5`^=a_9UN09~VlPy0Hs^fyXMgM((izqpR9X-+HZ>ZjW&dzuNOQ8J
z$@WUEAYxr#U;pLL{VaTpH$MvsfOi?~+mI0v6%p_YF$q&m486D5m-(GUdCdQL!#wKU
zRt$^P7>jiHiOfD5ooMYUeV~2y+jFl)6W?me+%GSvsp}AHatlM~YYHvoF%EyL$^dUp$FVxeq;#)uZ};`_9|1*)8b6g(JS7~~Ukivt+EJn=d0Sj-212hl+a>lW
zevQ$mR&tu&xRn!ok`4)&T;W9at}?=vM{k(m{fR0e5Kv
z&er;xB-S_KUo*_v#cFlkZwRz)tSLA+YRo&BVxc{K1F;fC>_W&$V|+({p0XUbkNAwW
z-Ig@8a~&HRXI2E##PkLes%@CE;$ExYosY#tw)NFS#!4UB?Gw5d(S<{VFex-kmQlH7
z=^(2(N;_AY1+X@MsgxCe776j3i=WpB?NC^2vXb-F>ngNJ-_a-*6!s>skloho%BhhN
zT#J+3BZUgio%JF#JY~0({Z@^$IF9{qw(8YYZX2m0^?2?`qDY)gv}tUQU4pwjuGFCm
zx9m1JGt21k2dlD>1!V_WCAG~z(Jxrti}4pStQ;+%-A2fpHF2mp+C#8|+Ql$58wTg=
zEll=ch}SRLNW#$%eC7eshz#4tauOfsN;t!j#s3tCo1U^Eg!0R>1=y_OREJAtZ(&m`
zTu`}beF$hy`!e_;S&LWIC4__XQ|ie3a!Id@r>t~Cf5_I?IQ1&mX)`PD?D^w>GKVJ7{Rb$
zj5mglDBk8GQV}G_r<~{O_4A-U^jcDg1CVSo)Io*y9nqES-*(ByHJ2y3t+&gw-L7rq
z6ni=i#(ac>hhOkGPvO_OIyz{1I#r^Rf6m!_=XyDf5WFyRpFhvEqaE5Ai9N(V`X$m`glD
z0%li4wbqi30)CT=JvOh$N6U%W`8AT`@Us!o4ptUGd_CBn3v-#!Djwz(^AA2Z#hagJ
zvu`fI6*3$byArYP3WiKcRx!+Jd9NVHh0$y-9TAS?qRWu*kSRiW-(QraKR24u&h*jk1L22Csqe4cUcYa=YpMWS^AvUD#IUv^6>fLVsmzt$hBGr4=OL(-p^D
z%CII6%9knC$}I)Zx%tm3dc)fE%#
zDKhOGekhE@-yE<_*yMB|^}#M~x=vyboki9H3k;W;s}@(ISfuq~hiC+=au#j*$W_df
zNHgZ5r!GxM8G+AwJ38{r*^Y~57h4G_Lm^(=)!`LFDCIQujOH47d8;}xLo&cuub2)gw=>RzhA6VD!wt?mw|#
z?1o(&oc(%jHA{K&byF%Dm3lKdFqRooLCD(bb|yuSJ|a{hLsfluFWUKY0*cRa$20S?
zDbnPP#aBedOS>VxW+v`SqnS&rJdfYW_}QfkV4NW(MX$Tx>zv56d>F=fC69mf&YxVn
z5X1-t1F^B&0o?rBxen==#_w$otzg|w!U9F&^iwg`W%Y1WU$#hi=JId}8=t47mnhEb
zyAWqa&keS4&ZI2+o38)qmxDS|S`BzeF6YYFCG;dxZN117BdZh%H}1U^ma+bEwwsGl
zPxpolyMENBE3Ol229C<%9Uoj>NBmyuapf+QeaEVjqpFr`Rk%qw5W4sA)J34~oeJHC
z2JuFH=~LzkHNu#J-O8@Ck@ur}K~J14xTY>=F6k|GHYgJbBk`)u)b!P1Z#86!_nj5o
zcs=17u|LM68rL-P#6TU|cEjkm=^T>19^w?=R%ep$>IH50|`VmaXC4>6D)p^@gKFwrkt@O|A0#iA8OofMB&UYIt%)LVgmX69I$QOd9glc&VXySI3u;ShBXrS+Dourr4LY@O
zsVZOe=VI)oI%4M|t+3e7o@kML(HFF+_cCPqxN!E3zrNhXJjlvB1`7(`1vM-6yxI3j
z;dSeu-#@yY+ORm={dLq6d2X4bX}DB*XX&e(%sDf81(SpQT(>jSxmZ5}BT8&@t0zl$
z23}X*WgN9m#kPM9K-q0oGx8~OKAU5KRWsYIq`%g*C}+87~z{wgwp`^4!zWaZ{_0u2QXt$yUELh@FE
zDZ{dW=XBAb$NlidwHdPCVYwYn1~Nqa)Au6gA-75``Z6AnH%VBZyv%L@4shN;q)w=@
z=450pXo>f!p=1zvXHE3l_1AK&U8pGy)LQ+UZL-MVUUb#2@sqQR4mYN+2TgZ&VSWNh
z3iRl9eR|s6(;BsA4JnzK!X
zKB~y}l6tFJpEsCV(760nJIw-uQ~)WjkaVd%3>_+@{YegWzB5$KpDfx+fP}fNnPPWZ
z8~5^tW?P}rX_Yh(@@X9ZLL7U{NHA1d#ih!PpAhYVKb&r3wV_jC6{kA6P6~O-7Xdx|
zl?csk^Qa8nv&Pd+Oy+AeAfqujDiUft6_;P!5afS(xA=6P0y4FQ8N0-($1RdGCpLic
zC1=ex?$J+?aXdW@A$KO1SK}QRmQ#wqF#qG7Tph*yU}bxoH@3wkUv_EU9hSb#tYt>mCcb$?)q_Dowt2-tCZS7i6V(Hp+TyR=jZ8mcjn27
zZZei|-Ye!9e>L`
z*5a(=2yddRR2f%kW~N9gqPu3Q@c6lpo96sU1*^HabD{motYh8O1!vKT7l}G9L*^8H>8iYU-Tw3S
zPGve_U4qUOM7z3>|S{7Ckmbatgz7IyE<5A5}%0uLrS9;=A<6w*Se(y)x5TW0W{>4{O{xj=Jq72VqW`>n7lqCA+>Bjnrj7miquPFA$
z2;|FTgs}*`xb7R56{&!E=U@~)EA~!Yy!kI`{yxkd8*NjD1bnW)qtdR=u~QAu?yq4T
z&Tz3^L~aq;_YG@bQ|If=s?vF_ck0~6MC#`T+jjPM7E2f)$)k6-Y*ur>v$tet)Dz_0
zZ-+l*%@tK?kbxGe$?rAOZ}!ZzxiAx!SCVNrE(xFx4&_Xv16I_)6Ky=FrdoPYE8Ko7
zS7j2FV2%*@g<8T*=1qMFO_kt!Qaz(+%uZUs)V0{&bnnsm=
z+t_p$A*Y1$T#sMthw3?>H3*QNo)(1Pp)D}nG?nvpm4Wcu3_<7>K;rcj@%N34Q+u+a
zajUS#nXQn%EqTinPxE?O3;)Z?Z$^?yw$fNcd%oVVV4I2Uh;m~8>9%H`#!$YkC4U6e
zrZQPdj985G%P|8qG&~+TlE~8C7jaN^4wq_T>TjAdl`W;7J%Wif=X=~{&re#f_tc}~
zKYjaxSmFUWxxaPH0+=6Cu~RbCbP$&_<4~;IC`=QvEQmamcK2hbmcT({7K{p?^Q$k7
z5HR7F<-pzizK!UTw^Hv=F*H$2rQ8^kq9eu~0#NOg6~2Zcwn$7Exed(wYw6C)cIFY=
z#c{x>?qn>fl!q)=kluN=C1WdyL#&FUdC3i@`O{e8jGwtL6b6bQKdmf_sz>CP+@RC3
zkf}SIgnfqYkD;cP0Wg_zjmCc-fuvhdMNA-rthmWk7;EjJi{tP9Cdu);X&qhw+=TqS
zvMy98*BhgRCgVCQ;>WO!r!FdEH7*KCZs%{cq_aSF+s+iw9YPlFj?z2m_
z$aNBtF&75=)VC&tFm@wKq_GdHB|y`S?ttiPwrJ<=I+{kB$koPY<;>532r8KVJG!wD
zT~8|-ZoyR=%bb3;vwD(E-y>b5G7{GpWoZv(3F;Ok^0B5YtH&(}{#xR&2$|6WpiPjd
zqxSdydQM&7-c>o?Iy5ET$VIjCf@E_ZDb%-6S4C1g+G)A{dFuBp&rgeeC!y|U=f@y&
zDx}NJIhn8~`E0`2V;R)d_#8{A7Toc^fE0S_ll2Bjvo;ex63h-Ivn61bsWpJpG4|GO_-hEvlCbK
zwLFn@5qKOU%>U2ru}%DcZ!h~a#@y1)!x_sMSoEm%g1F#mk-gX9R
ztQE=X>c6RX7xHjY3mXV`j6xBgx~{OQ=hl8HMT3M+>*S(aWr}Q4-PHK?N(W
zNCHK5xZm3h*BbGihO3i}fkM2G7%W^C0(O3#}e;3#b+B@^!p1;|_a2
zE5f(e7rTnmH!1=z@sKu{9#nQiA=vlOWYrv_;jvBE@73MkQaguKFfI&31y=UKuDIlR
z%FY%>HGU70XP@uoJAXdAxrJ}iHh!Z4aS<1KaX>c&hSP2(#GCBc?1CW>hOd?qm#1^A
zWwYNXWazQCCs;l37`RwhcO}~d#{y;z?}a;KE9z{Jt=f+)VE`GNFmyzuJTFg5k~J&I
zxb$ZdTzfUb|1Mz%6>ZHhm+}N!Mn}}-5)tarHjiep`E%6Ne~G_~pX=)G{&d?8@7C-z
zzjsHK!LBxi-^R_)m|nO1pwVSvbEXlm4>o*LoWlK4WIOft7|VU>Ku<`R1f6c~ta0=5
ze7>Tyc$!=Xf6W;l4gck`WL!EoxW+KFzOkleBSR7S
z1LItrv7=&cFwa1giFs5#TP#nhe1VV7Njg@C$c!D)!PWK6fKQPznzW7EvOGm?id7Q(
z;TfO0&a6!^G>|pF{J~)0qqUVKV&a?)E54l+)gY}LGuY9#cEznJ{@4_YA85`=t8$y;
zCmkC>>MO6R)LkJoSN0}1!`=VJ_*wPd5n5S&C>%&msvsVy(-GX}aR&+261C=PFkz>1
zPKADePS-(qU^D8Kp(=B8&aywqHGEv~B3Ij>AsbeMzpUZ2l-h5VPP!wW?Oh=f6`-5m
zcx5l5#+giOrHY|T7ByyZ#6j)sf8!I7)Y}L&gxIWR9xlnI>h(l*IB;$8VjcT1^j9ss
z%>%TZc%Wq8xaLB9LT$bavI{q3KG42z;AT%coxZXB_?Rx)xLm&Q^uy?5V-$8q``pp+
zY8wl{A@?7z4<2rNbDzfQ`ZV0!)!TUApSqnLoKT2PlWSD39Y+)Scq~cHV-sHg8Rov$
zE%G=&UJr+?Jygy2d9_!L51l{CB
zQ;E&(BQFkkTE`g`47$DNZcOcT?Bhvzeh>BJTDt{(9zBuUODE@g`(bBa`+6nQM)|UXkboJRLs%_xNq27F*zOV6{52)Cj2zWsqS}6mXVhBub1M`(%w^
zWw2viveI`%RZVV|2t4wzciLcuY4iZB@O%l@am*Gh^J(oBdSiQ`;dIJ5iLSm{~uEp%2AHQ1=kmN$G5cZ&52d=5?mNQi%
zLTm0*=h2_!U75qloppGVAVNq_NfSH&#zIuhvJz|DSz$s{al|)a!4Fz60TFwCc~wPQ
ziw;j=y%QiXwC8c_tWoA^zuNc<6_&?iyFdTe%H3{f0chLN5y;hhr-%f?9?Ve9E^y{|@9{-#ySr}HH;ePYsI*8YlG&n2yyGXrSINc2Ki
zmVPKRmBEPHVYPbMaz6jPa;lY~U9sys{+Y~NBIRY+xcxAIV^9C0%4VgM&?~9=*@Q=L
z!%CtKd>@vr>m_Z*#(mM5-xr}{2NM6xb;Q7b{U)$CmUw6x_out&+L_86R}-1Dj#<3K
z3VHYB%8j-z*eZ7;+qs*v+f`|1;B_kfi&bi-$W#&%tDP
znvp>HiQQcoXNc4j&|FmkbL0iIe~{|9c*MRa_tnqfDgFxq2Kn4D@_e
zkdLSh26a2Yc=4%|4qw!J2)S|${h@@cEI&T>e=5zg*fYPK@>h7dxx3zn*Ph-93tqIG
ziBHq8j+7ME^2uqQWc!0)Y|QDt6dZE&#BllSy)T7zhS$al%cM$few>~9
zIo?)FIMzlFnc!>(M52F7=wpx>WkWYi?!Aa;aFF>6y34s_-MedTdxhmNm?oYl3}8;`
z($zer{5KhPfAd#IP$ZLAucaii$}?+*6lM8j#wK;wF6@Xg!GIm0Yp@rAzxCJV<*rYA
zdmfQpD}*$(-^!IEvNu;7y@8%Bv^YE1Xk1K0%9{+ks572L*z(WZFfVcB2J{pT`|lKp
z@SEQRt_y5O+Xo9Yvz>A8fK20?p5A!RolPq*g>e~?KH{BS%u1;
zFm6J)4uop#A!zbr<4}orcSF5+6s_7)EhEHCX@ww;%r_GAB%9d&fIeil04jMk)*e=3
zJ67p@uq~~|lQqC2(pke%SuD$+(?3w_--1#`BSJ#OvQexisaAy6XoX$QL$YkB;uH+l
z0BFNl?U+Z#LrAKzceEp%K&lcHduKuH195#kxnYDwYl!swya!~
zihgc6ofoddi_=r>wLu<7TzaMz0$FZF?$W74R&KgG$>fY)J`U!}z0ylZT4IbB`5t(S
zsVN_WommD>Jx&hjEf-IaQ&hZwAa1JYg`iV9y_^OHw(!6_C_6k9KX+<&LjNBc$q|0_
z-E0PJyKoTB3WxuKFoIH;ISs6?(>URUN7AW9fy*@%VB@N;2~_I?rTl%o%x$KaY&CdyIIF#}G14HVXudwq^`tlkLKUbsjZq&Zd
zD`w|w=w}Pu(s9uyOGJ%1dt%t}}e?QZ=@#Kt>GOx0@eK+j6)Tu^@sM~
z;JNM^lgzi(mkv_Zh2jyNeM7Xtw>&zpM}J?NmRkoI5f8}g-^v6uO&>W4)Hkk?vh~s4
z?~Uq6{@`<+aW?zF`LD*;-G-H+o(nqvkTOcTrpP>1bIZqnp`Iux@-iL*wqFyKxpl~e
z&ObTA5|6)?b#VPwZS>RiEdo+jiRVdaSm~}7&RB5^TDrLV?kjB&o;F2U-k6s-trJ>k
z?Kgm3(s&Q$q@bn+MATHATXFuuKO~H{Nq0tu`N?>E7;s
zv{`3uw8+HMu(U{V6GJOQkHh8xJ@jhso}
z^%(R|Nj4(%V_VpaOETB`de$OB7e)$OI)j7Vf-W75$7;_KV&JU5ae9^QQ-?V*ot*qd
zjp)z+nEwUAdI5)F0yT4KnA(Veb1O@)Fp;u2I==@>g(D;ItIoT<=elVOL|pMYN1H-B
zP%&!?KE6Gm8%fuMeZ5^!Liyc7a(NDdv=f7`U{ib@zSP``$yz)g|5<}Ezf|2F6ZBVT
zyX6f>k+!kudLF`&^lua)7gX~;Ee`htD+=tQr#r5RsAX$AMzyoQ@az0vj-x7+qOm2mt$I=le%OEo&WNT@)l
zK)|8$c`V9E9UyRG9Fzk=Daz~Zkid`@qm7K9;g@=Zn4_Mf?aJ?DjqZc}TLbq;Ia%3+
z?$;=*&T8s_9d|bd9TPcW1Z64v;G3b{ojMB&`dy#tgOPr09m2A^Z>*-WO0BS7wEiG5
zXn4Q9+!6An6nOqe;Pjn{@Eg`AtWVH8gATquoF=wd!>P7eBt*wmh_#t!Cx~=8O)No&
z8(B*HV+F0Q>W=fFQs+RcC7^y_19IjojFsp4FTqT=j?s@mIofeYXBX(5W(36r2Is0^
zcctW(Ru1%fl(*We+9k^n*m}$iXQOzBAwu*69uKqqvJwuECzT_8sb`63GXwhIDkV~v
zf{DD<95x6>>KGL>jm5lA`elNi2cjR>3(KnyPh9Oc3_@;pJbxvOId_btX{^HVrETx+
zjTE>E0*y?)9R)Lko>X0hDUZLc9G|)al7#8$>8j1wzogfkMNkyq87K}k@ru@jC*R2+nod8uIYFc$_
zzNepzgWrjxiJ9&ASgYe!OD*-fBQNnQL1>D?=}AlQk65{vFZK3A86Q2E}q^w
zx^jH5Y1IVuW`L?SSfT0U5hr{cZsYz}MB8;Ml#mLm<9F6Frn%WR`>Dolf=Z%R)XJh}EI_o0GLju_3
znDbig5G#-@M8RVN!q7El!lc5}$G}zxK0?1B8UVKOZR_+)ow5gQnhu_)t_wg?udS`^
z{LTQV`8aPk&R_20C~E6wgI^Xri{8=F?t-Zuxp0sE{?UN}u`)1&L)Qj&9y&MpIck|s
zmS=mFb8Ed4v|zP$NXD{R*YNW*;r|vqo+stNsH8M`=9W>)zWdg=YgadTcC@Hq~En(uGQNEW-`$PY-5;O6}1Z
zCG57#+E)N5Z0dL(KmPljJUg4bt}8WQohawmqtZX+wqv^p7XUjm>rv+f17k;~OKj2t
z9l#MTc^&n#O!JQNm_DpA;kBbIrNXh@#c$+3QqaM?+bL_~#@E%d6*SI6c`ji)B~D((
z4mz6mPj@ZGbTlC|skoa~z3ZFGl2lU82+y437S-<_J
zYipoVwKG|S$7(+QU|$W#GR_oL%2m62I3_LzRdwfD1s)7+^_tBS9yU#N%rG5RgHGh=
zo=W|W7uYs*8_WIfs@V%u#agvYU|I~&l$KWeIWM&VDX3@+{u%wbqNpv~EP4ctV9GEf
z-J7l2NwR5bf#Cl0d|k70b4|f)oT7SES>976pl@FN@1qyUXgFwnkR)rqJ=5GN?%Z#_
zPfba=opTt{3;E~2AJuHe>3V++4_}1X5D}@Y|Mq<O^1Q9IZy`bfmDI%1q)!J
zOPSNZZ<@?uw=8UGx(QkI?t@^0!FlB>1-wA9yBhtUS9~S~GQqkKYXbf$tF|B%ud9<5
zwkQ8w%bnUOX(q(OyY6FXK5#fK)@`!~{$(NON6&ZprZPwLw|_+sv^oR06K{du1=+KuAeQ9`wPB
z{=Ko^V1z?BKIf?BBp5MO2GmEpRQ@sAM{j{xu~NZ=$xv#?i+}&-n)2SD6C65sodS&y
z%As%GKY^nD^I5YT3LqCx};4Uf(8ig
z?h+(OaCdii3GVLh?oN>41PdPA-QC@TJ9EDNlJ1_>Gi(0KxoB=+0q;Kh)UI9nRMlEm
za`Nr6%}=-Oq3JKC#>S^BO{mL%2(^?8$d?SMB2^u?UJu@`e^Exy;`gF_npYEoNH?d;
z^UJ)#KrbY(yVa12cfWUFHMBdNQm=9P-LSv)B>*@UViuNb;90$&eshs;gCXG905glO
zqQ;p#fU!r+CVClvR>rXd6l;{o)hOHgeP91yvk1I0fxfkd!@vK*x9Sm4FO7#l>o}lM
z@#9o+%ip}DdLHtJNiUpry&nqs2phMfKYPDE>;v;aFr)eaVf`=#{k)r^yS2T2ce*T^
zV=;}$XQ$#+%Le?ir7z6W>fT2o0z;LPcdq<#IjL+Sw%cu#n7fYu!5FF}qhW`Db
z^$CDaIxCEr7JU4dM2AZJjhLK$0YI17@i$
zJpLipDbPL7m`NR&@d|YOTI$+=0ge(FcXU7k49Ye<{>|OKS%8-NUQ}cPjJN=bH-PD0
z<)zB%>VvO9U#`6(D~birebgJCb;LUe&6NbJw;^db2>*kF1Ayt~y+m7X>a^X?900u+
z$wTY41C0N1mkd1YCMPGKg1(aq%(7JU7Qy^`JGoYz!06fP&O$$lH<^-0apAsrmeJNR
zdAJtrZzJa@y|3C6tl=3AS$bDlvNKPfEnVE54YslkFEnfpwk8LyV~_6Zlh)i;s@YFk
zfc-EiEb=(;+k60$(AS;ZsQMJBo)L_HKa6kLad#+5zI(V!P^m0pTSiHZzVgdF)bJqe
zuy(H_us~Ch0O>kqeM&d
ztebwa=n%eVVSez1Y5DxGh5j;vV-u7YDcXCh8Xxv|t`&#rq`kp>hAwsNn$!;gt`-e^
zz|#Nw!M+Wa(?$K9q{4Ogi;n=h0T@yN$Ihl$?1e?P?x(
zb2qXFVyuiwy7F2R0H9^`nLSI<5r>T@66T8eUt0hgO#{9G5c02c-P;6W7J%1Zj{?0jC+RB7j
zxL8@Y#Tfy4PxGHttpR2#yNE?wn#6Bj4f&UxI@w{ng)DT0K&3UO`yX<>04Uc}12DiT
zei+Lwdl!*B$LHu&;d;*gIhV^oE(-+WrHQdvIN$il1a`;pC@q
zmZ!3-8ee2?-5)1b(^tz#DI}E5;8$3AtnLC?90G~8iyd8-@|1^4cF}#(KKyTY6&<((g@cDc
zVw3l~96fS<%sWpX{z_RBe2rh}w1DsX_t(+vGfl9ijozq;vAI1*ZOsOdIg)LD`$BTw
zOb++w@gSlaJK=fB)Hb7IC@>J?F_JeU%ab@k)Gt~DG(~7tgO|-P(>PAD@I@D;Xz>8M
z*o+Nzv8w-m!r26|D!iNao<1PncDeIO0t&930X*)#k94w&@I4?$-UpFVt2LCmjDQ#Z
z>U*!=pc7&pkzHxM1^xIM;d7(
zvjc*)F1KOguvf(oG4L`AjEp=SQh7-6W3*GX7|ycrI;(6Aq|&HBEHF%=jeSxJ;J708
zWF34g%EDs2^ptHQR&33_osj8a^Lm`5WjmLjcsXGgKql2DNr#y=fV;JfJsJP3?cx24
z^Y5*3qo!BnMtNCB
z{F=|)v#m_OC7V`PzyHJ0I&l-;51oYF1mvQc22vSvKu#e5tseex!-hc>3g;mr(!~`j
zTnhzM6;ELdx?8SoC}iX>YYOQ>0^Z;
z=?_oC&~jSMVPky6kvf?PrhYV`YGQQtX^d!KSJ{LZuwUm3_)hXDiu%YnlWH?
z0AXHCNvz*QL38-51YkP3%fxaj6{i-Pe*)!^TdjZ`3%Qs41}`>G+cX+WfY)J+52TlF
z5Zeqg+|<@c(luR;a_d;Q6NkB4)t*Yh4EfX>l4*dr{E>DPInu*$*-CVbaO#|BMJ%Sj
z@v;bM%PaNHA9h(ubX#EJ3X8h#a<>-i?JE3%WfpCvlJ`9=^kNxmi*=flhQF
zg{g8E8$!o<@_(An6UNLboZHx{QT_)_^J$^+NRuc@zT%IV+k-L
zjSg?p_u1LatN`3(EV~WBi_=T6_Hm7n##GVpQ`2tl!c1C>=UTAwDbMf$(F+u-{>>~M
z6=@;pw6s?n4|lt6*8lzo8YZN^_lxOUYCtJ2P}2IF#qVR6-2inB;IUS`&V2H_Y78A=
zlfY|h{t1y6Mc=(-{1Na+h59}*A(R<
z7G2VW!!7YJEhZIaRnW2oJ@S5H@#FS1i>1;bDK?&hKyvt8f&Uj`lS7`mP)W#v{KquM
z&z}oHM_pt;u%?`bzUeLcI*vZU*{@>!)5aO;{p@%}eboVl-jG$9mQIusI#gT#??@L;
zb;PzW+HZ~djf5Kht1m5u26*pEbw2&Ih3VOLAZt1ei
zuRm;SGDmmAfV#dJTj8(vpXruu`df&(nzF_)iMO-oc*q%5=v-=W0b{o$D_R;rV~Y&C
zd8*}S{87%nKuVzYWSj<0;F%13gpG+bBbaRJ&~0=kk~?8UATQTcVZaM~>nVd3r~=4(
za!3m?$0H^QconNX2q*wRAm#?fvLtkVI%>1Sm{oBhCC^7|!%u&io?8-oMC0$}((#A5
zM^W%%b8l=IV}8B*uc(Zix~c~LfR}X#s=c`PwUmh}d=p^aJn~qpSFU1cYj`r($zYrQ
zxUw<3&fg-3;8X{DOjR@(3aP}WN$0YMZRllB67o&n`N^j2PIG*2!AiIq0jDGVhd{1P
z1GIEL29KP(q(V?3`3Zn_*B<0DON{0kq$Fcr}2#1fODMdCT@jwCTF(
zVz&^@T1A~|LzSW8VLhO!iGKVxwkNhS*YlnLacrrCer7EVprE{BuXV5N
z0eydB09(XM*EVWfSsFXc#l(kQw4ks6
zDBW}NlZdo<)sb->jdK_V1@`upzd~FUSo>6LV3LD3K#ujJO?ka}zn-%x(*0hyDwM*X<#eKrYg?pU(ECQqH+c4<
z|MZTT_Eq73Ra9Gc=UWE@*jUSAJZ>Nd38dh9t(n&J+^pR9euJ3jjflGbMI;XFR)bL0
z-H!5dFhh_e*#rh_esM8auhMF~hHj#sjPE-yYiU
zo@ARA@<%-c(KKcK)*LgX&k|EC6{&6;4nOheyI@PqfE1cb1$ytpFa5A+9rnM~2Yy4$
zyGZ5Xe_v6J_R%p-lo_eK^(Wzc(uzzrQ7&K*dYFYFicLKJFxc~(OWQ|&y%Vs|%b}Gu
z_#@DVCgA{yzm@G-S<1*HcI5OgmH4;+ziVYJ`ZZCCGI9ey+0$x4n(e?NQ$!>jq(d+cgkp@|oAuvg~s)V9rq7`*V+J0B=)U6Ha0`xEc_
z6K!$JSPlcAM`REVNKZxIp6b_JKmQ#p*>(#A?oe(Aj~w>6w#lw%D;_ma+nOr39)>?
z=u;c=!OM`a<6&eTo`eIO>wF5@b7bv1*)^1Z+Ab%LH^*`M>1~S48r(ma*_jiEmlt$s
zgk&?8bnbxI6Tp53aT|d5TM_TJz{V~3k5E?B8*!u%H6AA$5pVn&kp%cB!R_ytf43skZ>e(btIX9)juMmTpsO!6HB
z*s=#{z3X0{s{O#Dg6Y>|0B!e0t0iX!FmHiyS?zTxaDV?Qg99>&8|{FVz`-PS^^fSL
zXY{{Tx|!FNd?9s7>1o;C@$ywy)PvvP^qbKfQ$cTTtTU9nGB;N;ahU?l4x?$H9bfU-(pirZ6&d;VUmRfvjOfK;
zZKdTBs&8gL6ACJ{+QXGmQ9GlKo-P5mX25X=CYuT^{FARifZ^%*pBh=@4s{drbCir|
z#ivJ3%X8Mr!iQSNZl0GH-ib0>Nw!^SYbgbwI=|I~Z7GfRMEA**xIbjLCo0P)Xb6a6AKe8
zgMhiCgPgIwppBKSjkU40BOyC1gP@J2jlH6+z9F!nkg>D5p|PmFz8frql(DtxUyE=s
zeI)$H7c7H_xuv7AJt2dLrM{!Fkg=hSkufX}5A5%Kx-Ok|5`hA521fXnGlwdY`zJs2
ze|_2i{^7qA_%8+izoNiB%>P&0`R``_OM(AV;QvK|jzJJx@kQ$YgU!J4Z#Khkh0gSe
ziSd7`&@0q69hUphyyvU*);q9JJ9Vl%=1v{c%~v>?-M&@GFr)5+5)&b_zGDKRRDZd7
z8OzpwFUen%^~k@aGoJl~0JJTMqkHvE(th^#@w!>Xk*Br%62%g6yc{&)_Qk24(Jf9Q
ziJ_7$LKh{A-ly%=uC;zZK{FZC1n&ELx3qKvd;J)=^B*CFvJYY|`MDmlz!D36xKR@0t~dX>fuMO=T++2hFMClgHpkqgCbgm;v4X__ENHFm3}*>%NgL-+fCo>m4`tS6H5s?5e)&oB|?3tJk4asiKPa8
zao&;fB1iC~mJ}%EU^XhTL3KDQJ@QS{o2W=Dt&Xxa^RY-A**VXUwTvZnO5zxoGis2JidNtc%z93x5xqMj$GZ}#2I#p13ZUBQ0*ShazhbTp
z?~7r0tK2@enov~5`Iln8ZAU&z34e1)LQjiQ2j_>`v+WkKclB%fiXxVktne|Az%;ww
zJk4Co&eba2+idA(gEjdDF3q_mpLbt-gD||{%y||WjHl*1A27g}F#_MBYHxztiP{o*
z&5eyDk$7}a+f!$#Zw$s5k@z7Tt~X8>^pcF@XGTJ8(8$*Q9%goc47b-=SK{UbljRqi
zi#`%>JBfm{)(=il@&WxaGd9VEQkkq67iBsoFAvn0{uG)UOS3y}r^G%Q;CWZ~afXr9c{lP5*$Rm;n18qMkr
z8S&Ko#ci-jy!^fbzWjL5xXZ7hP1m0k_A;d&+U>aKQxKaY43?3(!=ze@P{h8_`wIQZ
zTg#uLndR`9?~WU*#Xj4_F~Eetw`fTr7Dny~@uRh-QDH3UO}aw8ZSzeDJbt~3yi87-
zb510zsPKE8kpDpBBSaHTzyj00=LVTb_E68|eYxzBwZ2PyZK~a)w5}W%KGGBTY}^E9
zH@KYK1l5kKx6J5zP)gsMD6dCwHFR}dx6DO@Z}lrtwZKX^zYz)DSV9Oqo$In-U8SIJ
zK`NzeB~(7Mu7}EK-@d3;>ToH&Z#kgF&GylTwgtn|I{nE6b*|yHkA)%6#Y@M4oPT1p
zVRY5`hSs>JM~l%o!$kK-VwZ9vX!muqI|4n?e*5w)H<6Y7ZDw^oj*QJgq3rga)q+#s
zdU>3b+t>~fRlgpWvTC#fEr(4kMrTF5E$4goY8-DwH`1vtY@2n6&fqT{8kZEss~z$S
z(I;O-I`yr_pVwwS4-D=2$1AKgeP+i8iF7cXt3|y+N}Gy)U%p!12(wj;fy$YMCi~HQ
z?7*Pvt!DDj#>DY4B2s3*8p4i06fQOn91#qKI%&yb9zWWB{dvClk;g`5Q1;x%wbyMb
ztE$=)&rvX;AFDjb-!N$@2{lL#c}+rg7%>Ywpbor~MhTC1yhG3q?C~3RLYK%I7Yc$H
zEYaq;!5-^`Mpv6m?g%%O4u(kMyi#XBHj4XUR1qGZ#Th*V1g-0WY{2Gao&PT-|#YM5o6mlhdsPOj^1PN+JxLt0#
zZ)R;&4U-{m%_`!iJFM+n)XuBCA6PXltHs~JTH5^Dh&}_A|1veUy7e;4+eCp(!s8!r
zu5z&dPLEFx9gh>^XHEqtq$(O=RvNX5a(F7N75bxt`283AqZ_R-x^E^*?39uzmMsZL
zl>^>gI`SRkxg(0zd$i+k1h?7zRFrl1=qE`ICsCoj3GPE{MKPJ-BI#iuTTIfbJ~y~=
zUIt-g(R_+*EadUl=<&Uo@xs2B5BW0qK(IU#@dy{xbcY=7e)PL{<~Wk{$D*`DEs~H&
zxss7>l`&nft6f_yqR&eHEgCaP9QZAQ^*YpPmi-MQk29byaUD$tZF{>JMHP*EgH8J%
z?{VFWF552-Ve@E7j>&RftI#uXWvp0d#BSuxT8TwS%!nD1L^wLdzQ$>v#!w*ct+eO&
z7PS?Mi0$H#)(P<^kCPxRHq!cH2<%-C*m~tn#HfB(auQ@7MX%Wcb1+A)RpJoGP@?4R
zNLy~kmRSi!F=PtbR6A&kRj>$bQs_hCLKPD&Cxs`kC2s#fp-e^geTagL4^nHTB|nid
zunX~Km4sv*iqKP1iP#v&f@`F-`tU>Wa<0RsMlv;8;dv;8F;IYG3|2&SDN($|bp07G
zJ$-8PQ2K|!!K6)~pXMffsg7##+B*5{q?l$Z8l|aLu(j53l3$<;NR*3M@~`OQqbYP@
zX0sKw1)ZW;6nr^!YI@`dhp=z8z5{0)nc(QDgLB`L+=CGc`?>Zmk5{6%5>;EkDsKAo
zW{Pjp-3!a7D@AT`jKW`ByYG8y<|hY1VZQ)$XZD=Z`=1ooVa2AO}L$kJR@o^
zGz7)jm<-Jlj?L3{+DkkgxU4(T#WkObo~2dmeIvGrS+;Ft@fgptKP^k!r&=@@YdN>!
zWnbfh2?KrQ8L&c17zgpJt?}JXcDIy6RJ1ZgFpCyB(?5P~{}_`wx}AcY$1jTbO=mST
zV_7ZemUDuX#Qv*7rp>*m$D6_dh9RchiYjTvqZSUH0xy{1_%E8OIJ}lG8v{}4wqn($sda7RsYWnPYu4F~
z?;t3@umn}%wxbQ1_=WzOF|BEu1@pYxw^_!`eP1Ziu=Z5N;2mM5_;BG~N%cK&;z;OzVA{da
zX*y(Z(Azing{;!awP|hk_zJ=O7z-5{hNyBOwQhGolxzNx{xUH+KO)B@fz^BJn|%s5pD
z5KZudDmP`1Wu;Z6#cHb+;j_9rAD*A5Zz8I6B!5y1A0}#QuU8h#b+uI6&%qU*9=ZQ`
zG}p#+QRUV3d7HJME6(o=&3*YoQba@moUMZci5AZ%_w$NCA)Q0I
z-bUi}ezNC~OK8w93J>}EovfU{(|vdmbabZ0u8XE+iIR$ip$xWMS1#KbfznjZ&c
z6BuMtkW}t>;k|)+81I$k;t~H2jV|1;(opIuYWLatb)nO3_qUI|3=9*|;YnIlHbU}C
zN#NTKGF1&B7`!i9<<1w|cez;F!LuQh?!L&%JzL&4+Yvt#2FIgl5R}!_EFSKm%yOep
z$T^@?$}ZvRhtf2RMJhzhK-z-y=rxkC#PU=S#3ar4L@V~^(67HR$2|%{ouKfYps`?@
z<^;M^4y{8SopF47T1PrU8Z$z4N(yp^+4Mm=TB6dAScf>;mivh@GkXqt<3`Ks9ek~2
z8&_S~v6Mg#!!SGKpQKqGat!YHXuEFt(O|{Ul9JKg6_yE7IQ}S<4!jW~ZeHZ_-o|b>F~`-yxCL@LW67F-O$xXQ^CW3MJ=}Gy_hYttYHwPcpSgY=
z!hX*_&+YZ&hb(SB?10IN=N!+^CcBc-Iss1RbI0XN>f@E9{oEbsefKj<#cq=zPpO3XZ5Jbs#Yb32q4Jyb$VYhFpV=@$=T_;rZgV5{$S(1a1Z1
zr(fxMw8TEpP*ey;XF0%f({s&wPi7TOD{r@c%fb@!MHUx@)nbbg(uFo~2e~g`hufIt+=cD$ZppyT2j>|z#wh+|-^m8lRQQ+U{_FA1q0?+ZIH0~5ss~{YDV0h=b
zEoY7-3(51K?kBs;zGg_xG6dG2=z#a>ML;(}SCrQM0c;oCS8(|xBRM*29
z^P&q1f&s5U(o%P0#`>|D1Of@c@$1~)y^@%J$3|<0>ILA0xE->Y%52o63S>uBc+1>dw0)yWzFJUf1A!{Qs+x)a4uocmX
zF?pirNA1{_IB1w6Ct(w@@aR~$N44Q6kk|+wPDu{LhUNKi%hes-$zK~$1XMWAr+C?7
z?z$Fx8x-v*2w6q9vc9=Bg47D&D^xDi++TT@lXY&VI0&JNTfJ)|szFcwb*$|%LbBk>
zIHWL1?ajdkjtb)%W#0RlP>fHa2`ULbf9+ETGw;W9%*8>NXR-HV(9gredefnHsM$Z9
zSTj4@?I8sadKRJ{oTFu+K@Ca$f-4Rd?A_oKaE;ovl**c09>LnlZgEpl(V=ahd@6Vm
z83^W%2(wXLX#6_*IFyzS6Ju{)hQCK*j344{t8rhzbDEPWKgif28gz%%oz}!JUoiqx
z(2({;m#2&>Y0HJei2!D_M!f3Nl|M?|A@Y%wun<$annnbwR42
z>yOt;VRTWYGq306%@Rx{F9_Ic5Ol465%4dEAo9YmuO#i#?Ljo@-Yrj4MP;8~c*i~n^Cs`~0#q`mnq>-M*N{UV8dY^(X7{
znQW1IxCjXi5(e)>^caKLVe!*`JABOcA*sEaJSl;^v{T@IXiv`=UuiBL566F**!q
    8X{YIVS1%`ul}uzOjA#>)TN4YcSoxwa-uRC+2)?Zgzdunn?gM_&sH;d
z1&+NAI)}q{HcyNe+m>LqQ=sCR+e*Xs+555R8F0UWPHGHqM-XED2lnI>F54N;5`$#!O1Bvar*Z|$L)4r%8IBNB+
z|3IU{@RV;`dKR25v}(ha+WWXkRC2t9o+FukFuet|SPg2>9bm@X0BjSFbOWY-B0kyS^7g#paEGyKwKL
zakj=|wPTvO^$D97(9oF$to?E~iO5t^2TfEwtgn$XlB{S7AifHzpt3{cVOqS-cS4=H
zVXG!PCramS@yAA$9^t`eU6z+AkRnTAPED2k{auBXB*PAImY84YW9-t(&IrX2!*c!g
zSdF;-U<)U#BWXqw8mz{9NI&+6<5&0Uj|-f*@EsI^rqItDN=wp5Xo$5(DJgpNb2ntr
zoyC{=`0R;Cn4vyB8Zt`_ZBtyZr%bVbv-X*Hc2!ZMi==Zw;E2lR%zRLnroNsv>L@tG
z<6X~wf-Ad{WaXfV#G9QFI%ctcA#chV*rYbm_%0Ud4Q(Jf#Gi&iie)$+W
zzNCs*Y4edU4)%ah^1GmGfe^_`YNtLy_DHB{B_}MGPZFJ>IEE~tlR9Wzfn`NgVU&);)m~O949>6B{eX|J1)J#2So0ZuoUpH9@B|3q$_Nzit@%^PuF3mcng;XQ;!5+yuN~
zn3CAhD2|U;^CXCn-0(eFiae_1*E_4r*(++@m)(e!H}AXq-V}mHb>(tuq3`Z173{;1
zmb*0VTct+95z2W8s0vm1NM$}IF$~Ua@--txiu2Cn5Cq40JPU(}$3EPz7rW5%W6187
z0+L0NNd`}3`|)Qh!~!3}sf?#Aq}MBg28;66V}$rm!?2GkRfBKDsggMpTj9JR&mh;TQVmF
z8IVKuQ#|+ir3?-4Bnj42WyL*f**2+V%KH-}=tnOlnrYC-cT}cSux{Gn3J&B6pVvFO
zTL&(mrc#3ROC_*;d^3cjGaccw#o{fp7eB*OX4Os=P!gOg=E3f$ld9@0zq}TuZ>YlB
z&=`8KpRI?QJmo01{S?HhBBf!GMRwiVFzTcxk!;OXr|S9A&CmYjrZGtI97bU$URk0b
zHH)-N@2j*IwiU6aKg_pZN@+J1_g9lxmuZD1r!T*Jazt1f{Y{}YV<_Z$^+Yv7Q}cK8
zuI-_gw8dn7D=_mX5CwS|9pTenX4_uEedPKH@EAdm@=?^e*BrbmQ^(36t`Np^CP^Tu
zNt$MyM?pLnC)O0Rp-W>Zb625L`2aS6T-pOFS{neol
z3l<(0qOHNgGlB=vNz~OEOIPP(Z(A@h2Ym_A(wZb{tarRhU}wl+oMbtP##_3s%Cwd
zCB81jBsK_BDHO{FVd?z-2~*Sl!92)RCK!*~77xYTDwe^BvpKv!OgRWMPiNe_d6T&UkWny`^k5
z0kzLGN2Ts2E^&r4RY1bfXO=6cDH0arRrKMS;uah6+K1)v)fo&Y!dp(?V)YumvJ!rn
z<7+P*$H3@ge9$a4fbu!)4T5L?cnj?_S17x4>nK*iDOWNP+^VI`4+0H&gM-hk#twmL
zI`%RhI%ov~J-d^(E#QPRQ5C1t23ex|jCkY!PEc9O$kW3ys1mG`jeE{dqQ#l~P{|fU
z5$nd<7Now+ztXnTVy1#4YKz#Nqw=~NJuO2Qo*!iX=(uYu9qnUyujL0w(tri6U)07U
zJV@8#hobTqiS#e`D}t1jf#EfJ8du{8Xv|)rxaNB$-Gh=*J0_jw`IMWYSpng(3<_t$j6a}$toACUEVm@3`M-Nj;E5zA#_Sge
zP!&wu^GdmX$*CYe+x^8qk`a^JK?eCV31)t@B+N6>-IV3$C4csmj+hrBMpGhVBFs~J
zoF#S>=V_@%))F|PpSp+oWO!D{6}UwcP$NuvBB
z>ZhKBDn^~krAC$VL`iKjC4Ok+2DSy%=FlBH>0aMGX|}Yh0ml6%3My-R@={ok(JsqP
zZvIhJaQ5Qh8f5|R(!_A)&#K>2_LC)>u$2ryYRM*l3Nu;YSXjy-$a}!m@&Fs|2NnIi
zkb2+sGyeeu4?z-Nk`^~f5`_oeM`
z3h&VNiC}Uhsh?p@3=XfX#|!Ka3@(Yu&R?*qh7b}w?GQY)gW90TA>5x9IJ>;tK5Mfi
z!tf^e=Q}Fy?O>BdO4g1_ro64BF`B5mdc}HTk)A6+AWTnjf*Y1l6A4JD*vUZUp@&qn
zY9(XkE5S#GeY>|?8wx_Vp0nr;gQR_9(I@R?MRsprx^s$R_IH#0^holZ+p$~QX8{fV
z>&-cNTfJMq4C4U*O(vsVanq+?I-qSkTWhoAHj6>9%Gvvs>k7Pg{ay%Bn23iN?E2FS
zQB)^=*0Bns?*&ciAuNZtA~Do#=zk5#Nsx!B97;OZ#OcS7t?Qah2MX3FI(G}u
zAQwZmQagH_DcUH*Ju$5o7Hz$w&&RMd;Cm}QnRu=Wd4Vc&h-exF1m`tP}F_i^`>MyuvOQf{c3LM2tap8IlEy%z)Ssn;I4nX6@XP
z0x@JEuHS+1M`bX8O?4q=`h0EpjmXZ|iTyqeA~SET#C*j^J6P?KD4?2Y3`Yl;Lb;W`
zrB{E7ZuMU6y440wvHO8gRX*7FZ1yeVq%i6#PF?Hbr2#b%2W78-nM>fz96Gix2Q;fT
zoI@3XXJDG|qOD^;Kv{`2xS^9G-kWXlY)&q{%s(tPmc*Q0>X)H|ysup|avS0mk#FXK
zUc^8!q$sxE+I|?%ibgb1j``}L&O|O;)ol07CkoifMm${7g-OW6*d^g_s9it31zzEY
zJac0x?sQQ=j$RhlC%pAKgDTdhKaZRg^28eE1dO&gJ6q`uLnbaVCGX~~FZITr%ua-)
zQ)CiGqu0ZkEg&=$=q3Gvtdfa-e)tvf29}S5jQyWxF&p#$!7OI^$nrlmi$~R09hN_#
z_}uAsin{lSgxL^hab2IKYb_m)YHz1G7oCEWOn$NuR}!NRdwW{$1w#=Rj8C|=el*~d
z)Z?vg`KjF@?3{wOcpO0->C@75f0VR=wLfA`SCg?EDLGh9#r%~HRl$^P_$1}oXgOUnGzQER^b_tMAqhx*tr@(9u`L`M^|H@h}$>s~kE%Of961Np}w9ST9d3-`5s
z=w&96RH43ah9Y9SV?9w7=23b9A6?SKEF3=PYp49`c`rA)v8#LOSqfIehxS~4Xzuyb
zA~?PAX0%ydSzB!}_*ZQ~fYVxb9LMWKpnl=Dk*w;v6Duy#5osoAmg^XORTAyhx$_I(
zb!qYUuwyyY%^p-JFK+nUZZt~y&}Wh(rRfyypX)UXPF`2~*Qa@+qU?=yGA3%91#v^C598G?F
z)&ZWJIKYg@I|aIxp=L5H3H0i8XfqpAsQ{2^hr!%hp+ImB^b;~ke4V3jQk+>
zklKWky}tK@&{V#5)$-_3+`+fpQH)q|Z85icUg8~f1V7Nu7z+1tJ|gW|O9SW4lBTA!
zKBlqm@4_Aki^qnO@!}xJ-CnE{kt2=aJLz*|neTGG7F83P!3xFrGlE@PwWVbvxzc$4
zJT2|@gT!>8Wvn;hGU)c|P^`o1<}?b;X!%ZYv_k80F&5Zt8HS=T-#epSANMvPBhw`;
z3l_#D?#Dc@ElwV8M_f$-S|nk@u@kvN8+&bozccxbccdYVDkwk99l*hG#;*Jao9>DA?rtS9@3^B!AWG
z3JOY?=SR+g)%n0%vqZFHKg|HvWvh7=IHgW#%WwP282K1`#|;sMZ6ac48QdMwTfO{3
z`UoQ_eyyucm_Iy@ZFjxu-%T!wN~mQGHGn%x_{*S!xGgD+jwpiYN`zqF+UN1q{wD-4cNv#TAM
z8tCY#(YPcv&vPQ}NpMBR+0$brExRe41%*cs6;U6fcCrJ}{K~>JBw{cWPDoYQ{mY?0
zT{zgnm?MRw&gs!BYz2}UH7#jlRRCDP=~@7u%v~Wac%=voG7(6++N*y!)wd3S>$2MohIs3bS?*ZL2iq%#M}MPZcP2zI#`
zdIi-~r^~waYlOYLqF2J?iCqi9HBsBku2%fAq0Gbc%56m(yxqQFU66r1!!jSpYo+E9
z?`~4pin^2=zrBA2sk=_QMyL@8xbOFax^xz~cxNJveujuQqUYl>^MA
zI0U24H&W+{QjadU$n76nug2g&dYO?mEcE31HsHv3%q?q<2(^mmo{G`s
zNs>opoITV}^ipw%FutpIui5{&$<#Z<)JWSJ3b<9@_4qla_jak2CEnGgQ6rfKCQgWs
z-1!c2;mJH%jWN>9IBHrR@9DEzJsRBOR`T$zJn?Wg3m$N%v=W7rE)ecs#jEBFB!h7&
znS51NhONyE1ZQ4k^(|r6-~5=1u9zpTOKFlPjYsDpzGy(yOLtLA)Y&FV-`>{b>hs;D
zv9~K({HuD2bc34i^QC4S-6l>$4@&8s=udluHt1>etRF*bI?oiXi%}KfgrAuH&poA8&RmU#&Ji7Pq3flmTl8}oA&3omQBh=d)>IsRk8g@-T
zm_2yQ4Cqa=rEXOsLXDNgn*t;O$ZlrpFU^&nvkz^j>3;N+*|^9ffzkb;AMVpGRLpxG
zf`zDHOT$c#r{_AjW78HnHjhhCUb(quQ6!1dzCWqN>NMy5ii#m*;_<(SOFCbn@##H1
zw7{wK4Cz}=D06#~nxU8KL$bKI!qZMip4;L@CB%97fSaX&Pl(IM2P0)|>Ol@2z>Gz5
z9RiF`qcm=G;Jpd(OFzaH3sm^hYwnmbE_8PZ)0`4|b}wWHzm^x|*8(?l0Cus5wB@m5
z47zmvhPE+V3$pgQ3$pfX4pKD++A0J3Op5w`W0pNy>Sz;4{y9t#ysx@O!eP7!lmVls
z(ZFl?K=t%OTlp{|mHrLsz23t;gifXaqL^N!bR;)))=`QcG(JC3@vVhh?)6c$runEv+ei&*zrjsWjVjb{k54+cOh#
zsp7Py2ImtgT&GZQxDz;q&)XOGJte8aC29dlK`#WsOX?Nb^YC~`q6h2{|BJS7j1H{n
z(v7W7$F^-J9ox2T+vwOy$4)x7?R0G0wt3Ut-`sm==3TSCKlAUbI@qf-q!#
zz9ouDq%!g2a#{m%74AqAY&8U3Jnq$cWA1%(Tl3{WR?X&4sGF)$@<@&o2d#c2@Dt%v
zD9CeMR@n9f46uNPHR57pTmpn0_5tKnCba(t^!5)2i0Rpx8U9`LR)98Qzru2PUb*=t
zwHjgj8_=mFT(JV{8PnP9JX1&=xOy&e6uJCfZ`8}(TE9ONxk51mw_UaR=?iF3Hw#1n
zE-nAZo43=|zIqrRM5bUYjPiq@Y*^YNn4MvhqFmK7NY)ar+f0GbyNtFqs)P~2{5b=V
za=2hE#|X`lt5UvH{0BNkyYxyM1*%WEH<*|Ny1h|9D}%<}V>GfSap|yej|^g`OMgSj
zjdP>T_velh$fmJnXM}0&!xtWn16cEU2}QCgdT$46$(T!`Bzkv!NCpB-s-Qgze7Vvf
zrBgF}YJtcE{UcSD^fVS~RtR%tC+e;1xDG=SL?h~0whREdwHcS2T+Hb_?Ts9Lfi@s~
zH;`{?P`&j``3ai}sWGI=z-V9yi$m9og5L}Ua6}NsxYf6irtn+cU23Z^%Iu4Hd&xYJ
zc^^hNP2y2bNml07W?*=Ic?leGHIW7B*#aTRP<%TzsHQB=qAo2ye{k&~k8+BnX3akm
z{Nl*Ujw--H#!&4A3CA$pPsKw2B!crKa!IE+$dDHkA|L&h2l8^%&d(=Z2K1`6Y>4?B
z#Prq67F0v!MPLDrLjrfzkDs5%SLGf!NOy(l0infWo_P40PFl46&NA|f=!o_kr~sIV
zEMYZ@y#
zqQK~neSBOobgw;BdxDOQJ#QwK%-OVR2x7uc;on07x!GDRtv_UOflQQcm%kN5V>|L>
zl$QJyp^?97oc(J=>b~D4`?)}*-59%HGuY0Dni4gp-%1Pn@`Z~~2pUIXZl({DMK;0IGU
zR4X%gyPK
zSjrn!4&9p;o%UYU)2uhUZ@;LgV|p&WY5K08H19UY9Jpv=r97u3XC)@X>Q+;Ch6j}F
zc*^a2Nz>oE!%s4r%)`=mIu$$19Z;waR0|Z?>Yv`SW@i0~X&yY58=Sls>MKh7{JGEU
z?LlyerJXGa(juZ{T4DILJ>>Imcy(bAL?8jbk4kan+v_>BQS>m)t$*zSLu?Bn?2xuh!le5lt
zG>61vYV6D%fUcqh7sc0nrnM(v6A`|gGFg#Ya$q3-?iOKjTI?
zHV*!So<3p|3QeX$$6G3sJ_gs^J!kA5oD|L;3i-W7vKW!3oLi@G7Q>o!2Ab3I*BZ|T
z-=+-HLG>s|1F>p636WTq|96@0$
zkh4iQEEbZ!yE#UMGJB3|a_-rr&=*Ih0jrj(7I9^lC5^3Qm7P*>GM(X>ACS;0
z?!LX#7A_rLgH)6l6U}SQP++FOb20K)und)Q#!!7W0!UEzB1xI-Md~o8UUbhU>JhPT
zKmA_C&G_A!n0YwzxdHAIK$x}E;)vZlxB+JCvZIWiLy2EX
z>>YNm4davPqn`_`RqB(y5>d&%4QL|)yrjz-&+0|c5}Qw$rgwIZG`8lw5Le8DcIQ(8
za5_d;LSZPyb24F0hoIiBLKOz?%XRXwRJ}JH_J;B6>z2CNzuEU5aVQqe&low)bppI(
zf@fd2*|s0ud;Bo}*rYOtVCF(EB7|opONQY)z5$fWf4rPL>f|k>dGBEJmKy>o6U)_Y
z?|gh!L9^TxiNKQF+58oI4s4dN($|{}VNL5@#0y>6Dq6OsR15?ePj{M?ccE}$kW;7M
z`E*-lJ!VVx@V(GCHHWpQml|d%L@XvyImxSM-9KB8Rkh1>t4&C`2blj|CLG3)wv8XF
zVCvluHnl{sMni=?fmc}LKBp&A462u>NicyU>aw7KEG6HUn*Ox0(++3I0bTs@YQ88{
zbn^_K<>*Jn2A!x)^PG=T-8q`tu>y!C0RHeb<&1$RD)J-))YZ<9WS`E}R0;
zb%E2m40OqYuBZU9-$6UKqn!SGyl4N1O-y<=7Pfyk-bboRM6a+QbR4K0k^$2b^4#*n
ze`|5e=0+~}!-~VHV^}~TkiZNbcmD7)VG~e`%NXn@3K1~R+?nN+@WS<%o2CIaGL+ad5XhGcLVhdOumhet|ns^IYg9+i5Q-
zI0-eXn#(kXW$Y-|6iO)BDHlOgvNQ9e-plx|wP_S*XLm1tF*T0A};!D$cU
zk?!KB;35CixR#GQS8E~XB=`AHuD5ARu<-?JL1^jn9dM@4o%X&Q`LLvswoxou6GLeJ
z!uk-IM1-iHiHrY=%X>@R_tKX!roXq%6ubF^8t5x55j;r?}^VX!Y-@037A)ITQCUuzPC+^mQ_rn0x%AG>O@gS
zG56VaUT`Q!G6ft?q{YJ~pX}IEu=whYd8L(QZt9Em*LHHnwmhXdoaY?7!T3c>sn`Kr
zy9r#TD}@<)`(qZ`#Vmllu+k}1S%`3=ce
z9^lz*423u75O<#|fx-j0UA^i8h?DSG#^p{u5iy^RI}D_J&1}gy8T>^|J9*(H?W|4D
zC&r`dtcS|E%U`!WYXky>e=buR6p5i1n32@PMZQ_ZZ<~zUW=$X#n5gB7#~QI?-IFKN
ziO}!R)mV`;~3IYme`k3mo87zbS!&)(Bdlpak@NHk5?E8ca8zjkep3%{Wv*cq`Vvq2rh7{1)co`SyZ3e
zW>sb=hmMV|eATuB{VL|cREFz>eDEQG4%;U@z-JBSSVg)`N9t+W+ACd&kWdTz+9r>R@@_<$ehF*ax-2VqFy({4l$U3IyD)()^}
zH~=XSrlgs7Oy|^_Tg6Ap(l9l~vLoV%cdmYN0eFE!%M5rezTUDp-vRXuN~CyDxcz|=
z(N_Vo#ejOn;9bhGRqj{Bi6;jTS96=PWGOR2t~*pmtY847IV!tbAnv--0FrJt{RmfQ3j1m#$4*pAg2Shae6<{F6vNiWI5_p+ZKjNxHFF<1=CY~K;}jT#Ta
zfC`EyK^YlFdg0r=2O`+-*lWe>gAINpHs$+~h9ZtAM;4`EjeHBk?TueD8g?+lB86zG
zfxzpT?T-MCl3|d@%V4WQcMn`zOy0UrEpYb;(XVGpF&Jo}V8K9KxXaO#oLk6>sB46f
z0Z31KaAlK!K{B$)3|xuJ&SW8Bl|*!rLgWfh7Es_#|2R1WY4SbVwAOL}-mjWQ7zhXe
z(5K|;;gQFaydg;~UkcK1nEn^Buzxqb?C=j2l_X#ccatO)zAGGGySk3hl9X7G;9pqK
z`D(h|r*&!cfzUk>FNeVq?u<0`<3e$EYjD)t38L0rUGIsTqu!b0>ihf{*j&=&c=Q|6
zT;-;00RE?0W<31BcE7%%*L-#5#5q~kM?I#H9u~^m(!jB(mhydGru&tAIg+9B_6>ne
zEQo56rL>{Bc6ORUc{T4N{1F-iU24EB>6^d0rDdv@>FKx92FQ@NP=k$GP0eiGBK~hy
zsz~QrqKf8Ib$YR~E^))QdT2V=C*7rLF{vZcMsNU{s7wo>Qr(&$vl@8^48pgd7dGv9
ziZu9EO~XM&34ViFJ_NF*fg=r1h}sZU*VH@_`EcXSs`z$=!3TF!QH(HTyIgmK3gdZR
zY>iA0egP}DI9{&8WKDC(C3Q%tIxyj>@Z#MTYoDt(%4u!aS7>D@!{Yt$G!?~$T+7ey
z#-gTGC(T@Yj-EAhu5fhiB_^rr=w0VK_MOMdQC`@xr3n_ihl*$l!O5%Jc0XjIdkKc<
z5xwSRki)$xNU6ggyVNg*!VyNgeWK@N*;_4zuAz}O>N=-;5{!=kl)|D(0p)D;=(JW4^smBDJZm)iTI`@m@n>)*XTTTVVHpxV71bmd!hz|0%4*U
zqk@;lajr;szR;$-n@5gZbf_4rg)!XBly{}#M1#8^%
zX3PM)yrIMKBP*Nz!ho4%
zfJ;5%Zw-9Vz9aBErNd=b^Q^6^H2@Ugy-IiLBsg7Ajwc`ds_2ddmrowK#DEcLY$3LY
z4B44-W=g`ca8Z%KI0Fj7VRx{n
zfm~TqH5|WsMFKibeo1o`8^(Lh6Bgc|(d6n*8i80QLP0@ww
z9L^niwYMG=>pL&5SRMmxL}pZ73dl`Y$R^C`gz;*$@jkV^nmkY-i>3`jx*Mw_f&GPG
zDqz%p?Zf#`vvRe;ZeAjB>2Xf(o1gT|_dlWnDX6~J20@x32eE^GP3iJpA%i#%p1S;L
z9rCLV!7%}wGiuxd?c%UCMEZftitA}dw|2enTM@5ns-+ud>Qg{~PHc3930P>MMA-~b
zr&4zq18fP_u-U3RZgJf_<4q+6Q2WMC=}xxAw9SEKkNwW%Pxu)_3~hRMOKN2;GQD@6
z;aoS97kH{QIp~oYy@^Nool#C0S8ia4q6_&YF8FXo*)vgCT_X?1E#NDkFAf3%p!(aa
zzdBNX*flHSjrB61)Sb^8A-XQ?L)d`3e-rlowi9d8Ba3AV(_i`2WTVx=NsA&A3K3UG
z0dKw67f^A4$pg}fL(2oE2a}TCKUfBw0SCW8uV|*a@4uynT)-7gXYN-_LLW9wMK7g#
zyQUJ5?t?w#Nxa?L3G?~{o_7$yh!F3M#<{jElNlmgOqwVO+~8&KxpG>6+f>4EiO1uX
z&OpT^AzuL^CS+!`9h7=0G0g=KBSl%nmIG4}qX!M?94WzAP|Uyv6B4?>4C+W|a%=fB
zcs^lPOt?1@Q3FY=GP&DrY^+34FJVt+2SZm|5ouH}#vaK#3ZUik%wnS4f#LrBY{HSy
zNZn+%uJ+|fZRuVpa*_Eb9t79O9YX%jc0iney9lboJ`yYcTyT2ssHv(iRcbNLnRi1v
zis}yowmo@9Zh>X>eb>aFbmY7;>QUz0@O@mFOwRuHqOgVsQ|K#~SART1G+DUqX(ojt{)e$EN!XYIPU^T#XOL!i3^*EpwVQ`GU
z2pr4GXQsbd<4Uh{T#LRSG2(aE@c<>Y)Ai7lI8Tr5BOW-tJ-H+KiFDx5YFka2Svnxcen
zSh!uL6&b$D`dSuU-~Va!adq(@Z@qvQnGf4G+y@_99)Td22}AlsAm
z{JO_zKQ1B3zLXz(NT|^t){&BMKFg7EgDp+A{&7?js-Dcq!zP1!ZZ2Ks1YU7Kvp619
z&kz}_DN(*7v3L6P7%q}SYC4K5*1H}Bp>*>C_+bRC!F?42h0s>13lP5QBopW;zD|Em
zL$`e>7^&sCgLy*EF_|lx63WT%>h=DHjcTi>HUEL`OaIJwZOFunzVGO@mRVbZ8@)1&
zqfK4Y?Vz%4oR_pAuwc>B)+^q{#ZloDip_lF^rpNiLJl;zL76nHdIh+yJYxYC^=iCS
z+5TO?$~W}hB%A7kZ%uu4{4tpa38s7XQ_Jx0paDzXTrvHuD($!jLO-7Q5;KE7
zWa%qzUuoMk*0>Jmi`KB6d`N8~c37BacJ}m>*U=XAwdXmG*BwkQyd{wEJXc+u=!zfD+pAF+vRd42(P;iQzguPP}iSf+qxB+gzjr{FPMz%-$dH~{SckQ%#Q4Vf8&YaA~&qGN^;Y_O#qPqtGTU(?WT7v+y@X#w58`C
zA{ss0e?J^R&&I&`??$vp6^ZCo7MRb7_IL=$5oM(2eJk%6Zn+Eq8s_){5QoFmU_4Nc
zt1z-YcE6B}L_XolfVrzLfg~j7>6yR1iO|=*qH;C&alSQo;z91uKv7EsjdSlO8cv2P
z7E{b6o)3j{yeaTVkMgu%8R9b1vG739xriiU1bEAa;3q4Tuy@?n8SvJ5mzEFV(lYlZ
z9GF
z8__--d>@mx?XnUPq?9{%RP`
zNYBw}3@yeRzL8e2ies}p`nC}kFQYb9oKq2|z9+*Zw~qD2TkxW`rg_a_Q4-7Ber0*_
zxybsI}ZWR6nT
zm2Fut4Zx3G%3di!J9{q&{Jezow;7zMab$~(B7D=75QmTrZ1=2lP-tL>ngS|0Oo`d4
zI(Ks%yEHqMzOxA^&y*-~sWqMH^d_iv
zoF-xnokYytuu?Xp(s<5#z3Zxq)T&Cii8nhltSU5pi+OS(!+9Uycc#y94N{~>WiSiQIKz68tewBSE7xgPHT0P73*khY
z7HWH>+6EN@p}VZ!-d4>5E+y3ALrhGJ9cH6ppNr>$f9MqjuKJ`@G~Vv~djH(v3aIl{
zjX%u9V#?P7GqDMJ0HK_1{yqhy9b{H4R=7rV!5^Vq7TUwSou3B4w##%32DK@S>p690
zymgMHW-4-2NtUeU#~b%8ji(ZpGO%B`wL3vN#K^S6H@;)^UES?(aQG&r+f1HpEiirs
z1IsCmDyhT_Xp@2q%xr0R7(bkmxK9`4wJ2eU2^1`u;K&BrBHF@%#Z^Nt=yoLe8EhS6
zssv;Lfwd5Y3FXaxQom(<9G2E?geFsXPiKjgUmAW?D)0xn+7Y!4SW($GUzY
zA#&U@Vu98BlI;DqC3pw`qYW0Ie#m%8lB|l0=m-M^-L)2X;I<7Zm%Kk)&wlKP&P7)}2Tp?QnE;L7#S0LVcL7()D8l^t9;Ivm4FP(c%Pr^>9~qa>MC;$2
zYo#0j`7{c*APOB}QoTFdK-SDi&%N
z$>^p>Q66C*F{s)+VJSJv7b=mfY_+B^`&U=4CpSJZ?B)m=q37#njH;|%Xr+GZWIqax
z!>#gd+lAHQ@WEf}hhe~+nfbZpBVz4Yn4$p7Vmln*E7QVRA3RZlsWBG8Ovjx%RR*6)
zsy-zceCi9)8{QdHS`5IpP1HG56F1DA%9NDeE;=x7!Z50LDKFaV?y|jiGggecnE-a*
z?X*D5QVoCujV_17$C?NlK!fNZhzVoSn8Vo8NV|>~-04GO%T7+~l(Z$cvpZz~D9iL7
z7t=yVG#L&s`NclA{^{F8-(8)hB%bg|vp&9JwJ$^(<9dV^gN{ol2UK3ZA6vPjgrn*0
zVCJd=qnlAE*|5?xjMavhF}Sz$FRU-&XI)8D;_fm{`8`XcJ}TwmJ>)97iqcHB;=t|J
zU2cfEr3XClsGOFbqOoywC~QmHpoEE-0BCqzEbaI`1$FkZg_P;ou_Jl5_ErHflS3h-
zn5TTy3D)qyNOcB-Y+{Lwa37gxf_Y1y8-lmVmlOu)8LOt*ytgpU7zgGA!eE{}q4uak
zgn^4f^Ahy}6v&A^4?>z{vH&~^A=Yd^)?sS35?b2;)%*BTgh{r$kspwj3yT2;F-98I
z7C7@KTVcTRih~6Ti_;cs
zz7iI#VG>6Y=*zeeK(hsmr2z(M&E)bMWIgr`ZemCR>y`-W|Ft5wFoq6_
zG^ZKyYY)$T@;#KwWOnYRFJCJ=khL6$II{K-3Ef#+2Bk>
zCNTv(vgMYdLhz`*YDH5!GPlB;ix(9624;!PfT6*?vkebNEUW;w$F|V5hhDv#^LSBz
zqR?R`ZGJGN&Zmx0!MGBCXH8f$@-J?}
zwfqz^UJnJR-4i?eE{{(h5AUBOC}TEc#ev@u<*MRvZ?6{NL=x`SiiqHce~t
zg8KXjHJN$Rpem^5YHc)#Cj7#3R;{<(%OWLUrNOO8vV5K>{_W9u+p2vkpw-O!b~Nuk
zvol&lE^y$A-HG}#L+%|fYIu8?#Z+01L=J9%)b9XHrXR=oUGSv
zcPZklKFjq$e8WwjKpf<~D)(%cQVj@@?!b$~l$NMXh^HZ3Y)c{JMS9oNP+!p3Gj
z`@B~0X{Y5_@69jO^R%>iK)w69Sws95g$^RQJ$>j1!r@!Go=j1ykp~s~_%#!6!50V4
z&b=*0CKz-wIOB)ZDZx5mK7jG%nfuuBOg_@i`IwXnS>>%K)u3nTwc_>jBqMnQF6+m_
zNT~4|#%R!oE9v7B=f)KK{
zbK5-wVkRcUjw*+s;nq~){<8Vu4wk^Y$tfOFPl~ji5lw0p?aM@)ofjtld8QLG5gRD)
z2_%0T#18fw0W>lYnTCVlOb1wU^A2qdn|N#|5}#8|d@TS4mB4D}ftfdO38*T2h+!QmV&zKK_`P5;hjcJ&nf-&%Nf=vLb=cP1Ltd{tG&se8
zGQUlW47Y^tuRC=HmY!fUuo3G
ztr~(AVtZL-)QEM09vG;r(2%xfAo~fk0j5E9rm-ZzzpUU;4?@DIsxvC7%q<-shdx9#
z^_?VNsL0^n7$*>UB?J=Jv1#b
z11F&lSr!{oZk-i}f6b0#aI#OF_8jUe3YdbWu04ZLIE?FMKfoeliosfe7F}MMzAHP#d(+J
zY*eYDYMpRx&h~8A-lX*oxqs@K#hZgX&;aV5)zcDmJ%H8`M`CC(>Fo>XNQ`4QPmaBL
zdaU^LR@du+dzGF&QUvp=dpzB9#hLo?5M-Y$D1Pq)wqnptwVTo_7OvjH8804}W0^0J
zE;Oy-_*JBi;unL){bFEHtk0taMUVxeL2*oN(+cI?U+xK2E=5@*&L63{rsea%I}{kZ
z!Exf)F0~aXM+jsi=k#GBBtjPG(Jrm$6{CHBu00GZCHWzJ4;=QYJd0m(oX~lYrO=E>
z#4<=EeAW_HannoI{Oy9cvFg(E|rbHCzFuB0X%kb(IKI3Xemh{s=clRBA
zLlZcDNW_+4tC+XJ5hrV__y8wH!0O#&;2S-7(-dPd2Q1Yks0#HOPQeSt##ds3SmfkX
z342-h#B(o2?Gn*~;%3D0@yqs~Bw<4JQ(3(S_`3U1MbQ4eXpCyDR`ZH^~Nklmq`T8y?;^`W`{Jvm1t^?{n-4%Nk`Q?0^Js>@Gp*u#3nSqv_
z4zEh~g9uR+zvuZCk-FpBA)Nilugtq2g4wZc5hh!
zrx&xn_vW&E2aj5AZT*K(O#lBc=Kgzh&J3(9|Ci30jq#J(`QM{+mKWjwziiHI^lZ%Z
z|3T#ZpLm=Hl00vGM1fqd?jeMAjERJT)q$}O}sXt|u7oTx82IXTTa
zQCnP8_oEm{NnRxKA@V~8u6v~LTmbjNFw%9!0Hkz1+>GKfHl}-=r+93=@2?Wfo!kNa
zrp5H#8j2r{+oLm0PprsIp5(Q_G#bAziB0MP=DV8;Qx~$_8$uY0{>=~g=qit
z$fbUdOqhn;v#+i$TCVaoXscfWh6JzAjCZJ@W4m4gM~lm;Hqkc`u+P&+Cix)
z>7TLU`{k?K8_57IjI-0%lCWNOUL*^TbQ-GD9y@Iw{+}b_tXO5`4HIBVoK7Gq09m>U
z7M$qdP^2m*-md?uB75}$?|y^QOdRaRPXASM^uaBi-ebQ*Wu!$Xggj<0Z_}2;i=|HF
zRb)fRymFk|w?CdwSaQEUd+uIYAV?h|`F{1MLKgNfIh*piUu}OoJs_#))>((Uo!ad!
zqU?LX8GU1;_;G0a-uuq0>wQlV!Y9W#M|q?1?v5^0S32Dg$fPF
zM|N_+sg85+^Kto{IVM)reWbIc&wU3cbJPhnMghVCekJwmnb=?yd_l)YqZtDESFHkm
z(_{1KqCIaDH0ulr`{Tn|wk$lSZq$h!`DeAceL-X?!=hSYCV6$6c~rX@!46yMXH;Jr`Xtz
z&4FzvGAeg_m-sfbeW-fJZGoZK!nC5g(ct6cNLI_Mad@8vQo8Nh+gB5(6ndw3pB5r*
zgZq1pw##lvd&(v#@o|coB^wv7zo*OmB!DZFbjQ&ofK@SZU
zEAd-b$9a%>Pz+m3>k&+bb=Ye>IFwSwuc@zI5#&rnN5`Sy4Y+g#lT9JCa1Y1n?c^W2
z-Sk^LB*YZU{loAu^C}GvO&3-!lF}?*CDcpR_!bU!;oqY=4qx+;1pE*y)fy$&?q=7|
zg3S!^eOl;5=jj6~JWo0<7)ru2xZ{eV)nZOBOa?a?{`m2(p@?>4=CG98=nsi$)sz7_
zhzu3q@VO=P6LqAXa<_$UR-UnC3yk)c-9`(0Yw!J4Xc)#I!O;hT(!9)5yAWm2Lj
zevQt9NZb#%Hm@xU=b}~jrJYdgU@?CZp(k86Gd82ex3=BT4D`D`kLZiXik;r^Vpi&A
z+9d&V0Sgy6O&-EIM_AkacM%=lBx~0!m0Fjbs}bDxTMAr(X=zoIS{DuN7Y}%NogJ*P
z&a&3CkmvhG=8zc~@9k}m@mIV9tuaIGKXH-ouwH_O#v734O)EA`W!>afh!`O48Egf
zXJwq7$z%QbK2u2|D{Gs$p7|DRm20fgK;P&oxe}Mtl3wWs6*qsPg_5|1`(|(=`g(#9
z#Ae0@sAYK7J^VV68PmI151Kd4R_f2_xQtzZ(+OxKm>U}&mt36TuYa<6y_9d?Qx%%p
zC$JiHb#c+@<@33KP#~h^xR(-I=tuT!x>uXg%nd$!zb>vigd((aY(+0;%xR?M8cl=f!UMje;Q0vM;x#Xk^pa@
zZ~96r#;GJ}f09?I6wj*Dn;qYUGSsr)&X3HjG_kc8=QTHfTgno5R!iy_;Ty}rp0G4W
z8s{!=X`rfY>GI^0JBN_Q>)xZu_dSeqafa0<{UGV
zHae$!b+nzVz!s{E_46b_q_B1p+MbcXFLw=;g9`JvqFjz!uG
z1MS#gxJ@X_>REQ`#4ErGYi?#6ib6_Fo`uBS@^v
z@Wf*N?V^>_W!_@6mdVA2e${IHsv(XydA^_(dgpHA(On>$Zw*
z6|v@AwTKqct4O7qU>t9IEWJ^F&}tjc_}{X?S-d|rk>81L+vwQKG+ild{|L#_un#in
zRL{d!5Si;-vZgNfQ?DtZw0
z(82sPT8))yRYgse(!?frw?}YySC<3|Tx`7CpbUDkOt$l9kpV3=wG@^4Rf>m*$g4Y+
zMC~6pUf3C>CXj!D)E7SelR)Lx{^M}%$MTvR99bIeD}}H7KM*UAR!#6*)-jj3P&roG
zn!;8^czETESoro&=2BeinZfUq7Va<>sFRg$mKik~Qd~w`+yY4j^NE57LOx$c;8$Cl
zys9Xuc8D|t1T>J(B~2mKwi=_QEx~sBcF)u+6&@FtekfKC
z{p}754?djwZ9cWaQ-&io6xUfD2c^kyNy%oPdBVko=BiI^v>*46o}-RNHa@N{G@x!0
z*NW!nu{p1wfj2tyt2!%xtW}W0xvT%q4vQx1;#t*)Q%VdAlK`2Rn=3u@;%IM^(uqLr
z6b4mlKnPSM*yIL*KT9oigbWV**2!R&T_nlGq~z#B6eEdpNG=$V%RjZUU2pYyh-|
ze?e$0?)HD7(~8*SQRqe3zW^NoM&tJH#2~6dyG#48pZ|YvP463Q+zb(PdH*W_{HST^
zVQ$^X#r``nr&~33Ol@T4$(trKh3fLQb!Yrt1GqkX-3|Q^i6Mg34IS%mm>OUChDlau4%0%BRm(WP&7Z4i^+92Hert
z*CaZ#>-TW1lgyf8s*;e#*}eXY>w`fvq+jj_OcIXgi|wEb`?CZBZbuRpk!D`=m2TW&C7~MuQ9=jp{Wbm2TQz|{RVl2uxd5X-C)u#lW{-S$2
zb&t*7*WsOx>+H|zof*A<)OrVK^nI*Lo6@O{*Qo6m8w2Mn4t)PXawdM`?-WvF<~R2+
z={(ND{2tnou~d)G`E!#i2TWu2+)40%n%38u-3Xnx!0Y^Q(^cU3tz!J#aNXNJ-lb1^
zO4BQUR9|t)Lnq^-oNGy8SzgX1ckcQCB&a%Gxmf9La@6j-KdVkuP54s#PoWw7nW4;s**(DJE*wc&MyDbJUu(z0st8!P+KLbgF-8OfpoZ
zRc3M0bBs&tYc9B&4EAr=5+(`MW~U}$mpxxX!LKVPP6U70tr)R3KkS%ixDYDdqL=1Z
z(!9O%dj`YTsndULj9b?ZOP{xnVy3S8I|@`#wLBpf)R~JHKzK
zB7~IH*c7l@7J9HtS>+yTEP~r!-K3#TbDCnQ^3UUNiYO%tCF7kFcw|eZOOK%#hshH&
zr%$WM#&zDabhVl3)J)gJ?YK<@n3z2Oy83$GlT#e(a*6Gu)*qS*lWh`VM%OV&rz{*s~rB)`Y)#l_oI
zzFG>mtmnfL;$3=2k=lx(oPfg9nuYg)@}3UulR%lrLIXAX`+VMtvdddD-s65}W-G&N
z@k3>4El|ML7$$3#jryMG*H8Tb8cMCHJ{$)r`M)DG+SI1HhWK{8A+zr<`A|2Ph4ihU
z^u>{^zP)az!G29%NWcggF^S47Wks2Oila%vGNlrqE1`pv5%u2sB{cuF5`lp3x~3?Z
z7sYU?v3!~W#^w`UsFSQW$d*n&JS%EGAkfw#2C5r)nvB#sGP5*f)?ZUw;51zKmtGuZ
zx&|TS=q@Vw#jaBEj+W**k~?&&C?7ydoG{Wdq@iY8cai~B)_
z5tX~np}j)`Ad%x&fqG<&XwW1R)4bT~>l*R({&qX@_!d+-(-3Ws*0kJ{|FQ5ES)ibL
z@`F+eLI$ds#BK0B71c0VET~#mDfL`JmMnx`FOya$hQX7T1x;+LMUW|_pdEpYjo+}9
z=Bc5MK^d%6v-d#eT`LC<@rT)z8%-ez#me(rj+0Sp*V5O@4D5XyirQ>EDf|)#X3aCH
zKdTm}i}2to2T8bB&JQ}w2Nmmun70xT+$v+8U5;cMnhp2W7dmu7ixbY@Z&1Cx-6Tb}
z!?1muz(98gk{>>VEhKBPAc
zT(gv#_XhJieVqI2JI-){MRa(~2P@T-u}6HK8ln!
zA}pc{#|m_KuSSm7F&PQpaw2X!_G^qVFgwZ`A}WyOLnmj?o-Z2VjnFc8Tb-8IRNfhj
zcg(jo&8oXw?N>M_-e^S`-uotF9`@fbIZ%@Cb{@=X*E_k%bKl~(AN@K#mHXiwjsFHb
z-T1b!QCv8$?afVBuVXK3nrlWiWHWWg1t6_
zwUGh4j002+qi!Rlm>wa_bFD0BI6GfolT*C0CFGL4%*ETVF*ZC-bhQUP#CmcYgjAL8
zyOV?aR+6AbFdQPHA)rpv1-JEYQdbxjiW4oDa1e8f-_N>fBSRapodvuf@xl)EhN$SD>MO^n1Io4htJB%C|&UVAsQ}{a*mPW!sq-i~?
zp>^=XWs>sGb6a_Z@LOk)XNCi`%!Qh0PS#aRv)z!;o{<-|3=57hw-HPtpsNKBY?Bmn$0$WMxt8_2ag008%rkR_i{U?;gpRj^xr6v~)Z$zqG71bt36e
zN};1%n#B@@i_s^=w~%~#vLQc>&6mfkWhb#@U|RI!(MX{JGj4V+Z7~L5XQlAA&9O9`
zl6rN4ier4hDfy*QDo?Gpb9B=B%$;K6R(o1yERr}`d7nZ^-Xpo`bQKid=HvKcAnKt>14Ya7lwu)ab~oM~@;7Yq
z~%;2
z%0mUS(b6ChRMbCc9z#Fc$Wq^9UPhTR>EP-|naR@TJdtq%S}`treHb73q|J1Jh=0zw
znT6enGRCd=;w@&kBjw_O5~kNw9Ft>3FIW?-QdAUR}$S(4suw7Im=U79I#MbqbcVP~J%
zRO%x28ma+HCH)v4yoh@Yy{T|vUwtgyL#M2-te*&l7i>1c+U=u=bcXpjSk@M;6!T
zWO)4e#7B2;&xk9e5KQk4vUZPt;4WvOa6eGDkvFLX6&$ugOStRJo%0x@MNoc+C}lF{
zutpZ1yg4%z5}i0Co~4T8q)ZP{Gf!NO`@$6IwL;tW<6*%W#?WDBHnu*Fim`gcenJ+0a`-okYDv)`q-S
zw|?)h32Cv1ve$)=j!u`({yb9w(&&zB|0{>RPgcp7C!?Ud;HsBj`!&fMI!`nMm27Z}GchtVvaqdeMkOiOFFk+(lbk~`d*^&*t>zF?F~
z=gE)n(8oo>*=Wvo`eJ<3QU9$=z!;U%)&abCkckEKbhZ#3v#2SIa2
z?HlD)v~IU1jv4KZnKzsGR!YvZ=dfI4={Aw@92(#g3|40*FDPr%RU1=Zhw0@k3IKc@^RE>%kE#Rj}gg_;2Zf*53d>j=*
z#>{+v{w42R{vQ|y+KGA&*V;=C!fvS!Q@|_ttTwJZNQS50D(eY0PIEPY5SWk(v6ACr
zEV{CBzV}D%mTOMGZ!j>ghS$2y-c7X4VlETr<~ce&g}JBv<4;+IZDfA3l4FmbMG_OG
zx;r+ITH=p-4->|BvF=Wa0y)PAOkVSo{R12l++#l6vL(nE<-&Pq9El#AJ>V!h6XQ<8
zf2Wj=)6j1(X<0VPOc9YkT69nQ?=|OM2_N2zcaM3g+8o=md=~7MTgnVJIy&CgmT~r_
zWL}E(U&s6gqDA0+mSz~8#2s*&>phP$pet~25m&jn$dHwqpVB&4-fN!x71a&U7D^J^
zcbyd&W}T(-+7o48u6Hooa7DEajnDt?-t4Ki^jZkUWoInDosae{m6C3XpRrV^??
zJK)?Hzi9JZo^x9D7i-i`O*b@Ut^8g*wY}zfUyXl|y>MdodvB_1bK#cR!nuzg+qG^@GE~C
z&`Q>PsiipuZSk_!a&JUMjjRp+F%ogPmm;3*_kDFCu@4E;%7|7K4XI<9{LjoF>xD{k`?pdVkMa_jBjmbMHBK
z?|sf?@11yQaNfsAb~_=Uqu<>cFh7C;_}%`CuUJ8O>CaaWeN-MhEw{C*2p$s*^0p_>
z(5DsBr`=?OI*Py@#?w2Rc4Ot$HdD!S@yXNWAB^eiwfmEC&+G0!4KUh??fPW)vkdC)
zScgKi*GaYJRZC242_Tz_{?dfjD$x|PP|BQ1>)F$U+b{Z8ETR=8a$8#%uN0qFY5_)9&qK%NhT|VOLt5QIpl_B>0(Z@!f9y^VY3M;!DDu
zhuE0FPV(lD=$J7~z6=|cg?&%F*{vTPcgCYP|9}sx4^jX)jy{GhxXFAU@#cD6x$}bq
z>0S6<^>&k6Xy+dh;h!?JbPL>G%p%NakIX*3DUe
zKpD7fA%8rFNEGlYVmsdGKP5q?t}()|xQT&dg5@AQt&zZP&|8h;3%@}7%ez6}OxCR5
zr{wC06OacQ9Q};Y;H&=bV$T^uooy3-|B>fbWBprvmHY&*@PwtCKsc3CYjFEGF%g*Jw_irHUI^NUJ`UGLa
ztSw0K%FZ9ocZ}Z|WEDkqla$oWeiwXH%qRBl`ENXlAy`DmJSv1}eP1vDnS?_yrB>9>
z_OtYTVo`75)ML@yf7J7v?^?xPcd06Kl&MTl9#hRJ4p72L!c6UTQ8C#;(dhdJYOjk2
z9!BL-|Bp432?w6H8-GrMBZzCA=ct3)E6TT*_E_C=N8it97P4!PHYd3MgHpZ}3^R*f
z1Fk4!aRw+!mpCY|^_N4?{Y$mGYiCwxC+oaA9CW>G5i|tnQ6@NuhyQ5jFBbq+zV^C3
z-FBl!^C{k%Z&I$Wcaawl*RL5djTojRnQrZ`*gLFAHzt#}V0+8|z7veTOV#$%%T0sg
z*)u)2KC^pxUkjRayw9=oxjd4qMD%^2AmhC$6AUPbfshh-w`XhJS-Xp~;SDT>u;E&N
zd-LqbqUy@$K~9{|W-{$^Lf!j(N1(1@oZ9kGN+3p-A15sV5UZ`@+&g1){B%7q-ufBk
zKfe9zT19OxvM#IY37yth$A%*P${AXaF;49(cOa$iUE=U%erm5CVdpJV>X7v;Md*z`
z9sa%lXz;ZSbP(;1ruo&-{i^g)Zo~g!evx;dzuB88CLoCZySCe`#S;JfMe*GX+uxCj
zAz9*>@Zdi$C?5O=SN$4bgx=~e>(G#J+U^wqtnvTzRXkcBXMbqIX-1uf`$ixR6*WE2
zPG3DEsv^lJQBXu3cX~#7m{uz)Ixc0}iAF|W%JpOOCAx{chP+3aj`>u<98_A3z
z%0DHr$UoWK!#XR?Yes#sL4_@b3?*xQmnnx#zpYLR3W}AZ$9AZ!5zBrqb$xHy&?dp?
z#Vnuwo%iS&sH;Ji@}_nd7A6p)W8I^=6#MY(7u&*C8lReCt`%oDr?aZG_!5KBeTE`i
z=e`_RLbr;?o%mndOSQ}VhnPe(^pw=Ud{BIREG#b8Gu)R}w3}BXSL5w*)=;6={^twU
zeOGfPs_jVJW@qofDg(aW2(j2kR2067pqwEF5uXy&{^pYXubOJUv~96JhQNxyoR{a}
zquAC2*V(3Ea$5Eym}^DB*{V3Lt+nRl3v$Qo$}x`|eiOx#@mRK+hUWdjDu+9VHWEaS
zU7E44v;4K(m(H+rP*9p4bwCQ|(c5SOekH}X=jqQ734n8(dZF}tK_8i&uAx&*QVBjN
z@0$p5G0(QQ_6D1g&~d-Wew@Dp6z?aMrMO#63`BHq9KLM_$Z+ilYV$K4-uhbQ2Z|Zt
zu@k19)@hF*DX&j61C-I%uy%xJ%N54{5V
zQ}4MkwSFea@wING_+>NBa8oo{5Wuv0b!fARc;zuimC95cDI!>U*E-hIh-u`Rc
zGUx2e!*RW=vzQ%mK|bGbzXmBUzhlBsv0>)XF9}^(6eRJ!Yr!NEMkYMcIe(8~w7@HC
z2l`{v$5gE4MnumkcmfeS1{%@=1B%Zf|f^zI~7y!R#z@mK+1Jr4(Q!8
zJg1+Y``5tOz{xHwoL#l@)2GbQw+{VEOvLIoQ7dvZK9MZ6Seaf>+@X{GbZZOU*~seqY5iTzcrSYJOxw-7
za;f|0#2$F>J36qfXBPpdzQfZGB$e&oL}BqY(8a16jDA-lUsqE85#0Sw5WYDEN)!J?
zuguTxIU#2n1KCf3kOkDhlxUuWNEF?gLNe~o@6ErAw~zM1(vSpph@BbbV7ajdbcFA#
zN%i1tX9J+>Q?h-xyY@pf<<98J!&NEZA!Ew+2QpJ=j9~NFz#T2q<|2eUL4bQ_|
zvadJmz6yBD{d~3BIRjtJ$gjO_*)}rrrO{hlRjcEj5D9_#o$z9^!>wLRyu!$y_Tayh
zc7BX?#+!)>2b^@QpreV_7ri`uTCV2pMeBw=KM;p#Xz5-(G<^vF>zd;@>r~~g(Xo&HWfJUD~4@|%B0;lLk(R?xzYmJNLH3*yIc
z@v7;Lrl#!+Ut@B4W^|)JCV=&)
zVs>8d9FrXIl#V^y*)Z66N%trqO5w@B<4E8+&U!uem^`iBrH2^+zB}8=>yS|Xzq+RksH|)nPk!D0!Dse1rY_B`rT%!-pUP@#ABQB)O
z6?M9FJi=;+KZ6>|YUmpqnkD_*E4~-k<*zAKcfvkl_!MBca|d1uk85nl3|9{QtrWn`
zUC;iW*I&7vUy0odj`qgX&KvbZG?S{9GmquM+O3SE7BYx4a(IMBp_I3!bMLI8gn()z
zRPH8!m3~w|?vwAK^&&fc5YIi)IEHm1*8SZ#0ZUq54ZgZ>&SEzCeO8&u#EAdAY%$#i
z9oYZUyGRN1Ec4AhXH(<^T(y
zNO~ay&jQG-4+l~l0N;Jr?9_dqUAgk($~O{HPVvRrW*EUw%~|k^B;HLv>(G|&O=art
zw_S{>w0Qr5L9w@i;Js-2QSJ8t$QU}>m1cGzSX$J5F!hTWGE0nG!vETOW1(99%~1n%
zbSUJUPnf3HAu5<7qVgl!UYvFd!j{=GdCSTZ_ae-0ki)RfYCGhL;r*R3LUXE*Xg
zoKmcL$fJV~vg;LOS-IyQj|TzbGyW%N^MGGg<6TdeqW$N3$gE=S5-2r41vg*Elgkb5
z+B$sSZ4|-2-#wL_O^clluEX7O4%#I}*tfqFUr6(yYHsOq2OSN&{Q$n6l8M(=(MatL
zev3K>Cp!;s4`lV%pj*xpIG*2)JMcOUf~V|Q0ZOYWAH0~%s>W;fEGnB1wlDxcBS4?Slb78f_rWe8R`E`LV(!tt==>pLK%;pO*Aa1F{@wXbRb
zFmzR!yKzvXy3^m1~2}Y&)RH%
z4)yOhIP}B6ivT7+0vx{j9Rg~coMJeXAG`vUO>CZs(P>GXsjJl6rX
zJh$RC9556CPVvUs4`_sQE{C8NwX3LdULjbGKRz1G60n?&`vn2M+sm%35(M^x^F`S*
z_^oy|-+8>7K$a+9SIjRdV5QYmE%tnb4>!b1z6QAWFHgp}ow|*ian^Ns$fj+`Ev4
z6!2AbP8(Ncd&DL60yg(urpwZv78+{n7Q&k-R7uk=P@~V;o>i${%sU~*P`nv6pjbfi
z7%bjX_|p=5t7kvn_@S`@)WB>Y>elgc_b?3dNyuS;lOtJXf3$<*l%n#cU3#52KFi}6
zN%9T5YDH-L^)Us#V&?VKTmE)|Bt)TcM#|gy&fV&=}zYX^91
z=HG8YqAEoKd;|9mMi=AV*tX5Gu9Z;gOwrp(2P_r98Ts}6U)wQf+|kE+yuW;I<`xeX
zrs|C)Ss}QSS{)3+#hbxXHqMeiI^S=GswnE!TEK+Nf_kH}9PJgY+GcoW4o6C#Wn7l_QcNXS%gSLs03V6r-eA-j>lY?kNbG#GMt7QzKik8Kt*8^igG{G5!
z`^+QPOF?2dNsJm@IHSq@kix4+`98D7hJJ_^zC}}B*Ulu^9sG7o2zZ6
zN~{9ZeO^bW*2?9q=uvKpPR87{%Ur?4r_#^78i)mT-5I~Yi9ZqXZCjGiMD-)TjwJc3
zm?%9SKAt(~tfsYo#$9tF(mT!R98}Af8SvmPe&TI934G06gd7^buosdSXtPHfpLz5a
zw{L+F#nQg+ego}40|)w*7IfMn$OX8?)gJA&AbS}EoHiIg7+J#Lv!T0Bfx2W
z<3$5its>pA(!VyOf*R+N^@K~NtiNl#<^f9w7-x0NI^I^G7>$~o&IG(<%QE>FFVP!i
zXXo*n(?WHC+2da<{#-ntq8(4X+-K_tv!my<17-Wwadu=n~nkK8L3USS1x?z;G`!5WnmI5*vQ
zzFMfpBNKuy-euh7NX7&KYyUCyT#Su%zQ)rzqhL^l_2gszX?K&Yvo&lJ{uL2)Al9(L
z_{J~Bu{1|Ri%E~$~H=Q)RUi`Y}!P9RD^<{8N*SI~IDUu~_;>S_J+
z2r<8=Ccro1lma8NRhpC};xR%@L%*-V$^5N8uSs6S7M{9{akYVAhdo*x@`8+@mtyUp
z>+d!GXGZnKWxvZ;l0I3butq7VH5Z?#w9o4vgVt91;gL`I`jlo>^x2{fAwXdhU-obw
z_IW=n$QzUW&6vuZ^Yt<>j>=3YSzAe;b>5{|xyFF$hT@iO{#kX}>y{(%C}|kvr@&4q
z*T!>Ab9R4G)=D-Nt?gUPeY16YV1+CniE9HOxbc|rYiemxU5dT}?bTp~#0>bnv(*9s
z0#nL9>>D|E;u&@Guv&*7VOp|U6<3p7x`rH#m1Obn=rWLh^FRm7bSmRYURi(T$t%hWX^PQoKbz{t!*Xyn+kRH5BIFMiAJr-)H!Sn9zH)<$S}5I
zomSsApDwhdFbh|orM0c5Dxb5xG9FZKFB(;?PZm4d9zQSSJujc1DjE}NIII6`4__4*
z)G`$;qiYlxC#Tz@p(c~wUHO+JO_X1>WsW4}I
zH8O8_exc4Gg0C?6svZ3zxemSJKCo#%Jp!f|?pQ)16~sku;FGS?mwo@##98)YDFI8r
z$U=SG9Kw_?(b1SDL;+YmUdaQydfo|@7XvDHVIMCE<%of|2}$YZG&|jkYw
zH3Qc)zdLAFf=vyo?DF1GG#NV8?a95u`&+6jbdY7|CM?b^V;ak89Dvz@dxd?6atDm0d
zaJr2IUW~EH)&W-0xWjI`cj{#%zQthx5@klcgTFa@u5CMyaO)gz7ys5NdYOEnEPK)S-)`32l0~kIJB=1~lt$
z_bC8M53{Ez{w0}X`@bCPCWxtM8(m+8k2003|M+?)WwY0}6LYzjU?xayX0F02$4O%f
z!!3JpKb0ifZsBU@P1zZA?cxTNC&zscPgbnX
zL@buehX`Vksj~)dcA>2?Mzt2aY4S?!@%ht$96coRDY4@z&qmt#&t$whh~&LwBv{Mj?=
z=U1(9cu0nVY1qobMWp01m!SK~NU-#Q?U|}*>1P^Wcq=1UYnCt7A7YcX&ssHJ=L5^b
zf=-($A&L5U+7-iDe#ks~j;BX_qx9j?e(IBMMS+A6F-qSpskL~q*Hvs#%OA!&by+dg
zmrIMR5BCZm7B|cMyZ@QEysCtP{m&t8c>`#&Qj&~ek_M8kmw;>rdWVex8GYpmVEqjEv%{HpCE`1dHbv7B1TE)8c-;0Rb{msF!H>rj2`|`TG>K9O09h6NAu}G#-Z5q
zaChUVXkn3qxuz1})}^F;z>_f!)WBDoq5bu};HNBlZ%W&FdP71d6-3K@+ME)B?F|}7
zRHk?ovK@{$S(cF`L`P+~lsS_3O~J0VD!|eP@->B=)W(-3LjdDmT*U4q$+npw+6oS;
zN-5206a>8*)ngml=Xq@~XfQQdkPiY?rv++=C@kL3`-fxgBm
z9>_(Yik<<0Qrz43`s;FQQ?4|9aBh(PK8%i-ZPBRX{&TA^Cgx#Dk?+-4U(L0P$A;f)
z>IsiKHE+oVh}183cOPHPpqDeZ4!o>Bo~?*?y63SiwKR>l+hVAn`{L`zLYBVA(A@5h
zA=%Y4vQV@{V9#&%5sy!w?7B@Ks*J$_fn(_P(x)-}(9g~4dVk@HKasSu*a9Fwdk>?V
zQEF@zKJLDotmV<-&fU)&QMN+_cE&Y@CI+z{^)VyTKSKJGk>MfFP+vgiDVqL3?ZFcT
zAt84erAYjR93xXI5*V<>UgM^GHM;_#Qq#a+YjRK%Z~1;M_PF4=jL+^ctVMDAC3V8V
zqFd!snHKQ^aZ+&GDh=LG<_Fnblgv6$c5njuWM$i>>2Bh){!VzPl^QKFE0
z8Gp1-GY9SVFTncUPXG)iAp~Qx5Dv19^%|8dphLE2`5+MgXwkN53t>^Ft
zH?4))I@4#%)U!iHIp^6+%Zq{{i=Lh9f8K}jZ_;3Dn+_c>nNr;MDNatO_1*|8Yb1vn
zk+~!H)}y0~$DkieH3|)PPQ@4v_`V8qwpO>uJ$R`G?7dK(Zak>5Ui(=>$PotXXd4}1
zd6Cd&PVGX^25sJGWfv!H3R&Trn=xKQLRI^)-p)!0_ZYX|B<=UtzPnW&W-#W)JAZwn
ztJfMnOWBd_x`4n-X})@EkX{S$Yq=E3aFOZWPvePnS}U;>%Oo
z7Bt>9!`Uz>GEt`Kkkmjf;1)Oh+qqc6qEq1G4{h&uQihr=u;Ilwf;iV3SI6U(a_GT4
z*!14@X2~$4tmoX$>a|j~%}C#la;JseK$Q9p&feH%+k}8joKa~y6Hq+W71es^@w5^q
z@MFd%Souy=ppib5o^EKvXf(Wcpn2F_?@u-Ap})1wJFq{X!X=V0=ro4>d_$pLQxwYl
za7lC=(K1SHMd^Y);|TYR;kw&6{RptTo}&7|*ly%`xu!Iu~*61q~^`0KR88
zSQgFgjC_c(g+HuG%hi9d!#1i#UeD%$!1fqNm)R*7;p`{W9iAh*NVBvoGqH@nL5E$F11zTv9!XdaI167{ycIbe`jtd+s9%cA-KRO
zmvEWrs`%$IvmE+xw^>XeEv?yx-O4-hiW0Lt{
zQ>oGJPN>$?MHis0EwlZUw`ViDHJut_1Gv5M3tPtUR@#(ZnEzE4X^&h;#cGq^LdHBT
zT4&B6Qs{9*9_W#n(4O9r!l~lZV5)!I9*#9x+
zrlknh{H8XRS4MF)9caP7^Hg(GLar*>*#`O@E%l`SD9xZmaTC--@l
z<10i2wl5=kP;iSg13h3ByZBxsnZ>0o)4ZDH)u62Zm`<8W|9a+|)g&9CyOu4TUS@BU
z!nDoi0QA8vnI~dJ2tAJES>S=XnKMt
zwh#2BH%d!idXGP?2sCa#<@N2bNR=e&VbOF#egg0Z*xrf6Vn7k<{b@bC(&e)93Yyo)
zJ_j?~Z-_D`C!2MvJY{!<@j=!S_7=w|jAq+I)+Pgd%`YiaD75!y=5|>x_3q^9UAYp*
zRUeuGh*@`!Q#~Pzb*W>TZJd}h^b~AFd3~z9Z{a2;D=1K61-fW%Vae+6vpTje69o=@
z0b;XxnRecuqH3pg(6OX9{QwM>ue+C(qbK-2Z{tX_&II1yS9M+cM_iokVw}*$_+l;Z
z*4~Q)G*$lsQS#^be+EAhDl9D_jY478RL)4#l+4OH&y36Y+fr1Q;YgX;$@-v_HmB;L
zBe`;Wu$^3Ce-pqha(~DbXc?tuwbDb2Y1(TRrB_-!j8M=&n#>jLPyA_9Zn|S&6G~us
zBC*&Z4>+HhV~#GesPMJe;b$XV^+fR6&7~?^P3j8L>E*xY{m}vvRq@C{Gr^}bYwP7)Y5vh^na-*^so{`D2>Cfp@%w<5&10uS!)<&jG7s4Du$b>#
zRR0(i^!n6knvW^0({E68tzRX~jK<$#kR)4lW<$p$dCzaXLU#Nd-efN7G!rB`rg;p1
zKg|Fr$t|T}f6lJtclxb4;{HteG*fAL9xO@
z!bZX@xEs~A
zH=up_4V?}Vzwi>?t@xlD8!5cmK`P2)`3BMoN$t23wbZ}X?(
zKJjpadsQ}QroP-HwrNt+vwq4<>7jM*Sy#0Ju+{-*>nlDx++&>)g%oH*G0wSre*PA7
zb%5Kv>Al02KeT2&3jOg%u-}>P;2b10>~
zxC#Wlyk?`PUG7{{-za`bXmIKnd^w-Y9F@>ntPXT>(@*}~puzhztxA8#tftmu^-@g>
zvsV#EeTCp%b@^m-gZ1OC$Ay08Z?UuZg_qyC2s?d>`ea?NiIYac&e7{13YV*bw}aq5
zoe6^?Z(D{eNgDUbbCr)99pJgOi&7R1O3*3q5`Egu5+aq2_o3^MW$;|N$&mWu_R`gg
zJS;;#sZc1(ralz6oY0L@uS#w9oKwM1LB**tzbh|!ZY903`YIMIwP$o!obWyC`>e1%
zAbtffKV0TE)9H_Om<5`7w{pi;W+}J6+p1m5@*KYxFV+cCIoGTdVeG2jc%%{$wPF+g
zgSSewt`GLj%SDlCa`cTz_nS8TC$rlsEtS9hHc$VYSN`B#wpF?-|MC*8Y}oF?b~~7P
zg>XR4qYk5&V@CF`3#Ud$SM4g7eO}~Sg}7&xzdnzPxo^}dlv^jy?tmf9s$aZ@GoA{I
z-(5r3SpSMa4z>iA#>Rt^gycs<8~L>OVpLQJ;;HMJW;mV
zWh9#}C8{lq04+6HOUj=NeKKR$Vv4Y-S=+6nuG0L+wHtAvLWqxIlz`)Zi^`^}6mjj#
zXKnb*54bY=O*N!3)91dZsqPc5a>LvIR}khse6P{2^|1XHj$X^ybYj-!nBnRz{Kv5e
zNmSmTsji@kNX=FX1$eeOTfpSHM#e-CGL{wKX)@3+FWnfyJ>#~JMk{)JOpShVzgouq
zT$1okck|WoBB(ukPViy?|zFiUC62jFZc9ZPAKqX{xM|wD-TNKXbH5$h(~$Ha|(j
z93Douv=sH*^JYE#y(GGQ*(~_f(8&$eZE5K94Vyct?S2+@@C1a9kLq114<2poVz$#Q
zWp^?EIOd$t%Cg&mSbQ~o%F8mhZ0mQ}i5ZEK5`|oBj+LqFmq_g$(Y3Rimg7pK7KW36
z5{l^f{MhLg5K2jjIehEpwHu!z)%GUzMsM}F`$VYVe@+&G>#}YR5pn^=wpFYZV*PJ@Y5LX+CTh!
zZb;j}5qINYcd$ehxjXN;$l6a3w{If
zMQwR|$YLATiWOa*<*JALq*`{Y71a${?@qkaKzZI)-;*Bl&M{H=K+n<%4(x>{I`>gN
ze>-D*U30Yl0*XIeiR=GTQUR=PTU7O4t~jpq@v1Mb8yFHC1jD!q+y!HXgs&I&O!$_l
z)Qac8km8)_a<|3kr_DUxg$opEA*~NS-uQe=Xa&EPe6vogSxz(TAc9&nHAj2@I%``G
z_Sxy{{SBh^=b6>;Z)Cbj^7Vi9(m!Agg>b-VN7TNjH`TM9nu@EN^F{rY!jBatme1*C
zL*F`eddMd-A*eZE&MbPRF?$+HQOxCSTN)Xtw^AE28Q0+RdW9wj}l4M!H}@&r!rXuj7Wfinrx(72F2TD
z!^*t?dx>qh24w15-<^B4@9`RQ>9zZ6%I)|{Ru`|+_U-)uXGWrU$A|Zx@c(#R@)hT`
zDT@|pak^puY+|{g;Rd|C0);!+x1v|=8^Hr-D-IrDc!PT@IAuno3hdo2k8
z#wzN5&+YgY=A^_qb%GaQkvOh$YF=Nn_Gv8P`P$gsNdiVUd_wgj^wU;Pj$
zmmSjWc2$RfZQJSxRW;KeDhOL3XNd*GK`5dyH4E5R6ynmL!J9)#R4SxJNLlt(XH
zjJZdv_E<@$)w;N5&(dc_i_SG3Q4iE${hd+2R6n}7B9qDg^?4^=MPZFtaaOIbO|1@9
zSS;=9KZ_kwW<|Z?b7`kW?Hl^fn(u0R5@5efYM)H*$%Tt&AcvlS$mcd=nLZtH8rB4q
z#o?&T1LBI-5)sgv@WI0*wxC&l!3$Pb-0YgF;1%t;2dGV0!|yT&=M%vGGg
zVXx@u`qG1j`4RV!vVZXI7>=nwrqL!yaf{^O)=u$zg+ft?_B
zL~NqMcEXpG?3n>Pw_wet@jmeU4wkh(yU%Sgr~e}g3>O;Kbdh2c|1>MQdp{B(y73#(
zxqU%3x4E>Z@rdfjulYUBANUOH33JVM{ppE6@Y*G0U%FmZN&x*qh?i1(qZ@2M)V7>=
zy9CrQl1qh4&p6eJhnz{Ya8`YB5+EKlcHkrA{U>2oQD_+B$|L6Y*FmugH1BwIV8Qh7
zyrVyUno)-sH#w{HR{jRF&wifBlc#CN#K~1^MG#l^(rPigKB)HDNyVRB!$wer;3S>C
zwf_XM`m4(Te$O54#)fBOGp}g%ZwCwV{WE$5MoubknEouz9CD$vbp!>^TW6HSH->Tg
z@hsLgJq&2KFHgUZxFLzVVc*mr6{V}_6sH@Nhe!*H+_DfZo#`8ucw4Zf>dx1OtV`i)Hao}8;e*@iW?nB%k4pQ3AA?_b)DrJ7Hm}OO?c)A1
zlOzw#MF1D!n=0xqo7#CoZCOK6knJkUGzojsS-)`%Gg`P`H#Ibup)9$_31fdab+Zi|esl`dfzPTKhhQN_9l*>iFSui(jR5GO
z#{%|%Nru_YrE((Iv$+RNKeB81ME&D}rGC7>y)NXBV3+~QP%)f$JXGntJgxI##{j)|
z+cxbSmZJqIvkXxLQ}lNLm$S16Tfj$l=^FH`gr4r@=J+ip0sO+b9W6nQ+*Fy5C%}V
z!#7|F4?Rp8hd*a!;U)!16xt3~WndM1=|;qxPfPYqOWp@65u|$Nm}~s0?|afn85P2&
zn}K|^+n27j=cB}g
z9EF{|%Nv9zbI@0t+qIgg3o9NP6%lIch^5ECkuhnTKm6b#tu~T)30XeC$RxR7Z_wG
zA`?9kyfPgdI3J;cN4S%UATkXv3U*TRV(xwL&x?~*=k+Fwo;m8c6s$WrQaM;-HY&}2
zWqVnz6lQrLIe}H59Uv)pRab(mWeuAH9EW41wQ~B!9gK#gpW;D;XCT%VBqGGf@XU2z
z$YXowdH0fLN3ztT7=g~8xk)?mC(7||?2PDV3ZFh@CX!|*#tEm3SCJMVdyAWXcXA|kd%}mX-_uu76{32J
z4itYp=UZmR__7CLhqT??ahj!y5sn(}KaU%v>v!vaIe8EY_IZ!-j;H@{Pp|lYG(SLMYNeHIVo%bqRW`o$G
zJs@q5zD-n8Ds|g9x@7Erp?u%R%g>V6c`4%ttv3f``n4-do={mTQlP1W@5LNUdbuq3
zG32C5s&MX&*uS7vGQ_7Tkq|g+&iXhGDJ+nRyuI=E_TM&FK81SZ){G-LZ8qhdJXR-z
zr&#dn>5_o1_Q9d^i!#0)0?%kw6O`JGg~7O7x-1*rLNO>(f45jjKRc@*G$2Dd+~w54@cRV8lcFIOIT
zzn-q$E9cP2EU{z8Y;TGm$_3o>S$zAsajUgZ{AAgFiQwBYq^hq+yP&C2)2|O0UTa(oXtYgg|3V>#|3W1;Hx3wB4UxMLNGDeZC
z^K17-ZCThm4MFtgC0kYY0PKM^vnJelCd8sun-J=7CCyZKBLvRTDeD
z>)>t^l$>f@_d}@DEYDB%kyq$SW~~FXG{E0~6CTwVaL~PNX6(Wiw;JC3?HTDLlioN@
z>8Ojr4`G0+P9{#!@}xiJ!}a_eDrHpE4()$L}D@IiAl4J{;c!4biuzIMY>?h3|M|y^-;Axiyw&7LzL?H1Ay-%58@aV9uYjKjMJWsA)L1pSGM?
zNm-F&KLw`i!=8<8o&cLGq^%tX)Z|XY)
znfqn1_JNdfuJPP5W&^Z6su#n_y#W+9_HQe|OwE}gZns-@YTR1qmIPy#X{;B%08g(G
zqpj4aWMBR5TYUz;lu4C#c??o1uZ^*j`4|;
z$p;O4UX8DcI=5_oR`a|XspdJzW!G4J&G?P$RKb5g5ou@@gB-q}%oqV`{#vO%jkRHkgW(|0u%k*|eWY#@z2
ziyp#``a~qBX!RtGmTp^5mNbjG>EMeWI483*C*|cAuje%TO6;>JCMVy4FpgP2tIU12
zdiJ!%j^g?N%4ix6)e#hNn(@_eRK~B(1fnPi4o#9Leg7AZpV(#Mr0wmc3Q2J??
z_P`4IWVclnd%Df@^?^-ge1eZ6b>m_@{v&A)l$cM!_)Yz{wFo;x!iu(&Ag{8jM*G`I
zmT+d}?YZO<{7fQU8C#8d)J(}iase{);LVM8?Capcb#D9@CY(=l6XCcap*f5}hy}(q
z&f9FWEWch}u8ZtPrX9o|)b-q3)OO;?;||P3Stt{${Tr@sj+Sqty&o*gG_Fu&jH&%X19oxg4$OVY-o_>Fh
zK8A_@4F)sykEFD|XA;#4ND4I*{K9s8E2j1Cdu?xU0u4#mRuU*@oL-OLEX
zMt@|P`jrZ}GeO(hdBJ78z3OfUNOZ<2;rR=;%YE^OD*n}`h4iCOarHO-_8nm#ydDP0
z@o0-JH>{aM9Ns=gK5Ik;R2!-|%ILkZc%}F#_XXo!)3)y-3ZX8tmg3-TaMK|$uxo+%
z8>0d%MB~%k$l)FrsNpCmO18nRQ&6hyKCQ3w;QlL8)pr?oqpd#vMQpl5;;3u==7nU4
zbX2IvR+Z09rO#-JJJE5X=KFzjlC|5Xe~mdppgmKH(uj=V9FqOhfrEp^Nn;@7x(i{{
z%v;%GFd_clg7)?XbuZ%AN*D@FWVw6ihY)q}YP+34uo9O2dgD`fGqUWQ%=XF|-v-E`
z>N<8tQ+YRj>2jx+4Read~3KSb3llHTF@_;GTx`$17|NNK1
z`DG)evb%A2MFl{cVDWd{+(G<+mh@(q3ZPrUB
zCg$!0+$Az%~-$kwV+iAr&q^`W{`P(|hhu*RYBGRTX8-0e1
zL+fur+9o0$u-yi&9%|05*@G01`4|BoI*-fmOCx(VfSq8K)4e@jKGV6S03P_VW3HL0
zzmC*C3|GIhS+8LKc5#j~QQ37sL`RZWgiqEcFNw}?aIWjV9
zIN8s7((Y05J$F989=O%W$uaihV+oLzS;94MXLl<3GiB=J?6FPSX$p;_HssK4WQ4LW
zCDDpPi%TnX%cl=>s7cw)-{WAaME?vCT-^G`a=j^ya{^NqR5T92m|$WRRc%X$9}f;H
ze~^T77>X#1IW;YYKEe89O9smTEDEvwBdPpCp%ZBuy89*~rT7Px4H6oimAUF0-uE2e!ZxLC_?9D&HN>Xo==?6ySQUj2^s#ca}+#iLc
zI%&YJmr9F@#x!)vvLJyj)Vgv{3#(T4ux)e@MU?LDZjf%JyF(bdyE|1%TIojVMmmS??(S}+8_pho
z?{~iUI%odkx|rGfS!+FcKlfS-A**zjiy-V;bLu*7yWH1sej>wsppxg7&lxKfh2v8YXDtkpVE@;dTSa9vT^V2!HPiSI1kT%=uK>RL
zD>Ge_{vBC(M?iyIFviQ{4*-4NcI~IPcxlY1JRa#g5hcuc*}N(DQOYf`<@O2{c=B`;
z9LLpOxhaJ1kS%!ub?7a!py0
zhF931nM*ye%``0o8C&F*7^-5b;;751%ny!ou35nQnUGQD;l5PovQc
z$w8k?)Gq_~1HTgHw{+V;zG64b{Ym4Y#jI!>M=L_vIl&-4_jT}t
z+kj7VDNFHl)o!dp{_MzBR~SDddlBa~ol?1q7KwU~;`fv%pvq}CA1E#&U8bv}rQM!~t45{OGfUG_=CtP;*||jYG1iDL+5lxs5wDa5uu!J?hj<>*#||RHU*n58EuE9%CuQmT0+wJhEWYX+C9w}
zN@*~IS?Jp%%ZVxY7~atm7VSq@3l`Fyf_#>dq<$8JaGh0Z?N@LgATBOwF(h_z)I3lD<4R-deX^ldW{K=2>z-OXi37hAB&&6c=x
z^}qhBCp;Ma%I7QD3oW5YKe;6
zloWD5nk#1NJ&&^I6HFoJ;qkF-r-5te4i{H2YT`=G={iY?*!4#QTh}@>)z;cA8vg?y0UR6Hcn-vt;^Dpf2?
ziYUOd^Ge~Kbh7J*+bGq#V-I2_K5#28Kc4%eK;)8C0|J#_)qd&6ZN}5}bY1BtI6rT~
zW;Vbuvar@TwmwR-Nu^8d26CuL`=`ey-QCs<;uQLeNJ4+_6!5WCq%gLHhQjGIUED!e
z5X2YXGmq5Xc28&U@JAPp(0eV{ma>&af2)Xc{TCMW$+lHCYMPc1IpfdSuAJd7-UmaM
z5pgGsQa{f9Z-^&~Ql|nBH~Z_bb{xHgn#wEPnCfk*dNurx$!aQ#@@j&N{))Y-_)L1B
zklUOGmn6me41qQjq)qjuIZZi+-=E`npHm+h@_-_!(yuG+lGuq$EqD%ew7S(Nm-Py&
zYKG0N)0uo4YVVwmX_fS|b6y9_s
z!*8_6mg(iNCD&Q-L>?xL-?7ojB!f=em{%n1RE*aVC|5DdJrN
ze{`d@e+Mq}KwM;^JA4&5u`qK^GU^v>5utPvw2b6tf5GB+STIYwt+~2-*ZM8XZ&su@
z?0m)l@!|&D$S&%e`AV>E=Q`D;=0mo+asCxCjLhue+p3ZhZ#{mOU!NvV;od;9Xiuwf
zgmnFz-F3FQ2}te_hO5r`LfknlIOXr&Eb2XY88g(f-bk>o$B2L&3U#~HS8?Kc@4_>Z
zV%HvZ-2L5Ae+RH#WG^3Rw&mwAT`CCs@qpa#J6)@yNr#8c!yU)Mx~i<_*>Vy`g|WpJASRGyQGT4sdQtz0Tm&;^
z;ggpEadpI?HKvzycOXZ#6=7!3)pZI60cr=em?Rx|K?VU_uEXhN_{>fzQgLVYu2e={
zf6+xop<(uIv!^I_ivwW?)Ouv%#$EoDB%oGOs&M7tom~)sK$keoQM!D_S0m2)w*3p2
z8@41B$!TX3PIUt1Q!^9$TdwuCrsslC@OjUg|1!#aZ?6P$P`VqC;OH5tsWp$iUAQ@*
zDjKGzC90tX>o(x{E|DAd3qyKAnf2s>$*a*qKXM>B5=7}{Gr-js99ZFTw*8~%kzIMc
zsQS`{j0ghI?zI$Rnbf|P34D>hN?N?APBytE3WuC#T-JzvCiW#m_(pObx1$nr)}T64$UNl8Jhl=Z;j
zJvwIk(ly(5dVG;w5ph5l{goRRZl;}-j_dCY4`6*}dxfwz31
zJIurjv(iV+@?>CxJxQZrU==Q-lxM{nY!F2hIcnT62OC*|)9olZU_C=;OC4MTda_(X5}FfgHZ3xLxJ;XmACz
zu&Rj!r=GuR``k7Hw>jTo$(_z1tU=j&yFGqGqI3D`)hp;68AKd0myriD!pD15YITBN
z4a<{Q!efk$+4E*v6i0)R3~ji1HBX%W)pC>m&nN3?m7Vicg;f;g&a0fgesKQ~^@s$8
zLvHuO3<~rDC!?p5VI|Z@sK3<
z8a9{>H!swe`a&obeJ7fQISG`yB)G5W;n3_$l~+WzKIW?A@LL(NWB4cOu4{>YvL#>>B2sn@h*
zTYv`IXmHlAw@_ohUYC?WH0J&O<`WOIIIxsu`P{fD@q_QkWXW?WbmG+3c!fDo$%4kY
zggG1ng~uY?(VoAVz9Z}FVT~C6!$?Niq0cHq;4C91q2kdfv3hp5&b|B>Zoy-o(FTOu
z<0SQCxX0fUP}?;x$)bY=*L#Pbn$^(`Te9F-)Z=cm^;PWPaf0T#63%zybIv3XA?2<1
zPlN{5p#nV7dYL>$>Lv2tL*cu+pS{-&R!~lA@2i#?%q^3?tB~W6!vng})*V4T`TU>x
zB+Fv&ZMOT;+rkAXh`I$Ea=^i*U#l%3lAtiV=~aBWxhJ~#j7SLEMr*{jSH8-o$cVd~
zig>5fzI@>?3xBshlyF~*Ul;b#q;2Z;JfY8>5pF8kCoY-~2J|hSK-&E!JiW};=3wl#
z$9dk{+eWKo)hEbHJ!^M(XGbdwVi{30Pl+(IT=Kzbq~I($-PB7!6T6rImk3Pur!HfE
z2X_gTz0a2^T!Jnq%}E58ac;6R)Y<)e8jhd&UzE8pG}YS*jM;gucZI`v5Wk%PWdX%|
z5Gx{>8MD^%s(7PAG|&f!d1f^>5u04u>Pp@A*rSImf-q8ILNp*FjayJ`E68Fl_PanL
zbzpTt4Hc4YS717lPEGH40##s(UvF~(XgoO$oZPMOTac)GVa3&H{&^tbOmVmw`+m05uvCqTYmSaz
zVM{};?(*;azq)TkWn2V550SErTEevK8=@C}iB3Yys#b0|G`dSZ4?f>EEa0OMw65zlD;$C8>xHoRNTL*u!FBjQP
zGnwiW`JNhe)A`7)iG==ko*D6%J1}cKho%JYva;l_5-E1;<$6{>m|p-!_ZkE$<7LXd
z^IsL5C@9h8GQJw7J6!q`M0IsH-`%>Uo?fP^W_^`@+h?#>NDw)=@diF%)3nM=QWpmg
zHRK2Mnd9eY^MkzSd8W#hVB=0yBoN3Uv&3mK%G@GQz~sTeSVB}tQyOGu%`5!9PIqV{`qME~RSc&;%u@#HDh@kuu}3~7LUReBoK
z%M@qGrS`?o&qJ*M&;xb5qBCXFnFV&g&?+#Q$mX*n?>v2`92Nfv
z0s-Czarn6sA2QCPx`ESo^N|DC+>8`-E6$X8#~>n?Nwkvl$zS;&$to>B7f}n#Nu&>n
zXBapD#m;8_vI*#0rY;`7xSx{%c?d9Ekh!UPAOGbProLfd9SHi;8^N|*_8A~l^&1c|
zETwC@;={l|rE_qFi{1aIg+yjz8HWh{A~y|M&{9lg2I_hDPj|j*dSLDW;o8*GO3Lh~
z_r5g?Y&=4Mk%1ii!#C_B-$&RVTC8|I`U3L={hEPX;1NS9Ss-J0%s47Ca2Otn1;>If
zLWUW8L%uBXb{r8fSin$0hvP1#CFTYz9Gd2
zE?78OihZVxKKA^*(G{BI3(N`NiodkDC90u`l%b$(u;M|}wFv%|3_tf_??BY^;r?jw
zv^+fgzwZ${sY}}VBBXxO$HfkWipImqZ8aV-=9$iSwHvZ3!trkemvgE*w&axZ*n8b(
zWYBl|Jy|_(`jq{3l2_1pMizrLP%$Kb4Pr$Nl1S8D{Fwjqf0D!qYdwMSh3Z}*-oG!k
zrvCUe^4B<0J@qit!=|3o=4Dg5R?3_fSNJ~UtR@`zA^m6T;n}eH-Lk4VE$i^1RX^F2
z<-WfH@m$LL?LSYX%(p+^j7Eq9%3|-cX!I{Rye?S8-w&JZ0O+#-?gFLwB^wPG~U5vtm(Sx$I~rg`L4h%#7+z3yqEMxb6+r$
zn=Q06&Cy$Y5U=f9@I$y;^WTEKuCf|=YSH@|FR$sAvC8LF^XOw&jJ)f?7V-Rx-NIj}
z4(ovv&nElSh
zDY6yT6j1X@kLTtp;Cv=tO8(jc)g?J8wwS=7Kyai)=T3=y+^5s??OhGiR2G|_FNP7p
z0!#DZfS#t+&HALcOXtK1X^c7nriSayonL~0*}qkaC7kgzw2jcJ67m+aJ&C?(P#hH+
z&Q#y0U~a#U$4QN+M}m3;gIC6EbB7-?eoyW_IJYtsj=CRr06?6>30mi8WclN(xk3QC
z{G$3#&jFH}7NUN
z7X(;3)D#qI-=I2g^2cXKyU^YwyV7m3P!SCYw^dHJkcFp53?9^Grs;@9X7Wku*INXA
zz{XkWfbejIH~M@P(u=JB>}fXFT-YvM*8fqi?;~jG&B*1ek99l9!&;sc0SaJD;t1&D
zx6qC-YCqf&1Fnt4uSqrEZCu5%ohi729#3Z4tqX
z_Xfp(zDDjNefqP9x?toF0`i=HYtU$Hgzu}{!_G*(yA}V0L$46w^L0-Sw{s3++7`?>eVY&!(Zk2AX2`7GbM2*vaMjLw4FDg
zFa6f*)u|(zge)bHzta^r#5$KDKbeMx6`Q1ImoBwGv@+lnjQH=SUjVIDnrI6-2Pj)<
zODb0t~^m^9=cLJMz8J*|MWCSJe*DC_+9_
zlDg$e8tHlEYitW_!8>#$TzFvqXKA>Y=-|W6V7#7R@Sq-
zCQ{#wXHW@s%0j~o*<{$jL(JUsh1c898r59yE!xZ0cjP;dTxP`~x5&pvbD=Bw6=2EE
z1<>i%oYv2nIVtJKtBQ+z4ipGCU`Jop2W>41HHyS_u8$1Zpr&s?_P?`VyK)luW0H-h
zObh`nrwfa@gA9-c%<2-
    *DaN|YA&a(a^b|r3>x+f$^=HR}wi0F%pNlRE$
z;$>r8VncS=T{P;ZL!{|A{lxt4mG8j92AjJB7P$qaE6lN9{|#fkA8jf9c@gEN<+3xG
z34r{7I#20?<|KUOirz(Do`6qmpe+12++K!o;lpv{VW+@R)H>A}onzgs?%k37{Nh7-
zyKA98gpPd9Pc6=6#+6+jA;g~Km19Cse^Za6hwF8yY92Ueyp
ztqc>aI)_nu&l}t(jq}A#R~DQeA-eX{m)0BK51d{sc*Z-Xcw4LDPNf5UKb&aA8T#hW
z-*qlOOwAEVz6J_&IhCt}5rnH@#(z{DsMWOn?)t%dVU(~(cMRlSxbi)}P0B>~h6+vv
zb0PT$h|9ZoJfDi6D-XWv;l-&NT>6d=NaNgk!ShClc^D%qA_36eWjGHoX2l#
zph3fTR41lhf4DkzkmWp@TQ;b+5KjFr`BB%LexN2Qmi{n82<1iJso7S<`uNNq1#*9r
zUjCevX`T9$EbV5kgrk?qk#47S!p`v;{T+y-cH@pGH9Lvf@?SRTxwb?ba#99
zlN?EWN9LR(gr<-W5taT~&czad8t{{KuWKyfslROY{cz!St?E0JEC|5
zSdlxk8L}JZ*Cy7vU_IVv)tgj|suTTZn`WFz!^pZ(U5lpe!nGq=68({yW@c;NC)xF)
zL_6I>TfajRU)9_FY||)kA4hfOV&(2=LxK%=$WZddc*@?F6FC;IApenliNxvc=ka$*YR$mW+QmW|=W&+b$9t1f#qM_$Z<>=01wIDp4M#^W&eXs)-Q!5b71j+9ct
zYM0wCWh-R3xjs06sx^#a?+&W;yfViC79j~>0jr-dm8C02)mzcVs~mNhTJOlI_wrLML~
z@b5J=@4ByNi;;Z6hr6b!vazlQgU^wSkm}hPON`9p<3B;Eb@m7P%Ri04C>nC1v~%jN
z4gMTyHow_heMq;7-T~c9)_oJ8AfOX1$)wc*3Q<57o`B)4l0lE1$YaG*$lum|vcmO^
z&omP&h6DzBMUeUS9KJW=IT3mn_wcqa8`C)SEIir?0S!h0q9yhX?D@hSDp%52mw}m#
zxXtNm-s#yrv$Zv-MysK5{Z#++^6dGao#ap-pB0e@YYK3pbN|&Vte9vUh1`^73(o7Q
z!&<3&%~zt>&Lv*os&j>1%$>Ik>|Zjt;quw7o3>7~C>iDSN%h+zoI
z)=YemG;Z_3_B<(Q2Oz`P!TtlUAPE{QkDfbk3aBE8R(`l!AwB&QTzhb;@bX|4I*OX1
zHDx~2^2P!}eZ6U*r#l+iOGw;l;$&B*hoy^Ty64guM2#V8!cS*4W+_?4d=o9v+DI66
z%0J1^#BH(pEsu3`&fYa0nU
z{pe~7b@r2>Q#sLjv41(xnvbq?deLNJuk^cfQK;btf7&7N+#IR_T2#+kGhhPym!!ds
zNRbpH{7=`D>c2#@$ph{E2GKk`V#?}#UHCqi9P?I)>FPNbsj*)M6I>#L_ef$a_$OuY
zICM=V^nh9H+{Y##CbESFbo6r&-`H7hd%2PxHbiMbEyZ
z;4UBF-hViB5k*fxP)bp++L?Ei8AQ>*kh17r_6>h=UuHd`fUeV0nEQkD-mi};pE>rW
zz4&!i|9jK*;WgfWG2f8R%8tR(cDE||!sP3JFGKmY9|0tge{heCgPNR+D|Iukku};1
zJqZq&-9N9&8l8siv4@W?vzRgOTxJ$SCfv^KKR=dJ1XsQ9*d7#?+rG@8e7tgnsZFbV
z{3zWJ4mHIezkOx|c{0cas^OuvbuYBReAxMI9`WcFm`e(D+V>5-rbJ|0!jdb>){Bz$
z9FyN=`0(>rR4oQCaFwGO0lvvl@LZw)QuU+4*UnsSAD@!$gSH$3n9NAL@djnf8f>PXE^+Xw<1t+v02pny1r~FvD%MgV#yKO7@-2OBx
zxXMz+hQtj++)SVsYx#$9fXn{r`hu5zm&(;VLkPj&eRL^-3oG$-x80$B=)}~=-AHla
zd^-U%YCZn4&32K0q1YO;6b?>tZ=2>C-F0+0b3;*19hOTaF5|hUSu2c7aI!bto?Gwi
zV!yO@rZEdJf#6ss%D^}rjv^z|1c%TB#uxG$-R*hxAhTZgVz=Uo&PaHLTaes)@33O|
zz?_iR=ktvJEMp5O_W|sghjAr2M$~*TD#Ai5qJ46W7?~f{%>%pmx5*}VuDlyXy--I6
z;pkY5bEDq6L2W|xv)gTfFG?@5-h_pgh>cQQSHdQmNcHdljP~5`x-tySDgMn|g_rYQ
z#u|o0Xd1C9BGUJq_T=Wu_GVe#;&N^t+H+U0akEk!B*5d)HuGSGCX;LQetKVvcLYAd
z-yO_(4~)#L6ds^vOcmSMvwz;5#N~UCnOMklUlssK@V$1Iw
z%bzY{c=Pgn9EHQ8<;|GKLI2QzKcMe>-_)8}T+Ys;ZCaYQe;!Y=^1MFH^}sb}q3~pS
z_(D`>d%&;xuzl+PJZ-LS>Z>gj0bT!2J2K0M-Bsk5tRvOfmI~#~%i7nlPhzGbE+Fd0
z7d?7jB@KS=&yR=6s!(JKH0$MlLvmg`N*$ahRC!Ca*!S|_b2paIff?Q)mT#$~giP(!
zU<5-ZBi{)c9bONw_Y35pBNT|pw=R-PBfmJOf=k!aOb2_9)S{ws+pIMVvETQ+^}0(1
zI(S9=_}fgo_=KD+n~f}|aiK$a3*ED|&n*K>^5<$K$=s=;K86Iu
zZ5GE0RhR45_NS7%XL4?O`Z2yn9JW+aLCS4>TVNKuIM3T5{AjO3#(rY8y}pt4EDiq9
z6%P4%_6mSy;C|+x9(YFE2wmR$EBav)2y`w^`co2VMf&sP!r8rr=Ith40cQqUOaN$=
z^AEJbrE<yyVzX3OC7
zEQjTJPGe7E_nNkT1<%NqUf&4RU)OhOd8!(RP`AZ$Q?#K#fizrO_Kuz5P;Hl@+cp3b
zh_&iQ$@Eyp{17BAmAauug|7WiLNn-=7NE?eiwrgZeiRYZv?HiP+p46Usc0;pwWXNN
zSIhE(E9C0{2A~!LqXB!HQVZ30JSA#{|FA?{0%DtI$i5;uGRW=?z|Er069l4EGFrPF
zpWh-NJTAZE#Mb59wxmog1t!Sum1HLfL&-Gt!0CTBEso@UuBbPqHh2pV%xu-|3`q;?D&U-0#(4>KKL~B^3ZGA(P@*LWhT0n^Lzc~Jq}K=
z)Nu^naQBg>P@ixB3eJzqsPBzcSvAMaD(b(+Iq@QIncf9|mB@Flw-JUBo&oc8#^M|-
zA>P4@xF+!;g{sjgMh2#I~7>l}9a63tlu
zhg~Sqj|v9n7k88|^ebw@HfQdsAQKgN`!(N7*TpGj8d((WS)x5EP+kETvqNNLCJlDOc4zz)aKn-V
zQ(>4)XR<*Dql5sUqd%k+zU0*wIf02HPJSWIB*N*w^q{Z;
zKYt@RmA);bF0=tM7A+jmbWiX;?+E68$cUkr%^ya@SAbq`BUdR_IV&gs?NYPxsA
z7iIh0fqy-_ewCC6oo#Lt;ipVmzn1b^R<&N_eGSv&5lf&q0bVF-p2)g1VcQ*ig(d;D
z$5aEFVUMTT{{A$*$(Y@)Qk~Lr{79gJ8jOTmR(~a+6(|l)jGYU3-)rHwagbnw@9K
zw;w^4Ebh@28@GBdskyW6t;$`3wZ~V_&hzEA274IMrX`qFUeiPvG0m)Z&%pzyD|0tr
zv)rq&M|>wLIZDMe!5rJULoQ~cI&?(p{I$iw)h7l@h4ydJ!7^bBJ+JM%_%PouNcDjE5;pYbcdq#Vxope$c(&!g<5D0_4D
z+1YG0Vpf9CX|cGjm2m7Kr^MEhu;1z-Vm8L4xR9&_O$W>~$xj+t1~W_EmSpcFs*!(C
zuQUtnjw?q&kU#`fbcqIyNEV>5mxmrmL#o3bJA-iOqC8~gj^KahmK*yub3D_CXLKfL
zgAWgF`D2STG1b$CN|}@ic*rUvrV1>qsTrH7=mcpVzT3X>Z1=CH8{B!ot(qwgvtr7fL#jz}bVH!6Hrsad;;OXOi-oTEV2u)zSTAZ(@WM8$P
zv+ITTbs=wPt^5CnFmI>DnMeC{HZ_E!aCABe3PYt1ALURAYzJ_zS_WKVzYN3{8t3|d
ziWV|n@Cmi;i5oa#YMNRE;CeOlFm|B=wS#TtO2v>1rL`H&;>r}Z-=DL-+7oW`@9t{>
zip*L>UF7dgV>dOfB5bVGt-t3V%^
zpX7Ul)%b)<65$w5AdCRr$&i8V^f<$!Oj@-*NoZ+AZ97w(U!pmGz`#<6(%|WPdSN=O
zo$mH>^v8kBx=rr)EMsZr>2W6=uLr_0O;Xxt
zmL_X(_qr?&KS^3~t)Z0l>se%Gdtx8&UZ7v|)?68ANL2!Sp9xyLYjwbH_WM#=nvPiq
zJNJ#VuYm8Dnhu|9dHB43MAW@3Y-f1>RSZF?-SdwB`!i>??w~hurvz88|7$UUzz8x8
z5{&7dLl#men1Y)xCd~$*I~(ZmDeJE0O^3h>=HIVBqXDC0HGcO9Ur2C$^jO2f2tKg%
zF18Lsfe!Bs8B9fE!cXSl{nOA71aC@>m9PLUc06XS(YO*wOm64uU
zqF?!A`tbdQAC)0+s1jGXWgk*~9@J7L-wP*vX|hY9uB_|wN#RD%_XHdg^@Ahjpqtd_
z-{`eiooQBZDywZ_w5uGrdj{v)Z^BZa_9&_zeteA~34EzxHx6F$D?VvD;SS9@!Q{0Y
z32w_W-|l4M^$)>>6pn2oQ+AobBr9szMqkR|8Th&;H>TOB{zp
z#lQ7TWu5QdDOaT@cW)IGeju?wEGT+8y33i|TH5Y;3ZCh)U%lIzmBgCik3{OEv&7G7
z{7Q}1>%6V5{#i0K#t0(~rNr{DIyHs1yUub%j{CWscI_=`nDTMiz2oSE4vzEh$Q)vFR;y_E&xX%{MFV`{0F}fN{lf|o2JK8jk
z=H*(IuA5o8Mx!4qLgv;_kMc5p$}Nh_4C~r33|)aA&hIsi>XFSuACFTqg*+S{*y*fc
zk5js?yd83nwf9a=YSLJEr4Y)!Cg#U=&bd*zRUsRlr|gN8o&&1>%@+@|2?7IN>B*a}tYmy9rI=Kr*5l+JPZ@gsiu?te<4Dbv!G}Zjr8*3o9j&c%|q@n#vT5!cVv(CId7j+Togq`
z3w5T~sUpJu9`iEkXx)m2TUt=PEwqDotW7;EF40A>1NT>SoB#iqo?-tnJsoS2I`t`_
z^$sl1?YY1Q^CI4N{Ku^TF>bvXwRv}Iq-&#yl1Qj19Oe5?R7<09(=O1qTYK=Ov3xzX
zd*ej4kx+SvP&uL$H-e3?Owp&>XYK?)KEtU`T%flri@YTAn_E-vd!xE4f7exIS^5iNsFIDJ
zcDY7wQMoN2Res-CHb;ajiEb*liQ%m11klOuai4f_)$4CRUr;Obc#5F=js@5pa0s{~
z+&7?@92AxEBI?A731}OIS(BLJdm1m|OSzL4hhlxeCq@sb4}KFemQUkFgc@ZUuqghk
z2|;brdwuT>2*>auTH|KXft^z8901`1$R|+Ov5Wdl;grWVPpLsJmdhTb))jIw_
zcfro>x0|h%>sTAxpNaf2k{#{AyxP_peE)h9p3UMSOmU*oK>Q&X~dYma=~vPQwh@rVw=HIGM4A^=x&P$p}|m&7_%6_Y|vx3
ziJB{P66GD`{gf;rkl1rVWJ)+%!K0*6%$*YR$`@9^(Hn|F-(ujgasvFmx^qks7-Ld0
z69tntwJt!Sq-}$zxothTA}vW?o;#^*$HFKZq{b0^%az5#T-uImn`!z0%jOm1`vsBE=a@)0fAKA(nA;y_2})aC8Q+eoz?pV
zoy;g>%f|PudQOv9IkgJkG&^Px$<%MuB9T5Yp^NLa?V
z1MNK}q;vD0g%)34N(zJg@F3|XB|Iyv@Qm=)fCZnDhA6z&wl5`Fgh4=DU|n$U~I%VdKMFzv?&Iq%cNON49@Eyic)ulcTCa*!_mBGAZR3
zgyds6Y*VP^GfH}rb>HdoI
z=-rBU_tQa|xBZ7mql%P^)Cg0UQ1ithYY|>NMhVnu%c}XBo&@x&<*MI&>CZO5_!1Q9
zrsS!^14sMAP&YmTD~tQ&c^ZRj*pVo?GEm%c1woa<*fd+=J5|0sBi|>Y+{I$lz`Zkg
z6)g^5Va;f6&15OL94R%j8Q=7of7Xd#6#tGC#mljU^mt=7!`;Gx;Y
z5wga&5qQVBmv>oDz*B8!&$4$vq2qd|P{YK^-&^*9xAC9apog-q7rMS^=FtnNH@1FE
zD7HyREoFj7xQ!GsWdGLnEs1ge60bizwksL_UCfGdh6BHRU^TTR$1_sw8^Yi^Po6A(
zcB06jgUu6n@{MQ30aH(=AMMFFBL1uFYzW9tgMoV`EA*(8^v!!$ln|1VvXuE
zuI5a+hUP1P$4Uf^5y#ecuI*N~L(%DjzHLHpxds{h9h+R#48#A6euw6-fKcN)XjY(>
zrQG^^Y!LEHz3u>flSi&0aquPAzWAp|a(vpooQ0%$?&mHHOP8zC#hCq;*HfenzH0L;
z;?~S#`z?hgOz#oQxEA=pm>5-KPXUsu!xy>oZ;t&5rp*rs%10HN8U$I8&~=GQ_{@$<
zHJEyt6Ixe<0zdcoddMtwlnG}Q6v@04a1m$=tB)wi`j=2(2@(0
zIoP#+bxHXup02S(deC%Ywo
zze^}YtyPuuoQIyslVYa@yPanaoAQE@FJOQR8b&JGKM{0)BDzb{eM#rvJT`b}kyD>O
z!evr>p29@J5f=-QB|Y7=8U$g{^*L>J%C=FtHFMG=)IYw
zyM9sK1RRNCzBMy&y^<|VKGj+$D1}iZnJq&)n>jSP2iSQ5d{Aj|SC&HBQ0%!leDlXR
z#|ww-PV4skPXtM3w3?^HY%9F&ATxN3J>eQLD?&1f?Ye@d9cg)u$LGPP^4BK_7xgOc
z_;}dQpy`F%=9~~h<
zvpDm~o7|iC9EoFvg%pp`asktXk}c
z?vwG_Z)1PKBAxbZ`_cW@tJdc}@THQ;pU4J{;C?)+!ZqmMDd!FUuqV?xmNsLhiR4H`
zv?91ARFstsO3$ft4XrY!jOIvqS&YAy*~qch(cR|C9vwwJNexg^&!cVZ{>2HtWu!)i
za?C(qzO?<_5xS`SD1Q<>G&IwIFvE
zrQ^6i(kTO1;@aqH_y~`=9sVV(U%q@)v|rUaRf^lVY<-=7%D`s5GBDfRjkZKU)Iu6K
z?ASVr;Lp>qrk(fGRd-B-##`KU|EU)xf50H-?lVKa|FFybf=bhZh?Mt=b;T40%c1K%
z7I)@m^6=G>L+^2
z#DWm&d|^&UWb%w}1(LUz(RGax;kC(^l7q!76K`Eu(AcKu(SF|@Q-#5Hche;KT?8a1nS22&tl=mvN
zvyS%8QoWKxf4RCf%gi(@P`o!F{NIS-5N0p0ldIlbep}(9*4I4nme|k8#6PkH9EW@`
zvsQBKXT}fiyUfBHWu&X3Y)Mlq^Y+PEfz{k1O;lbbsZ;_tq_#mtT00PGh`RmSS)7wj
zKxaTuUC0|*ro2h~<1*@F=JauJ55X@YYH?aAtoMi-i(iYh3BUb&cR)zSicsY*knZH6
zCk2CcVC;g(&C1i~-DDZwrkR{mgn-2&dU+Xd!3aAGH9xTWP3*%NpV~FkM3jQANScPX
zw{j1Pb>tvbyDc~~L5S>XG?C%C?0BLZe*{MdgNQA}P{^a=z7=`ZYWk{Xav$u`G$au!
z-TxS!b->xc)mn2tq`oiB-AJ-#Jbg6n1s(`Ko{PwdR8BPLBo`R{qEXH}U0#uKwLE{b}UpR5e=F=OG;
zye%0N9U#NP`r{P
zRE=oK>jiw<=cyvutH$Av`(ESLqusXj7~(^A#i;Sj29r&G8do->gT4=C?_8D6C@IWhw0kjuq-o8o-&azm^{6ov(vg!YPvp_1$$D&v3f?RxbsFc4KG(gZP764YN-%
zk(Pz)VA`zs9SlsXGA$#YY+s~QXxa~|WNI`jQc3`3lNqs|A)Mu}s(&(LRw&->&PNUz
zerj2^4&K9eKKcV*5GpOF^&O{fpo$?;JgjBT&r^mpsMSo4LyXFIu+!z^N>DaKy^*tO
z7V$ZTCAVraz06X4X9rGLy=Kb1(}}$J$s(eh)@vt*pg!3t4@)p8`HTL;Pmc?>&1I3q!NHSK4xvBl%hx-t;z39ev
z8~$7(&28Dv5fKEF_p}48d)A^tjoI*Zr5>SNW5nv>vYHr}qqC7<#5Oh3~_9vxBh)ye=u}Vqk
zn2+f99)Xvaw|lNXlEd7={i7SlT(fxqoUCkFrhYM;s%%*y6{yCE=a4Z<=+Nd$#XPmt
zTHB6c=E1?eV8v?V^g(a4A2FuhtCe%0QD@?2@@u*)(+tS|36Bj8%%VBH7qJ_ErAprE
zzyoBv5fc$NgW0KNkCs#s${I~~z~6T152)uWE$b_Pe7BpEdXB{iPZ%{&F(75Th(|wY
zHtvW}&Z*qesm4a@vXp6yV{!t25aoU#jDQa3r%d*o(?-B`uY^Gdhe@n6Ff=v%H&OLM
zzw&X~5UQA0Wh#N&Cy=q11^KrQ@^)U?hIj0kRDP~_)&4r8H$C?dD>EtN7qWBihAowP
z{P2Cat2{=BXoWKHXo*Dq5nt1aQ~*}6!fv`t1h(GlTzUt-k}iL7sJZYkTU061OF~<3
zWCO`#(rfQ=o=(lS>QAT%VyYXA6%1lftzxMk(f)*<1OUR-H9<1%uDSaTix|DJ;G-ql
zpceOC34mt)iaHNd>>}RKbTh8A&+D20snv;LSW!MG4k<%M*vxW^8wexd+?H?bQX=L8
zO71p`pDcr}8!CFZ5^pMXC?Qf_Pr>$d7UMojzwW-s$Zo%;sle6Yix#7i^fpXDo&b%~+)GurYcj5LQ%VaD@W6
zq)EhDADsl!B&YYwF2))OUFdBp(62X8E10DNuFCCf8Z1P8O$q7qGF036fCq?q9mi0c
zs3FYC2@B2NaIuECFus90`@iV|`$qv;AqiAHN985FU~oBCwV1;H_+56c{>I)ox
z$7&T+spU+$Av!%klv~+Eol8y%{5P8vw%r-JODZb(ls*t75r!olL-Ao-I_N&7`WBT7
z1Nx%2LQ2r>_b}AUGx4BzldOu6(6!g{(a3eYY5cOLJY4_WsSgmyS|;_r{4_F+xy;Er
z`TqZ~_10lgMP1vtg`$XnbW3-43lh@O-QC@(fJjT{fV6ZEIiPem4Baqv!_a(Ze7@&>
zzu$HJ{`BHFXP>?Is(Y<K`JeSwk-^hRmm_W>Y=;uK+q0wQbXcK>D6$o$xyk99Y5qOUb0p
zFaDC6pOyU19{SU!`jiokM23`wEr^b_Wwj^%Iv_@-VuvM-Y|SL}j$cN32+vA^>M4M@
zwf;7ID~*N{hPJBCy@#g3kW48~@d5ZNZ{oj#6q9~6x+(`4&LML$OJocn1Be=VEo%;pqd3Zru|>&)zo~K9WKf4$$c;x@-Zk-
z&Jf5U4Rq#3u+ZbF6i9^Ld#096)@KxzrQ;As2Lf;YhaBPw{Ko3vHQllBq7oP|MRcM)
zk^G=L0T3{IxcX{5jnu56!Vu9bFP_Fbor;>p5c~JR!64~1)%QR^LEGzfqJ~6qHJY!i
zK`qUVn{>pO_#bib0(KqIv-Z-$-*$tAYKkgqZtbOf@#jFqn?D>Ny2q~yzEuTD6f((1
z&U!am`&hXZ(-Qvo&;h~0Q^DzxI--xqo#xW8Ij?%g40Izt2?YAa0gSTbO?R~6OFt#u
z$zk?u;MdU<2E;!CKkUf-8tA42mSoaj-I?Nt9gb4`GJgrL4^(`Z7z-;eVh10r^4F~Tsg0WBcPPg
z@X_(8N&Gbky+-(G{a{4DnoJ$=3}O-*ZqqoC-o0w&H?&kLDh2-eQtNX@U@|s|_7`Yv
zg_x6@HBkpX9
zEUbvkLudW2=T}rWFPiW
zRB#_I3Z*ajHn(@DYA5tfL?!Vvyt~OSPzTh48Xe{=n_F28^_mEDC-AH!|F;UBIl>(cnEFvY+DFd~_#pm!Ri-Sl7i^JL!eY
zhk%b99h0p>hs+LK?ZiH>&|iZ)H3pZ1o^{2Gq$Iw%gX>FwF)(;^Mb-;!jcn}AYfSQt
z4UwF*i2w3%;}kdf=H2-2am_d23u?=P3aaOaGUVOImtNa~bt)x*YiYT#(m1(DBtyAp0CqxyV6MW|iGE
zF1L?P_<181pc9>5EO0*MVG3D@`10p7n40Pm!XeSzzX4=Ji?CcT0cD{Zo4^AMr9P+I
zyE?>8yAKd)%CE=9!V!k|h*e}A$K{t$OHd;7ku%4O?yEyY)yW0ie_fWs8FeZ>{pE;)
zG;Od%K+tVuvcBnu(JmyJF-@t^oS|1)#LvPU^W#9PK*yC6Vi
zXD;^wp+Q;iNN!x+J`SYcY%sAV@TfeXMuONe8q3$REXLp(Xs6
zgk$)5%Kw~CBABXCMZJu@ey)e;x1SUFTQ@kbcDii8`=jGIXawyZ_C+hKr$(A`k9ye2
zy~KU~RT+W}pvr@C0XtFZ9d?%0xhJFIWN>o@({juyz=4FNK$n2;oXRP6U6-T;ov+v;jy19j;rcI+)s|4
zG>93ByFm8X4?P#>39tpqyh{7JNjlbquL!p=I$H9EQ)sVd>``1N9yjldM$(l3Rzs*KnHlLW=Yr7zAv)_SK^6KC)+4YHQOy-|)CY
znnhRJD8hgC5Jqu`-x;fNQxvp)sP#+(Q_kX{Zm(>soMWHws*?{+T!4pP^1d^8`&~jp
z!_wNYoX&!q?HQL!yTDDwK@<&I%G^?mwdHOVeJ5Rr^N|E<4>{B5&b6m%
zUS$bb6R6IBK7p$Y|F8m$KZDwMq;q^XubFm>o8JV^M=i
zb@#%EHAebJmzcA$m>|lfez}^)(Pr+PIYaEYZ?UpiUUP?v!56o&y;AyXtLJ?y4b|Jz
z&+J4pOXeGDAyPb^?Y3Xfq@c(!q$0ZV&5c(q$@RD?_?Ub4o0EMp9OsM5#0;wn>u_R8
zo2BV~XP6<6XoUupKlVAIJ^0U?oC#bSaR-)IJC-7U&zl$j1OQLdb}Qyxi(rC{CH|;Z
z!0?S0VPOyTG5$l=><3MNH2;
zN#Cb7cNrM#76lA4#X;`Bsk|St@@;kKiOKAEFdlVOTA{V-_S=8GNvt!zYm5AL@`eM9
zs;)SA#2iD*U7vC?Jukqx$<+C8Dc4aMZGgEHq^bR!<+@P&BRuOe`o;B35pwsUhJ%=y
z$>FLYT$FmWQgZI}kv3x9S3=Kqtq>P_z*DaKZl>1{n`>Uy6s{;Q^%2lAFF-uC57KbKi7s@Rof6
zjnnz^VUm!Al&{%$cm27`3|k*(;rJef?JPdIh6YpAUsvpgqlUHEJ(C7Ij=W)9{(8=z
zjnd_}-1D;jW#t!OHB=B?hyP%w5S^mG1)4S80V6?vC|??bF~iB
zVLh6jpoib!-DRz(_1rj9&(&eokoIim`;nU&0Un9`yU$@GA_BaAHVG~tsXh!5zee2W
zNEB4#q1Ye93e`2L|H?q#f>@$kvhA$oA~I_!_vSt(L+*cw9r?4lZN*pl4HLIZzjew_
zjr#&GE8Z}0AMgx~XzBAivbfAk@Y08+vKvJU?5Is`c~joa%a#%cpim81>;pit_5yg=`^*bW+e_R%&pX0vJZ8Ydk*;fB+uK2WSc~b6@C>4~F(Tw|$
zDMtI0B>+1692goqi}~nyzyaBSVA|_FOv&yo-^yF}Fr5Z}ZJ0)U6gweukoKhOv)S^%SD1K|Sv9GhB}L!-xaYnfLeRND{$w
zkfh-KS8;w7%PTs&pI$RYqldmmFka9bD|}H)L^NTi0cr628=6l^Fu~AY=LnEPhtD0E>7oouRG=rEzb1VU%{{x>eIv(q#q7QD;7`dsnDul
zjiR&v)MQ&om*YLsHqt4v3d77P4B-tmr1`49lBOT%WaL)OSzmDI=ysq+rwmkuBkt#m
zaKuRO6$a`EC+qfa_|g%2VTvm%Kkb^ck`}IRRN=XBD3I^@B80O-9*lS37ukpPVI4zOzi~elo;Ju4b0hj
z6A5*^5CUAJZ^0v}v^Z%8(Gcu6hytn)f$p1j%t~7*-z4gXd|ZEMs2v~HKOX(6epHn%
zxw`IXi5@UcFEN_Sk
z9Ho;Zq1q=|LvVxhYS-C^xboa7SJZcAu79!sLS*F2NX=P&?^X@$HX275O(I~pE
zD0^@1Q2*W?v~-|E%HNhtK+OPN((zlD*ujFE_JQTcz4@Dy~K}Iz?1`9Vjz=y59
zhT5jw|AI-u18Qxzp(7EP^6#sVmh
z!7$YM1bJ?QbxbN`hvw1&W7gl+1PTm3T9s`{-eBmAsq&41n2s6mGqFK$JFl;b2N(0N
z2;pnL#R2qj344mC&7C}J#JHuw^<_}_R}#cmRBv|r+`jQgN1;?Ss1FmhaF_Q^CKqNM
z6rO276Fq*zQ~eB2ioIQI?ZpCp)WhdGXD46If8mp1SDhE||MHZr2yq{?b%2PWJMCHI
zhwa^`cm_aPg(`TTIbMLZAAzR+(q{aw_^ws#hs#02{Cc}GfQkZlgjHEC981C}h){cw
zX2TkBr40>R1#gZdA^c?ICQ}=zfIUX2Fj3mzQt4PkJkvLdK=W9iSTMYUB^_9d&l2Ja
zN#}ctfZUDSIrlNoUm|>Hfr;FXdi`?MU7cb2J(ryJe-P}lRb*5asaHIDwg2=Z3JmTa
z^!BdyEvqqroEzQ|Ut;z5=YfZpl=1&{ktQpm-Di!J4kQhyWVI5E{Ui0Fm?SFF{9~+Y
zb&j+viCF6&saqC%2(8GF8OovNG2VZOsU(?SCN_~vk5A7s{8P5R+2g$dz4s~T&bKYg
zCqZJjnfTCT<903x#jA3_Cn((Nk+frp#pbb=A>g7Ddy3fC(O01z4`W}GnYbr1CKKb=
zVdrg6l+gn*o+ESsR=R>t9J)O6tkO3IAHa{isXAD|Rdeei&%7#_Z9B1heJ)@uX21gM
zTEln}WoM=3ODOed8CMGCCsQ{7>oHX^9O!yy@aIgY=J*Z><#C0Fc`u
zcb2C{nbK$RFb&wkYLD?}(I=8`&w;bBA+y~gvdZfS`?uN_+Mk;wjkQ4vCbqigSfKDj
z@_!3>QznP7pr%nxRb2?KpeGGG@^gx45J}`l`6d>JUui&@lBiO)r~)K7ahr8lGq+Qs$#Os}_u>mi;S8
zCENI4`-~$fN`?jam_R>gr14nG%74-DXv8u&w^el%0Mh+Nv*8XgdjK;)z_^kSwLWFB
z-oM3jIO(W>VhcPO+;R1HKE{sk;^i{>8@WOCocSdtweEy~7_ePWEECb&2i~I5mD)Vi
zpV~XL|j)X=~6>Vr9F!N1)PCZ){k9fA*UR=;FV~k7$mUio9quQ
ztQnO92weY>)oRl0H5(#1*H8^$o88ojk`JE_dxC$?!|tO=u?S|`wIV-=<9)|;ZxlB4
z_)DyqkLFkX^>Y!=Y6CO`fkJ-|aPWsW+g9oLCJJD$27*vToV0(Lcz<9#m0
zgV4}Fu%+Jd0Bn_vpVJ1aovexW
zd0zf36m?$4f|?}^gWA_?qpYPke&z-G$Nwodz4^B0%8vmspzG90KEgX-Gp=(HvKEeb
zS5<(EAem*-)hr@(9v~_=Y)~E`)f<%g@bX&NuOOAuE(bcM6NNwvs`UNVJ_~Z$rw!ze^1vCIx;ZnFnFdHi&6*=jkRg^6KM0V`E>kuIc}-Rduh;+4V}Bw{
z$wbs>SvBoml~^OPe0UP>Fgv>cx|2Zj8=`w-v}B47&~};5Aw#SG_~=x>aH`+AZgGh9
z>YIZ4{{FYLzHpnfOYFOk`QzNo56^>!ffN^l
zQ}R4UY-fL8LIXh)smX-XLG#LhtkY{MVSequF2hdeT>LQxYXU462*U%vnoL;9uI0Wao0l#Aw;Qij
z39jLV!sNjV!bcu=zrFw_1;K{#i&6TA`PlF4UFyc^K!ou~+;p4~;XkhKGEa+dGIqWr
z`o9kvI;BP;IVss`T;|jOs;j)x4%Yg+m;vXclV`{u9r!3k*I=M!-_|R~G{9_VcFwyl
zo9Y{8aNv_jNi*xAK4-PeQqeLtu%;a{@>PTwt?f%DF?2yC`m|J9g>dw@0mSr^$?lnt
zX=2$+HOJE2CczIPasS_X8lQJp0hLOcu1P-$W8d?U$+F9n%Y6Lo8z=rsY+D?F
z{!crDR>l6N?vHI91s1v^
zwHnr}hdSWB=MFs+Zb?q{?$%o6w3cCpj)OJ|9-%P{kt(Xds+6!Ma}VFR<>;EcH1e>9xJyw
zjnEk+D0{=vWv+={h)@3x@G*EQ>HYVoGk@Rw?HkJMPwDX8|9V&dQIZr)$n;UhKrKne
zi&uni55~Zoov$$NEH@y2AA#8wZ*j7^N~gDNTqOxY;xKUDuW-ElG_V!1*(_tGHslTI
zvi4Y+e0pc@!Fz#YqgErVp`=kZCUwTQkbUiDUh8S&(3*p?Y+HR=_;
zjXI6&tgH5~@F5K@duGP^IpQCzFJzp83f3mO(%CUiuJV-;U@Jt}H*FvYmbe(YrNrC4>K#2AeT{svqS%iIN{YE|)?T8nz5Z0z
zAWAW@o9}zrVC{xNGKiAvo2>G`wjgt*(vt$bYM%aDhxB_qQZs@I9q%P(R}P+jrI>{v
znZ{6}?+NdIQ}xlgBl1@2
zF8jL)>OXg0Vj>3-1}fug4^K|ZN4z7|B)gEr9L6&AoFwd7G-M^x1z2XGTL(jrUuE(a
zYqsfLg2Hl9|7lX}kNIEJsbqx*7#5Z22_466z^gdwta5|vNB&m4sG&;5UB*`c#1zb1
zk>;%9>q!q-W#9s>gr_oLJnHH)O%VL2LL?*Mx$?(aJj3HwVQ>+`=|xzaEhXpDRmq1RajWv8i2S_T$f{sO=@#9W8(%Dtnj_zjTLTC)<
z2XcD?lXaNX5@m2{*2TvcBIa1cb|z*hpY65aBSiK#(;rxT?=Q}1Bz=4eM*Mv~efxdE
zVsa=F;9E!1gNHEfPYuT`WigGx`sH&<+93LVpn$}Yz#o{V-Wdk_{3%B)=$mC(MVWa-
zwU&;mUYo2Do)Y#Z2R}b6MaQ>=^quZ}4hMP6(bsHNY88Gis|{{~hqaGF=fyyD9fydn
z%`=E>*j7>2gHjgj!HeUz2@G{+bF2O{;P`><&-m@1D0=>QAa_g=wIdtfwB&CMd+ro>
z;_;K2>;1J5c+`rN%0!U!Y7&tGp`d?Nd815j45lMc4up4!sk78EK+P!s&Io>GN*<=g
z7IimlPe)hg>ER(U*896+=puzj2j-Ju)dyfIx*$S&8C3Fh)*Ye$*8u019X^*w>mc#;
zy&e}9hH_&Q_4>gK?~gCOG{L0>@NGmuseX0}r@;-p>howSf4?PYCe_URV9`QE)rUwN
z+lZ>R#6FJ%)kkyft;!6bswKMOU21A-fF8m8$dMhVxs3K)i51)9^--)t9fb?WtTgPt
zi#Uk^d~LaBZo&MXW7cZeEkvjUM8dXXMK~bdT@hE$AbgX(et);4wZwy7poJ(r5$+av@_U
z+?^A~hHn?sC7;dJs-e#%Xyf)j@r#0E;_rbfZ;p|mJeZ5tgChNvO5_m%u`
zo%dPSqo63HEcGe0(}sd85G8-ld~!|=mkg6UHK9>mp@MfValO<(m-G9G@o*G2>F-iN
z_~v=}P-!*+(*5`BG;V#K2RC*;EO*VCP&@FTW}lslic9+(JKy{dAvn18{Q6MQlL(_j
zf6;Q>ps^TV&nl%)*cX@G{=c!(66@_>r)Ciiea{=n>2PyeNT@@D!w%|d`J~7A#(b(;
zMS3VdjBxKU2P8bC(&jZ^CcxJxm{z3Ez_VV0gK!e(`^EcK>CMV>4$~%il~{oit%lP=
z50g6$s@1GyNw?PVe$v~#U!(KBxFs8pI#eTavfi{b8)6VQ(~z!NGY-1Cmz&t-KYM6x
z(oH;#)v{Y_&+edDZ!Qi)+AKJ}_1LpdjTm!k(~^OphtpGj_iH0bZMr_PJzpFSgRK~0
zN@v@a(9esX{W+yx!s#wlGL_vAID|&~Zo`h)jNCJZw={KmUfeVmQ}+G9$0I
zaqNy&A3aN&$~{!Q5Tcp=P*B{q`|9j41uGuE`*lw2r7;g|ite?M=e(Ste{Y)S(`0)G
zOUs+8e%4;QNr+on`lo6^K0)>$GCy*&vrja$OYi-lb6#h;Gv+Fe?|a4<>NwM~BZ$Sq4%HBLU0%p2?)Z4U)d>DJ`mxaZnyd0y$sBUv0={7DLo
zbKg}8cQ=GAenji0WPM0xPU5I$S5L&Ee{|$wHV+y8
zOuU4W@dKg{rdH;QWh1E!T+@HM&*#z$lGAI
zG27(lbKtPdPZ#p(;Eq>aFhHi;4)=`vu*9j%Z;QA}Zy0;l=;YBw-uDByx%
zpFiG1<(q9dElP&7=?WF>YGV!I?kD3HBqT0J_bwWt98E2$jH&Dnr+RLuSr2~4Dh`BE
zHJ+Cu(^TEAdZssP8R>AMt3Rex!BY@nl|rEg=Hi17m5Q+e_q91bU&n257ZVqfnA7tI
zMILTIFbgfr2jANby3%qKIawXQAT_s*ZIL#W-e+Uvq?$S@rF?|RA}LhIf@u>pR2BLf
z0ge;p^@4<4K3k8T4lGP&i~GxiVG$;7EJ8NQbGw9tYRd=yYd?~2LdvwRpqp{_JWT1r
z47V>$?w6n*5S`Efv8RZ79K+85#$6~LqD34nvED-Ednr|(BnAV`kOV}MhX_l
zF~^0QC^)k16i88AJM;XVJ^_txHb_kB2L0#kP9{0;7l@T#UeB9NN2}d(x?vfEJ#<$F
zqNU-3p7SZ^Wo{;SC*W%vh-nM6Xf}LN*ylAw$5Gp0$^YV;J$}Vr9^ciCjrq_Td_k^L
zkJQ)+PiTpBhI`jm&-MlmSKNs^Y5GhpcW2pJ=|@N7^ujpilm*TJ;IyshD4_0{hHW~2
zH%^ZcUeNROj766vdlmKYQDU6*v1z~W!RN3C4<-0rC5KN!sgL==x{;82ljUmqBD+3x
zm`=O(JSM&oH08u%T-EIDV8>lk|2;)lpki^BionI9&Cekwvi|ueL5H#BMj64LxV;(2>)vCX+t&5n(!c*!tH#^a81?C?k?$>jf@*5
z2iIyr(c0z2CvVGr16(?X<>i%mdOnsiY1b@`X@=r9KBjoBSq5`c#;}7H-I^S+DKYIg
zWzYA?WTQ~s0@@NX7%GT6ud|}i`QxSQ1(2fvwWVc~e6NMgt;
zRnnr6s;7rh4zp1hV?mg~^k|iOuwGx9o@*}cGHVJ{u}C2y-<;6
z9euzc{F^oI>Z-1Uy=EShBh7I5`GjK2PoG*22RZK>0%;iS>*m%|T{nzXQ`4TVnwGp%
zp)a8Ox_TEn(S->fYWKq?^Pzl`d?apjPnrE%PtjXeZv}JLetY>oL+uTbc->5m>^fz7PxFtB69r^$nz@R`D{(acBv=(&x=(A?xqwjeMo2t-_1;FbgWlnalxbhPk$FhbNDWqwIUWeh?zlNmy0=f(FO?wE^vd^li59c
zySC@*p8ZY13l!MI9M0O
zCNM}PK;j6`Mt1CAshneVR)Y%zN&XRX;Tz2gyo>^99@43oBy?>7YYE`DQJGkMWb_?@
z4E9ycJ!@&0+0XIs6_kvg_6KvpRDFnkm-pgxHuTz_k^y5Jh;)N*$%d
zj=Nd=MXa5jS&ewGD>Y;-c;|3zfT`}Wd4d9Yx}Z>4_y}|u1=hMZUX)R9s;+OV@Mt2F{4K&N(L@+p-$cK+%EC2k`HZigKgWy#&|EHrem4N#*CIcTP
z?fZwnH(d;KQr|1R1lK0L?K|9~o7m-(Kc*7G59~W1v;TB(0YMRwa_~1(G6RPy;^N|i
zN6x=tuP2V>*Q09M=;(rvc0X6&SGN_z-WEDDr0eCM%)Tu|Gkc%TKwM`zY}4#-T{nMM
zq%XqYoV7IC?2pyCa!K&sy(##h4BgonwsREup!Q3Bc{UAaw8L|MKai7FE^9S_b;$4NRoc48`s-jk`6XyK9rFlX%s%t1@+To1
zNp#&;$w-Ak*wSqEL(tlqy==%yUinFFK@d&bwwCR6*WzJI#K7AFP5MpmI&=Z!VWsoLXiUn+x+bH(n2ZVQ{=d6{p!lJRZM#*l4@suWFpD#|C>Tc5#jLa!@CKys@zu(niWM(oZ5`WEA3=7n#rF#grx1T9isf0
zQK1&<9Q6
z-rXjweaM;P0qLJUU}g}b-E=Y4?qEx`7S){`&7^+vCoqO3W)F=#KEC{2rejJ^dvuGLIdvrnR}Pd_M+2Le|);-RvqQba^N3X(`@`JDCXAXK-$EHTNtK=zd=d%M@B5})h_AwwfLHldtWpxZp
zg2PzT+nTBSV|!U;622rzzA%^d?wmf@WbZkH2Ysqaa@U`K2!ckq9r74)Z&b8!KnOZka*!
zx&qjAZz&W_J$6H>wVz{H@Y%op-97|lm`a-35~y$3->a*h2T
zjRnHjU1mD1c5fQ>Y%Y4g**zC2j0sotdz_;z7sbi84@x^cw{T^_;?@vtx3bq=w2p35_3!T&vN
z>9qJfm3d?A$Jmn8LPalkld`|yy_393K2XEp-|c5|sOkK&m8iqw)n3J60~u>9Rv)A!_0Wg1-=`pYV7*KU;lmx!1z<2zgkRO;x_
zP;Fu1^33_um?hLjIME`8+1I8g*MK;{4b~WNBnXdTl5+6!fuib!zteo~3>)97c*8#e
z#Wdd#q(8hTDO5Nz+Tqa+Gdd&`vf{QY8^2PwRf@XZ&lKj1g4Ss2`--Xmthm5G=H9J$
zgZS;HF1lf@yLn*ggZ;#O080I}@
zE-iyUiLGD?$}re+$4816)=(;ReTvV@xEu6vuE@PpTnp4+j}*g>E?!9J6I^nstW(8o`ISu^tm>t
zuZ}03t}!P*`h!@f9z6RX*FlBlkt+=%%6r9&i;*WOL$<;rqmR+)BJd1Wu+eqEEXZp(A)GM3}iT
z+g8rF5o1-(Js?wsff~3fjCus0?K&M_bSrO(#(#+Mp)T9-NDr-Y7CU|A!Thk5_M)}t
zb}wnf*HH5A+v_UXLh}ccvPlzv6@O5E?w=21dV9E!cOoTd4*t>U_f`Xxat6XnmG5EBxAbc@BjNgkM#H>zw~)vln2
z@pqBdj~o6#8~(#a_H3#sgt5G+&%*-3Id)cWCg-EqFH}b;yY7ESH|Ff&6kxVUD2MZFL8Mj%`#dsWP)9&lAJvHy{iHf^`
z%fsh_;p~skuhMOY4$VDkcFZO|Q&2vAFNTf$3@04v%Tw%u<><`~)k%T9ouc2c9j&oX
zo!>~1$>r0|-^++0?~p<4`l-|R?ZHQneMz&6rYh26{qVEFcnc1JBY~q9jv9`(RF5Ss
zPnx!P=ZyrIN|s4Mn|plksj9WyWCW<7cro^q#QxTMQF2ryZg=bkhWUv49d2*AMlxwjk1Ez(NiB=933*)2Ci+QWKpHdFX
zcMp(sjTi%Cs1vtQ*dc_j^G4QibnQUm9%TfGkPS^nbzvb$_5)0_I$yY_Wq-VRif1j%
z>$&+IJLxjgRrJpQ+Dk#E_(W+nZ2~>Yl^c|qq?%V>EokWq^!Ij#t3ax~T<_pY0W*GV
znB-dl7#Dw}gq(2!3jHGBv?Ail}6
z2-?jj00(yDqAv08GE8+53|+CF9e7`kW1nsHbGCrBwz(a6k0r8hNM#v>GqpW`NLz^f
zJZ&k7*h7xei@J}K?(uc5j^>PM?jcfhezv8M?;vc4`nDdl#AC85=5i(FqUh_{r(h;4
zVWyIsxkc>=KXt;@=H}6_Y#;GyG*Nt
zsgEg{3zQS%5{H+5hg=6D@RF*Dy-+)59P>poFeey2#^)jE*F4MAG_sHhxr9E5s_La-
zOG@x7eR)kK4GrB~%gvLH4skF?&S|@ze1qM5g(EO}v;1qq=o21(rg~BJ$(gl%yqKrY
z(|8P2xI#+cNwXgK9hk_*9lkxUe8P(?!_mj*td~dLOdQ^&D7Xk}nX~Te(3WuWgO*0C
z{MbfV&O0R54qf(!Q_rjh>S5|Ob1{Vz84@kdeIr`K%@a-pUFtaI_xdJ2%NRNvd^WnQ
z*eAQ3AZnI|j4EMh>i}I-axc!w?rw3kn)`$-8mt(qR4?1JZMybIQhy}Yt7e%RQtrT$
zYbvvYMk5T2hfP1NL=!<5(DhlgKibV2fhrMZFK;+Dw~po;w0P_X-+Qk%ypjlCH+e>W
zS9Dd$6#DY^-Hdfb%S~I3cQb}<>>uD=`BpGW1s!wH^`&ZYhh$fkM2ol8?hZXo5#l)$
z(U5yQoSp&UHL9Dq_4HNirX2ANaEF*kJKO93pcy+a)H19Ja8dcJ$>JDw!bIw0z&tCo}0WP}afi
zv8Pk3>ZU7(D}kDm(`@%akGa3!YUMN-%NkTSTd7~+x}>i0_^|bfu!ewN{CC=Z!yVD0Fa{~h=I?&VEYX%)WCE3GSb?;UzT~4yl)!O#
z>KGa&m5MkoSI5UadW2Prb=ErLEu!>zb3C|BmcMV|;Hyp0`;|;sz0(sm;fd9>KAV7f7!}8m?bCQJ<&yBA~=u@9k#ysu3NoDCWYXQg+*~Y^JOgSHaa(u8-dSr6Q69NjNq-IAo{086W!S9#1fQ~!_QO%sN+3vQqM>k!)k(6tv>6h$zcMDdH
zP*ye5jM>ztg%1|H?yg1uB$oMHcy@Y{L9lwbh*Qs>xAuBgUoY?dDYc|G5*zZ^u1?+%
zm4+$R;*>0DQZFBwK!WtHP0_cio@neap%g|jZORK+MK1_9W$1DhbG)=z%}w2n2o9F{
zqHa2^P5)IMzFH&wf2jJ(u(p(mM!vB
zQ0%<=k-M2(K_3!UPW9STh=aA?_ikO3Vh8g*O80E#A($A39Hg$Zn3q5xu|gz8NbIHa
zAz30+n|9HDOS-{O)jGr=q9*66#qW@CgI7XMc8O(RK`-3$-tS>ygMq-yd=~TSb-KU4
zPY*qLF$*NXnhWzjRAW4+Rp%p3Y13xT0ne2Zj3W7SGfLFlxZ+hr*ORK~WWm
z9yra0vSZ^Dd*^B{jbcI9_KU<1pF_<#zX^*n^ZW&0NRHN=;t^-?;jB@(o{1qGwX0
zn8aD*aVhU$tHx`0>+=iQL$ZckmlZ;zb;zs*K8}~rD;ZNHbmc7Df6tUCP9x&=>y>P~
zpNhpp`hZx5RZsvTDzo8~CtGtY-Pev9D_UkP*k12%zBq
z9u^}Nx+<-;8+9AC+VXS8_K>KaM~AqYdM$wQSQ_ZvQPu26@R)K+-Az#&_L;Zdu{mmM
ze=r&qhKf7m>CWr-i5Z*TB5*GALB&F^qqz0>Y&SZe3kgXWVYR>Kj(nReU6V$xSk07dR4M^TEWizb`N{+=0#TX0y(WO-IR)
zLfOIQXem5hPgnOi`<^7di;xsOl3WStB?1Q>h*QW4uGZD;RlTQu){ws36Szz0$+9?P*{zsNrf
zk$pkXz^%zgi31S;_S1n5AUv;#>P_2o7f+i_R@_k)~lt
z&`oj@0LnKc63gBaZ=wj{c5>CWr>NK~$p8G&BZuLPe@<}Zj2th|LFb|t&sJW;Y7l_O
z9(SL1>jJD`Q+($-TmV-jMtZuEBr-n;a(8JsOMMbh)%CD)-Z`PE%c~xFL$Yi6>z7uh
z!p=TZ%GUPVza}m9a@N*rrRYRNU}^2NmXVIe(i!WeiZ4n9eDk2h9cS%QkyI^6Oy!tX
z=jLY=0&PaV1e^O^w{`jV7AT-PB+o<2Qe_j6SsH%wzNL9wV@Ax}QFC?kQ7u2OtWk-z
z`B#m)pwZZ+C5U&Bpe(*fL?k1~Ld|9CR9@WEi^}UvV8%JPYpe{S&Q(8Yc}NX4D2rc5
zUj7>G!I+j1CWQV@6FonSFMmUG)D{qV`fW(Ng`U)l_#e3+Hx~}PjuZ22D!A=uHA39K
zBhq#sEY0$RL^{o=!d@^$|DF3BdY_Er<7(J7pz-}aHq3WEu$mj*;P=H=LaZdO^sIdn
z4fSKKk#Vvo-y~i5@uT11gX+i)eEhssPI$XwhF~}$xWDGbjR}<2R9rHBpqw)J!n!9M
z@~L~qmGa`f92Pc^j@g@{HP?rO=|V649np$=iPSqdpS%e@5(Wow>4)@>E-@zyOh>-S
zOFkio!s6Cw>2k45BqV=Cd?d14;CejEzYTddcA_l3KQv^e{`69O4J`mDoUIMjg&5w(
zOWlp5*qobBm>E)Dh*4WLwcxoBp@yJ+xg4c+bU(C6&!joePK7h{610LjHPL8w&V)|2T5AsQ6Ac9*zNtDTW9v8turJ4
z3nA-&MEqcZh>F+2U8NS~tzCu!DC7i>){^42}o*(n!Z
zkoL1=M5pO>V;^4hW$uf@-&<~BI`E7zMzoHgFn7$Md|o_!6FRC8A&N<6VRLkAx;^4@
zOm-_Ge&d
zeJ86^)W5
zF&Yz}g_QK!Cx^DSsXNlki_=BK+$y~gU8BM_-?6G*ECt!C
zYdNUT^Xq13#Akd%%fiFNv(vNQ>3Jq<_Q6>rFMitBwTNe!gx^Ht-o#a;qR>3!yMbp}
z@4M(`uRiS;u&upreCAZMY3-OK2U^dC_1vxGhQ0HucXxJT8SnQnCFms7v?-=8f0oZ3
z$yc}(ER7|530E>PStCadf4A3kW{Mv*VQrEHgbc1U+|OSo_tbf>)DRSbZOGjjkOQ^=
zhdqneT|b2(d6??NFLAW`+O8>wi}(@7(tY)@?H}ZSRMDR<(AzG94Dga$q0v_klv?`0
zzw>4djgVg)ou=yZa^{{&RjtdaS@pZMMGuQkiHEi?A024vWn&V|eNmE+s{bTE1c^--
zt7?lZ*0PtJ;bGYgnA=#)%}(54ca)b`-vf1%u_Buy`?0FXY{T6VHLZermKyw<$+K|^
zI6jQ{DE!S5zark`Xa6wH%)b2U+wKFk2vVCEqfwfd)c*a$#@eEOYUz#UA)N|^;r36f
zR;u14i@682{ag1JiQ>hn0m@Od_VxYM&_7fB6HNA;uNz6;FBS?l*!=iu`r(3Y6Lb8|
z0Nd170VRuQGn>UTX))C8{WHY9jAU-+pKbPoueeNHieSW0`TJJcez82L_0IPfkbCZ{
z!bcZ){f@MFWW1()V|x>;Lq^C=_6G%~*uS?)VPh5X%`HL+B-bs6TlNY~&fNW4swU<)
z2hLjg>l#287ZYTIbJ0;Al!fXfP
zK?mv!GDeq{BiAg_sZ!;z!qoMsXo#8ercN3#8q?MLecbc6X!3?B!~JR$J)Mi}ekGC8
zAfnrK^xvI(Z_G%}-QO4FZ7=Bd6;D>1TFPoX)sKk~BdPNT{)m}MXO_~qgF=1Ly`+)F
z2ET-u!0)nl-TgM1wX|D_SEnr5SE+eo#>>i6UricVKb^w4Z@=TK+0vIXP60rFW;Q#w
zlq8CAXZrW+VDdPYY0jThc2i=|{_ftnM^*pn7*LN&&uNXrI11iYUKo?T-|x@iWetl&RLCipm0uY
z3Tf2u4~Jd#)>Z%Fm&u5KgHNOzZ(teqhym5;CQ!u#X7)OH2Bf_i24RWLpItKIDNTCo
z?VKi~&;@?)3l13w4v`_YbTvsPRdI$nrrFZZnmemwOw0GX{cYB!(!j-r99qrYoqnWN
zNYu`rG8N=kvTg_3lvs}Io==f7tleMxotM1tt7K||p^brc^(?T9Dtx@)d$mD+@+hfE
z1YEc-6XXj40@Sa)UZSMRM(XFQswYZ;lW51h-ap+JKLqa-TPEc6@aG4z&%g>BZT!qk
z3wIRKZscI9!b97!y-Bsq%@QRp!qC4{B!hcZR8yPX8GbaP>I1jo2fm}?S-$0_w}tOE
zJ5rY?19E)7G2wG)WT}U#$YOtl+&tJVuEgq%&wSN1ej(P7M@QMor3atxXEU!{)xT4!
z6UL6*#-I-K8E9>7@EfQf9a|`)OZa=nLGgR-xYMHFr}{bIBb^qgb-9>yBBGL9*+6)O
zRE7NJw3k{aTPwY@sBf@{v^kIQHY#^uqb;v+PZ1H`$G-Q3#W%u`WzaWsvNj>8m
ze7dWp#yH@flv>4BP-hNMrw+h)?=Z4GOwlse`W?oHW$LXx3Gh!!<))kqu(2fV#`TDF
zp~z?4p$r^}(6X=IP*v}{IgawVzL2wv_NSQrb5}9G-ClRs#}a(btaTJCV|Rn-&pt#>
z3F>KkSi|Y=)-2KsW{SWvA!Z&vg`oOuhtVObXuC?oYrBr9Rn6LAL-vIdHEo)ipTCcl
z4xf8w=0=y2bhhhiRUcj7Y|C%y#m#@(no0R46!LtYS*6wFqdb+pKB0Vh-aUtkaejBH
ziKQR+^aq=gTSOC9QDEja(hCpX?KJH1r8e8=Vn3%BWwsx928MY!&=vxIHW+PP+Z|=x
zfimNvV$HniU+ng6Gj}+9Oq*}baw^j5gNL`wG<135?)hB%JG7kd<1gP<<(2ZZ9w>Xn
z`?Qi>9we9r_4o?w9J&%_zNaUvJTrP*hF~dKkh&Rb9pA4Ld#*Dr7i=0?2w4s@MX+&a
zM~_jxJo5#)Ia&Z{C)3dktccPg`vnUf)eHJfcLxR_WErU(J$KJPL-@XB%3AEOOw5Gi
zjmMjLwl6YZ^H1v!y3-mE4sG8!rC^A>qf=7~uTfeamNi`AIa_i!
zXz}V-*~e~gKRgJo{`srJ|HFZ0dyjq=&R0ElOF+}{;G>#0K=uu?;Q3r+IODyb{aUg8
zV8^PDmFjrO#9NuFSjV9EA*fGc%JO@gPCkndc7m?6Bc*4pf!Cl}G*}T^HgGaBZt+EP
zrj9qz;kC9Pn`(fo*f1Sf_*f7++@sR=B54
z{ZTDldzTia)x!hms49@(F1VwlvfkpI-H%^}?8V|Bp-jJTJ+UrU_bX|kJ
zz2ZFUeP(EA@|y|Go72%6-r)eyN7hW%gMi+&$a&kjuJCn8V!a&2p3>`ozVtJANN=}S
zZC|)HBlUDP!yM!2Ce`kjkfaq08~mkbkKKyk{tipG*CZFK}Yud3t_nMgM0;9J#4
zMvz+Kzz8ZlE(|6st^8aiT~G1M*pSp^ZGG#gk>|%A>1;_ux$GW(Hq(jTo<7Vf{AW|m
z8vjlrW>~*3x)&&taO#H5CXTnx6fH?kM%$D&T39l|GeEc)!d=HvoUHz!AHtqG9wlss
zV}$ORUPiC}@JH}$Mf03u)mi~epq$2skZ?gu=si?k
zklRUV2N`RDZ#<%_U$xhSxz^fyyUcxUVx6M_TlZqUcyi{pQ3F2Ex!(C1^F7Xe(DAE;
z9{C*>jg%G8K1<5hC|c>zQb{gPh*N(Ny-#H{ml;Qp_UKi8M5jzBhs$n6L9rZ#QDtLY
zX)$%$bZ7q)$Oo=DE0fSMDXT4<(-b<#)P(x)Q-KvNn@Qeh?zoZ2-?!I`g041dAxcG$
zSo$N`Y<{}#tx*)ZdI*YYJqrdvz_96SFZhU6#qBmC%w~O_Z%fV%c(hhG@DUes&G0BU
zYjFgCsrr}QRyDH~v9XG6%*N=`b(oW-FTYQ_k!oIYB`U5sOke-tu_FtSud{S%S9Jxn
zI=Z#@-!l5y=A=va{rcU&TiA-86kLw7SclXC6%_}`LA`CsDK^Gom@rHsJX-4;FONmM
zBLtmEUf$3m%;wp7M9E5ezRw*mhsXuY1#^HYF?E2ClY-7%KQtLXdvuIgdulIF+9GkN
z!y#--(RW?3wfdlcV@U|!9DUh#J}h
zY=4&J*Zj-!rh;)e@v7ykNR_zC{T-ngWYT@read7Tw>o}1xZlxS)?~O()MWdhV?`2~
zS@r@G|Fq3r7BIfQaW-*R6CYAG%A3ij!^2ETgP>AN#G8kF2L*B;SDtdeJYNXyXs%eh
zh+5F26PamCW%oDY;jr?EXz<(#;l;rr^j?k_LtV0U7IBW8IH#Md?paUwwa_l#n*i(2>V2>y)0(Dy^Hi34&u=(G
z@_LfdB|Ccr>`XE{r+5dJ?H@i}nq+1OKb%Q)m@lxlJ{3)NJ586VkY^R0i{57eB>-HVA?kYChSjL?^4@`JAVlc*SGUuVpn)i<
zTME4MDjBl{*8X-quJfXP0Uf=QjVs>fzy%jb#A^|f(V`i_RWLJpv1{*-s>?e`=;dx@
zdXaX~7MDJTQV2;FP
zCq>ZBVcT7m3*&t=*OR?aNFncZek^2MCi8hbZuA>1sQqnrY0hjhn7AX=bQ2c?hh4q2
zyDqA0u_-^-xxWc|^;dOxa(Ni{vF{ETYg$jaNU)XT_aRO0<+SRiWG9y$RXx7=
zD+T4G<2hPtT6TR71<{N6Aud5R>8TzyFzaE{ZQ(o_=JwoS^G{TjwPcM=pYyq2AX2|n
z9O?=$7TQ4vS1A=EeCBB=3GeM*bfLza+%f)Ha(zwcm8(?5`7?WvvL$)j$L5GXj|aPhI8#6
zj1?PYVi@ekW*%}Gt|V0kiW~RRNeW{c){+2}Gn%0QX%J7*oD>2;GZE
zgI1oXX3JYkrsETh9Wn6uw?XeAg$=&>bWsjX&1aer)ZcFIqEO6~HN#=37J5_k7-yK-E|R;l(bd0w)ON)AnL
z{$i3ttXySwZ?$Zjpy_Vet;#jj9r!$t+|&4sKi@dhKnOlCfcRf^jLrm)M{Ne@6hXG5
ze7_MZ@>6AR|KzsX3q;kgH!|rFB7>^PUuWQT8{By);K4~qv+43Es3d8AVJkN
z+oBa)Ja(Z&CpTP#o|+z7E|g9Dx9Hrwa&a`TrDcP8$|Tp@_KVIIa&sOmZn0N#0-{{I
zBO;Q<0f7H)tv#Kjg%*d%ho>OajdCO{(h8hf>zF5D)9k(xx{BE^Ew_JkVBMYFG6K0n
zmXNO?rtPn7&$oJAJHE$(C2kb<;AO@zCHsNBxW(t^eZEjCzIh`)4bzrdfsE=Wh4Br7
z>dthNI(ylLqFu~#`JvkkX#z4gw-^|Q)N;+0{n9VGS2`OWD$1HQLdj(#-9}Q)#EohS
z=qGPa%#%jKEvI@fQib2m4q>^-JJq~*4ji+P&04x(9N*m5&{blm5`XQf)1!S_a9tRJ
z^=KI>_28uQSxv?Tk1|~$IcaZ(TOm3fUu5x+LR_OvM{0jFckYlev)`R%Gsy4Z^5S%1
zab2{i|5toM=6bY0np3gXc+o)>&_ZT5J97TPwr
z-hJu{hTH~$VAZ}%^w*#ZVDanvkCy3YGT`s(dP+N5U2C(DPoIE4^R63uQ!UJ*-7>u9
zO~EVUFo1{^pFc+<$2gC^E>V@w7wK9*+%7e6Z2!8pD~FoWX)_xp^+$fG?~zI`*2q&l
zFINcnFUVH+)y1x7qg>@?2hNywH3i3W<9mX(`ZwB|a+;RYD}Y_?SqAKsT@3)b=;L8Q
z)ymJt-F61=gqC#qTkN+Wkgay_+Yf8{PXTI*ApPWNnqlG@)!@CYq&@wMNNXAk+uaK0
zP4#UNKcldQ5j5nnb4$v@@D
zn{O7*Dwx#G^O#pLh82wJZQyxe-=&zfAF4@T77Ka87hdNvC>cy{RR5v|0Yl+&L*SlB
z!1&_X;wm;MdO~B^)eH~y-sc#z=-$*a*PVv*UeR)%;o;N@sn`2R>pVbe<=CjJ2WzMLl;ybr?xHUT{pCl!}%mPpR
zwdFO|G7`VNR7y5sC4`7ld!4<|QJJ;_T#lE^bZ%7NP|)v@=HzT02wmOZfb#bFNO<<5
zN#Fz68rEP?Q?vN>@5RZM+q9%Co7ap&2G&}FJqt}giMdtG?-p72GD&f5?q{6Y_A6*H
zfOe`SC>p+J3&qQ(^a(
zKHnMpN9oTkx6NnA(?-2p);EW@T;B{(S81JAZYo)Y=xWiBSBHG*s4A$9DzY&sH*kob-*&zA(5DE3QBm7il^1NI&+pC`yo-uUE
zX4t-Fl-Y9;ITfd%Tu#DXZDl4~3@b3P7&x_|UaJ=$?(A?^6!oFk@%r8C`tw6?55FMQ
zmoIu6b3RSozD%uH+tWJd;V3JFAxlJdX0{85vA(Au-~_pXcSODcxObVIwpz69cH3!r
z$%6b8=Q6DESpRWoB6_)He}Gfzt-x^e4Y=+3zS+`b^*QN!U!jy>5~=F6{Z`DTrz)&b
z2cQ39F%#>wSPLW>VhIGs&E7FpnN3>HSw^|f4=F;FXEuYz$}^4Uqo60-;X1&#Rl9)h
z?cykv$Lm*`&-D!ds{Dc`;kE<|S--84G6vmhbb4)^3+XQ_TDBLhTpe&I3|E#`-lT@P
z1NY6>xp$%5?vPL0k*5Pi5r}*UQgHR2HtiHAFq-FfK0QvKh`fPe#7+43W0^y&-AL-$
z+)G4r@REpwI7`*p5-cyi^~;O+M(dPS!|an5dvZFuQ3VjlA+FZ*o>#|A)$k*kkGyz^
z{A~=r{|{5N+pW7P^|w+E6e44PyizdDqkl8SobEW$*2dlfe42ZT-siG?dFG0!*Oo
zVZ*ijLgX8pGIBAQr|q?EXb?JUldJXDI`n!jJAv`kQ|IiLiognYS!V0e08~&qqI@tb
zuIHpE_D*x-@hcAu4Zo4b;afs2YPqzNV3}fmx5PP?x`DD|jl5Nd!q7J-f1jXL!G_nj
z{d(NzkfM79{f33>U+8&cnrhj@iF|rV8QqpQ{08U!+ef3H`UQ)IXn~2i!vy+@BhdPf
z_~1>Tkcy}9xDDo2&TG$NT=U69`Tj;O^RG?m#-plVMlUk<#N{Vv<&F-ISQmU11Uv-k
zi-c|tV_iBkZaxhObwL*KTeYc+55pfE26~>U(i4iyu|vSOEOd4cd67dv8Q)@mUasV{
z-%47`knSDdc15_~IVky8(|Qs5U=*k_R*vDImdsakX(m7-&BAnfL3A
zPmno#1D+o1WxXJrV_?gZ+CYudLW%|S(J&^u5vJqeYva-wL_&>~Yw4#vfgbUGZ6-Z2
zsfqgBp&{T&%wGGyT~~GWbTK}8QT?;ab`+Zfyfr(b*_&d
z-P?n>d-T%2b8a^S03W~m%GrL1XPTP4T7KAd+8WO7VfebMT~mfJ<3`K8QAJ}trb&Iz
zmctG0+i3#^-c)@@$Mtb+D1+|y+M2(pmSp3?H3FAp19+HAnzkfBtmR`G(lE~|aM*X^
zq0OL8fpzJJoFAIO`t1Z+!`?Ji(`^KH=^;TJpf+Q*>Ot*3cUGrpO3XhwJ
z5UP-1RDuhp%2zJxy!}U4kysF25Q)X2z_d|&OTm;Dx#L*UX5B^3(Si2ZZQvpYIW^yt
zYi&b;&1qoaRto3)<<
z9i`rk;7Q2~%WXZ#=-86@QDd1NHZKbeWLp*+bodUJbJ}#xjA}G^c61&HTJ;!rnhZZm
zHY`;)#ug}Yo5H3vv^|yL(%SCTYU4(J)+y}hcCPbTp&IxAE>*)C5*&{In+xi0^eV4X*6I60j?J$TA>(BxDD8#-b#ql#@t6D(k_Xp|z9V=w$mV
zis4B+O6jiLSFhC_AuO(Vd4mFAd5f+=D3A!T5Gx5JND?<>c2=}sIoZCCm&DpWJQvx^
zEOatfq;3b)cYsR|NJst(q{Px+V(Ye=8`W?orNrP6$2$nR{qbwe*l4rv+Mp?8tn1Kg
zH5b(rKkf7zKYd0YSA3`*m0PN2msRw;GBGgCu5qF`3O_brA_^5>?!C&Ym5G)B>m{q2
zZ~JT`sHqeWZZJ~A`o8tWu-38`cA4tTq;I*&Dk^FnO(6Ib{ft?*Cz+Uk{r
z{ps6rgV-h-MQYOFeteIIHMG&+4Dx?#H892O$O^+p_Mt+p;^$qYs(fbX;;<;Wc8|zb
zEvuVK@52!q2*)r+$)bd7Bd
z^^=B9qr?3edVwX!YCPIbwlWiyr+A=`>@$m-wT&0ctF5El#as1M0mk@;}=9Zq9xyH~Aud6M*hoddO^7P@!hF$XTe)aID`v9pDHNVts
z!itaQUFawWn{J?1hHp8_hsVY;%DdVnC<7X
zH|z^rk#9pYnOw~3R|{GIm4cg;zhIMdp#7hT+(%x!4U@^XGCr$G@yrbTx6|p^0-S#5
zVZFuq9-@YN#9kN3)B|6-`8-#96D9Hu5m$vaXV_|cBW!t0yM}TZ`h9V~V>3PJRo^^S
z4qQRTeLa9Od_JV@f!pW$Pm#|3ofls#B(<3D5@_IM;^zXV1$xm@Tt1YZm!4WzB~}TY
zNW7ajgd84Bf%?hXmS%Yf5MA(&${J0T@J9%${q!DuapnQ6UT0y%+?{|cD)P!~?12Hm
zeaHZjGZy504%96~>j_#uUU)c)RH&)i)ESec;IKpfk?WTBV+(}z!jnueAtIf9`ygYq
z6FS{~u2Q}DxaW67pW$E@xOveBg71dddWbFQP8r$bJGL(#kUg7{S$?cPc#6>)YvC4A
z`fxMmwrFdQSKzt3OWe#HRpFVuPnxAuN!2gRV(IsO0X>X@zXv6kdOYU_j6YoWRYntg
zwY`tfi>~R7CT_@XeVE4Zo`Pqg>6I9#{;#qb#)b;fcnwdkb@hw4FECDljdyj5
z%g)}*xamXXo`A*{(&nUBgV!ul#KT5+0ts
z%}dZeM9)OsO5$E#ZT#Akb2dLt<4oom9`91Y`)!y$=Aev62EAc*4>Biht3@%-RBz0U?>BZE>E
zNK3|MBMr~5S06+|X4EM3u6D;?Xe3HA*Vf~jUc=#3lCNha+gj%C2C#<@&@cP4gIpJ(
zO=z#SG*tIFM%Ogusi)SQ+v1lp2xQ^?K)OivV(uM49q#sxq<0&Kvu0LS%LJLM)^I_s
z3OaTn9{B*4v3e$a#EdXXfR)|-Q*QQO)-^SbjECvq1Wx$F@8vJM9EKm3%ekHC`4cpb
zSn_yn3_{$dxRUP9zW{A|%uw@-dPfS@6f*VfEGtI0d-sGbG+ziw2A>1`Y>g{L1%5Z_
zU%q7?$0uj3*$-(x>fy#Bt?iT=-3-qc=^sEAf2H`=ZeCu8xd%D&pVN)Zw$j!wDz7f?
zVP+RIMXU-0ic=OYKRAAaYM5c0XMFCoKRWQQ0%w=IKu^?`(-M<94gx52@HA4T@9+Ba
zy(+(|c~avV@O!y-aVHbc<5e&nk1wOA&&gHAYS1v^mbf5W5+=H$@)0X=s;g?BbGXpk
zd=9*M@QLqBns1v6n14utb&c*G-*er!+r&RM9|lKbd#6+7c6%9Wewlcqx$ZL~>+|X-
z0+At$a&x-j_g$LyQm6LZjLgGmzL>Bc2B1!_E^D~GOIjxPbljIg)Qf7K9T_~pyMI~g
zm~8?wYc$S>My~UrN6QB?I+-w{sXNFeU*pD>n>jioxbtook}gdsDfV!n5gomHd?_#I@n2cAod{1$6uIO%fx&03GqL^gAu;pZG!1wk<#3gccRu=&f
zK3FhVzQ380um%MTzXG8ksMeaV8YxO%5^yXpCq2P#&x9HyGlC3%SE9cOY)U^KBao)7
z!F3sWkQXN={t&Uke&aly1@cdN_W@&_9d&$|XyY{r^?OIHhd@>O$-||b`B`^5@HeLL
zH!i8&&gx`u-1Z)5&KdLbAd<=CwmT^+r(XC%AGnym?Kfs0->+$RR$mIO!
zl&{@2mAO$CD@dB^Hc
z_4QXyr{B^6%F}9`doA?Q%ad=&mYqwB4~!yB^ldMYa)~W0OZRxJxLtq5#89cJ;TP^X
zruk;){|Pu0{c4jxbfc$r}<|UHAQ0qHq!97EPZfZ@u#*^jH~mxOGr?@>gV1dj
zFEaS}8NBQok_ot+v?uQn9yJ$%4cO$_%nWv;jv1ob97LlwKjtbmhDz+GK<~X-f=tFU8f-m9Kdh*0~R-!rTn5
z-a!SCE=WA>5MNDTqr{Ai(Nh37<#Q%l?@GWouZ&o4Ze>o+?#aUV2fMeuLBN7BH_L;B
z5&T7I#%AyeSA>fc{x#ROzZ&BOGRe57=M;t-5|ZedEP_cqr`^LrhAAb^?D$B
z)XT)2e9)fZ1p(bt-AS;~-|%NIn8LB&;2#(hAiLS`E@RI8|6z{c%Z@Go7l3c^qKKUN
zj}rVBc}rbe{RPHAV}g;cx6m=rYg$l)PQ_m
zj3OyxWkpnkKl^Rw8^nS~u9t;JbiSW!C<6VR9PxU%OSGYVu!XVrz^gE1bK@vn>f&A|
zC~+Oa<$&zgi`p0vuk&%~^S{!07TiU+)~C{c9o5JC`3We><=te<<84Tg(N$L+2>>s#
z`~>B!79%6yp-}}~9$?}kJHMBq6560ZsYrPL=<2)vrs8Mfw7?)DcSiZ=i}8dOeU%pu
zZ0yzCepP3bHqwue;(3Z&Q^ShbEGVWy$h~mSsD{GJl^TD6MT9T#?ZuCu>mq-3WQwSQ
zZZQsiRnQBJP;aSr{mh%opm}sxxXOk71uTPl6bQT!)wful9XSagy2HL%tA_RI2mJ^m`+kz?OY
zT2YOJ$$jq_p(s$qU(8Tywbs%kF|J73nrti@N2G|~=X@T411Sl!Z6=6p>P1Hql@F4A
zIO|f4l#ed^@vOT5^1y2+9rGik>~C2D*$9&YV;ODokWM|Ijhfig4&@^&0KTsh>`=5@tIeF}K+TskC)xHrcoYjX?I
z77i}Pv}6u1k-VA^MD3#5&9V>oeQgnYo~o}d!cQOcujE7D=c}>*`TAK!dU@e8MP#bY
z`*ts3rmy4;(h~%vN5U}RK-L39DHR}Ask(dNzV*h_DTl%u07@Q^I)P=-U51~$%aKNs
z_+7WQj)2#I3Thu+;btBFUsO?^P}>My`s1w$!m|LsBSATh(r&$`(9j^Fdx@lr&~1pp
zV#JXB&=k6@qMh8ycTYWu;5h#8p`d?MtOq{+bUUF?^5kajzaT%vGZ8dO52xqr$wU~p
zn37NnXG_FX|7dN*RGUc!YCrRSYin&nCt)R+{}cuw9-%<{C+qYojULe^^JsW;8?lN{
zp(w>muCFHu1rJ4Hxll|1(S9|AllsKY_G#$DQ9XV7obX96rYMK;d@G(%3xY;jv
z!C2q*HGisC9>KZNPuK&*n6XBHB*zaj5v!7Hdmkuoj4|g;Su-Sn?>AW05ut{3PctNB
z+ZEXA{p#mGXZqYYDjrT92@Ou(^)fc%Bw
zQ|HsQNO;}Drj`Ev#xn(Pyz5o98$L_^+
z;w7SzfGRUn+nxq^-HQY&Y+IG~b)ovih7rKNp<$N9{m$_C2qWw1-az^>9yGA!q}evS
zY35YUEFYONY}y2aGstfpGnX=dQR~TBWoeG}btrjV{B#i_hD7G9F@7X^T7e=bBEwbn
z^KWMkGBb_8aH2u{FX{hw_*<8F0bwHI>(yNA6w+n=?&Z9mtkmk*+rW5uv^Q@2Sc`S)ZX{f<#&ird*zBf``UxwY49S*62=$qFu=&Vv=3{T0*T9Kpl<|7M4_%RkH2$p*)#kLtkWx+KyyG;L-}~5vM|6O4fo>^zO@YDT%r|-t
z55YoobaSstm;6$pJP0$bh{)c*N}5HqU=g<~5risq!m!nEJcA
zJx_S3hMtnc(civyYBY$|#w18j{kOs!e|oce$p|qHS}NXdA*ybnquUfhQZIkDIH*M8
zXxi7x$=t1Afq%>@M2q4-{~Ino;CWnZ$|?Awq?!Gx;#n7(Hdhv!M}48tvHMYl9-nE)
zRUV$oQP}&P^Rl68GQNB+MRJZ&sOtTrEATpicvoI!iaWU+&U?~dA8Lj4ujnAnh2v6#
ziCVwfjoI?l$QMszW%@U_o|b}xb{_@z3F0}Y4{t4}+nMjrdf&c+8k%XNPO&EHLuSQ5
zVg5r8QSPhSU71hzI}#FglAH%mBVD_)cH4Tln+&zkWO>~ScTq4j{&!`E_@1)_onG-r
zyOia%-W5&eEH52`96MlXXM*~fS#{pQ9sl!^v_KfXKQwI+)_P6z4~Vhlt&0
zMu738#K%}O&`)r7?02Hs1$(ag(ZR*7FP(kWB244-Xv41W%d5%)RiBlWKQt{Ya0zR!
zObm9N>OXiX_`%6WnuedzUG2^R9=34
z520A*^%)g`OElGS0r>UyX|^g{k-v)RP-{=A6xkM
z6wOn}GTl6)4#sBG=WS7c(cTBD#4Rdg-3M#_YBqBmRs$L;NfKS_N&os&B*%kO-*iu#7q)G(xjEJKb3Z3I~gykDk!#=C+
z%uGKfO$cVyZQ?o)mxn*@O?_089tw7STlK2v3ff0}#39~R=r_}M*jitop@-_lzqXS+`v$2^{^~^=6MM7*A!9H}p*UNU8IEtkgjFj88S6R%A
z)l|8V4CX4IjBy)XS2fByWN=`UBX3eisVZe5B3DqrJ*O7ycd+=YXt%tqwycnn5|${v
z7b!Luj3y-|?lZ5b=U1w_2vc@PDRtMNr)K~(txw*92Dr(X>FTVCNSi(wDksa>h|Ta_
zkF^m#Ep(1bWo6xK6=+Ah4=mL`0JTd6x2JO7VT|t!lMTS*43t1P`RHA>``GC2>2yE4
zXx1r}2z8PD?QDsLm$RksemI;Vvjy&(s0}$kazDIGBrEc3&$mV8^J|-7`*@Xf{i}hd
z&U~G?*j4j^H}h_+8-HzKS3pzW*v&XXLrKJnMK*Nli+QCOGheIkF`q{fH^zNh+R02D
zw-E#Q=$0zjy)Y5Fx!HEEFF&6r;5>lsP;{b3JA2a=ODdqBI6+kyX8+D?qVtybkp$*q
zxf^AGPUd^u|7t6A#(+OO+qL0_$eLj6Mn3|9&_|3N|Edr6y9psEABD2+#EKFwSvE+U
zgw;>G;KlL{`!BDAY)EH-P#YZVa?C^BBOp-Jy0zzj$
zE+w7jC7vEZE2eg{Gv^Ft^7yV`QJq=IyBSP&jfpGY{W}uJ1dm)OJ$GZdT|45-^KtZV
zFAt>Vi7sS*4?`eK1V)#2F*c%$TbcA(H=DVjo9#L3QuwWT`%v5uklTRIKWHMv3-=5k
zREPJ!l2a~8<0gG9F{4H1Xkk@<*Z$a@gbkJ<`WiCkHpNIx9xc-#o^WJjm99HAK1mKR
zRZFPPZebiW_S95HEmbY}wB6WpSt{MR+*~QLEM#c6EWHzx*s)PA@R-C~q^w3#DE)Uc
z%#$$W(o>C###-a!r1=tm?~{j)7+`c;$gdXhJw@Y5*yq
zM6i2x`8I02Ab|k`v9v673Iv!9v(aYXhU1a^!9HDSVn-P-8|+2un%Y4Ap9_`snWXbo
zL@|`y6FyzL-1}1g0`H|OAqS>x4&r0re&lfte0^3|=R*f`n
zE>0Wev^b?mhRnuVQ__8~Mx6qL1OJJDAt4FG{bQi<|Br#zk-p>1l3q0wZMXEe(lW`>
zs?@>W9pE*;qgHIkNj;9pH5s>#d%p4Gqc5o5F7-6%?ieW#x0v^tWis1ud-5e64c@9H>W!ri(cHPpZ+yd3A$%6M(=C0;pX5u0etR(U-6>t`5~&`RVb{g
zGk4gvZ2EGRq-5SvCYEZb3Jgh>#^H9e4~%+au>z!~=0InAnNe;@WeSuBHkrg#pO85f
zaB&5edjLhv-orvs-p7g)E0nAspNqAN;v&FPPf~UvlHhlTC^SS4@a>7c;sm&C=Ylv1
zCOzlV1K&BKe(YZuV%taQx(?kT+fq@NY(GvG;BuB8K%IGT6R+XZ(Z>x=FEOUep6}r+
z*waDftJ^tjsIIC#^_)1*cp(r7WyR|xnJw5eHXQR}|IfkucnmFoKNmNwuL%Q|8ekt~
zh^jAL91c5Wqq!~Vvwyz{bl)S-EJ}Q%Jg+R#w`u5r9eicYRnN&Qy>Iqd(mGd+vj`t*
zyJ6Ut0#fFSkB}nhoTnnuc}cgYig8>c%)K~99r(lP4?IS-_x%0iNArqhFY#RYj2E`V
z|8o`j$1EZc8O)q!wy4d67G2*_Eo)u(e=xz7Gc~6D#;E>p_UP@#=*<`fAIimWBi-@pBOh#Yx;R?mIIB2&rTReu59?=psN8PPnX?f-S>VhTb5RBBUtDL*Y?*Uc^b6!0HS~a#{*;D^NrWMQ1HC`
zaeB3V?oB@fNZ@ppvNeGbFHtTMK%}#Cbhyq`_jyy?>#ee()aO|bR-3g8v2!7Y@r?Q_
z`S_s?>n!{bsM<5Dssq~WESiS9?RxW7@HG9t3kI`bez~d-ondxBTU-XfBVy*Rx8|fj
zNJ!Qy)@j}}cdp)MmOM<}sY5fxCB!}Krq+GooE&-4wN(AnMZ~Xn9xehxFY~8K{>h0U
zzN!xqw?%;E*a|N~aI>9JoRPOovnJQYcE?bu*yyhLgm{#Q(?GJBC-%v~8oYGqIhCZCexDwrxyo+n#7*
z+qRudY?~|Cx#zi`ee7?4?~k{C_3CwWRd;pORp)hHT39DFj09^6Q`1OQDsgG+DCy=+
zw$#J?HvCVTo6?f_CFl#YX=hcPd)cLt{Dwf+yI*&-_3{=HYMkR^pATl)l~+_L
zgiQY?ui>Vbb~`w42L4w$y8n)WkqS`CF=OUD86=-V#dQv=fw1Oqf4+1>u$zuw;`WlM
zk)@J&cNo_BoZ{J_OD^2cW?ko#O*^2}b-&?&8W7dm6X
z4y`uU>-IXgcWURfI6GC=IimS=x$<6zh2gr_A)_C8zYV6F=C84qJ-*3C{^VFL0
zBunO3RHhU!pSDTuGE1Cair&*kq}1_@YYP#y-`%;-2Ud)g5#;bcD-9d-<;)ZR>0C1{
zsA|#9wxz|9>v<$D)GbcX^M+ov*5f0Zq>>hZi)Qw
zekVpp7ujvORR2|R@Tj{yDMfv+2Z_HS(u)zUlojOPJ@Hj8i^F4SMU<=4^-F9zdtUfG
z&T?Z4+X?G_53%C^FPR10@Wt=&1@uT-+{D0c!2^O%1Q*7ey28R@To8nD{#7TB2}oT$
zz7z`!?%v|FA8&mTpqEh?w_&{CODkOINs
z{g)1yp0WGZGxINh@x{{oHO7ARrQh-o8TxhjDTK>K
z-?alLG{Ie1pbOED4bH$p!-UGc(dIL>+#HKE52es-{IlN4CS~eF>Jwo5ZXNPlhp-=i
zkKHupI(|y2_c++>Uy70bL7qlgTSy%=a=2CUwW{wgw53fJ0*7A(78%qEJ`Z;rQ`0O8r9F%hk$F%))H}uadWYu`*_Tbx-vo
zYGcu*0#N5=iuu5n0jHifU9Vm-Eqe8gdzN@ytCzqP)3d`i|1Cyi=AZU{H@4y=XGEb|RBMmddJW;Rc?dj4oOQYIqB
znw2d+hnYE2*`F(ZfOZy@$Vx`JT@Kg%WQk2qEMEVe8ftqszu!NzxvSx=+wA2+ea`MK
zeXse$er{mfpOnUB7?$m)x(e3z%60ZC0AtXC%MGpjy?n8`&~<;Wi~nNM0vftkHg>_)
z=Bi};7QU|*of?nJv-jL)dRP>`!k~=5=_Rq!rIYC;juezD)7$5=1Xj##(|1+1&Ejdy
zlIT!UD~kzNtJU``3b@g8T2A>bv%GkE8G@@+N6&lPEVpy3QG{kP!CA!Vp-Se!XmfP(
zq*33T_brn%H2N}Tgx}|-|L4aU=iv9qe_mk=QlY!EZ6f1?-*=DY28IZvz;xf%S7E+8
z0u>q$ry#ZK{;FPwRF^1-*kCqbBi)hcwdrmbwmqx(D?gkfH?P#<-E4M(&ScX%?)Y!?
z$X2htJh=@Ov0$!pQWqwdKaXxs+2mh0vt1TV!fuA20yZQn!GF-7XS#C>HI1J3hP}Nc
z`uTHo{-eZRdz?O4L13ld*Ncww&W>AC(9%#hL0Alv7?WDI36)rCpSSeOkOsQ4{w!=r
zuw+b9^VZPT@|Z?PEy-jgbQxI1(U?oTT*wTOU$kIBwIwI9~nyFxg`J6@Z$f7L#a6&>hU
zo$V#pYu+7iis2I}ZGb44IlioFE&O?v%9TLz5@So>(`ONqV(`sLvQjrPfQih_uf6Q{W
zZr8CrRw}LB86ocsYyyx%GM61y(yUZsE0mW6xs~A8e;85TaP6L6>Kr+3X2i{`>Py-h
zI=`rtpr|yLaDUCkjI*f8E*vC$j;~6JUlZi#MDFtakeG<%Fm30xm6-|E^ir0&i}y_l
z_m5Vagi95giO$RfPP)WoQW{@_M$#yAibN^s5(
z-;s35LqxAn!!38v6s_*oN4C(Ro*X-VSx-4gLMK?$>p1eX`h?h^KQTq%c3larVpYYs
zS}9IrxvA_ypMA1bMQ1duTeJ8+hcg}wnw$(8d$s!S6cO*#ty#e}Th%O+(mH=ySMgZt
zD;=-Pm6V4Z+bElBxJx}E@8l#JywONZspX(Gawzy1ASZ7|V`Kf2fPtgVMrzegT8+?7
zW(d4Eu5JkX?_t|&S%1v{Jm}WwckiqWOpn8#wfB0aC53Pq?f~kS)5cu%eRj=PYq-cM
zjknwU=+(PUYEJD6)oE1=f)6DB4
zHjdlm#RVJV%lh~{DjlSI=i5f(!Bufdm50IGgtE+ZqQ@T7(%bKoOq&BV-6Ur-zKa}h
z*rj^#%#RhC0By0&wS*8Dxo)K^F)*QjbYy{37y47l?zwR>(lQZ=$k0xwpW+1E>=U&Y
zwu_rC!>*4S_dieR=aE^zvfUP1dr7Q5#uuQ|d_x2FC?;!bqr^){Fl>)07b$3b*
zM&|SR?gB9QZlX`eL2_s0FEI}9w&;!8+Rmd8bz1m$oOCu)G3FY{VJSYLokF$7QUAc!-ECJUNa2^~a1X|+1%P&@DzY%a#wB;t@8Dguy
zwidijMuppsx*B=-v%_IYyOK$=d_hv>bx>?rtaV{z?)<2*FCt8BQ{y#JIXIE(E2L}d
z0PmxEIi@1}UQ)T5j*?h)B6TXbNn@Rd`lK!t)NNJg>WkU%t-g+FoPPWLsuGpXarnc|
z_sUJ{utFJbHG|#gX&$!fpBlwq0$RNMc`{WmI+w8&@Z|MS*!|w?P+!_hf@(091E2~9
zLzo>?OO?9aEphIQ#p%90x}o>gQDb*h*0gjVAZ+By^(dDd*Z!DJ_f^sJRk7<}CuZSf
z>@d@4X*(B43?=oxEC~3_e4KmJxA3`*$;_}@U}4oXA;NA{EnjjD{e?6w0^_i8VH)c2
zuV9ix^j=-tQ0UgZZIE36sp9I6u2HTyc|GJ`_ijY|-eA-Mq=!b}(UqAk}tmQ2m>;fv?fF`zi5lCc3^@0w>v0
zr%ZFXTUM)I$4&Fj{T$QuKlA4~Yh$`PAV94dce>iH{_!eoP~OL3x=lw)w($}5vi0Y(
zDdQ81quW*Q)%(@KEieLrHAU}F?Wr0!cD8cq%V00Op0W-zLWfb}q`arWaG7_e7B3yA
zIKqWcm})7wmRGlgz0P=5wfg7lkE`s(WsPiat;42}xp6@A&yKFq&ZQ1aq~-f>1wh
    mo~J&Bk%BqCzOCW5PMoVuK=kPXX0u2zO6
zT9a8kdJxqy5dtw{fiW<5&?WU;2%0UEPo`0aa@xqY)5cqVPpIr9roT*15bmU)&BUXF37H$hXF|TeruT_wO3MP7BrLsfFsSviJN;JNo^H
zC+R5&Y5&U*@ePp%_4Eh+6)iwOe^(I@(BC_OfDoYnmGRM1p1}WU?b%dUk^Z`7`~Dxr
zp=TIG1Sk>6XCLYxb6?LMB?m*01Q5R?0<`R3Q^QZ`@|aS~-}?C5(*&Gk{(pOZ9_85B
zFAJfb0DU9VNgX6<={xzy`OPm@MmA%Vq$ih*bW$H#+u*O%zpgDDF`IK)C0!Ji27e>u
zrQ@c	UoQ~j$OY>j8WipBAydGrLLlORlGzCzqFgaNf@n}z+OF?Fda-wehSoW0oK
zATUfp-wgecSsl%imHUf|L&%?eJ|g))l-e7kOZ}$7N71MqKN-1SxTV4WjsPYbJ5IIt
zeaRm|7ez000|Qri_0_1
z)=~9#qHu(zIOsov?LO!5^jFdP|Bdfo>c7~QPjLa;A
z421uj@$kUVOPko5IhzwQF>^?`9|*z^4g7BNpOBRN}_
z1Q|>aaZ4W?uYl!shwR`J0LZ-JGc{2rO`dXg{=|f)OcY9uR~1c=K-(brwYB@yzIFBa
zvGGHJ1}Rb5oG?L*7`at4Xo4;id&PNk!$kBCe4y9wi%Z!)WhK@Og3m-7VIa1A|&Ayb{vzIWrDhGjXpVefIpE$pF
zb$+pwjuG1vRKYa^X}BK_qc{SyJmO`^4;8#ego@+2#Wq~)$NX5v$0}#DpZ+Xlf{f8w
zMqbc+N*S7NuZRGZ>4!SB&+BzQ#M@b^#gUPL%qY$b3-vP^vJX9dg}F{m*W3^s=5JY=
z)r~8CmlR;1`P6^vWDVqBs$;GZjYbk3i9oZy-24Q;^GQu0?T%>$El$MOfY8E
z3078(jw~&WJdL{~@26{jl^mqs9Q4(Rv&?IMa|}gY!ubQBB0JWY_cf-DrRK&oM~stoY7uEZ^x}uOJ)+
z(l!dx1fR4zK0yMGvHQHvJ-Pjg?8I8n$A%q~NlmCmAk3&d*p&Rkf%uOZn7kbzrOG#t
zb`a2)gvAR~-d2XD1ZnhW?;AkBN4w#wKcr0P4|wF?Ft;UW0CnhJE5E|S1D4Mf?rd&Z
zD4iyeKu6s*mwt+MPf)J)h#Te|ziuueCO`(*`^Ce4-ukOdTI5T^#EW{Ul9>!5dmXW-
zVuqzU+^6I@b5u)L@gc6vR1Xqzv4-6NVG)YNI({1qjeG^h4g!Y-7Fp+dO?b!-pCdX4
zUWPvGD)P=xl&tH3AV6X$u;AEx4@v-u^x}Ilhx+ad4te?*qMgrip>NT$jfTOYwoeSc
zbES|>uIm7Q#c+RTkRGBNw!L*Iw68KWkE`4-X^%@m*bc9sk65qpVX^}a+@Z%Y40N6;
zr4FXhjbBnvC39P*PO#K*p~jRf0Tb`pXeRr8;&_7L^kV@&h4&8w&j^@;pPSpYDmI*=
z-YAFEvmftRn+g7-*Qtdb6jc}V$E+Y*5!G%NC9mX7bR}
zfD-2xVm(JY;**NxkWcv&
z<&&=rj??^D#`OFb`aSzN$cuo!{;rS(@VdG!}42K(eZjDxYj3Du(fvDVI(8
zLy17r!1S0slnPS`LhR%y;XbkY7v)3%PVWq(hNBm*!1Mkj^Tfh%RcKqV+;tO>Z$)p@
zTq!D5yglc1vySpmLT978XSuOmvJ;`H@>pl|4cU#JK9IIM+zB4yh}RKZRw$E}AQ3z*
z0;_h7&xMaxLdZ2WtJEuB1KbL-{M7v-67VpJC_XDYLMQMycNe`oZ*M@6$(sdNDE`b`
zle{z`r8w8Pm`LzUq43R}kp9T*#@Pu+xky6EVDety3#SU^^OH~?FbTNucc`;8E7+)f
z@eR;Yov2tsZox9-H-srsN=Cj=$8vt6)d2
z^L61wVM&sT17+m9%(oDC2&7~|c;<3Z{XwZF?4Qp;QsmTw4&CEju7;%!U=oHhOx+G;
zj(4_9Gkdxf!^c|3o`H1Da>>rl#*nt4Q9Hl#_v3e@7h-=C>Y>>j7e$oD5Y)dI2XGUv
zdKWjAj7F!RLwY6>wmkU&y&|e8z|3xzbs!v2eUyvh)zZz`(It+HBzrDiMiNpUUDy(;
zQ#cNQUl@{gRRU>`kNiX%AgUjhz|Q@1BjOSicBHz!jfHkoO}jG5(P;cy1jgBADyT@J
zSqCvuw+>vYG!A*e2iBbeia*u2%F&f2b5{cGrigGLc0>!?1L-)-@Swx|&2p=C4{9*h
zx1Zt-9h?*)orZ4G^?s67B>6bN1OYkMR1U6F#^Z)>0qg}kTZ|XJ(_i_+Bq%pQvW=a;
zR{(99&Mq>pBpgx9ZCItefSom2;TPD<>V3hwoPF-P+B;$J7&%I>vyjB9`uw*@H!{l%
zIys#KYh3VdVW2RQQ;!q$cTFUO`F;inSNPNgLh867>Rl>vZGe2yFZVv~S=BDIJl+49&PFAx83N5
z(=>FHH~;XR+Xl?bvIgE#d7f76`*0W(X1QSWEaz(
zu<{UI^tN-y9+gNgoI}1xFY5UR5wvJ5e}Rw!qY`p-lEfkfHmSZ*apBycQ0cyBw**Oy
zG1gKV6J3owTQP+)2s(4qwa_kOiiqOVo{-|x#2;uSeN2pA^BfX|IKsl7aIyeU!A^3v
zC?HDv;%n#@IC4h#RtAfF`4z6<&IHh`B6VSjYZCO9$ZgC_vOy>_rGU^V5<(Wn{3Ikh
z0cY07AU{Ok+fpf)0X}pj?q85B=!K*DB|R-fBC*~y`a$y{6(Y=I_UQ(2RW*d|SQbJ&
zI;OkWa_CWE;5HwY^rS#qh!Vc>Q@}>iXjdEk2Ho`bH@faMXfy
ztKFwsOu+v33?O(=a0q#xEEh(vw(b7;;tVw^Wh{AkEzlqk9B4%13r9L|XTrwc+0}v@
zc6Ie)U#kvxM;-CR@7t<{n@dNB4jOzQv`jz*t~~FGYE`B7Ql=udv8S2^
zf@!nn4NOF2OO%i!XqOXtxob%MuQH_iv(CubIJcm4-2?O}8Ock-{2+nYgE
zzniu}j5h<&WWU8Kqp%h4qEg!Rg`7)fuEtH8SPVq8;QFL64WD9{Ct1npQNz2a6C^#}
zH1mIA(Ot=OeqyM6TEK&Vu6DkMpto9E#oK5OZvUYvlsTSfHKJd_AZKJVP)NbsWR#JA
zm~u4lSDj=fO5dSr)_@<(3Nfn;!QsMD6XS_}jUEfDl3YqgvI(f5{tiZkz;>qP`lB6%
zy5O`4JGrD(p-3Y-a|X>~{I)Wgkv57V8_y+gO&|;e!#R=!#Z=OBs0%jof1)tl72%Km*iBwm0zBdt!x%_
zR0;Zz1z$zw%ZN#4&*NeLts7zsLL?m#U?#-miX@4w0^K2}2tIRVN2&vw8Y>{O6+Y5Z
zp^(<>`KMdPzDY2Tg1O02qw*U|{#Vs>Md%Bm^~k{^MOBMFsu6}ovI^8IWk(#!fM#{gtz*wSo}XgEUojI&AFZxr#>XC@qJW2!m=8x9(d2K`lpVWZmbCjU`bIHcn$XUFzZR!0
zjzZeZ_vJ|Nhs;&wN0iPE8;#SD)I4_Yvnc{mlRpCMDxi1u3Rm}ya2fD|mkZ6?f20}K
z)bF(r-ZA{hyhnS^E-51om`3ho)ilj@4fCDN@p^i$`XC758wS?ZU(Ju(6M;!zr~ETe
zg>Yy&No}Jq>VIz-y`3}@#?qFbppENW_=$k+5~=flZv~7j?Ejz5;J;cfCo}8+R?E%S
zkh0qs`_=_eySFjX*Dmmslw=4UH431qQMWcE1U7YyYhSc6Xd|sI#E-o@OQoqJNv3J8
zZ*3@(dLY!;AC4_@c3LUS%&nRmylY=CZzeEqq7>OlTK{v&KBumL+3vuGBDp?!VSC3I
zzxx`y?i*9jMqp>1@f8m4l-+>WBMLYEHA$o;j7|Eh!FtFD+mPnQGasf$ewr5v?#zo7<&V*4sV!l?QY#J#sOSUp+
z5KGX8?B>zYHM>Sa%IE9|iM=lo=}B@igP`*wa0R%%x34glZ!bCOvXZ-ZZ8=h7$hSQ`
z6!>&Kke>spsgwR@qV%b8@w>#qEqmx&_OSjGumEeI)JZ{?~_W7guXAc>C
zqok*^N&Ou};A7WCna{Srj!LC_)V@t)yijqZjh%;3t;4So=NOcVkz|ky0SOe&ODPzO
z%t&R7n=v<&t5_&7t4swH%RljY6P@t?F
zTJKT_Z14)3Jm?yXBJZ~w_RC%57D4zuX!2b5{I0d194Atn?^wC8^fXDntbU9U2C5(yCzg!Vy7sCg3X)d
z$F;ImG;Z`uF3+)?Ce+V}tW`j=CpUf)9I;26aHlS#XMTmtl;GIWL1T|pJVm}9^6MrRcgAP*-
zVjghDyxnc!0G0@>F(qtAumjL=;8ozZnG?%}qNFZncfRO$n-5CA9?5S2#1YSMF11VxSs2Yikh
z&vqeh(5+8E;Hjqk>STGYj~v0EVT1>1+7Zb(l1A6_M+3D_pkE-l0{xynHf_YE2yD|M
z;dM&LFUG8NFQPkIVsa}qqJ7s#w1-Ow;dSfkOam57F54wto0iP$OpCh_zKX_;I$!f}
zEC>=R#P5I~qk&&dzGCYrJRpr!1Oh0=dM4#Co}#(ucDCZ2{jukK3w5e^hG8l=L`k>}
zt|$zKA>g0$6{b&kl*)}AEwo)<=++JLnx|#BOe+hK`crtJAJW|58E)MwZW}3i-1=6>
zgM-leh69`;RSFxUP+>`WE{|m8#@ox2f|~^QIaawxRjQ|?0^BaaoPK~|&U4slw&wZX
z;Q)g=Gm}-KneR|=j35qoQq=gdM{%I+nV)Pp1tJ*g}0;MnW)A^O~cy|;R5Y(a
zUPs2#zfKc$E8hPzv9`p79)|1N&TnvyC*j4YT>$RP`7iJzrkgi_Cvht}Xp%FBUHHUv
zHe;EwBj2~TPoz=H$p+Q{ycsH|7cpw;sDpXpP+?T_7&&d=UsSAQR3lFdSf_)Qn9R%e
zdQ2iR4S(d@c?sO^PvzkaxcV@Kwe3}L?ICK)Ur(R{Dg?-w#Qm@%ySLR?tuS-iQ?iD3
zZaHbMkSSEYn$zQOMLRw)gNFga^}#Z6C2`_hqQuG$h4t*M^(pot0S&6Bi#manFCNHn
z!+6to4U;NH{RV^_!CJ9Y`+?g5_WQnSZ3*s!ZCE6#X5I@#h?kaEPC4`9^c_W}u;H6s
z%{}rmJGz95@YN9U(5De>5HG%2yIM&oAfcSF+jL#6oUV}2NrPs1c{=P4ePCIt#DZ5)qt1&T
z-8S%4!3LheE*w^O>mzDxs9fF)Sr4-T4$GF8Yvh7!w#+dmHCj!CkO`71M1=I)N+EP1
zp3jaqp6B++Ty^1R3H
z+Z&t=0c&RYi;ltfu)XEGuB(6)lgTf<(M5aa>!gesoR1rrIPkH{aMS!ZeBQhZESL^E
z-U5?IcXIbu1|~}ML#C8_i)4<&)@Tm%&OSjrWew9QK{WCBG2>54nfXSm{!vp1sz(;B
z6=_{$JZ%vncTtosZxa+y8mzF@HL`M&O?B{tHKDTp122-j>kvzi*U8HOx`KWnND0z*
zZ*L6@i@0i_Q|78L-rE}v3mEZqW<`ARAEIz)ixjxO_bU^KXlee
z>ibMr9TRa*>h^Jda(Ei=+|atPQA1WK0i#VgBiv`SQ=J4!-l;@(>|?@=1^y{@mksx~
z%&*{KwiRrDBC_EuydgD17+(cdN|nxcJ%E{TMXD&d{QM@QilPUKYxDXEtzf!n
zA+0VVIO(~Oh<$VzHj^klu{i9=TZht{h(#OD>13=&Vj#%1hn4F^XIWo7pMNpJL%NsY&?Hk;HebEA{uz{y_xT
z?Kt@y(YiNhEN~Wm3#4g*SFceNUpe
z8YE1!aUP5h_{EzM`l)xS(qvAK3==y)D&JRAG39VR;em&kl8=0EjIK-t#%NEb@eJ#^4<{3@I%v#R_-WBOMhlWFx7b^&_EyRt+WBR&#V0Gq
z6Gf^JdLnsCMvDX}K-eXdu9Sz3*f!)Mx8mwtZcrTr5N)e%>!zDzes(x|#%2DA@?2>w
zY8YXzzrN}gotft$KDMRZlij+}1E4pB#A0Rr_NkIZt@%BL8DnK84^>dF;u*#wPHxFHXp06;6jr~E0<&)1?>g&XcCG?UY;lI
zO)0usbglF0
z>9pFFYHZEiJrp6Ywz(?_s~Nkj>wI`V+#y^AY3u*36}HRVSoJT
zgZQj!YacIS;X1pNwEkVu5@MFKn0oExOX;mT7v%B+tLCZZqSD!psv%W?5Qrs;lDe|U
z@hVUtqftR%(NS*ko0i)W_tC;how2g6wGxNWN@^psEq}<0B&&APLzH2NfrB;}r9iUpZjZGcNR2PTGe!L2>_x6=a
zw#*jDxWl@ZoqXEm{za9gstB34dMV?|(_%`sjl=jfEPLc0LY8rrg+)uH2Ag0>C|GLw
zMLITf&Z2so;4%*`CzS>h9Iz!Ldm<8)DcTf}nE-vkp70agAyInWy>u5-14qyTB>q%g
zh!8O(;fG(+71TQjZ_{qH?+FuD8c}K*7vF-wrZp*iSdyn5zI%>lw|!!&B?=i
zeJ9kDwkM$4)g_B7e_2Q7uEB67P!*)aP-3BXMcZ@dUXLjaBes9^V5y%WRVb51)gv8H?Bd*cQ<0jeL68#S
zc@Kd!cY3;E%r612zproaZDlf9E?ZuSh^$(~FK6xW8Fn)&iKe&+Li
zX=?WIn4eSUz?_yLxRX%?j&P{pZ`vK5ZO_i}Uk<)07WU^52)?u!d3wG#N^dAyLZ!0{G7|+c1+wM0%SP(zk;rHEI=8&p(cpiZB-DfK2
zOSibY!`XEtq3;FF5vrgu*`|le?GdUiOMmHYgnG
zNlM5I5uJDV?@(2=TD0X66M_YHiTd|YD3P^T1vp`0X2;KWG
zg6Mw_0<{cri+24%Y}GTmHSA*V{AD94ud{Iw2`W5!2@}^`>1)41Uw;WBizg-8mp2qb
z$Msut>cQgy8;~R4RTXE$73nBg{$wLSd{C%U)+a|{(%aT09BM~Sm_z|+Aid3ETpR&)
z*xt%6CPei+ME-)8u&>RZ#S_}A5!D&^Q=QLGC>raDW?H6hw|&h%pA{o(J2IQ1d3Syl
z8glEu9i4Txb7(`n>D6?cVO@n1=
zR6?{uDb-O*L5ZA~4!FY=Wjua~++39nNOx0VkCAX8gn|gMHf-ZERVu|?IyG9EXYB#O
zqPqWPvvqD3R~mdiXXN^VN{dWvyOm3#w?8FD=;^pA2g-M{V}9L&$kDYIK2V)284;K(
zt9flNzYipSkT(9^@I%|~2@(bbZ`5JPHgVRHBb@Lz0j(d=rgb*O)dr|N(<-Yfwzl8r
zMSeJ7IWz`a8Tlrk1Z0EBB9Vk(DB>QO3Zon~5F}WbAf%DdxnW*YJON~&6mbx4RAJBC
zWleZ;K@va1(+yV2>zxG`l2*7uU<~QUMMX~kjoxom^rr~K4`(3yXcssjPWgL|z~(ypD;sO)Q6H@CPgl{7&^mtV>$AH)HEqqC#XAkUx?Z%#hW}u`tif
zH9RsTpfzHTP!v78AS^`Nkfr^}S(;qJ43~#jP(*9B)Seb!<*kLVfR5bmX(HVd!;r4_
ztJp@ch~0b9;%9cz7|6n?O1xy!g`s@#`VSvJ=2mDVa6f--a|~~DIG4sCZQ&$G$IIi2
z)LpNhUGA$S1qcjANo1*DnhBLL10VUh!DFlHMq{9tTs5(Z&>=2IZM9Oz?RlXJ8n^X8
zZc6XTN5Cr;PYpifV@*+$HN>Vy^^5Geuh&s#tn-O>Qolx~v_)goA-LCjchIR?=7rdQ
zD-M`tRGW>1>pf;5xTS&(22SjhjX&2R_9qK%Ew9sVrjPk?>68c4?5rUE;G)F_gtsjG
z!-3y06+kq}6R=+OEi>koh&}WONiNnQulvup4Psh`iAM)q0?sjH4k8D~4)&1b3E=*~
zjR-1)`ibi-3AD!hCUIq{Vak;_S!21YAJ=v)B}(5$1vC+U?3f2$SYSe!G3P@+1c}H`
zg0gp&W~$V1;8uPQf3QVUhmQl>qqg8d!XZKnDaOtOI08k^&Xw=CN2HOX$$aJuL`t{|
zX*XtqSpteXs!xSQ)@M`eDQah%sd-RMz1uiEUUej&wGU`jiq@vN4GIMlqYgrSgKxiG
z(3m*~V_1H3X7xOIFMCyY8ShRJt>@mm_V?snWlAcmn524L0?#y=4WX(c8t50V-MV78
zpb7g1xjfiaehI*?k0xzV*RgF^*KHb>)6tC!iZF97bxax2TrNHLLz|hN9g3BytpZZZ
zi})R8H77OqE$;4?Y2bXvso?QVO+-9tL~x%DgyzCO*N!mto-!)d_Nyk--Y9m&y1YM}
zjAz4QlMsf0sX_wLRhe9@!4aICEhDUnCkHV2<2*ck%T^P6kPzyK?Z8H<$wp=mcM>)*IKvZM#+I8#Vs@qp
z3>4v6eLnX|g)v8)O+`W;4rgP6SX6zwBN@8)djZ}}Ij_zEb{K`n{uB8HWoWQJqi
z46OVz7x(#6kLizO_U!F%zHs6{Fs6NI1xm2*h-|$g#w#B(MC%kFq*x&>YcxLDyGc9xF0&`a`0vM68W{Tke;ZEk9ll^AEGkrR>5*
z%$Oj6XJwiS<2$M+%gM#!;3v>cR{s(l<#>)P`MdzKGT6XI#$Uf86@>|wf
zXlfndhLIqVj}3wtTHEJr2;ghp^45b(s!d7~wi)jaW+B3(2}0Zw>pEazA`$y=N4Jyq
z$0VAf63#v;&qtAw^|@b>?haEGOT=w05iLm*g&Ikg{Xq+}gG{ds+|_p`8rDY%(`z0y
z(K($fy@}JVGFrB3{0e%ym98$URqS{C^giz|tnBe7Y-p5pnGBi0qa?eKFib}20o6Ep
z*S5aiys@GX)acI(^_iRvf7&xDoveRi-S4xQ1H;X%FZ#5bk8h-yM_LPHdiC`|6<1=S
z_?(cBFw_%Er6GIr92xyeRB~n-3`(16Z3~T&!DQe4h8NSGSbxN#RR+YFj+vs4Qbc8v
zV8-&`o0Ao-Xc|57Cn(4wN32$#D!EXdo{kctAGN+_CkW!Fbb@X9iM})p4$B56ejZw&
zghS2m6$aQ$=e@BWp#YOVuCTlak}908;3H#mHCwWgid0jIcO1zQMd=w4pKLYsgLTrq
zmuYpD9pyW9$fyRtdh^Wg(n=2bzQ>gs=G_iE_)NlP+W+hmn6
zOCi_MHig1>shib&5&
z`mxZIw7$4{7=G85N+~krAuW6*!m{Z_A#=J^;FXJKWzD&}C?jwL6}O0L>z5I~=(m7M
zu_P{cg^#11zxSJ|a;c}oa56Bpvi!w%-AZ{&>)O0v?bd~pGt*8}rc{*|e+KqntS-vXT-1#53G0$MOi!4R+mtY`qoY>AjKrK^+W
z&BrZr5G7$loLwYsMy6>!#rKvjBK7=C?H3I8zS)wGbBzIsr*eOJ?w0!<{%Nk-80es9
zIyBzd*sKT5JV}epZxbQ5xxkn(WZRYr;}3x2g|7!WN)a&`+$>;yIOG6HQ{ZPaC12E%
z(s~+LKimRSZ8T(S!G53#%e*6|B4h%#Vn4gbSd2jHYZWWO(epa+gDg73x18_(I5Uyp
z684wO%t+g?%tSbYfeLbX;rtTGBYt-IM@i=(Yh;mY%^>!mIBsG8a%5^II%B?XC&#J
zt0^|=Y6n|Nf2g-RYDcxEFGw8`g>b$B>yB@np(S*
zCKEk7#**ZT5n*0M+Y^(DlK~GP^Q9C)jom)rP9d~&4LJAB6%PNASs8ushn}u5d$-UR
z1$I0Z>xK_j_9K{VN8*2Y4DEI75EKuMI=3d}9pSrrJ7RE6@H3CAuR~N$J_$HE^o%7d
zC@i)X@xQQ+L|d(aU3j(#4%KG({3Rt&lL$exChJG#4l?hzGa7Wx=ZAxsM<(9PnH?qp
ziVPket?Gj>Fr0%HX6wkc1Eq|7=x*``aw3f#^$F^VS_lUL=UwN?K*~}0fJjW}9({-_
zfpqU?0wpXKY=Wz}M_KT;@B+%j`0Z$~w!#Qmf^d;EM*~c0jJ(o
zJ_uHs@<}Z`ft0mM1|#!^(`F8rvol?azvfPZ>%IM5rGslk#;N!m`@u9r)m2(3Z}@Yj
z02>6SgJpNo>rHHMbUaW_bd4SM3ujuVOD5O}g1FwMs0_?E0+dqC00XX)oflvwKX?4Q
zRglfwGO9PmJtn(|7OI;1O>`H6{m}UXV0A!YA27A-T@Q8`4)(nGtc^3CbZb-_1}FTy
zeUfL|pReM7{R=?T!uxef`#;mkkISL=grclps?1`ZJ=rF6Dnjlu+6w**G9Fx({%X5_
znG^Wojg7{Kma(t
z0l#XRM^;%Lx1c<_{ZBx=#2C!rikAMCg=xX@O%j9KVlPvp*6(!T)nC?JPy)j}ol
zbtj!P>b6J_n~QtI3+p{YwIP(Bv&<6PRsOi+EyS%jzV`vWQFNGpApC}|@%2%7KEL%{
z-sxwMrTKxG$#1#W_-~;wP8bEh(2IR#1_NC#P};p0qDQCHF&IRi&@qc`o%MQP-O^SQ
zou^a~NJtL{sG8$gZY5o;RVOYl(sp)Pxh*dX5H;+3X`jFKjLf$>b*bsCx2>SEeBuvv
z-t`eEDRf4(E%RTJNX3EQ?IagA#0T9|4k{iGF@lap7(tY`$#4gQhi{~T_1^UmP2J$W
zQnB%Ti*%nAQg$^5&aV0$Ku`@VHP)?D6`XuoXK{9mJXP6cv{;!j8MCP%Zm$HYdh&cV
zHVZjc6L(Nw)=_cynu0z4i&9PN2dE}pYW-;5AzZXetEQTdJMeWXUE5@@?6yEy+;7Hq
zfTpf0cyvDi>B&)M|BNlv4lWCA$$NgT*%2iX35aRsL>Bp)oJMCywg-gW=cuc$VH9IU
zJ5?*R6PtSb0o^@MXRKGJA-8sm4iMES{73L{`RGI%mW&$PJb%3P_0&d3~=R;
zH+|m*OLl3lzH+$s=*HGn(4wijIP9D1N|LIWGLpOG-)AjeMQiLyg7N&$yG`Xfl8DnV
z^IHO#}()~xbdZPp9}KR|r~KMFWhA!Fb*{AObd{51dgX
zl~&wrwXfQ8-Q__{OnXiZ^$sorBDjra1FnJjQU>H@bQ}KcStM08!lf6a+qi{p1wWJG
z*cTUD>zHu3^5YOCR>1Ptf5t~xwc+iIH*zl};NUBTE7K@+?F(`xOCmoas{fv1+JNO%
zP_9$QwTxbA9g(OIJ^yLa^9!W8dPD$s$om%`!|nU84ba35id}63hmL4BwHreUMl?$_5b)>tzZ6=+DEzeR4
z;laACqv-tRhZvrEMbYtmiaGjZBDq|EhRiU7@uJ3IVg`SN3cT?#F$hy&s)rKTUDP6J
z5geu_bgqLtdvw#`A*kWf{fC~&97-cbIFl+mvd)N4CBL*iZ^rHE?)j%)-?Z}ojaZo&
z|8Kt#MnXmw2Il{|1Tqn_GcdCM_tzKu{NGbTc20Jt|1GeMc5_iq(pl!yNkE>1!vkR!
z4h&)u4|V2%V`gBO=$TQRBZZQL3h)c^3!0;ZLY^b0pxoG?c*vLQoCW|YAH2_W&$^}?
zpBkT50a;d`4o>xFIM9@91sD33=1HN5>X;;;@=LR`${>M2gmJ$aph$%JCT*gw1^PZ*
zRkuQGrt8zNCkr01y(Q_$xgz_IFTu
zfA=q6cCJ1@FLTt1>>w?yi_=I!4h~OkyZdG29>@bozoH^iaNkf+R2_XxFw_8$V*wsq
z1G;5yAtG@+@Ud)0DXhI-SeO5pA?9I4_uf4!Fk#mgk(CPV*r2RczXZb%$B;axG(~Mz=mIu
zV1n)W)q4aw2q4J%&Ocu@xIpqs%mDl~T3?LxwW3!%IsvZ=5NvYkB@=nYR>JAULkyg!fKi
z?)?CG7;@_VyMCNMG2(st{yabzfOSC}1o9-m#=WeAdwzRXkopL`08(?!K>Yx7wWqf!
z&&^|3^W??1d{4WrFv^{!B}K-*pS8KaFO(DoJ^;Rg`VRi_;p`9qXn;}xB(>;(zp};f
z0$&jV
zaPPkq-)gD9I#<6siJ4?b?^%7XWQV_m{MrcOyL=R@56@3+c%ZmX&;0p6S(Z?rM*b;t
z;-9}S7M}by;DU(9_B57+b?NltSMXc|Ihz+Zi<$Y$`5?=(sP@)#j6ZE%3vX?OtsMGPdp
zw$hSN0L-5zA2EFdhyVyb@E}q^{<**WhzI~@kN5~M03Sc0zm$*{5&hb$Us@=u&p#o4
z@SHFwz(I-8ej-A*!5k{VEIBrXzebs-d=B6_EJ0tpsO$&Pq3@p`9(IV-2cOQNzFr$=$%)
zvN;B=ulw6{U#%e9FR7wTD@!IG7Zb9QxNZ<0R;(cs4qyj{kmrAkg(VyYGEtc_CzE$5
zd@((0$B%pb-1ILoqc%3;nqqYk@84%439+z1mMXSMu{OPu$
zY$~42AiW90WDNSfU2l9o#{X(oo*S9z^f-SSSx`1TQBB3E*p>SSW^;yn6Xr^?oDTg#
zUP*rjP*D0?gHt`2>BBdE7L{)(^SD^wXY#w#Q_ZHxSWvSAGQX?_z2JS859l0j1>1dv
zcrwg!e991BvpweJwLu$38_%lI1Y|h?Bz9%ov+*Z|_7~O%O*Z(PQ%H^bvYn{0Z7Xrk
z%|J||;fA>J)#JD$(fzd8PE)zh`r@Z6X^$yeUxt|en=0k+r!}vU2-3k~t!yBk#Iv+W
zY_{tb4LvoTxFS`4JrVLrzb8c{S^YtPvupY2U+GlkoAO=5KdZam9NKapk3Tvr#(`F$
z0@I-?7J@XUp|F=K=8IHYz8%cNL(5&0s94x<;88-oRey%JIU~!>I`v4^>sc(ItHHp4
zHld<&<{R(xwlq+IspX2JoI*?OwsVqk7Do=uj42_~Z;xFABsx-($BkpU+sxB+t!7cl
zULb!BwQE
zYG1c=nO%F~&UskDa)}Javi2nLPke*!)iOtsxp*75{Y^%$PuL8B6%se1bQ{JuJ|UZ&
z^5E=w0UIc6AtU?_w8-E3lB*T(bjllsEoaG=JND!6d)jM)yOS%1w$*x)dV$>F(f
zbUG4wh2>ZA(q`)^E;iwHVaqCltcnu7QKOf0PP
z1{!I%DQTXrM>{;T_<@E{v(MDhVChrP-dvtrn^XIh9Jlwf
z$#15;3J^mr)wmY&jmXBOc4`gDDe9vsLp9Ki-(yM8OUrAvx7eg#c-Be#H2^7eqb17s
zTkKGcXIM>u8oBZ;{AQNyqS)9y0;5mf7v%I%G6?FBPc
zaSfS%N>Kn~{tG(O2m-nWz}#X(c|67^Q=~^|T}r;7!#DNYs%yh+Bk%V0LAAJB2?JSo
z_i)+nNv3)k<7WMfAU60=Cbrml52<#PiL0a6$-}0bAX($d^T>V)JU=dvr&~i2wC9^K
zj;`N#MSsNO+1{V2i`eIymBzmNPwt~^TqnZ4?c5CJZ@Y~|>*7rmss@~ELw1L}__|>Z
z%GS?3dMVIM`JNYk-}1Aoci+_1N0y(|5mCNNaF47
zoK?j=tFX0mK6Q_~H8wAngvV=7ieKGzDO-ir{lpSop=v3qpylxG8eGTCuXKilEONT`PrVowN2H1@B7%mr<)A+
z6U=Tsb%OQ%x+=&6BotFshpI+yA2gNTCtdAqNU?y0JuSvy|7UcoAMpS(uBDKPx`a}EQ8g6Q
zFX2nq+H8HuLQHsE$&uzwL!nK?St(&x(s{#i@Ces?W~_TEH3U8t>{00K
zP`+@)^A38nYrw-^Inc&VwoZU=>RcWuoJmi;KZy$?J?9#3#oEA(DT2R#w(jjM-cHdS
z0cNjjvwg+xlA&uK;4ieW#G0${IGsE&o^{xTaq6<(vuy*6uorn3D$%Nt`NHerF^0}8
z{&G+~QjLM21`V2?yL`3vK<&&SYS6H=>X_c|H6T&)mL1|0U1>xg9(;81O)%Qe#^LnQ
zoI=>-p?`;0asut7n~w}&cQO#IRD5=ZR%-Cg8hL=mvks=8xIJHSDjs63Z!_+
z_gwX)f?s)^ut;8Lf_L&V@6ecN8*U_2%O%@w&tU(x;N-D-f_V|7RaIEHXvOjJpytM|
zx4A8Xv3S%qmITAVjB}cO-wEnxVH)65mhz}kH60zQYFTT0t#Nk=Sjy>f>*gl?2C_!<
zw_I>@)Jo`P_JVV<*OjThFNI(vAdJd4k1p}|5T$LA3D
zc3XG2aen9oI|Ljz$j1`&cB02fZ!P8#V6l^9fr-wCE7s}{DC9ngc8`4=Py$AOYBTh4
zlYSp#>UFjygAtkEXB33V?Zl{150FHloIg8_n;1%D!PdB2#m>s+e&Xg6SWp)HA|$!%
z=h7M3zS{(^)($iLX|(Sqd%VMsM6`>95&U0p7&;E!Q}bKT=3
ziONth*V^~DGjL`;Wp>X~TijQ}CfPVb-q#Z~X>0g;qxBlQ40mB`uEHjXY$7i
zL5{esTuA~M7CH+6EfR)X3&J$!0ZPJO3^aWDuI`XPjxTGbjCuG%FYWf_)dUHHfT{yL
zqE+XzrEE8AyIr?A9k~RlWpva>6NV&P!-n@74U_gCxVo*=?(JrZE8d0oj@W>L6;kv*
ziqSq1rEhM%&yU+VEqx%|qhi-*#tt@cR=(6(atnr!s@I;rx`0|szk!;zrmvNYc=K8G
zYTE+5KkIp4R3fKL8H`q!d6F-jkdfd3v%_kQcY5Rvy?eNP7Ayp=)!rv8egPfg#|g%uTCDcW#mwe>`Gvt3d0Mb0B6Za8Uj^88xN6_NV<
zZCJZ6Gbb5Nd!iB}%!1yd&(r^-a<%?Ist?BS1oVE@GJlzNdCkoUHZ6INHY%g1q@qYg
zfio%t3T!9F>KHmc@5&a&U`ua&XpArr9S6>RDvea=vORffk}$_8p#3)!3I8VK
zT`LuR0M=R}NX`=Gm`2;au|q+Oa_d&MS($LmTO7rjbLHMd(&)>TLMB|Gz&YmWCm;72QrWyyP+veviCxg!x9B1YebuwMiR45MdJMZ$Nm=_K)$*L>n1q(`DH^j&jOgbPF;C-(2ELOD6nx*imB4XFJGF3!29>6gUC&P`BCC;jpOQ)zwAM3U
zbkC-}$xbI)`P%QtmbnCJo?ua&fXhR)qkl=wX_wR%XohO|aj*_O^Lf8wV9-|Lx$fS+MitRS0llb7
z90W4gq^m(*dq{88F#_IiBu40e#6N}cR1{uO%GrxgT9DysAu3JYT~`L$TW5(s6Ik9d
zT6$XHQD^l}vD2^;Y)dWL0P1NuTCF@Nq{@W{OtM?QKXX{G2@~oN>(KNY?x4r{!J^Ma
zYNMB*PWLowfOeCR?*S$p_*XcP-TBLuOSd7I40^RbR`h>7Dl$!}x<&lj=zmJA2{%J%
z1k4VjV>AMQUL;jZ>x$V8@rP{f761gqWjwDw=Pu;mVOm228I3EYz-
z7{z2`Vb?QLUPdc!0AArldCQH>heMYRmUBRj6pUQfp?P(278}sz*gFyRKzwZVNg!YE
z$rFDO541J$Rc8GKIYlI`u?Hw+7%!G+^>Q>etM}vrsE2$H6CTMmp2fcs;;+hn%)jds
zsY7F^{jcPct{yc6kNm+dOnM$2KeJW)-K!rvI6V%q3g#)bBD=}H}%gC
zI94F8QQ)OPn$=UTf2p2kG8NF5g~jY3hDlV%bw%4A73UknDCNwAa$Yz2a$6qYy5f?d
zOR2-+D&KoKOx=8p2p{@)Dh!QAoD`xm;^>4uqK+OI8PNU2pR=_&v`M2hP+k_kwZ!sBXA%~o
zRd$hs-Fr{rS=)~fX}$Q}JV3XgF(_2uhFp0swz#P74Xw`jX7drw0PfH9@+ph%N-bES
z*sS1=*>M&qOZpakCSyD9Q>~4Y^gvLWwsJur8*rgbXJG2gZb(k1)n-{k;1bU~#I^6I
zSi?f}-;2Xn(y6wj+$?KaNOL?4Ja*R7a0*r)_s>GJZJv)a1n@u?#G&D@W!T8Q(?5&W2)KfYVrSLVp-$OON8o64m~a9@9&-
zfLxsqwnKh8_=<|p(JHOA29&zr26?*u5hHDFkbNijHxnJ!)6qFzm>_zEERvc7)&8#C
zoVYIQVVbSkQ;m=g1+&CJIm}c-meHkrc}D<@=)1@3q%S?=DN`C)bhO;L>IPH_-cSXr
zKK4w^q8PsFis!99O>KIp>3bc3$ARYApo79=WbQ_51q2^Q*M2eRHmT7q3$AqBi<#X1
zw%DM1ZxYvK$8jYdTH@7GeD?T0MVzZ4?g>d-Icepvq$|Zr%2XG3pVfq)JZJqDIQKt0&+<R~$^f9R02s!OiAy3~c
z>EFcW78A@M`@V~9a9&Vb_W5?`dYi-KBzdIfb(y%y6p~+vYLa*c?()vAX&m(OI`)Be
zSCL+l#*!jDVcQm~?uVp1vRzwy3i(KJt*OiX>zsBfFBX>`9coq*4dBf$()>fz{e##w
z;;nQkYNxtWwb=qruI8tUFY>ti8A)nWg6f5+A^iMa4&j+~XTv3A1WZq%
z;(LG8#wGxuq(`^;pO)V;8F^=LvIxXO2{NEdp(LjfwXkz);krKN$_>W#%e3n=_ehC`
z@!qK~Ps3*h%ZC8G3EC;_gGk*cv0TJN>$#)?s?kOLoZ!}1M6{MSuUw>$e7(0m*1FDT
z%Z_L4Wr>krJ#n^4JQ+5uOVcZ6G?y5(Z2S1v;fD3vwHg+(?UeXu7PmkAyqk}~t)?~o
z*h0YERSEP|BGEPp;iUAp+B`ZJu#rBu>%x0Bdsi;=p8nkQnl*o=9JN%l)!A|_ysb2>
znN??A5eXLZ&yg)qOc&sdl@AiFJ%6ejr~gpL>7x0HmojE*uY|ucE?g*wd&Ol4Ec5l&
zcE%ai-6zN#a^Z9`HHJ$egIfh|>WQ>U3#xyBB~{hJpv4MGr|7mclwG8LMf>0~4^WXk
zuu6|k6*n~^b9%WWObTRiFR=KQ#6%UF5w+B4YYqK
za?DIrA^`g%WDg2&oO>&_it-3-*RGh(U8xI{!^*5{T&&Fm8|IZLpeR$k+jhjG)%W_Z
zM^XrG{53+}h^Idgp}+ZQaNX)J5IeXlGIY&CmlV+L{EhQCby$}{EJ}~L|4uzNoNj$l
zY6-$q^==>G*5t>LvWYj-ypJ^Gn6PC3UeE`(M)All*aT6xZd?gX_tbp6!FF%uEod!0
z%k9$P(%-UZLL0~MMER}WyPx&*U*!A12#2xBaWn>S=EYeb
z1!J3QGt)+{5O$=I2byU%a=4EY-fic)*cXdlJU%NO2m=YODoMKyQOG3i3^Kz8rqYRQ
zN#(ibj)sn%)%#%O221ursp9yF^=kbb*)o!EwMH=~aKT{QrTQm=fEJ!16{gTGEK%oD
zC7hQm*YNv6KNs3%Ufw%SlU74Sl3t|QE@lx$rw3Rjb}(>D93r*1*=d&vm5YCtOx@e~
z>d@>ivp(kktOY`a9V0$wHD3Vn^}N)uoAEGbn@Z9K!q(3xGxwW}_wW%K~dQpLZUV^30Uo*K}6fsEzTfat%)tlDKZp*+}QxW2Em2M$|
z;lRSr#6F`=xFtt@tu=&8=IhzL7vHOIz$#DWAphVnoxFPUr;^Cr{M#DSgU%R==~%nL
zZ)VXXlL`!-lHt1zY7wSO6^VPY>J!PNo$`lu9S<@E@qp|GCEXt5znqOv=~LLg5`?<=
zbmLI!*iH60+qThnQ`*wfhY4BeiWhu!E`;(zHaWzVQTc9*9A~ZL%12^kiXx*iMi9v8
zX+yzaTKeeyVQjRIovlWp*|D!(82J0-qqh7V+vZ`qs$yM@p=A9d_~rvcl(J(}QhPG^
zJ;ZB#lm9$KpUEmP=$0QBPy0V3wfFOVcX`LCTIO$^sG01~h#_WR4Pf`kxpz9|g)Q4v
z&zVAOs1_~cFO>bo4DK3+q)^4K1bBPXKmBzo?>{;|)JE1g78XToR|z}7Rbz#KT2iA2
z%WQSERb;o3l;htQ69Jkh-{@jmKWv=VYoznu7?6>|m!!545FE+aJQpUW^Vp4epF9pA
zt1csKfrsqckg>f^%`2vth?{{j0jJRr5Uw~U6tZ(!CmA?in#gKKY`r5j>O^HDR;}v7
z96=8Djm1wysp85Pf&)`Hc6g;QmQ7$-GBOiv`=a
zxq0xg*w-J}q@Tia|K+nwH&@q-jyZA&fL5MGf)F?J(_74>BiwP>^!A(a|FD)(Ob^9L
z`AzbWiTU+8=s^GZeu9J7wtGA;_HFMbPfMN^@%zJA=ZK<*c79Lp+}jxW>j#j^=DYZR
z)#f?=mztE}ADWPX{l81_j0BAT$h)lnJ^i1>c}7-7HunFj&j0@e!=`o<e|vj75X21o>#8Y~}G;huYj#<+{S>Vu^z<
zvxG=#Ss#fBILp7d|AkiCdaCLRK!nt#4tFj$CPf0r+6qkH%+%yqxVSLU9tcpUs_!L{
zA02_ue{>ir-|*r=)ETI*vEhMG5zv_bg`kwJ6tiy#IUmFk(9)CBe+{6iuh*Y)Z9QZh
zw3~UTo#-{IPhE?^Sb1^1=u0{lA
zK8&j$lO~{f{|f+eJn7sD4gf2-d7v|(CYIs_767ITDvb&(=Sk}jRt+4Se(CQbQ9(gD
z=^#Wriqc9lkbe^fd9(E@Nht0$nf@d&icUedI;u~asQS6XBw}J
z6qT7j*V;JP-OB|?1L*fMhMHGnt80gP;7`ilY$^aswY0#?Ogare5;
z5AjtsBCVg}#t#+F9t_a88m3hKqn@^iii(Io?%c!!{2{jqb--7YtH&*u-Y016Tn>D6R_qH$8Z?e4fVj@2O
zv=r>Vkx@u}A8bVmFDA&kob4=2vmw6gq$bbHB)*MQ#%7_&&`BhX-&VG
zgzL?|9vx0qzZ+@*WNfn6-?tsp`K1xOqbr!bA9UEKm7w4G&y_bmOF-&GC`ri)N0?tG
zT0c;U^{l|@sw*0REA1_S{3PT#Zup~G&ZQd{^enG
zeK=Rauf~L}EdaE8zLH)9crJgD7rP-y{e)laZV}3#7
z26KJU99PEzkBvip#icZKN4R`|dn=uM!M}~z|68PV@&W%=GIfoACzJ8X*YWgc_>t~y
zH>u>~H1hir%;4MW>F0L+5q-tp0cg|TxoT=BxAPZ50jPsr!^2zOhwI1M?xnTX@Au`t
z7fSxy$LTwWh8FJS&h+5u7>vID(E)I+ZMS1;WTb!pQ%`%*j|c3h!*ge5toJ^T3j+Yi
z!(SS4HMPG4h;Ze++5pv8QKT`YPT70PY?*za;~|3HN~3W**gf&R^N1kfpzp=s`ogfT
zkos4wcZvT~QKusQwWYU}Q_<4Yc4|rZ)!$drH@HmEQ178BqdBl4MHQIRYy3R+_z~)P
zg8j6S3zpve|QNN0dd1(ZiCp$17u~pXI3|)
zCV_Zg+7Sy)%;1k3W<)k>jv+kFtZG;3uX*RZvhg`M8G~Ng^@WnPLx(sHHw?N>u~1t%
z;_0=F6tVlnrj;O8uoru!2}rdz&v>gt&{Lw?kurWBoWC}l!bst0yK1=Wfi|b#oJX3P
zHF-dkwuK)!X^RpVhJGm>ARyU@QSAOSarZ~SnkoA{DH8F?!GX_ku%IoXVDZ@~uZksG
zcec`DWZ5$JN++P~7|x|Ee*#7c!H*{HnqC98;Hwm%#b10h0+n(hzk9n!d%^~2C1~Ee
zg?YOWim2W}{3;`w()#nAJc|Zp?*xnjDjXpDL+OKBIPskl(P%YTw$fts6Nb*Ak0Gna
zs+d~sGiPGYX%bh@t{{bn0venPy9tkdQj6G;8^gX_AWp4dzL+cso(!wQ#Y0&=alG-U#?>QR=?u73H-%32OoK#9w{?4DOep)xQ)qq)na5gg?X
zt5HVr>h9jZl7wr!Jeq~^X8Tx{lGpLv7pSU_Lk83f$~UYR^r0lxwGJ-LDBULYfr#0UOSV;phy#k@8Q@#
zPh)OK9J;sFqcuMHTQ+XYj$OfAv!Xo?IG7}vv&GrTs5>c$O>Dpa(
zGvaQ{`7#jn5L2sERJbeAUMUpmu{5i@>q(Zi9VLz}UP-t`zt&5GBNJ@u;3o}h8UK0)gL
zL`5En?^cfC5l!CUxX`$JRF&U_?vH|`^Do=Sq|R!0o6sDZLO^6!@8iSu`zf9FnvmIQ
znf&D|rk(!$D#3zOIBeC@bUd~wl@N@F%u7z%cIg(F9()TCz5^b)*U@rZtK^`@M7%S(h5?)u9OO
zsx~wl7F;s2CFZCsX{`S)4T;N2fz~&=G0L|03>DOj4(b2WLK8hVl
z53S_ITbqQJA2T%+#7rB8C~b8pg59sY=zL6iVxsz~If5Gil8{j&bYPS<`YC${1)d7x
zrBqvSR!zLiR_U@YZMCE}K*Ma;NrJQZFC`y);L$Qqc}L5{gsMje_v&v9XOJvSn9gbj
z^=*rlT`uBld2ISa&r44v?iI-b7T6OFmwYox>SCUcmENZ>3iAxNU{p{(YnfBNwnhoq
z>xTDxD}pGn2R^h9@h7!-S{$m6!x=Hv1E@%%Q3Kx$TzSerh3S3LS^B!4A|T%t&*ESksM)jT@c_$exgd=|C*dD)Q41w|U~G
zuAUet=A<49Rf2=TXD=x#kN_<2)byo|IGcr`iphF!j|o`Z8igX9t0wn}Jy6#RvO9JP
z3SujBul>onYj!Hx!#Fw`MaC}e1hMOpEYl->4N$WEjJewCO)jS*7VLKcsg8ybA|sb^
zY$$V(yX~2EvK4v<5c4A%2!nW>f>FRxjvD)n%@^<{#VPk}L)gkDI=jLn!fg8oXd>;&=|-HOePA{f
zJVTcy81@T9EHoid3ripFd%GIUf{6VHc~qjHgoYRHAFlk0V$t(2w)F@(N5M-|%mBsx
zcf9x-TT4%}b5P>(Dajnw89g-&t{O$ZR4A)|=`q-h3=b^D#Yl01_>j`s8|N$98pms@
zT=_iC(>XscbA~Uy%Zb3>+4Fa=1hUO-$9dOm2k_FvEfOr
zmBxk$7@v0vp+;&9C2S6A+F4Slyw<7~Bd<}OH4A4d^niqrw;GDA5m^`DMOmydY1AId
zobQWZfOrehIsO4_K?~KYRdFz0TUA^zHxKiRAPjZ^M#C>&pS$#1z=_+(jN*mCTYdSC
z2i)*ktPo*gTQWh$Q_Z@(mUZRv;*;~5r804k3N4oz{l~-!I`c#qxR)sP(??xF|KWrw
zgnoHeL+Y%?TPR$&`=bo2KNz%Q2z-zzr>ifwEX>5he69Yx7LTR~1Nq7?dZe=MCA(_c
z$L-~U$gWO#EwkZR+EH2HS?6?Qoy$r~PfoKs(dA~Ud{D4vUko5Cvus`9>$*XMO0?Ch
zr|hHEV|0+G-XO=P6S59v?PZXF1c@Y*=rs((bolT1lNkkRrjsMR23qs-CvlL*#G~$@
zE_x|^R6{t!2N0zMuZo&LQTVh#JSVCZ5Gz46Wy+*sOOVZuCe9Qj{a~i)KzQ;}&Tnt{1cST%VpLr`I(Wq;h3Y^|5@I
zJ&8urMue(4VF2=za6MEfaU;2vhwb6T@^ZUj<}=Bbktb?u2&tk$lI)XDIO6xlOgNG+
zM?tlV<;b}8QU|i!pI7fiKQe_n`UtmPVTv3Lc1?eWQY5yr#Ft?x@^T;bJE@~pcstYl
z+K5J0CK|Z{`HZ8>Ayj$d9ri!%CxVXTUOhu?GTq1QGo!k|14s>dZ7RsmouJ$b0vP;y
z_OOL~)$EYa&y!Qz&jMGvuT5fhvZ>b)L{|o!~(X3$J^fVrA*DS
zRyq)s0;uq#2cK@4gM{*d_G)S=idb0u)?yZ9Dmcri3ZUs8>R*{u^)F6O=Lk!unGR$n
zx>ChHwx0-n-uLWd@q4ba#Y`mwkJ|s8UQs#~ypR|YLGbjHle&)O_SracL7qNtSSvD4
zEj`^1jX{VH-$OxXhUhhyK)tWFDAM@2lzd2L&J+n0y*tN+N^KbP=QnRs8207(jwDq1
zN%Pe_JbCCiU87Lj9b=g}ex5{WSE3X_{
zdCoA6EKbvYYAx5uI$}?Tk5|AWw_C-rKm6Y!2sHdcHHfEhgU^|Gcx7@}nf5bPB`a^w
zw4SiY9sZ#%h+YRLl0b?^W}@m&Gjab+z+U~Y<+1D5$Okz;<=H{J#{<{1mJ|HC$Vw9V
zLHe2`!C_CCNr#J`UcR_ZU6MU;hJ{XknGW>FZz8;^kcbWIj3&AQWzL)L?
zyZi1LcbB%S_D@>fYAai
zYt&dVUOHy*3x_`(&*L20{tw`08#cL^fjr6YD$SkNjY_v|(33n;RGgKdcwTdga3|4r
z^|^fxMVD;&IHanNvUn(|DbAy3O4>A
z>|+t#|2V1cRc!&@qtRjABwGj5o}Cz&6|D90&KfuSEUJ+bj3my-
zs-!>M5Ka?0(jX4liN`OM)vD~fCl;$za|KQtCI~al1g3#@t!BJRT(5n={z{BRADGI|
zpD_9~67MPv9yd?)_qLVkC+8kopa;eh$@fz3Q-T*^uuofyFM9Mytk=g~Gl~#@8nt&T
zab2-AY10=QT5MX|(tK7-^-!Jf7*G@>PcN)U;&d~C5tN6uan#=9*P(X|Zjd|R8DD3Y
z^^8oIAs~=I*9o4VMhyNY_ET5lRqH(&^U$9a17qFcJ570Lat#X$RuOldW^Oq7aav(R
z7SJbj#5fIz4y?_KY>XcC_lo_2?-&XrtB+qw!VGwx{ByLU=eS`z6e83Kg{p4}jY>z8
zLf^v3moQ68{ikUs#V**ePFyQKoG3U{I2C3ZmYjsRJnMI+U@}b<@y3@4oG^{BF=%G(
z_9Hw3PawO)yG3`JOx5y8yR)QXWsoo(Q3!Q|1v~>id${8~GWAv)OKoMN(s7k|Htm=q
z2(c%?tIgP};UYkeOX*F;zB5aJK*4afiTL#z7Ie^wc}qEGNTQ>+7|ci3o5-7C$7bzM
z3`fbzsOzy1d4?F^V`I+x`fgGM3r=(;1WD|cQ2;HSWALbFn&EwBzb;z~)($*`l&|B6Ie{#Lqx{Q;MV4xh&0vy?YGsk@)a#vKhV)s5nwZ0A
zo<{S2V!##rbDBrB7LeQgT6d3Is_z}OD1z*BQ7YaHmm=x3f;)jy9onZrr3wQXTKQMS
zh5m5r1~}|7b3NH>dqxAs=@srU;Wnuk&+{+cngZ3h)Dm))m6CF1hlX@otto21igtXMYT^km2EYQf21T2lm1*Cy
z>+`Tv9_cYxJZGut!Z&k_Lo6Yq+1X7n|H!U6Ou+z%L+B@2_NLoE{_x_~xGbK#?)|VIj_e!vsO~BqJ)kp~i0O}JI
z>C@AyN;LM6GKxVu%x>w!trQn;w}%q;H<_nPo)*A_Q6vh
z3luI_vsf=WPjLE}|BE7}%XJi(Q3K9?+1Apf=BCf-VW#YdL^BO%_4o*Eo+et2w~6mU
zxFp?51kQC@B4=}YM=!cyr1z=1Y(|vz^2?^HcPIph2@zK8u5R_+(*~ly$~g!k%0EC_
z+~OAKeSix9xu2psmv$vh=5E0@10Ea<&N*^fhQ~4n3jYZ}x42=Qa#`YNH~SMy&*u3X
zLl99Dx5#gU3Zz#@x8wYSWJyLl@^la;9$qAfe1+~&XT
zu+-TE`fIo-bB+x5i!}8@vPlmU)K1&5T(dZ(&4rNq&teDQBaH3aa$hP5DTTfd&-_BRzy@Nwo7lMnZ9p!3*o12(+
z6&rXss~M*1I9g|Db>&uq+v2?8rg6fN`pbi1T0@zBfHZ&ET;9-3oz!imC@mglssM$Z
z>5($iU6#N=L^-f5_06!Zd|pVM)6SfUl@VZ-E!2~K{ro0=S^WY*VrhI@8UCQP<=Ew~
z`hkZ695@0dX;PJfHbPTVXsGdO+2>POzL@yeD-Eb3N$2!jbHHht7jLQO=v~SOxfG+j
z2nf@CzE0i;fY^4GMA^1MISiAD9jTz^mTZk;%u
zM1Vxz8Xl;3I_;RI=k*kxyDFk`k=x|+hF3{_R3{P+y)7>bXzffk{hHoE1{?BXEJHE0
zuP}02%@Gp4?mZ)`v6A@DuxGq27OE;I8LsIQ;CR@A|JpEihHFlsz11khx^6i}HD^1*
z5#G!_c$PlnU2mV-K`#^tnpA1VsYtU}(QC2Bh;JOr=czDc$FpxayU||nJ59f@$?dEP
zYy894B>Vf^Dd
z~{tj@R%!zO08ofkTgrAs=kV?%Hk@qdXr`Dh&
z6WDQUG#q%~%Z`mf4Y+4ZZoO%*qB%D_x#m04m3l_anK!s{RVR*EQ3ah~4d$s@DRPN%
z^**_X6nTS13cPNW{bkQa^~J?a=Wpo|^)E6#+rOo6^gCaZyKF8a+Z4A2!Q3L2%7$%4dW69ux&Bddi-W~qljP%~!QRDzKGBFY#xb5QzbX}km9KNoo}P5p&_WA}5fg=Qwu^};
z9(~96k_B-9kOOrC;?2QwnGK{n6Y`hMMMXc+W(QAD*eYygW`?7zwYMIz8eZOn55D+b
zBU^@Id1txdcsLRstqsP~TNA=J~DnEQZ~;EjRO9G-B)
zn;I${H6Ds|82FO*+F|DQ&}Rtsp!ICN+{zvv#k3Om+Jnp9HR}T1Iny9)*R10W_?0CEb0zTM-JCdT_p820K
zQCbHd%J5zB@^bdDN!5@*3-WEsa1h<6VmD-d#ZbAAqHau{F=iQ?u}II*Up+Iau09&A
zJJ(huLBHJr=HY7HB5g`i+<}dmMAGjQ`a&8M+*H`!yJw#w7Hws-yo@Av0;^`b?<|c!
zJjAX+5IdC8%*R#{4AWZ}n2>x_Lxlaq>K)~%45k3~@tTA-ho!K}&(LZzhTSxu+OpSH
zVVP}oE+V_OqT>>n-kP>%a$3Wxtfh^Ev^xeBQU@(1glc%WpQ500(~@+(RV=AQ83+tL
z)PH9lp+n89-?@W-EqcY?EW**$3tcPIZ_g8J`!u7HB-|BZFTp9bnYoJ~My5Uu%0R>4
z#pv>5N%x5++L)nQtnp0Z&JUxG?6!2&bR6YU>cir1FZ*&k
zQu?hw6kuNhDRkr0jmuX$Rcqt7J#(1}>3TAiKZF)7#k32``Z!|3@>TupUK#dNuJA!x
zKSS{U_z8qR)IX~Gh+_$)p-x+MSB@%_2i4?GP(YuMA&(2K;Yoi5$6O$+f+V6d|K5Lu
zYZW|nw7%RSh}g?-mqSp@k~IEI;=I7s(sYrV>iYH+Am^(!>Rvw~+Ug}LC-Z@C%ZIvA
zP)7rMN&@)?X)%J!%sL9bnB8j|=a6KHf3p@#laqp|_9%d}_&c%%e#I|>Ge{zqV=-;0
zL9(l3{>XxmG6X^kbkwuvIc>#c1KYQv)qw+I1a;!aSU7uC{i2VluK|sF0=j#}JMT;#633T~i-?7YR
zkB?cs&A527x&##>cMv+MoK{h-#Ul(%;8$czl>Au7aShsu4<$EL6(W`8ZrY1-BqaPf
zA!UE}BZznIbsAn>R1o_qU|8Pc0duMe=7Jp@)F@zu>=T%-``wv(IGWuSvISNArAL%`
zfr~Jow(`G=UCG!$CPBj)-Ru4QaDj1&bocFeBlYYMQ{B
zJ8d1*R)79x+h>ixe#!2STCFmpc%Z&pGQrrlW9UDk)KYY@$fnXmgU>-GX&ze1iL4sC
zE#&w{@A#p0S6VIdS*GJo$*Uh7n3a?-FTq~x3k_?Q|I?)^xtR%A0EbyKo_;o`PeCML
zqz{1xMB6@Xam&QXn!705*od4x7oZgQ2vT3c%aMO<+7WDcJ|TOGgfm==Iqybfz*#yS
zmnldU#MZ^iwh3%9PBLZPR=jB~EjKmvs|rz|_I5)rozpqtU-!8ct|i*)ox2i;HO?xw
zkIb~WIc*DJh}WHJQz!c|tU1um*@Lf|HeQM$l|<5+QHG>G(qq`+JaAn!iya(NmDoh}
zN+K1t@rBTw7vC;l1Z7WuWcq!R+10`ElVZ`Sdp#|3P_*5~i^Mt7`}xw=3U4W1YrIAE
zug2-o3k_x(>MC9u#xx>1Y_&4P`c%>P4iaPzyO#FP_;dWU<^-`J!IkRJa(hN21l{${
zuU6(g$$%B0jffu{3djlY(C+UUvT>;5^=sq3a!4OX29~zMXS?@kKp||pou-D
zl?Q5Yiv7J-Ak3F;o7ohwUqV4yPbCDb_*G?7	HvOh-LKKR6cfu41Hv9^=Ol*R1FF
z>d)EB3Kx!rhgAIuxU>Hrg%xk+_&|1Y#(o`UB1Dz-DFLL#tt!-NtzGeKYIa7zz{LeO&mW
zmO|%Lq|*3DjCeuI9SMvzu1K6?KwgIe>PkAaeh?4yLDT!fPpr9JTc3Z6S=6ZoA455@
zDT*Q0WP3^ZWwosy}@Euuxq;{X&rJ{?pWR2(kCyvV;lpk&@G#|A>T&89pP
z_mxA^0j}{zhp3QU^Y!As5OuRSG!*l{@|wU2A}xtuB1E0?os+qW6%Y7c=3~xB+C42$
z6?(=_nueJlOGVQc<5>8|REVzgu-<=aJkaCWM&97q_ss8g=^EWkS`*te!3eGb9sF+O
zQMs?2%F7kUk?zIb?N0s8RSpo@Y{QC1#=#iSx$)kaXB=Xx9%;BR~5VUsxpb
z>fSy5*B(S{>6~7e_d%--ogbYoZQ$N3qzFYW^O-+qtWB_Ycyff$WA{u)P0~lyonD8E
z)etr?6Y&lySQ%Q9Xl>gc<&sP^GU^ss#8q(L1}J;1pFCe+)32d6%`8XF>m|2Rm?&;<
zUg8?EyQo`EpJp)!2g&2jN5WVe-wHYg3zw!i^RHh2m_Z^{%g0GPfL({K;>oP0(bsQ^
z@#K3a77!>fzZ(S%n_l(PI{0}PFH?@x2qajRMsubVYBSy1xN*Byvp8yHOYv!f9y!3d
zjCahzacgxuvHH~$tv{~j+@D~n#Y0t9Auggo9it0RF1CzAa+!qJxJ-Lw73}lL(T^+w
zf||jUo48~Tmfke$TCK*>715LbLT;*a%Hi|8IkW7o9EMLpOmWB#DXU(*WT_#GdtUc=
zVxJdQQAsbS=qCuK83x$?)Pf;O5+lR$SVtiTG)dZnPfGFYYJ@Xpk2ML}U^iBS1>*IF
z1C>`LT=%M9rZ$Z+&PmNnjY-7YoNDFPr(&EKW$D%j7P|cyC2?Zk`*x|3grH2;w)Q!k%A(>ED5U^Xb&jYG1|51+t7&iBc;rq`;{{$4ZO}EV
zeo2QGir<9T4|Wp*6Wa35AD?J7(%gwdJr>hOb?af*a{!>x>^gk
z$X{&DUN^F&au$$%+<uulB=k{vB6W495NUaQsabBhjaBQLVHIv-V77RhG@>wIjD`_{PEGj{`4o1t%~BdO8NLFnSC!wHqtwn
ztPLz{4ubaZm&8~B#EIrjJ{54WQ`FV;5BNcQnXMk`b#;bo#2ykYf+V*SvUhM_O{6tn
z(<#7Ob>Qx`I#e8Y@!1|!EW)#N-Hl!bH1dnBxcd&8ueO?J*8wC3G|X{35nA&sg)K6%
z0nc4>y!IV}E!LQ8T(cwTW`I%fC~Sk5<}n%a*-tccPo39u1B-WLa$~GkL(So`WfzN^
zJ;OQHrfr&gv;5uBLPa;XgBzNbEKb{Phjaiu%tW;t=v=EylV$Ig4GTSOhgI6wD=1}N
z2<|gu4XjiO^XnnCpTitfCg1t)L@Xi87bS=NTQgQ{bz&h>LhF9H!ZgH!$FHYo*t`zL
ze*T#J5!MNQa6K>6iB}g;QV&_?)3ty4)yLCxaJ$%K(_5L!`>AKBy*AEDz+_DS4Lr8
z%w=1~ct|d`umXNPt>|?nGvkC0VQ17^N*8>V1v&0hI;3uokcoD;VZXhT0vICoLg&Kn
z26K40QNgrUaCUJw#T)_C%7Y{5DxD73o7A+}6tmT`c+T=;qMx%ewO^hS6$-F3&?Qa0
zQv&-=(T?PHla?Qay+I-FInni=ZV0P+6N0T)1`0dhZtq#vtBs7v(Xn?tC#v@#&u6H-
zNwSM0v}iK-QPG)lHMp!na_kc4#y%rQqjzIQq@`!3VV4eCFStfN05E
    bl
zyaeMRW1drJGV-1q%2`i!f!70zAlj;hJ7dUnKM{&=SCsQEzmzw=+(Sutd{J3
zK`h$pZjks+!2cn7;
z`|D&Ju&!|Gf19~q-wDN7yt38$(-A&Q@lep0EHtOQ(
ziMZorZIP9ULtKAs4YFI7SPU0*PhG?gG;X^pIPzPI=?ZT`RDcemTwN2ye)4#Kia#gr
zTVlH4;i6G-m4jw~p!PaHZ}2*9zhBFs&X(N`nOKMxwaxAD}x
z?x67ft2-ja?5!&v;fqk)5%X9U%|nJ}l0;f0h5>igJ$v6o984qa1azV)ks+-1Q{SUI
z(_lE-JIpsu=68Mijx@bg#N8(LN3T^B?T
z5|%0$5B&L0Eggw5aMVxt!wo3EgX^eVB}Wl;YWWuexYqF(}hHZ1fiQDM+Q?sS*0-SN2X#vbh7X;up>cLI_jZ
zr$kQU4AxKPD5@~JABy8P`7I3loh&^TcX36bomvOdrs&Z%oGpyBh-&%#3Tu3!@LU~{
zK}xylPF_OxCj=2Ha*+do#~$E0?&QcNv3hNk57qfTNwpAHQRM??9M+`;aed}+C%WJE
z&psqL&vCIKrVwIzt*vP)T{(NZ7okBYCKW4!5>dQXXvT1%0SLB@tdFXLV5p!{bJ~~)qn3jEa2vL
zm=ycEVfY!{RvzmJ1M-j_&G12QEHj5H#*n6Ss;z@1cH`5*kp-tDZe!O#@i-*U7p-w*gtG$BJrl@Tg9)2xWBo
z9B?5msh2Y~QNKPQXjC7-^iVF;8a@n)o_aDj$(4p^dV>vdQ9Cl+mc8Y2An?*{zra}7
zqG5k^ymsKl0{FQ*i>HIA4ADw#bfT-2FSU{-nbCBq`t)T;pQS&6F>UyERYq+$va4CR#~(8Mm!tcnM+Z}jNmyE
zic^h!rFvOJ_t>JMfph+lb8C^4fm-<)3ttq1%(8E#7tS6g2hwGRWgFd7$WQ(W}uSRaFd0?vnEbXi8NSyc$OXL&E!fL|{V^_`_O4
zpp0dc6y2J5A{a%hafH<6r~8ps^b;!bjWyhfRjAaEo3SuwM@E{SNLiCs6V$qkHv%B`
z+qA$$O0E^kD9ZOwDJuLM)9}9(HAcp1jBU;AEdVRepG?I#0x)39+|(}df>dm0q2B#$
zMaG*BdPis%OK8`GVj+5{PQv6PMF^b){p`Y3iN;UHVSMVtX#6w2{JL_Xs!$TXMj~-C
zu^O@Sz&MiiAaf3~?W}IVK$y5IM*q=VR*dlDDFw>;{kNRn1kzg?qpyhqS{`ZSGLEvN
z62A3h4d#9{O4D^Vm~@&!j5H4s$i?*4o<~$N6o94#_Q_fQj+fY#xl6VM%)uEKU#_Rn
z4W0}k6qMFzy?VP>G3oEIjl6)3xq;EB!Ew8j!8Ae}6x&RsyJ;Ytf1
zU;!l@?NhPJ;uYSdCq3@gPZ{SjD=(dslYuE^X?6Qtn^VGQ`tFRj4HY%LVuVp&fOS96
zZg??00t?R9?Ax}3CSD@&!`D6m!x38DGXJQBGY$zNl99=i1y?*i;=41(rPk+PDFVNG
ze_lTHWwH{r61BXfo~%i1?e#FwlXc2P$tn`X)5R1du*ebjZo>;b>Un+1uyl6AeJDjC
z)jd>t$2P5B&#QX2eg)Z8Rm{q64s!secC+q?GEf8M9?FLsv}53U0m@Ujh+UfxZC+hR
ziN%dl-lZyCTBZT@8jGOXA-@zR&Gq}Wu<*)Vsw~??ah57*6!~03oU_?x(4**9H&!9p
zsgIXQ-H7($@^>I_#Oj;KF-y+#k@T2T)&6Lt6rVf|fi>f5l-jEL_s4NqF&+Dn_L)Q*
z>9X!Oi(J#Bf|7I_tEKxczDG4WB*vs1=Oc;1#07ln_bUL{99`PWt^G}nyH5uG#AkKT
zRM4=)`xI54244cKl-H(g+(bK8^sp{q4C{_a%^6mCzJ!$nX9d{fpDO+MZICvtcX&SJ
z<4@CKq!Y3&-t8;w`=_)bDeQfox)ZtR$5l_W(D8+b-Liq?9dM@Clz!-43_I{CAWwnp
zGgXo_NPZd6@w9nV8%6fTzY{(O=*xH*UJq-0zw<(<%}Xf}sG|1NBMaQ-SwZb`P9}KW
z@S}S|?admn5_Nbc3a(E)!G++i1=$Kzr;%+TFT@7;6s(M}j>y%=`{nPGbbQjwr1BPp
z23qa3e@ad93HjMGa>Z(vEePgry93Nj3w3M}Wg=PE(I_5Y5xc;(SX`6+hW{l*fAM(t
zsq|%%%$~m@0#h%{{sQuB0Qf*skolS0G)ftV0%nzi4pQeByF}v}g@(E|hY#*t57>d|
zqKV>q$o@T>dd*lO+E~Y^mJXH_mwyu#TV%`6_fM<)uL0&n^;wF!I=
zQk}W9Fp2IBSGBSK9oi8+YvcqVH!
zIOi7!B`j+hBER6C)~;q5CQL#b24p&DA%>EigpQ`0u~m%b_(1(k>MHxQhvp4-`}brS
zil%Rrl0^s^7N;^8zgWJI*n3th!!?NqwN#4%b+>+G)1)3*JO2>2!>Q2
zJGZ7`S?MB9&d^0%)>R=w?-Q|8zZ#kweudItq>izA)bd^Rr8*MKuS+a9|7r?M2;a(#
z&IgJ@_M?5|9x4BFxjjJY_<_WGD0*U_na5h;GEm}+w17nWf&syW0{uo%w07tDGb&2`
zWM{S0L9&zBFgYINnlHIv8o&6$
zop~YyKfTxpCu^>{K3)?~o>budnE?$#w1>d5+~RkL=lB#j9ZF8p1h1E{qVG|@LU8QF
zcJ0W^ZtD{bYy_3GW%#}rqS0Lp<6)7rP`xdseEK(7<3UxTEYujX5*h6Td8~>PE0v?$
z^l_UmTi9LUZi2*X`0y1WTiqo_w#T|h>Z7oW)FWZKv#qahy@Pi@|E64~WjJ(WxVof;
z8oL}#BNK*n{KTP*LeN5>wb`nzn8wryWHQ&OPWXhEQlI_Z4p-Y3>=~
z7n$Zy+iQ-saE?6bSHWcY0bJ&lZBfXoI#&uLyO&1R9HgXg)slFfRk~!k_awV>5fPZ)
z*)E|5C0|d%`A3_9-=0iY2BuBgUpdVjGXW}+FY&P@ml=R3gg
z5&us$iZdfeMX3W1(Q%;8MNwh^TBgcJ)}-851ZYf1ktEd?uR5Ohi0BOCip&Rwl3SUFyg32Fw{#lf>#*A>!+|I?zY!91bec%e!ApMMqb%vt;!=~
z5=U*iJu73~MNOtV^9g<@w*^gL
z47-PY0KP(Ub6Tq(rz}x2psw`zPELGeV~#@-^{5U=h6wQat#CI{@3%kURmm3IX*U*L
zTA8ZWr+(eSAvQ_s#7*@KK1$jjl(E45T;`nqM1~SJ)`dxD_2aZYQr?T-OgjyvsoU7N
zR4D=i*2tMOQ!zTLQTJ`s-0;Z?G~CXwcv+{`<9ya;(aPoFVJLEE4&Fg=llw7C!
z@W9jN(I4m^fEv!j%TgA(p~UGz9Un^jYjFcLoM^t}uTdR{(@PVVCBnx-D#L1<*Vs?g
z{w}40_L&O1JVM(f-(eHInbs8S;1IWPQzknW=T
zP}|)?;bWIRyBy~_Qb|Yhfv2H{=P50OeYmVYBN#F5rs2$G6JJK3iu!aoIQW+-S*!{T
zWRvN35w+bXmg(NkHSvb3fib&ziOSQ7!le}BraU~k=3rM(6JeTH&)Z+;j
zu>`@`*&8?ob%!g7^?ohlnMjUmk_STNVLW4a4Kr5JuqPLU8shD)4Bac!zs~USd{KAQ
zYhG^=D)GCsvELs^HwlehkZ7&VDHu(O?4qGiigY~>lGnZdZ0y480L>v&duwhSo<P%x$yJ!ur(es)z7Zuas}L>eQk67s`9ALoM5z(rD{ZcBjU^BP
zaj@pBK=}=~a8DCPAAmFDa2-_JGlYc*OO+jA2;Tw^SymCF^L>zuXrWAe!oi_(XIMBR
z>S}z?nUiii_g3Zn`qS0+O)h^Hm7E3d2`YaeKYq;M*}1TE^D~M%X|h~~i_^wUEu}nC7l0m9
z6Y2-|QOlqa_deAB+Xg8(W&Ibr^3BLxsAa#M(rEU#4p)Wv2ve4!GL`&66RzoYEkGF(
z6-%;_&vt7TX_UqP*Ty+EXTmT`IJRxuwr$(CZF^(ewy|-tv7OwpZTmcRKEMBArn>v8
z>Dh-Y?=Ymf+Dc^d(Y{S$C>RD=XOJQ$+D>Ap6W!=k&r7o_i#`q(veTbC#7LmBv42PU
zmV64iI)xqE*iu-OI){z-#ZASSwed3*n%^|IAWZF!>5-O=he~GjYuMl`3Tf7lrfX6M
zpLpID=f&sZ+IXvxn20LB^FeFh_wtg*&#(s(Z0)jVQdL&*wFk~4RBJ}B8)Uk+D(W0ZiL8AL|j1rq`#@YPm{6U>DqWstuvdl>bFjWrX_mY5
z5r>$$Kf&)$hrN}cmk)^&E2&hJbU*K(m1k$5#EFGOlwcU`fiDMYY=rbqJSA=
z)UBtAgRnXLD?usvD5C7NzVkXJ<>pqbs9oy%2Lg6NvELQvNLzsJ*|C!j7x64$v*QZ%!hrY8I-1z|16*&F0Q`L@Z#NeF)r|o->mJ;TXXuuaG=-fex;#1e2HlWaW$>HHU#8#B_SCub`C3eO3x>#_1i_5xgpE_$kz0wJVa%Dfi$_I$D9AJM{+lr(j8qq*ISW
z7`KMsWefAezFeyOTrHBPUnc2q&Tju<2d=q=&+&nO_9z<=X6d_YBr{i-qCO$Vff8M(
zm@r!tbTS8ovS)cYvcA~(bp;pTo+!WO3$Zk*Iur-}P9)23pb+<9J*4Q;anYuy*=cV}
z787I(O@rgL)pYw>&SUSdA-T>dvG`c&j}}+>A+H{D_d%?v)HC6)hlj!Nj%asp_zjX6*)#n5NzdD7rNg@s_TC)zZhXHnN=;8lOo6Y
zC?4XsE$wp4sg<=un7M6&eGTq>$6d(h9O|2x(XiJl-*5{S;5#jvj2=&X{HuD4Oa^g=
zP6k+V&lbUH@6h7#OedoXeer@Zc>S
z8xT1oU3~0NS#y72B(-XHmie3FQSY*#<4no1w+`Xc+~osIyVJptPRjq
z?cVNotf{g26E?|hbas|^=SPD)u>Y0=GH={5abx0|kS-_?{i!m_*%PVK8VjSGsIR%Jr-Ny{ZBIM9P1ssz1no4mo>6xt+RK%DB}QDbg#4yj
z)0AzyVY)yhg&O7`or%YErl;TR>!QBG``_^C98o-nn5PHGq*kJ>taYfV#WrqOKk8=f
zSOOLGd@;KRnL$Gon3fAPGW~c6b>nH~kXTsh4cE$4C9okK5W#*qw>7@>TsGy5Eu9$9
zIJ=YquW&Z6AL$WMv$C}qc
z%lnYf@-O$T05ViE$v1U)8&6zKW||*IxF@#`k|CAUIr$jHFZ|bwnD-$r%-_^A3`o%7
z$Ce{to5^lw^+@@Vg!+KY;sFNU0|tvP-Jz^FEDY*f22mG3e77HBdvk_|K`@5wu6||R
zO~t*RA6V~86P6pcIK8upWzvAQVe5H@JS{=)5LUhto;zpfr_PX>PZhdJfmADqdCs2c
z6Bd4_Z*w2)7GOL4hk;(i)%2`+&*>Y|UtRXa>n!M?lBI2rUenaQ@PpXcWo8^DoWht%
za$dmsppSDL?W+SuXQ9>9C2(a?GCU%9NI$VqzevYW6{LxF)W2iyv{kV=K#
zf5+p%juH|WO8`hRm5yoNO6-glr56?Wo-A7BhEt1WBrt#aWqlM9&QY@X0dKWK3f`0p
zmw<>Q&1Z;n*k^MHobjo`-WsPa!wEG6@D4*Y5N&1bX0XVqks|)$arK5Xy!3DC{bk0V
z4UuPfve?o|=Iq?4f6(yW>$14?Zn<6xH3M
z0#n=H`?^S?g2IRR0jO;fF?AI8$Dzw>r4mDK8R}7kEA5tYFzIBF2N$3uZtqX@tFT15
zN|%7qk5Ix+G#Bo`d$Wq=SAC!1${rm{3FXat94NTtk3+f
z9*nSts}SMBm>HI>@WU>2h7;}ps(MHVf#ngQ?g1RuxgvWhrzXkiNAN-ZB!&i;ykTrm
zo{DqJ1Az1$u@tyK#~$`2EJHn_FIq<*Sc{9+kpNo|q=-wI?}+Eu5E0`X$whV(=6t1!
z;>&fnkJ1SVF`rUE^>sV(W8V;A+fa0s6}}rZCNp(Rx_%%sjGh#!=INendoP83
zE~78FRAYT0Nv9eXIXqeVoFSFRbuay|(+~OK9RQu0rNXDbIy&U^yJjjKCUuYWBf}}X
zE}1CQu`_ViDg1sHmi$k5o@mD|F*=MVw{`GRR@jHQe=`=~c@;zX9#`$_fVGl7Ten(PCjwZ#%T{i|pM3BoDFS0w(hmJ!NzZ5$
z>nKt5lByob0Py7g76^i-|17$WNNkbdj4eb&Q;8x8W>Ak{-`A-*C24HutO>bp1F*O?
z!(khRXyJ~42{(gjPYUG^g_ClM0>>?uBSsw0I$6c?udqH%6ZvgAPO^%9FpsZuKU1fz
z5K)Wv`EOM+rhLN`B38-YYg-UZ9@M)@SYb^RjDpQHfY{}^!h$@aWeJnRNayZa4COh6
zw7j8w>h3IzbL2q4IWgHinHwFN#!BM4>(+Q93Kq1ym*NwkqpXBxmuj%M1%D!@+^m0kJj=Kv6O~fil?OF@`(JPevql?>DQ9TH}{aNIH&`eJw>rX^RAPtg)
zgSh5h4DhVODsd829&aQJF9D;_D${&0cxaDzSa&~-3r~A>oed81k;OHl6!iB)tu5J0
z3k;ZqiW8kiUg)B|r&D_mwweqKAD3X*kF_J%rw0*W0;edLnCnXDbJ!yyKEd!nr<87p
zPp@U2QjoSA=yb6|xZ~~Pq$$$bXvA0Zsj!0Bt3}an9MR1o*G=60!E5!K;-34EDZ+Cs
z4RTCSQCNJY7WVHyNroc6%=a;t^vnRhj}rZ6njI7*QhEb49Ia^@36caxGh*Q9X1@Jv
z9`Em0k3SD3MkPFw0`{g$K$^aES!GZ9aIj_hx-%iIJ4hLyr8~aMF9?i>$!V$GC}&$x
z9%`~t_9x*uiD5^bulcSu$_!p4JmBU#q$&e&7^#mcj<9|4iu$y<>Q$CS3l6)y9cUwk
zV;Zo!08ex;6l|l6bWzL7=SygC;Es!%RU0v~zhvGwzlaBaY5+ype^Z}Md!exLyOewpjXBTBE2pL)@E>`&&jORM?$4NC?Q?Z
z50UrrMZN~X)DHWeRe^d+2qJLYz@36Ojb+Y9%zevg|JaY=%SGPC{qVy;iHEWHQ!u%W
zs5}@@Pb5A|_}^-W*T2i5AN;?4TmINIz)L9DFed7$^B0DWG=tUBg^}s`2oYHx=tQO&
zVig#SnyRA340e>{b|0SV7lI<|xnT@tQ}$^|K2g2%MSSGkL-X&8(BQlxg0L@vE{0|F
z_^3@~5N|*<<(Ax!;Qw)2ikN+oX0z-MizL%exATHg{vr`YRBBUkI>^Aqr6F6A6F=>`
ziex%L0GV_-xLvz^elBoDK~s_j)`E`BQS#!sr#nm$P}}M=)NTKqX?f;&x!hV*D>t36
zL;E2TS_Gv#O$Nqa9jPgA&bidBv28+rUkdDRS$sQEQC!!sJcCwRHmnp{GV$E_7d)D<
zJU~&UZ|Q+Kc_UZa95Q~Y7HOcG3+O74%5dfm$+Q{c+i7X_Frl>t3HRgvz7wO%^dx>{tM$@N3`P=baPYlY;HcBx-KujLE`p_75;#f%arg>TCea413)=xC9
zS|7g@)Ys`?x-*|yNYw;*5=YcD>8yQNUc@-#^D_;0cBuNU^^>e5Fc0L!;J9>?zib_SR*etT+_9*NB_VY*JN?sz
zGJK3k{?fpDcBuB2Oa?N^EXXuVt5Yo!!;6I99F9jWo-SUo)F`!u)7`M;0xCZW2XDd3)Ss8*qz-ZxA
z0DpMt0
z1a4NVJ$rGUX<7i34vnQaCrvobYfaEeLbIX0EGUB-g@e477I;-Z)$88Ds{ogD%+Lpw
zqg-`2&z8(&4_ajG@h^ixl}s6UOnP6zEI`qGWd13wgyM;+%CmzPHAWm!?}9`oNathu
z+Rr{_7;n_8BMvL7A7l+bdu1Vw&bJeM^!LxAHSfLnQJK!-#TrQq1|ymtk3Xe^oPzVvk@(&V
zvj?1lK96c2#}x;ZN(rG#I;syBvO>+c6Rr)<5$KN%7~(~Fwv$;Z9sB(ec=M}V(Tao$
zXtMDWZ)#}`0I$21+65log=8I+YAaO1wrPwAdcOK6Ti^U8s?uu|*@kyUuY-S2w94`M
z6h|3~V!fzaqk<=qGodERE%FM)tW*_(;(BPx&Y|?77)~q<#ba4d1ng|PnJPWd^jfWJ
z7+mx?h3gPBy?Fw=j5EW`@5M(u+KCVF1x5fcNss-Pm
zte+d2kFkqFA%m{x#<+N=tYflC(j?bYdEMVKooW5t_d6Bg5eH{3?>rc_(6@Q=BnY}A
z$)H^1d%Ukq(Zc&&m25%Tjq;nro!8VLTUF+y4)SauH|K)i4y}zD@>XufYG~QDeQ5vJ
zTVO0eE9*|LOCzRgG#1JrvV~!*;p>{Qs>nn->NUfoKfz%F@2-E_@5Q=Wjl_k3%gOnF
zJfOgmI=yg2XjC!WFMR{KYcmw9(2&DgmWk-tEKl1}uoBJjZV~KW7Dc(Nf)FVFy>K${
ztRJ&-aN(}Ey^V4`5SFl-Lhn$RtuxGI^>tNo=j{B@dUFJ3DZrEV7Esk_RlGMb3D=tb-;k1^
zRxjWH@1KJ{W}P!{(Wu-2=}Rq@9=VK$Ws?^x2EJKuJ-5}v&e@o$btpv&mIv;GJ4o>mrY&_zw)8@fJxkV=g>F2G6-XWw^olu8nYhdszg}{f~7@#mti{tI8
zLn-IvXtn*Fbb+iqvL2?iHM1J;TL@ZVVQO#IW5{tS_bqK!%Y^LRJWUyN%!T{VND=#i
zxNA&llH0O~=Nx_5v`7+_oupMz&()eIh__Fdo9?;{W4rf7r<4-v>lD!nWHx+BJ-7=6^4caqnSAu7BL&Jqqis|HUeZU5hW(Ar{E6ZIi@E^HBF$Rbj>y+y9Up)3TrI)1^`?Y09O0@vaE1gpT$;z1(>%l&!R_XLGw#VXSVFUB!B
zSEw|L1*;P3gTyd%x%=ocaq7d>|HsYk>bARmiMcsig5Y>kbJ}SIp9zHBGf^Rxe!2(s
z_3jdz`{!Wq7#kY{z!MF29nF|WxU3|>PAoVx#sC?l27P7}(c#(hhU&=r;1}{FL1~<-
zhw<-e8s*&>koqW+iyjkf@h5gU=v6EJxI@egeW?`81$Fg9lFgt4vJJlX1}X+NGPpUF
z8!QyN^MBg$Q^uCrTKSsmQ|Svq_Uub?vZWe*3zV)*;d`z+_@R?AT-9t$i^kSw
z8-F;+I~(nfB`|EU^HGG>y1Z(wRn3yOV^@>d^22ph;C=X#
zkyrtEktNM~@=7mM9`zt^5`${#A<7tbX)S|PGX4Aj(@XKY1@92&<4ETem#yGeC%%K~
zmgE}_AT${srX#eAYRGe^@31#%6BkP>9PczLiN?BoVA1{{`gd#=iI7BrxxK=(!*f{)
z$eqT&d_NRAo0}QoQu%~ct$RHX!*2GU2>)8Mz3E;G%cOk;`)qoi_1N$1Eri7JrvC@P
zrWUsrVVoPHk7#9lC!7s{5Y=r?1!0LvjOZx{$gr$>-eCyZZLbmK=~4s~TcXr}Xt^f|
z=f}-oALqKYCKbMhX)6f@-6vh>|Mu+Lo;~D)%g3CA@XZ>|?b-$H69?AX^mA}r6CokL
zL{{4n7KtyqS3Hmsuzy!U_+M4Wv4(_xz-$2Q%EOvKar@ndm!Ng&+mQZT(YC*d0^T=9B&{
zM@XAE9o6rxp{1L=Jn0Ed?m;RwH*iJ&y5n=A+jRtl9RZF!N{YWQufKY;%v}OPQgi(?
zb1&m1w${
z&Z?B0lmC8FabL$9!O(_TA>PkJDZD?+!1%hgvpg-6+U5PHYOO{G#9x6F2L8YBtOr^-
zq}vpM2+9zqP>Y~jvhkM+nb!P!1f3N7OeX5lZ~G`P@X2gOlUkFi3=k@
zCr0%#`@@h4kOrKHQA~iMEZOReomfJ8f~($7$0*g0e|fbC3M>GA!Igm%IK%{Bs(mN|
zdruldftFqQuE4#toWFVR)PE4n>5SQ;rwUa<4lIA#nKLS$-m&5PC|5Q#AUwF*<%$O=
zIWJfzWFEadEi+YGx2z{fW*Wj?grU%UBAp#=^YxXbQXPH{a^1E?6lYq+>|TzuPx`#4
z8`&4v$RnJ6OdEf6%LcUr+ktrf-rPW-R`6M9!LM;I}ekxQYSbV*^K;w|AS*DFC^!te6xgiFMa?R*!$$E
z6dg;Q!&t{RO|-Rxl|gUsZG_6;$@Os|a7RJ;kfu26rWncg-GJRD0LLyUle}mDsnhs*
z+!At);%sVn(v8ynOZg5avwe%uRF=8#-%Cq;eDq^N^$qu?N1}vw^Eyo9WzMya{9NbT
zntHW+)Kd4#rE*}liOPb7qZOk}^grnlpQ9CVNam2VT&yb4U3Y~Cxd+0-R#l>K|A4cP
zbN`n3k7&c17$q=is~)?F!6OUwMc+6>vjnPNgJg*TD}gP4J)^hPBBGgWWbfeTW@F4RxgD1k59Mhs$}|~
z%(|b7CO8E~_*pAvf)A6RgNo4F{VPd2b5P1%LC1^(
z6&3+PN-RA`3L;yh)?I?dJAj!m1=?cfEpZ*
zW;eyZN%j)n;*8SSkI83k!YXf<)k|ZzgziTy6L~$V0A=bn=}ioP$8!78Z7Z$H=(`PY
zarUXE)cuoMV})AGPiE}GBUr;+Z?6=D;Z>qR8}kVk8mpXvQ%oR%<>8-YUnH_62^JnQ
zg}zlYNVdVJ`w2UGT^3H)qvl0coS3;4{Pp+CwNHB>1MlNMla1lMt7h$A9z|>*9f^a3
zKhK7lZnbTr^P}|U00vKT&^;4L?QVq#+nxMPszn8lO2)nMZNc-k9%5~As5qoi>pvio
zq317}(}xORN?PWU=nmXF5)91dZ<7m!N|;qC8~8Ot%NhplEIRSTS;rZz_102_{k6-x
zl9Zi8j3~f@MccM*+qP}nwr$(C`?amtwr$(C=Vg+=nB;F}S4e7AN$J)(t6s0CeW8ED
zRQ(ScN4qKg!uDMVPO3ONd`ZIZnNUTj8P-npE>CbFCszxp+fd3QO5o}ob8l87LVajf
zI3{rsAoAZ#N&7uz{*+TJ#xM!keKi|XAlf95KL72#b;nfi0X2$}c$gD~Ub_0>xm*|1{2$k%zB(
z!kmz=&0Y{Ij`@{>$)pi=0WTC}nK+H4udk5=h;2LQnRuE?7KB->MG-Ad&n_SpUjGAH
zW@Y9+=4>-at9X;Rwq%zriK4k*La;!hlik}u8io)Hu!2JOq@Pc&zCG8_J^MfI+>Up_
zzN1o!)`raR#aOxv65sBFnH?e^tD`Y^Y>_-4|Lu-EZkBz}y&@#5QXZ&69`
zRT#$GU8>VvOc_#GCv89S(kYh!2hCoaRN*qFs@-yn?~E#8lNqD+YR?;E_IP(cs>f2)
zj%p4p2eneRPJoW}20T6114Qf(T6Pe(!f@^o)I{o_m#p9ZeNZ(By@cfyNivTQny~MI
z#$}Ss!O%iTzq{u;$w@M|0V>Z>?U&CW8^8%A#eKl>%U>YKJARl1LgKg3wtQ7sX7c43
zTWmVvCXO$uKyA&bP#KEFtKXhd{JPOvgdp|2NP3?eiQZ<<+XmTmMkUty$_sgN>&)t7
zoE`8Y3=0P0zhbmn%JnJ%4Q1?%&-G$ct)3k!fr}fzyv`>k_^=z`D{2k
zx?&8pC$Klp$23GXcOp|v6LL*6E?b#F;8`+QWN}x|af8c!MxgpJx%#j>m3$(L+0DkDxZZ8P|EL~ni~!RB*O=-@17tF+W!eo)q=~+w?wQN5B4h|i1(?c@_
zj_``%ll%Va@X+v!_RZQtd9Uf`s30;`o_os#01=u?6mLmdlPM?l=!mnK>HgsOkyE~$
zf9urE$vr1&MCXwSN6V7szaf}H>O
z7a#$Yzv;i|ahMqYn;wUSk>x)=oPT;8c1CvA|5X2v9*2XK?f;?2QOMCBC{DRguX-@Cb?3*5V@?cc@?0y!mqGtovD9Ua*u5Nig9;2M3R$Az$tv<6b25$W1ryYornmAwq
zKory@RRmB71tX}fu6X*?g|zrC+YJFDsCqj2oqj}n)k{hIafbyE3qYQm
z*|}oB{hBMW&f8gE&zu>W-wA{>F&8p7evjuPRc7`M
zU>=(Rx_hw%ZwC2U#1OYGZgbHu
zgkWH60{Kz1|Mi}os{U;lhnVm~LsnHOO7+9$vDU(z({`^u$I-$0iTvr#D~Mo04+su~
z?Hd~hGVs7wQnV&EOh{eMw(@S@y+&sAzDx!ln@Roor^gwv%2YjOqgLwoOCV@KY?&xV`>>;nMU01BWBD|_m#{K-t&H5=JIX+!hq+zhY*82Lzh
z5s+s#-rEGv>FVx5M%Xhrg?MuL-TbG=iA%&aFflv3`(HiIP5v^!Hg+&T&M#9h<$K%Z
zZ)^TRzOdB0n^3I4N?I)daRDPvH_h0-Ee6HUFQ@VEAnA#PC1sU8Xvy1c*pGcmPikiA
zrT^%+d<4~9n@ZHLcw=UFMke%}#LiUz+>^fUclviz@XzKPv{tqT(5~ANe)7Yuv&lPK
zy?uJz`YkuKz~tD>2Y*kSCQEZY*v2OC+S-pJ;I7vlzt=9-uQqLfo{ox$rbNVt|HyGZ
zfugl=LSkkMWI$vpY=OOb`7PAw9UHJrbv1zdNSh)XQ0Esj34j-ub;6)FRNu&aKe7Q#
zWAJBpqD?h`SO0FAKOq}H_ZH7WP-YH34C4I5+M1+GCE%(E$iUb2ko*n3%ci)c+Xk=hE?g7WLaL
z{R{ioPYp0Y0KI^4>i|B1&^^FYNmwB=-D1at5`~g{TauG3Kg$H>$mqg^`bIqQiq?q0
zzsJ5<9!rx?LMupmle&MkQz4cGVDQ1ThO)=#VB;*hueB#V2A!^`VoxFsDz{NTYbr?I
zlAH0k&_u1)3vtp#%UOmQP{8v@BBQD0kT1yc3-#EXs6oy0Q3@`_xu53H-Gy6PB^}fx
zhixl#ywx49?)6M@|6@BST7}uGLokClX^u@7ckJR#orIAjBpMZ6vO5nw7VrA7y~l(|
zFBL;fS9kd76xO2?1GVKLU&YX5gdHiR5e;2y8K=7gKkYJ@rLlz8WSx3{t@w43#M3Kx
z9+5>qaOr~EV2J0Z3EMEg;QL6Z9pxiEX9ivMH0@7(i$E)-+
z!iY1UC^SCjsRGw6mJwN!o`BQ4)Ab8;0mFO1$XeT;Cgkp(^`4253Bsw7dMw@;gg3?d
zZJ9I2F=*9&C0;wcpoTl51sOZ2gIWu4uSb1uoCy9>fcul
z?-$m{D5HFx7`8~4&Mc40itVE5z-{4Z&?S4&%bsd6A>WkB5HXfdkTsKC5M}qb!SAcZxz>2~7>zxelXf8dT#20H=Bxg;;8+edAboqe
zMS4(DBt8c_0X4~90lm+(FU3@{_E%$~^aaz=OrO|vw=R9LA7uv&nftCiN&=O(!QJJy
z@hUhd#DFB14rWjqD{NV*@>B38AL39s9JAcc>czoWO!dov=8MI@*ccsOaI^fLB>r83
zSTPkbLKbQD_BHoW!^5NSR8yeAE$$RAG&Qbs4b}raAMEzyer(2>oZDz{J58233Ncgh
zTP?eg;?@{FWtespqd?RYP{e0KeFxe}1Zl2DiX{4JJRBE#v%Qgl7IrmvhofFMj}3y`
z(_W9&g^+iMs9sY?cc(?^etURVFP<~CnK|imu;-kdWa!^_=5`rM9N~fnlgv!yTJ276
zA#77uEgYVp-PyeA<{2{0)SU>#gV1P0Y2ll03D;BO9Rw3T!myvKx?Mc}YSihd2ExUt
zbW!&CQ)KAT$K`8p%@>iJ);YFT?ye!tH6ELVWt(9KZp{<6R~`y$R{C$8j_G8nbyC!Tzsf)sA0T$J2p=)|LC$D2vMZ%iNX(Sc_Q^=w-v*614*b93Q
z&U!Zafs_=cn<$3)D$_mrOg#T$u6^GaTmh#2-RsMWT}2U%(m#L^v8CaFZR!@!5+b=64euJG
zl<${QBazCZ|AfHPt+-sSb-n1%GhaDF79yO79DYQ=4!5spJ%nkI(yd$z7xxQ
zh(NMYi{w=OUUHln=6XFpBXT4=guZL;lj@ngo!oD9$QgattK%-|tJYCY4=d0k|JW~w
zOA*sdnDz=TIw?-)!$=N2Unr!Xp|HOVaJHf~DYX*_;zc&xI~+J6?i!l+6+u8=M`w^7
ztLS!cVMh4h6XnvMY<7+j#bF0owELZvMIwmUT8#usB-W&uyeZnb*7*=S?pF19lPf3=
z?|Va)XmsAkA5c~!`uUnuBQcuB+zRvVv-b)#rak{&xaKZ7i!P!NncyN@E^OtUpzOg>aEeS!6-Dhf@T&=IK
zGlLXLnVr(tO!L3Qn>3|Z=&VY8>Ju)L(BY+f96-uG#EKytmq!oqah@~jt9bpc+>mk7
zPb-Tass6U`FPnB;WreLu4;N@>kFmB`A49$t4)keV`OhGu$97C999R7~J(7X~1~s_V
z5W}nEW$q?KA}I!f74yphs-nG1BZ8T0VaK6zn^;etTN(`*MQ-DuJX65iv?R(j&?6OgZMV0)RiOwpXoBQ)D#2PPRe(1aMKuhjSiL5biV_-
z1DTXR4m-p^@o^(3xm=FGj%+okwcG
z4LoW2jE8X_zF^eX3AHd=8f=Tg=(sFfY41mOwt$99cr7@^wTnl9WBZlzI*iNO2(>F!
z?_w!K1t45sXJ935S=7XK72)9OrpPb&P>a$#jtBP=+M~#n0r<)@u*ODzyJapJzw~*p
z%#p0-S86%GfW#1L^+g@FXx8_;B{jV{m+yQ*pgM4z;)8x@>V)Oenhv42S;j`Xa{k32
zz)@a^!iz!t#4PafiAVk1e`$426`Y?Gmm8pxG?-%>&RD8q~+&3O?S8RI>SeU5{=QeflA`Ndf(<@e@(
ze)?Hq9_?tXTZ50{7>!%RYO3IP#s8d*!a~Fj&}-sVIBDO$T{-vy7fW3G{tiIQK<&B!-u9?vM_2g9$jQvwTJ!!feX
zlN({_92r40$o4j3$c&C12>HXvUTIp!n-2F}kG_c<8Qj90cM%@ariNuuK|l${RbS$3
zUb5yQR#*frC%R>y>pi1+jj20IDa=+P>_MgFCwcuGe)XD#`HJHf8b){+h$APDFLSmh
z4`r6j-yLMluQj}n4FQDJAnP{B)$E^PPY#73?f33=ti&&N5+cul27EzIE678!0q|%j
z5EPsRMpgPYjamFTLByX#M>*hAVE6BmU@1DOt75zP`O|=&6*iHG9-Q!*2zG5&nAq7+
zEjuD%5`NTj2+613QZyvlN;vIh;cRI`lHG-8Oo*mf&C)Uk5%cwXuaN;}=cGaPojUoi
z1r1G@`mZ8uA*}7AzQQ=Ga~zKO*p|sVN}Y;
ztGsO=Ms)h)AY8zFb37kX3i!D$V%|cjU^>S_+o#RM90R^Y;
zwfAcbcI?5j+tIo`0R8=AMT@3W52{QWeh*Mi(seH~q*Dc?xQ$=?#d#z#Ei+DJja?#V
zq0X_4$M`o2mTlnb5t|p{Vj0mM+91)l{#&-y;SL*9K6+4J&zM_Zp95nE?q0Cl%%s}t
zz`HKUv|E~_1r`OH9mjs$7GmuEbDi9U-1jYR
z;=h2}zAoD{x#TdaC(>hmW-pY3EWXaW8?wK%zMZv7kYFsYf$4{AMBz??I@n&iQGlge
zk|6iyX;yJdd_NaK5rGTw|1ikrJJ{Cp1mI_1xkDx28eW)Hl8B@6%IyoMUJ=h0136&q
zat#y;W3laT7lx>zLQ;?!-re3`R9<2~nk-K(B`KagQQ$B{opmGa#e;Hqf(hpkRndW1
zTd!VPk`K<{$>vXN|5UYc_)eef@&X3LX<*Y($~)gHbH|4>V|$af$%MDOneU8y{%{CH
zA+}|iw^mOT-YOKd31hp{-5yeGGHU&=vHUJCA-#vbk^uq1@+N=oBA+!R*|Xz2eLzBy
zgn>wV1?U>4_8lD2SiZC`T-&A<#LV9
zA-qIs9Q0!;zV>x_z6SG(1%?*;^2B)(9KYG-ZH&}Bo0|=efGQvlLwQ|Gl=7GQ@DcUgk*Yuf3##hE3v*Yw{Kn--Hnf0Ks{3681peBNF?b&
zw=@+{JRh*N8q9-U&+f~qb`BT4$Qu_a@f@NX+aNTW#4&B8O&=*oU(>SWn$TqlYGQUf_nq_T%kp*Y(-~S-^BxCyOqnk7TaooK?Oa5(
z>xF4$8V?xLB#WPXE6g%b&vxb5MJM-jp(Uo~wyRSQkfo4`N8so_?s!!{v5(oJ`L?{7
z%_*#^ZODL;x#9F1Ii2`YG8g`BOQ8za#JBQYL2trko(m7|0zuDNTVjGc*lQ!Te0|-0Yzsnjn`ZBwLG{EV7*wA4oc9efG6S
zzZi4AW(I*tiMEc=<5Xq~s+O2oESBAI%8cym*yNI{X?u*wnu~uQtt!^`Fh26g`dm~3
zM8$}{8O<&n*ucVHge+ud*GC+LAZZaLuU66HBSS%Msla|KML3oQk?{sQBF!%XF&+-r
zgys*#tkSBATBoYQ(Z)VrJ*Fxcf=-`EwDl%)AV%`CqYd~eRdslF%Ngonc@nG`k~TJI
zELYOhPMKc^*z4pH+LR&S*rfd_!R-HZeL;!?K1G8{-xATH_X&zLvRJJ~SpW)_F(V&w
zY#BPX9AyU)iH1?hrcz97+Jbkpn$DDNDJtO~Q+CYvHjVH5Hf(0+q<5v!dZSZ}{JVrL
z=dN+I!$a@}TyqY{G$8VDM-vTHTF`Xed48f38SrF+7oEOl8z&SNC?4REdL_L?%YHL^
ztz_DyUJb{=vJnffL}@XTvtsT!0gdS~a|IA078bmww;3glq_6D*uv
zfPr)yg?ZFdL5LE{%^M4+BV4!kSm-h&;5Ni-JJ1!^_34&OV*|5UtmU~|zn$kx!;m#C
z$x#9%X`jo|78V0|JE-&9dGK>JjH-pV`N2Cc*trJ1Y`op~?<7=;M2nhF1CA2)#!xg?
zT0l(^zJDlXf0JHq)|lFgWkm_Ltz@>k|>57Ko3FYqi+dxwF#3;#1X*KSh`RmJ!d
z62YZ=>k#-M-U64hqrqcwjW>tml7sX54-d`Wch8bmceuyl=WMR%O?k<_;8=*5dK+xj
zrRNB|+z6xufc?;M{;ri!8{UUDceyjFM&{qce|cV#VI$6WRR=E)*P4X^w>-pd{197+
zK5R_G($;3oE9k?0xj&A60x%{g_KLi(E9V8~o~qU^Z9Q+(3NtB0r30VSZg$D4hMfG=
zN)4n8FM6SsHpWfA1BV0&grAh*4vf?w9i;@Jp__c2oJF}!42pPG(>I`x{TC$N04V})566$Z0}7h^v;yjxM=hYXH$a)|Q!P$YyMJK*!qo&eWE`$hXQ)VdAjbR#6X6
z@|95bTNp%Ws%9qG&tc?Dx6vWV_+7wkbRThXt%@{sE9yRfAMALS^uWv;v$naLz2rqg
zbu~3{{s>Q>X{Z~ghM?~gH??1}AVf<^Zy)JM_b|$7Ah@KQhWs##q#V}0Ou9>W3b4*w
zmCfo{X@W*_H8_5lT?oVER4hJEU~1`qfVE3=%|(mrQx(1>paBG-(0N$CYP6PrVu`zY
z&&5T{I<>??cSslg+|Rb!3)CVgDn0(bVGwOfH13mKg;3jHo2V;zPb=Nx$;JAiJhf1_
zvl^Ep||`>LdEK~COgMEet$J*<+6};12U`ULx@6x^|ruBFiQE-S*qsY(fsjF
zj`+srwES6~g{!~&Ti-;^V6HA+|N$h3C|7Y&YTZ1<7wVBhbI
zzwAFEi!*(70mN+~qNd=^?JT=-Z1Z>I@PZ9~Bd~!4qjj`)+3}tDq*I)t1smHy2nWsT
z@)ddIL20o3ZMz%(eg|`B7@3*#Z*#jS$|nP2d}c>*#G{E@Y)4WDFxVehFzofK&{3SfE9`l7s-b?gw$)WcMl#P=f68zQh!i9-AkrQM
z8N`;@{-#tx_P*L(X#zG~f>Mbwm|dM^>PK+|N;@>{Rh+V%AA!M*AOFPgzeDi(IRxvY
z!GRXx2cbJi!X@|1)V8z>3qmNMu^Ugw)w96NXK2
zL>TERQS#UohQvq=CqL+i+}6#aH?MbH{F+3daMDzK{TP@<#{=D&S)n1T=m*O=kLQLb
zui0(GAZW!O@F8K``PyfYt^|KQS0`DhkRNBF6_*+aWR~5}?YZEs^**Q;`4i7~-7%?t
z(Vnt#KArvAX6Q_xawT_EuG3VsqC;TXvyzGmni2-26?p8<>Tq
zZt$4|U50<^^vkn)*eUgdhm1uFu!P%yi(D1L*Em&V1@CY*3
zWaz4n2NfxMNTnc#dW=pW*Y_Khc$xfoWP7LHlP&W-CUR~6Znnp;u%_s)mA(Z1goBl0
z&M}u?5>t1Qp`PvFfWlBjx3>D_8Z0&Z@>qOx>T3@c=`;28A1dLTe`lutx+ju=0w?UO
zj^M51AY$pwU8FpIlR#VXZ(-${gq2UxeK3@9z&nAcF4wP^dzQ(Ry9%-PJBpk
zY#2!H^I~c>=5c61ZK#-4taawjiRVPG4_A9yUIfhyMhUAv$dM2x-Z<
z{HGo%4Q(~b7LN}SVl&dy-5CjJfpuE|g!P}==KPaVOMzf^I%{W)E$fp13onA6_
z%IddNHFBgqRa25Wa)Rilmd*;hX;Ub+syvGUB9Ebsx73`RyB71z;QW^;Ao&1BbCY`T
z3gFO*%2=Ps+fQcq*M4rv>hjUgyyD_w0?&y}$YCj1B+)~Z3^(5ohzHw?i02PGI2Y>a
zOzD7V3+5g}^nz{om2BO2J}0JG2|&~qP!6AxEPvLgIXTSRYI^36RuM);af=xq;FERr
z2~|w69T3L&icWHG+#nLx8}D>20#Mi5QQ#C!TvJ5F!9t;2_ZwUkoHzaBl36z+zu@Hl
z&asG}e*&Dtq_gba%;<}*{Chz9+lw@*>tmORa=E($84c2%|KVD99y0vBR3put`YD_m
z4g{Bv^2chFp7Ih(uKFE#mRPAlXSkj4IKz5=Ilf!u!9cN&8grP-$aC#5wmyw{KY22&
zT0RNw$ouC#HYQ_H5?E~6#kk<34iLY3ZdNA`$ZF&1b8M(Z1A5BVCJ{>^oi1JbGMgJkXsPoeB
znh+9qKeLr8Jp+#S;?a&~p#_Tm??%;-O#?l9s%&*a^y5-$c6uzY_0BtuCDE?kspkb-
zbx=R1bxtm1*_Kr}K85d!T@EWjnm7F-?Un&Q3Sto0cFyhdv$Q8n1^1!0Lt$pfGT^(H
zp8cJ%sd!t5^im*=;NDOzYnFEpIO+$=Q>@Zo?4+H>-Es0^z^bcjXQLolydy_OlU;+r
zrX3p-s_$1Xv3M$HC*Poy-BrXaiA#p9FX~m$?mb)y{%ii6t|jps3}uX%L{N80KO2tNj5
zn(MG>{r68iNaqcn;_a3cq4P
z#3{^13m#B2P2s}jozzO;CJWnV9Kb{?v*+b?X$!|qc8oXs%6_#&xLM2dsy{uRm=Cb7
zj@7ewVIgP7n2eq#5bh#xP{IVuXqd!yIgBXV&m)S(QK@`PNF8odL}7zD*UXP~Qg2>s
zro51hjx6lUbzg-&!{rvv3EtBTQ`4{0j&CJWWKiREVe?I|%%+gdJU$lp@qNkjP>fS=-9|Un@vD1FP>u#~pAlt1Z5H6Z#t!s<@F;>r=Xn~E
zrla1UfqCawZHN~l!~_qt>Z?me|D-^xK4h7^W*)l@yX8W`MiX>zDwo$m
zW7d{$rC0bmc+w*qDlFft#%T*Wo|QgLEmK-3IHdP`S-&V&uB!;1k~s4ZLf)fT6NM@q
zW~i4>w_3()Q1ofp@{{@2-ZCgf2x1P_wbd}MRFOo4zcvq!97O$g!}S{eV3PZXIB@Jh
z7-(6MSOM9fskX6eLEViMm5LP2^FR(FbTRXeolXPo`=*kxd-RLO56WQHjQL!^3vsO(
zAu0R&OLguoPT*MZ=B9nbGUCx(s$&0qeKWj-3jV;UU@-v7R8iB5yw@VcH%4>lx+QJV
zDIoDp`M@5w8{5Tn9_@ej_)JjCQGYyf(HlQ$+T@YXMlD?1?8Idg2~4!lF-4Qz>D~|~
z$G|~AvidbOZT(X;rKbnE;s)lz(#d`%1gNygprgmN9*-D}M46#uFfCW|lzj%@W~WX6
zC?mG#c)-wN{k#|;w%2L_z%gfm%V5%5zdyQH*NBho%kYMgWjCpj*>pegB!QdIeyUw<
z)L&?-yv?#G7<1sNr9y~wKQ~b5dh;aj+H+hr9(sGPo6p(OallbonS1HzWTX|r0rjBq
zH6`r!s9Z6MsC;;7sLFgmZB~C^ipF?Z2%8a*s13wE7H91ICM
zZogKB`FTF@9zeZ=(b=k`kn|A(bEB~;0{m7gQkU-CRi#;|NLUuY@Q&0$?G|pf;wG~+
z*x@F+ahw8u%({Bn`p)(PHM}ntXgziXWc$Eo89gW|g=1bqV|aWcR)rg-@>ZHMAT_VO
zAQ0k6VaMt%k
zd{1DIpvFw$gS@JS^6gI{sY$U8p%VCYl6O<-S-XoWjP}(IO~p)fw3NV~v#8vV2S$~K
z8Qb_;wn?+M#!YA7gSHl#L^g
zNcc~tbc>O{Tro-DGGZP%Wa^Qv^n_z-a_agfLHJGv7zF-hXxBJ>7DO8EjGPX74od5jzygO9Fg9f^$ehX@DV?%ZgA#y(KSZ>CGoHG6^h@sRM}yO2
zCS+->nJc9tbQ(#e)yj);K!0>ZB>=o0WTAVHEr&z5P81N3o{3|7Z^`6^sYKt5`#u$h
z5M4!h8~DC)v@Q;ZB*t4EDHg<80hs_BTDnIl9%V|EPt%+yp&>jl{DQ{XF;y
zyb}Sj;YM1&LtT(>8KVLg?|rS4u~q9YRsnuj<0Cj~r`$mGE6Q3A-d~!I38!uh%PYjb
zNL8F3Yre#b5%ns2hely@bAPO!c_Cx7s6(S-+sLR%qxnYhoacltn#is3A{_qKwbrq>
zKbOYFQ$8%Ysy#W=o@e$qAL&>knB~tRsz?a?)kf)$NbI$f&e`q=TmOYbL+p*cdXt$;
zXGj*iEUXv2)MFy7OWAiaw=GXpl_k-k90@(;$1%+BGBDEex(>Cob};{H|F$5eXM_d(
zfH3-p>{)51ii8;y^$zqMy*Q98riD(+4J~~4vDeP$*Nl@9o;GeHtM^(vrJ0H!E$Th0
znw@7_8@mD3Z)!UqVXyDyZQ{>1A=*faX;gh3iv$D*tcRM)MxE71u}(H;(9oXR?ShnX
z_oBtS@f#^**{dvM@IFo>$(sX|3mBG$UM3%MS@$joNAy=vR50KhGBc+0q5&s&
z8;&>3!X&2p?#w>c`1%{s`ys8^a45Xp6G0M6nyg!Hm;7vK@lJ6cH;d7X%G64#^73A^
zTHEHMk2l+bq&S#`fZi`}rN!~9$3NJ$XMW<9cI--}970meG8})@*0Cd8VW6?1+y*>Y
z>!np5b!Lb4kgV&q-r425mjKSEuI_~;hO_Vl6&d*2m%ovDhf{Cs>Cq(!FKV&GEtCS2
zL~U56Mkcci@S0EF`*aE=;q?x`QF@4CYUJ#f3+X?EDU4d1{teV+4Z-Xkui)k`YsdB-
z$*|I}xr!vMlWUSTzAU<^Lb%BBRXKtd$?9uo5l`CGy($Gn%W4H7)5Hc&Cc5=cjT_na
z+O^EE+)6q2xc`W2XGl9g*mmlj*jb&CfUI0}(JXnWP75Ny0GSt}6Qe3f8%TH3+>ie}i%zeeIc9Gu*{j
zy-waY@Amd{Q5ongA!_bj>51;K;YnhWaEKVlD&+}v>D-eItMsgLt1eDt`@VszQHv(UpZx
zrZEj?`XyZMnX+XA8MGe+zchR1#|xZB&+YWl5U2&ldxT;%&1XpgTlf^Bjg^a~cVl!L
zfm(cZYmmIzx>o029cQ6}#KAFU7{)?yf*VYZm%u@g8zwz1Pxq0a4uLR;t~{2*uozx9)c
z?q8~$?XRp@C53gqpgMei@E@5ppL5vg9Og9PLzlWUA41l>M1}^1S>r^D2WFh#%x07Q$Q+%J_MJe
z+@KdQv(=tPbz>|@u!JLxx-ldw!Yu`*P
z+i7_35Yfu0NywRBvgsS2WC%+>saMZDPMg)~3W*RCf4;alGgQRZY!HYY6VPA`ESvdb
zoj>c-#cLi=!9Jr!MJpm>`fL|gGmMmTC3nd3ZlvnMI*6@uhAL54ZQOn-qDfzG&`{(E
zy9!2DladI&Lt$)c+*L1d=Os`t^6O{x0@_?mH~ev3ecmG$%!0GNChJwE21M5|TMZVM
zM*eiz&UsOj&=-?&I73s(pSvD|AN(Gk9@&)?mg9SsAP<@hdaEJ4&>Q0*gmw1B{MWRRD_N^3*U%eCeMO&1)2++}W;7gNeTe||lE
zozs{$t=Bz@zS}O(HTDome)5>h4c;MjPq0IANxKfBf{)btj`_SJOqx_zAx~gKuoZO%
zyfOi^bf{{e>RsFxN+wWzm7a>dd9HbYWb$k@1PWCoM5GFubBhX5R0UeOmfp!-Ei%IN
zT&U&P`rL)Qq*rx;FIMD~VcS4Y#*U@U>;Q=6=JpwBuY=Vuxtu*a=_i^aZ(v92%^3Z@
z5;OM0I-%{JYWC(Z&ZcHh6D&>#dq!|
zp1$iS`kPTQbw>|RXa_Fb-K*K=T2hRD>=h0y^7@L&!CnJI?%lk8V;iMP6P77h6dr!K
zjl^o`jV_a)TX^5#&>J%PQ?h(Jky%S@5->kpgQ@6^nhM$KJj#ciYIqnAQSV-1H@p=Y
zWk8#|YP-~nk#W0T75zo~`v44u1PyIj^NoVhlX1j6rP(9mnl0q3j7Mh~TxSAxQ?s3M
zm#OG2WlYBbBg%>(S!8f%K4H}?o#1KT!5r76Re({~=VMs31H~67^-Livs;11nLlq-4VgG*9a$y7FxAwN1sBIv*6Au8*|S@
zk@=lHv0e+W$DwbQJlV?g*Xq<{m2L*C2{&?W1BT@!Gmu|Mzx5OpqF@mY@6x%bv#mNxRrJFC8+M&~0X(dsifeHTz
zZiRX>8M&e+7;6f2Y=wv9$AR4QdNIu&Nf$gX#emKUYfCYj+xgm4Y^Tq}s2P$kPm^cZ)$P2ftcr$FGc0261?6}MNZhKs(
zvZG|+L7KW$sL@Fv>v7jL{hcQN3+x8=kB};oU*O7s(Ys{SvA>FP
z`AYnTP18bMBhMt8vZmtT4w@e_|(rFj|C)A2DFH3r}=!&$vGb7ow
z)6;fI%yiVtmChb?Pmz$T24$N;o4;+^JJ_9^?9Ts&X|}TeyB#@kU-PCiENx2xC2E!V
z1m1M}>1fmai1TUu)p=r`TQE(4(u1^xv5;Umj~4>46~8XnCw@XEiC)4?>SLO+s~m;_
zU2^YQCgQD9QC0MAFN-=(ny&L$>WzZf5}1HW5f^Bm;V&!$uehSvPrOEJ2fxYM!ra8|
zbZm^z-J;dFBVkKza!QYuSz34aV26*vfQcdT8_sOWgf$BK5hBTWc3@j_7t;Y0c5F>%
zO=XQlseZp(?8P0#Zx)Wuj~xwawvU&ji^yU1{-GfYHq4(9c-r++nFht?(r5?vyhq3u
zmUyW*V~z;X^w-VfGzPRGtHibZr-+t1WJK){mrEsrAGKVTbCMZ%3
zKc}b&^wX^U8IqXi2;w>N)UC$$1Ld<8u8Rv3i|l5$j$$`CatnEGXkVK7jEO22n7kCh|5sz_t)`C{#`
z8!xU?5gX+oDt;MuvpM!CLgOr^>F*0N_ICcHs#g0@QjL{4wivH3u1Uhx)B988q7%7{iR3wpeQ~gL)EcCq0L^54
zzcW{)Y5g99;PDPncO@`Vyi;#d)Lps-XfoQ)3Z14M4rKCymwZ3ERq$eff^6L3!u)pr
z@?yb_YYAX$>(vx6^3Pj!xe`d87Mv_a>?dt*s4tzI4Rd}7*ke=f=Qd?$gm_~k`5rB0
z#;^ubm7$j~j4pe>u?qavXuIjD;m|B4c%w+XvdlL@t^sL}ETUMWe>>g#?i|DDbIu2`
zHZLN}#bhz54!oE-CJIR-DN_5qAJm%uTuX(?;nmuMMO&9B&a2da0a=*!2DPyOD?3{nL2@pF|7Aadfc7lGif5jxt$Zu6Z@A
z!j7n?%hVA-t}tWDZ=uHNtSWGQbcr)MHWp|JD}9BAY%YJT+LJdp#0DOYRJ;?pvYJTq
z%Ftx=71H?)*^?!>fs{qers161TIhHNs;bUg4_Hh|ks*SGiky0y99
z6B82dUc!aZ?ITNXFPy|eaBUX3|Jd$S6=bTzrxfrSL)r$W`0a_oM^-t`erJBV9rf`C
ze_Z`CO(GW$gxax=E4;+*x9(GoP`2gw5ym$XZNue4SHdC781pikWuIP-w6>+-|Aqu}r0q9ZbQ#;{Yavw+-1A6`I}@ck#Q`>v$E6l4*vLCqA5y>}Nwnci82y
zC;t2oW9Qf$3e=?A*iKGt+qP}nwr$(CZQHhO=frk0=iaKRnyGp}%>KH6KtH>8_gXUr
zCaRkrjB$sJo6vLK_h|9y^8i2nW9>>XbId)%sw6ME_KljKv(eCe#mGKTBB{lJOrI&%
zxfNBS_xv6{9if)ed^s|$y=jLiP
z{{YwkW(sc(Lg#-x>5zNGK-3rzsBJeCb66EFLniDs$Sf3+Uu@yBuO`vAXqa4x9NkoZ
zEt9tgq8o{2V%BE6GFUKXLo)dujs%;709HAK5#4sid@j^LF8B4ZAlb2-jZ@3Uy36ea
z*pe5#l|f@SNr_A$sZA3k*K_!eA!GO~_}@s=&9cJ(SXisPotQnY=m${*b{|Ol1xYEc
z@UlV~`^+_6?eJ!=pDi;06aa`Jj1!cRqsYkUpld|yU;#I~Z|+-RPui6ixTV>%>&Qa>
zu|4X?8F*m2`g#*jY1A#2&EkTjK0Zw=3Q(9XtJM1>D1OQ95?-fePV;f*uMyHfU%DJ(
zvN@z`K?sq`vv;`4kNsA?G!b6sI#jyt#A6IsVWndp&hMFulcT0cyUpSo{Y{pk@9E9n
zI#|&+g~yN?SfR5kw`bTE{pxUiU~W_B&Auf>qc&K2J3^~wp~w}1*z!^}^m(B2**PAV
zK~3AQiOSU}?T)Vh2rE0R>%fkmH;Z-qt=h2YVcIfH&^9USm($~7W^X)a56(RpknFGJhd
zj>uqbg3lh;3fu_sUQH!D?n(Wjei^7n!`IskFWonBX<})%qJha+T;+2N7W6($7eD;g
z33EV{&AEYmM>WRfI3R;yhy>+?K$y%m;f2su(MK5FZGgs)e{#EraVhbQyJlh6yY!s2
z8LFKDJZZf?7ZziySYXYLo~nLm2eNJz^{RI+t*YH6^Tn6uBL~H>k8iy`>|HOflD1d9
z(rrinMY5aehvN5TV0F=S9avk`iiNa{<~nrinXy|!V}w?1>!5%!1CTR;?Rb4Ovjwyi
z^iUO}k<4M4k;(*6Uk+ZY`^4fzvFv%sAacBZX>OB$jgy0i#w93^Wy;W_VB0y?3tCmI
zYEv)}@Wz%~RQfgrk0Nn81q)ZB_w{}@Q7YZ$EQ()y;kR+bgfUa|p7dWswG#I;v{x5^N&y_aEb%L~HbW8Le1ApTt_FoC9S4L)(1
z&&V>Re@1CY!aa8>{xX$KT~w-APBb1BZZpfmk-ZD+9&yH(>k%P&uY~fAkr4RIaG?Tc
zXUP*yRmL+}W7@H5RB%Mq$ab&Tpa9p1(^r^p=uTEgfQci99)1!m3iwQZI;oe>tm0?m
z=`@pnZ!D$nVR>@ty1)33n!}xHy{2ee%Dzg^Dvd$)qtE)5sl9QfT{!l;*2>YNa@elN
zCD+Qa;o3#Il}_I=;mY;@tU52oTdgrVw`a$%xIVuDXLr4;J~hwrO%F8psuE2?@W@K1
zG#6EE*%NXMZN^#EwHxrUKfJ!oyJrF7F&J<7NvkIpk)gV`@@O?j&DY+6#cX$Gqfb&y
z+NvsPaX_Y%l>H4u;Aft9%&Zli=Zl6K63%=M?idZt&o104In_JNF%f7sgF2{qg#4sK
z59o*Q?_pJaXN5v5Doi>MJTh{AP94Qq<{}pEbq2xT%~}NXz50z)dIFe7*_!{X9YV>K
zy6{;Sf?)+n7Sb#i5WqK!e|`BTCE1#AR9cARkzSj6aiUGtFj$aBc3d|Dh57hjSFpuScU(HB5Zg3QZ)J^p
zon~PaF0VQo>di_>)3{hbafav?ftWk^{2WX3_RTm)qzop>J9lj~gGr%H)VrZ~+={}h
z@YAcN)27Ry&WQjkrDuwj;P~LE`Ifgn`<@dA|0B>?s9QiMN4Vl)@E~
z51aLT<6A)V=|RM&_eGHd0QY{9?Y}{8yOiuZuj4VqF8GFYmneRURHwpEn%bYB{yG9!ttIXP`!j_^SkM
zq$uBaavX*gD9Hy%OjStP?d&QsH2WgpHy!Yx<#vkhfADBSV`T{Kj_c&3X%kc`CBj&n
zHb6O);Gy9yVhHqDMI8diiM1BaX7t3773!MObDkM9O7XK%kS%_)Y$BQny#PLYs
z7Z14*R7t^>F!-SpQ_<38nZriA}LQV*>O;`Cn
z$BAQ7C{ucEAAw^;+;T8n9CdY^TT9KI?;^?99WDG<%=CwGi%V2%Bd`%+ru#-OArNg}
z7duZqVw5!=wf_-aS_lHoT!6^vIV8E!oAAH&k2pbbOrz4AV@*o((~>?f+~nNga*)jS
z5gWrzocH#@ZV)@0693lZyJ&HV^L6Ah=X@Rg?ZJPYhga47Y@By5AA(mSoSh#@@%#sop2a3dB*(|*<
z{g3iolC>lEv*#AQKXaGp&c;uNI}PoxbFO;iB)C2SaTBEhHK_;L+mIlYH`^xfmS}0y%V)YM-NBSHAQGVKYx5J
zM$Zjhqu-<#{C9~UUj+#pAe9ZFZm5;ldDiT_bhzNS*0!?I02;UkvVZ#KTcEE9iMhY~DREW4sjmFfdOuW(9nA;3h_*&{oFKf0W@&a#LO*QktB_FMi@kFox2
z8mB?@kwvtT3j|Y<*&_d&TeZdE+y!`16W9|{NtJ}Ctuo1NcbQx;eDW71U9S%r1NC9i
z2*!Dyf{{f(vHfmBSm}|TYw}?T1{aNe(JZga)~A6G`vH$&i=G=aMY*_XzVg0yH*>n7
zR=B>ua*R>kzuMh382TcT&QuCc5!Kg9y=2QY9!JMX?IUW-Rxt}#%ymFcXOnM0k=CPD=L@v-WL~gwf2g
zbAY);vC^%JZKNNn$3_CXJMEdKq9o~uz>&5fOS@V~F*P)NxVy~m#@nV|$qjsq2Dd^z
z!}Weyrc}L});ULy)SUL9VH5g9p+)BG{${!73clA~VXY&o#XJc|RI%JKfVnnwMz}X*e#a1Mo8eLOSGqo~@bkEwUi9?=U(oCMAKA&WSbIZJ-gGW$
z1{ne5UaY%&Ai+s;lkoNedjYgQom*pFWxU5tWP5iW$*U>~H#I#L^-bM%$xEqJ50H3e
zW)L+^oo%*l>P8n3@4k|Hhp8MAcv5d0gdyWqQEE66GW!uL>4AqR^oWk9Gh|QI%e#33
zs?RB)JL!P2P$nRbp;r!sXM+RxB@K{dmTrN_X%7TlcyD{LEXRkhaG9%?FCpPCDTlDe
z_{U|nT_)3%`3T3>d>`881u7y*2kg&rcqzK!==twq$xp(j7F7gjGV^0;8s6hZN+q_G
z8(!wJQon{!Z->O|J1MvOisPpXt%I0_7x&Pel5rk31ibt`FUBj97G_p>fGC$qrIAlo
zZ{mr)CLl;&kx0Iy4G%K;<(qJw&3X{Wml@OVom`{dSc=Q~+mECe584pp{!p2xt?LTm
z?0F(im(LWflo0eR85ktWPn-+FqPV7z;?MlO@YY`lT^bk;+~sVUv<&60dkHKfKkhWf
z?C5dt;C^Pyut+YG=Z5cY<-jxf
zLw+Pq%V2TvW^n^S-eRBNqzQe>S*sd>!8tWEoaLi0@n<^({rfhGl;O8AnB0N{wBhmj+JnZtQ%_RZw$)n_2@0ge&j+3
zZbRUzOJY{fy^FD}g@I=_$4Q
zr1TObbWijR{2@&YV2;8U=N+{ih(5tQgO_X~CBJRX$5R*9h=2=%rQNhyY|Nhb|geP^;A|RNniVA1}7MODc>vd^lFxk#q
zHToKWnQe{wOA24S$L_?kNZ{wi&kesF`97~LgvOIK^nkiu
z7`fCsBhq-;nhCOv-4VSVaf>M^KUPj@$6a-5;+ai21Bq#`iAtIw^(*HbIw~Z;wG;d}L^xOF3U0D7J!AMvcFSou
z%yH1OF>SWZM-7ty9s?)4IHxi2fcMy`eig`Fg5)eHC^+gB1gAFBb!LPC$s;rEMisnK
z4O>W3Cye1F{c>+UgyiKiWGlBGaNM?H^(90Mijm~=VFVqr2a!mscbq3>;DdF_11JE&
z+Z6}1_DOsI^~&RZbja;;Cnng+jr&^D*FakgTAWV0u?w>sh<;|X2(|ZyQ2Q0XjYDpL
zs?xaET~~TQOmJ6x)VEH`QDEL)qaY~p
zb?I54&ioW(yEAbeJ?^cqCx150KWbN4IM!!i3~S}S2ag;D#I>b16Z|=_is^)T9-%2a
z>!l>AEWjG%5DjpWV1}KFb4ytlqo$t{pQ{z_ckH#lCoB}yA5=&d;V4m`Pcy0JJv=GB
z&eoV=+TQhbj3bQGo{fBl$AsUR$71VsF}k+XL!OU``?sPDRf!^Q>~bA{N0-hCFtN(HETEj#qkHbZ~Y0&
zG>A%eLy~keEw(bN2`|Ija2C75&*}r}se6AZr9Uzpi%mo4k^|CX*=A5q3LH22!KZ<)
zH;;&ZT{T{rnb^H6)UJie3VKehEumPp2_umIdx5*xIyU!Sv6AeX%5(Cgu%NfbAqT_;
z5dU{1zInwXCw;td3l9j6^?nLObSk&&$a791GhBiwqBvFhX
z5^?4VJGrpOf)PQohdqyH59vpi*DZFRM6n{n)#tA@?1fwgb|11j^V6g`xvzOl89CpR
z$`^7#OLpf`5$ZWy-Ho<4oY+h#qOk?LetiwTw9BE}IC(
zA_F~tNZk98YhN_;c`4TTut7Xzh!KZ_IF)J-?-O-s9TC|%)5Ox;7O7oHm>rr>jd
z9YSfy7kxrDS2FqDLke3M?5erjYB1Y*AM1#nDFWByY2|SF)6ioz&{p@@I01LgR?iEF
z2EkCO2_bAX^=w{ThX|qFJ;Z*t=;@tDBVKNzC$>lNwyBctNDworZHT;LZo)*{h*Eo5
z9M=E|U>IgfY^=|`WehgPVA=u-Gm;gb8PDUv(MaaaLva@)G|=@3ezkg09TE@MGA8Rk
zALV6@vLu|V03|tBIaR0Dw}06gb#XrfSw^?~kPEjP01nWBcs^vWys02KuJwi}*gHxokyN)SF~qu>B?IK^
zuuai>ZYC@R=Ax@2RCF8_4a61w9M$LEJRLMg-eO|-JD$Cp(FF`~h3dZ=OK$$`5X(m$
zrxREH)gFPrhLl5jP^bVQRj;?ssZzc_4c=yNBb2^PDhXz*V4Zr&t$mSUBSit4`8iVZ
z7l~)Sp7rXu|E{TV%nPQ70w8(G#
zMxF;1;G&DEoX8z)yF{0HNsh(LXnz6N-6=NSwYtvCvTc8v=Y!zxfkT
z*26DFyy~T_YVIUAErOH#fuLH%#X{=th~`MnFXLo3h=6vZ195~J$vO~6I*2Lb`IM+<
zb20|2y^i(xDqgKOj;g{B);X&l|z5+%DqKWK)=W>XSes)2b|_8IUEdXytLnO4U+F4IMPZ
z=AurEwo*dy?bRER|)|cDV;pi7s`HDPm^E0E2e2BLW<1TBv?QTNx+?`591hP^PeWEaWvf#Kb{ZAfgEmVGcXq-gBF~1xiG)hF=d1mvd-;)}`nD4b~
zI|sv2x*$G^2~cxp5RL(nsOwW%fBDyjJj9c)h$~@WiWZmD6#ys1d77KTd9^#3ym?hH
zJYN~xVo+}i$o|lL={a;FN8+*b^rZaOua@h!R_xn>TKBmkjp9U=6^Nl}vNMm%ja=w6
z=>NTWTC2LO5+xG{-^5XW2}kNxx>}JFpB{pww(ABUIV+4bGK#7CbVEx!jV$#
zbc!kd)Q5;RssG!j9oPGD&eq3L6VjaB@ieM<<^`L<(-Q7WIdEl_M(M!hUXNtIknhf@
z7z;X05_|z)r?{b;yuyI!d81;CkNM?yBYxZ0gU3w@o=^QBj
zx8eanh+}&DSaRa_y=hhg?fI_@xOG2|q6dyw?s+uo>H57S9=m=@Aug
zWx~wLM?7kXPS0252_}}9jAJC2qze0P1tb^r-WZ_W&o^iZ=J@Tj5*NRcU_YvP+LBwZ
z(8hC?kCV%Cp+DA=z+7!d9>=*_IlqzOqplRr5=T6tp$;;mnb7Em852%F$(+7#i@~PA
zRxbUEpnzKjk2inN6Q@CijNRc=8hjE6$2`EIk3H~5q6^rg!J?`}O2Kykglw8>W
zOng5Mn5=mX6xL4|pysY%(`+S+zkdZ=v5BcIikv*{NLS9?=rO;il#@}c^{7rr2xh~k
zxQ$W5BG}(bv^xb|P0k)-soHhGQV}i!1b(GB*s1|fw5cP$7B82q*tHhKTOWn}l
z@>9-*(SV_b>E#;w5L*vLoj{f1h#{O3|}|pF>N>
z&lOXqX@*L2alqnY@8rhRc2ZmkEqJf5_huzx5mM%Kfu|!%n3t~g0Lys8E)MBazruSA
zkXx#zr8dCgwB)Q6&Ysp0;^5$Q5OPoat%`aQuP8H*gKXX3zS`C2T2+j7YZ+5HG>&G4
z?qvmsnQLbiKnw65i9glIKfV%RXW%hfbn;)27G5#Zvyw1oU#>a?luFg9#%|w*nfpl(
zqaDCqB58T#{>bZ?HZk18B3T+Ti##RWTzARZ)T@WR_-;D5vy8n)ctIuq(w*bv(%Mu9
zRbw@idB!6asnEtKi$_=t{Tnwapb0{d*IP6#gvnA28_3H^agA)yq60ekgTA_%{hb-o
zo4uHD+$#MkKT6Jm8xOXW**j$IeJQWG@iBIPdhk~r$ciYxf$sPzgd`uQh9ZOBAF`93
z^e}}`DiuH%iCDiRadM!ec;xIMXxN8T?NOGj;P|K{MrTvreY?$a#_}lX6OBQ_9DPC(
z4R3zxLDE#(|J@CA$Wr@tbEcqu#E19SA#Z7sx`@?OP16UvJ*n~Q57gBHW|0K2N@-k@GQLSTldY_t4-
zh=7kns~QlrsD4y8jSymOzaJ_YU=6pGb`s~jU^ddJdVHBT6-pZfEhh^86d8{YTV+MX
z7W${VEz0jPAM`6zASWyh*qdm*Q0kv)4?-7J7~3!WL4PFZcuM9Zq0&6HP$Y@zPS5SFj;D7c^ol1k@cB~
zNzP9vtHKQOm0Uw92(YADZ(0>a@FuRhU;#Fo$t(4$VUOI)n)R9aL$uZ~&
z-q=sT*@l_#M7Kl!1UqhD1Z3yA$#HZWG}^^6BKgxXxt99Y$^I^NK(Y{s=?Qk>=g5$r
zO-~u8lu}HVUkO9>V-tyEyWVMy!aBrqdet42si-5)L`h4UuFfaEo|LCh^oF56+iUPi
zs(qG`t(k9zy5o#Xw(Egyuw3Z$u)+ZRVaq1`LUA>pZAD
zV_hgMdvJ5f*^GmK8wD74DEwwhFRdR~AlfI(bRg1C#rFao{}ltS@RO~A*Z2GJZqqh5
zw)V%SYwX&U&7h`Q)xXtq>u|}`eT_L|?rt4FL<{wcFLk}$$0QFi!heTW**u^Q!-=wx
zpx$22uJqd~mn%!yP`OP_U)+8;%KLZmT-8is7FD=Pbrv8Ev_&cx=V7}m3?@I30qbWY
zV^w9;*}3DC`42bAayq2F+nc(DgUkb?4?hi0=mZDgkgVXld)JNuikzKgtw{Z8FEvo2
zHp64(DxA4uRCWhyI}KXr^YP(P&vLlbCQ^!#c9&zB;@B2f1b<~H)}hO*?7C4g)g2uj
z)l@Lb1SamUwnYryj`qibmyro}MWBCHE1O9U2?0vHG7*
z^&)gqDI3cNMoFTZvixr~|2S{bFL91f;pZYDq0T=Sn>5hxfQ84Kea38C
zIYt+gm?ZoBjuzfyENL`c1Nf02$9?IrG;MW|ts?Jf^r|hU`?Z18t6e)AS2Eos6A7lSiwF~wvSBEKD-utI{61*N4ry8~
z$=eEZ!9!wy;TjM@@ZC?{Yg$Wkq)Tc)Txl?H<@V5r!PpwlDHbg8u&Q|$v^ogpFpx2N
zCR}5M#@yXC^FNZJ)#$(aRld?+Z!y^H&xD-L&t@XLfc|ExNC5nsmFm-{3}d2kt&mYt
zb?V)&^e>MJ+JRASae{)r>!N?YO7%k|rr&s|x`H!Bj)HUQjyo>8&QE0084IJd%KOTt6m
ztMVS+=OW}S!!->v@vPS%Dqd|^A_$_~50|h(%vo5V_p~-oP_XwIoW)^VaLZvFCKgb5
zo5U3i!pkG>rmm?sBLm{-K+S_ooT{T|gdcbih3T%0JP(Buck5G2kHAU>+K%x>jSd$r
zT8<0G93N_#<$V~QXpNH9-k*~dJh+9lhN1yj1DJ1ppm-LuuyW+|t=+|)fp~(IKsSCd
z25epnsjH$EzxJv{RvSl#5R^^q#G3`i9S7)p9uBI
zyWWziN-;ZU1JO(?M0v1Sqps|S&%|HZv~w{)nBRZAa!qD_d9KeqK66-x2||lm*w_*I
zqR_hf_R1&TQy?{&+-n|GIM?}FU=~H}4FR*Nx2;!#}C?8c#af3>|%XvO@x+Hn|8oyCI
z9Jf_lf({fch;1*(aK2S%8vVr4k_NF8H
zy3qfSh<8WQ_-41)f?3Z8!^eT}x6d^|>^Ens)$jQzV(>7_=u<7V?ixlZom`SFW#)ip
z9WzhaFd!=zt>zGY?Uz_7l{5Z#RWWZ1CfX>POd%9lC7VPn0>LS@2bX
zakj_Y`%}?)eJ{)LoJ6a-wX1^8sap+&(>85dx9#;c^po9UP7)c|S5$d~t>388l|pwW
zlech$_^oGsyR@%Z47+2et}xwq+<>w`U$tt5v1@GIiweaDpx86yPpp|)bm#qa&njQ1W$o(Ae^V
zN08;V+4sWW!Jg*~JYlTw6?EqM{&36VTCn)YIxGG6Q??qDeG_`wn?
zMW#6&+k6C}i{TY2Y{c@fmyFO*U%D=+9`MxG>pawJc6bPVkCg^{kzynOsUfbNWnfg1H!y@@bFE%C;llg)2Q~pi410P-`0>L67dJreVSmY%
zi)tvTQNWQ<`HtMX^~&nk%=C9`i6xRlv5GK`C6IQv{Ts60!yQ+t$p8d@5W28sZ)g-kF=Q(hsrd3(_7^a#GjLJZC1RiM`Xt
z&vWzQ_0Qex8_vd>x2SZof~g^W2+!Y!rME`Nj~)2_E+9*FkO*Rs$YQ@}u?I>a@j?!R
zs+TIRaGpG$&GB>$duc#UkP6*092i;5zI}vgLsc&nS~a_hIdzb33uLC`Fn=ieDfn@2
z35XI8rAb927Pm8vBlX|~5s4IEhlrG^CQ-M}X;od8{C+PewMr^}%R7OcOiFow18coq
z*KR^REMy%bc{AeKCOJ??wyk_xc(PuCno*Dcl2
z_~hEO99CRncoYb%%8EmUfaHJmd?hvaGigr+58!(
zI6umm0*40*)6WiI%wLpykC#Ui|FJPa&giAUP^cm}>q4uiXFdQzJ3?
zBo3V+f8#Z-x-&=47|EQMt|oqakl|XgYmdPvU|{G?(+>@1U|#3<7YP;&SPkFcd&nxx
z67_dJpyez!0%Pic#>g5G3-_GGMxS1bzR4r)@F}Vkd`m`LX7IK+S{*A%fcXvI}rVa7~MWG5&BN|9%siXQu$5
z(aIfiuoDIS(|n^E)auIkik5k9J|+DPmzBcR6S&1Ua!HUi2Tcp{0m1S*BdBCtbb3}$
z-XH(%D<5)i#i{k(YgNt0UclvW?M1it6G0xey9lY(yZqr7PF37dI<{J|t8dz1Oh3N@
z>;~KJ-ChPW_LnKGO|YpQvx?a#ms5+k^;M)+a4>}Q%M7oc{UR?gUWK+h(Kq>7*al&d
zP96$^&g;+{e08oDP3~28whIoFdrmzHoQF=+byvtE>%g|Wl1M2^EgevkR8(P#R{V-P
z$xC4(p=T52La`_dp62cqv4X=<1JDSx+ErqA0oU)~_vNI|IC3UZf?Z5aeQ+YuxTE$!
zTCRH*jN!MzzP5iV1s&?kL(`;=0Uk^~Dr53-#@>web;OTJw@JGiZdx=cd<8S)ykROV
z_%KO?ciy8^awq%b(vZ(5ZLM=&6XW56LmIovo5}dIvJ++WlO4p+TssGnI_;r?7re`?
zP_k+1QkMbL0^*W?gz2w%9cC~#!yDr8ZJl3KsWwvaRWY^ekxwjdn?3$JzkS<(zfVSd
zc2OI>3XcexH*K8}!NL39YYUHePKzeWfwWZA$>!RLJ?3!i+rbt0RTVH!;aoVW4SG?^
z!K6HMHwYollER~Q?aQ-;Q#`@>tYU-&Z%{G_+`5g(!@7RQY
z)aSz*qwd7Dk#u;KF&N&_>ne@_c&6GXF$*-EWcKDnKoK{{rJuI4f9zFKK$*}MD%KUp
zacEo!+fW)@SAax1ZNOJ!#?Bf}NPaIGREI1dtnDaYZim2cma7Rsxi6E+9c66q-KIh9
zbZetoM2s58#y=>Q-tH5Fr!M^poaMl?CdPiqR3cV@8He`0hne-)%p(X0>#C8J7KgEF
zv>El&B;XMK0*Tn~)K{duz)>Z5Q3Y7~V^WMAxs%_Qt_n%gKDq|ijoEz=K&tbLfo#av
z%hGN9MHRTBP|Hc$h(s3+U=ZvXO03%>sEbNtS~wdu1AIygl-YTFQreJmHe}+GNJ3~t-Ztnkf{xRS)urd7)As_=Y8^?d&{VyLN3lj_e
z|36T4{{&plCXV=YqSgk^Cc-90cE%=9yu46O&WmqQ|BT)Ef2iY*!2}au2fgU137gCH75LE+S
z
zW5ao0Y=l6G6k*O1WdNA80EF0Z;K2xIV`TtvAn*u!45bM`0&t{u1mTE~1VaQA4n9i3
zt8syd5zUxT{5|`7G^xVCYyp3}c3}wpW+sS8B&l>i$YI1^VlxC<0bf%*l<|HM!H7Cy
zR>AQjMl%5Jcrb}l#@~S%I0oR`a)R<~;^j_Qe<4O0tM2ygrzdCG#MD9#kc{Dla_2AN
z2`H#yZ)r@B+bkY1tKV
z_I&&={Ve0?MS~$i!a;vbrTy|XbXt)BW>G7O`23~<@X0GN2*Le25(!iI1ueAqF;=^Z
zPsDR$hstVeY9Hq@dTl8hk2+R7TrElb3I%4CL`0wtK~RhrlqDw6+NKEap&;N8Lr5+d
z3wX0A2xXjXA+z~TbJ9FJ{&X}1_f9}uPxKrphYY>~RiagtC<;A)pp^*Hap{{pEA0s6
z3-{GuhdL>8b7&v!eA<|VYEu`WAX#WZTQ}Lt@7{DA~bf!3lqCsOb
z3kcYFBT6V!s$+y6bq$nMFhChg1J~8hSa>z+T<~~`V&+6%OuSTDFM3)s|FiIaK)p*(
zA}DZPaG4Y$OdX7XD#ieqNsVp$Kv5)zw_oYq@*bcTN)`6fP&hx$tpxz-8hMi+xAF3M
z^znK(2r_Y&Qp^)2`G80m@+kTF#p49Uxb+|ln3*uT+~UA;8BQe>CP&I*8)rU>bS$`f
z58FTmUSh2Zq%aiFZBc!MKc*;7%dnwr`(UA>k*IGR-yPY`_k!Az<0$S(`(4lgqKWFH
zQ=k^Xw>NN&?&bi-^g?8W{U`e%G5GvNB~1a@_!`nh@7#dyy{#
zpWgwn*4xoTp5{EfgoKgPEs&SO
z*-q^*J7^mYrHboU>)3$BdAL+Qb
z>D{FWTxVOHN1B2%y$(*}?lT)qqEWs~@HI4ZoAKY0aZpIyeHOy7sC&+TH^CjF6Vtdg
ztT5`?Ecd;;B=W~p6?i#H3d=XG_X{uC3NcO8x0u)lmehJyKW1K!^;}m*SEq4^(pq_G
zwY1G$$nPJ6mprfCXy-5&-EP``s_ozH3Kr9i%Sbln`WR&~p59wIZFpxH4$t(d%qq4oESyT1aXZt=*^wlBgeYvoVCy~Qj|5~8nepY
zay5Q54$;2I43SxIyf!que^-~fc3DRLsQ7+k_pz@mewZsNdB?{joH^Zof~9F@(+HQe
zgf^VyAD=+^BpA5~KIsaPO@%hKZpe*Yl|_AYwl#Wod)k0ub$^(RUviPPkHK`CV+N}L
z*9tJbpsB~SgtNI{XYKo_@wS^;2xuidC3r}=wYA*|9xa<&%)wu-!+}4xX2ZXbNB$m<
zXK1y3{p+fwUwliF%vEa1XLQk@+wVn;oEjba{>e!{a;g~F-;@*CJDah%oV9BCKEwEQO}5@4a@75ut}3Y)^p={2W7)DC#H{aiDta@%
zucR71n;qkNENwmEkpuNLmWVNWUv`axxKy$-IWSlyo%E|op6fG2<|w`QmW%d2%E@@?
z_OjhWE_g#dh@pLzKxU(($91chFH~#vNRe9q_w@t6mYtLJX1Mc0=Fiqv*2y}zgD(3L;3Vz_w`+ULFe;`H6IDq~%8YO8Cn
zvia&EBn$HD;G_G!VzF?f<7JiW!on@?;qHERWVI9b#w$cc6Po&axa4P6D>GMirT@6q
z5<~XlLX8{SY
z={>~Xinm+Kxq@t8LdeHMdaCG2wc_FzKcC#uuN=&FHS!P8~q
zuCac2Vnny~|Iz@hOk@&POvwjP_e
z#SH^Ti)HKMk1v^>*l$CY5sai4Yb@P4Mfgr$E8#4~tJ>5v@|GvGNx+H`(bde<^wzJh
zg^-rZeg+beYf@fM&evEOt4lWAq*T#bjb44t(aF!{gZ4=B_U_p$+4T+0QOjJ@-h$k(
zlOFi3!E*4ES;P{mvZm}grPW6_Z)@>n7@ZZDnMHosuNUBJ<$?kmo762rLD5>u@{BDN
zH5qkd=dq>q%k>UJ8)))o`bc#n+Pd%k(-Q94m&R#BM&`z=F1HCNP1ZLSfre7+L;LL*
zs|XpgoL)HHrp6k}@92H$9-ty74*h?Mmn{DwUNZgo%$1pxspMBvDeE)LUA1=0k3?w@UdPbL*l1%OK!IRYUMBEVm3-^bFw
zh?Nf)2JsFj3AD3si!h7|1K2RAWLz-2TdS|dwJrqfD?1w|3BSRp
z4e%#F1$cRB1`q~7uo{q3`>!?tY&qFB1Xw*k0f1CI&in!xKnvh`Ak065@53f1fQXF@
z4-Jkgt-Aq^o0(bpWncUF-}Ao|jsVBSEY2r@_&#v{g}JdeLw-64AhUa^!1)ExuY6Ad
z;Jn3EQrfyL
zz&ykh|D@BL17eVJfm;T@1Ymd|R|Tawr7HXO$Kf6@c0P`VcT6fP&Tg
z0rdFOJFxoqTVTKb)fvBp5J2OdU7fs7^1p3U&wMsNGCE*TevHUsdVcw5d~9IB?OpF*
zHmtwj>#FkMSnIH-FKQwP{a{xRp8B^h^-T1~Z8AEVLK*_PVrfdSwLB1IoG83F5NOlT
zFI*o5xy{*`==|c!0R7Ov+~>-;hBVMFj)6_CL!Rh)*yy}u=*ufvpMDVAfI!c|Z$JOV
z*f|7Q(y-mOY}>YN+f`k*ZQHhO8(p?-+qUi7lP}^9-oZc2GtbDgpV(_J`Z`z)Lz~CH
zpjK{Xw!9E)!??vXuXQ+Wke6}OXdu9M00MLX@o6czufmtR6CMnYeChkd_pdIYT)@`;
zv9<($e_(CxBERQ+PgU~0dNJV|J?TfFk_v8
zxIX>xAN(S|0zLI*i2`3-fH;4J!C#dEyZ~uf1sJHh`oX@_f9jPK6Sl+>0Fc)Kpu_hf
z?CRllhb^8!q_KWk0`Y6x!uuBZz3oLdJNo-?Z2ZB2lfc+{8rWNCm8_WTlx9-tIIEedvU)X0zNzr5N^)3|1TqC-(bz2*<)Nt8o2K4i5-I6
z`imVxuKp`K1Uc*vc5q|MFLv;v>l-r$9zg5i;KI+FDBzZy9|1iNz(W}AF0k`&7Pj2f
zdmMHxl&x7lH4Kb%6WHdL!Ph37BdBoAZz@#Tds!Voc4`*!+_l{-Apg{spB*Uq$vZe7
z!~F+#@a?}l*q7b31+VsauN_7%Khf7}=LP#pK)VpG0Da%e9sBdH(B6PEs2ApAUmdt^
z8}85RtzPNxvj6?;^h>KXCT4jWJvl$Vf0BIssevGoJ>~t|uUhG{9(~L%I`UWb7dHRa
z8UVr=v~fst6F@T(e7(FEzA7Gp9*auA1jU4lf+&?qexiRvGUu|GC|z2wApF0xPXi@I
zR0T$5DTygo`~#ArWgmvnfqUyp_Lu8J+4G}WRv*AK02PnN`8H5UMx?p5Ib42HDLZ4Bc!~T@=
zc7JF%OdcwAontJuv_2v+!B2k0@y6L)BX|8H(5O9*Err-I|}p!l$LO_kjb%rk48>&DuJ
z5zhwRdd{3l#N9L6B=pGW`G6LTcEwN(^2B}iX%H8|3X9Zk^uN63ZyVRnS_uhF%H!h9
z8Pj6eIW%iUSX0Q5nHQTNz9+9m;yov+U`f;SG=HoCA~o7|d><3f!v$z4M>%t$=1e71
zDNxo)2uQZJ#$?pTzh|;_GqQrTSPGAS9#|D~rY#6&(%Lxzj=?~NIgJk$
zq{be2g^o&fpLK*Q3l)D$6!PrTro5J?v2IyksbJN*BWjU(%8Y|EstZckZVqSbj1unN
zH(wbd3#Sf7%|r%pY`A)rB7o(emPgR*%aUQN^{1I2Cu?1kLS5rYg-`U-4!z}d|89a3
zcrNl~l^OTV5P$TT<74FIp*aDyov3(>X1RjjFBackDl7GtV}14TDwO5ae!TmpT>UY5
z6!}|ekhE_?M89AZ5EJ{@zxrW;O%{XCxjq$?C_yv^PlG|yns{zhMjA$CBP9KXS-crf
z1-OC=(aVNon~_1HfD)eJr(!SGm8bed^e%~q>3oN5_Hu2s@=PRZe{8>?sVs|v*Qh8k
zUghERds04odI-pn^Lb_uZw`9EBWIFwS{WVg4#-h=C6|F)Ntuu99LglwL_f!!(kj2UCr|tQ-
zZ~o|FPx%@=vRm(4gFm+dg0?E(Os{@>8lzNCOU{nU5vSHtnV)Tv(7USGne*}ac(J>(
zl_N;9W;iRL<+2QL!uI^J{G3f?7{$Sj-YR`loWHr7;V-3Os^CJ*ztfhTk?M{f&NM9c
zV<;840l<8!(D}4l<17!Xrd(d@<;Ph1FqAQA((RG9
z{fxk(qM9-sfOmpmsB|-Ntj#1%F-iS*jS}G+f87Ia)l$ip+=M>40go58tnB$55(-9|
z{sJOOxj)yO10Xc1rGJ8K3Jj(*N!jN!KaZ~s^oA;vIxk4!Ibo@z(luOnB_LKilj76y
zRDjWEka|SU^sqAG9d0&-3{+3DO>K?-BWpsSIYQ
zHc?TOZ&F*}7>vJ0N$#+?Q=z$a3kvLnH&fg?zXlD&j5^C?F+qvWzOx-v0oNg1N0T3i
zXyNyx&Q0h&!{nE&_D#`X^^85oZm)pYJi`3pTE#ewuOoj)3LBYvh#Ah`^|X_y?T=8C
zb&nlx5zqaKU-L1nNHzpiRkPh;?uZlLmBfAsljz_5)3j|t4|==DFAc{x^v*qTjVgD%QBi~
zJ3Jl#rh0(7I!j@TGJn0bWexx#mge{%{AyOFD@QGxEvM3)*u-6Hg1jxMyX6~y?n?I~
zBODmNko%e%RK^rr(R__P6RxhwKWV?~Jb!Pf?6LF_x>hLx_14LQ2poHzytLdL1(U?t
z4B>dQ6eeyILB_(RmrdP>VU{)Q0NmlVTp2GjDlU6
zvdnHG$iQ&^b{wK@w25GxbPoz^#nP!hQCB6Rb(X^K8y>s13?N)==8`aTo2gsrM34*_4JldM{xvJ00rv;bp#kZTvFh$`6ZHcOqb`#9^QxtOvL9!t&j5@=8}
zpQzUn@A{|}gz3P)L$Y!uwYt;)`IE;+iEUkLv1fmPERF^sM%)
zbXV(j01v1fT~&_Tg_+=t$=evqUQ)qigt^OmoQFd@R9}#IglEZKF3ypCJ->*9nwP3j
zLt$Jtkk8xrF!9FUKRhnmDGlVbSo#8ykb13?#oucs_|V<3=S$JGMF>7=IQ4G#*?)BH
z*;30Xs!csG4Sf6~ez0J^uhXygUDul3xc*)^HzDr6_{57Mx0}nGYpA~QG2tyuB;$fB
zO!+CBcrcxEEb0!I>FzxDEe%xgKyr$wl33(BARcMju4SQ3DX&^=nFQ5M`t0sT6+)X~
zH9Jv{*~^0Ez__@-vsCM={O>b#V9ie2EJnUY+MYi4oWENOA%*3t47fTQg4E0rLJ8uJ
z*Vz^Qze=5$FBeAl?t-oIBS$C-W+#JS3oRl}4tX+eGj{^2B8Lw6{|+!1}?YJG8)GkeWWNek2KLs6vVCuj5o1zig5m1WrQ;jCKSwvh!a!{2r}
zMWT6ppVp-O5iBHDk|WHgaOJRR49P2{sxz3vP*n*@J?W`rx`IXC@l?ov@`ZQhbWke$
zmOXJ?QoyN(d|d#a>*}GVGHT11*cheB+6VY`V`~h}fAi%HtluX|o7Y-tU$HhYoVzM{
zGSDG$*h!>xl#JI0rJBg|Z`4hn)0kD<4@c4r#~Jh|vLxQOEV@TCE_WH12%bm$e)B(5
zPThBjg($(n9M7_aHo}w!)KmfUwpA5Qow$%H=s69%s^kbuuL0ZyFjUyl6(QK(_L!sD
zc?zwVVh`u30L)U2TaA3*oT;ZF$g$nS3mo6pT|F&P;WVYdz9XaCz8wo3GK7GJP0%Gf+vHj4D^)6Yo(Log_f#IU{#I8AAP
zLqRTxJ<*UL^`4u|JtgIIkD(c?dT+PVa84g;x&asA!8v2#^U)$qxC2e=!M!$(u)}VZlMq6P`%w1M@KLCN!+Ngm
zWbOt-rP0+TZ!r|CwTzTvtm@ud2p^YOe4e&gsbsI@b!A+duXry7#-dJESt};T-bFqM
z!w7g0Bn5Ps3k)m2E@QeU(;tu|MYA&7)g;NIfv*IpQw}BIq3*na1P1w70?R<>?RjAGC
zZdLH@hp9L|;cUSdTyJXnl$E}i-5C2=)X-(0{Ztcd6NidLsTx70ZB3(-K
z8+rB<{d3v6d7(P9<^S$24(u$t<@6^sor){3cv9F4M&>FWpn|Z6hu5{jAJco=0lyjs^Od^H)nH}g@A~5n_N|hXn(q@gQ{1~(H%}p4$
z*vof~Z9RMa3r(H{;VP?GsyjzwgCk7{C&O$L#l0H%U0DRIoB{5VKI@52UP}pDmlJ@y
z=zS??&syH9!&b542Cnj2hga}@v(2{5-SgLH3jo(bRlZGcQSQqoW$#b;BMQWQ#5;=$
z`?LA6lA4qrPxDx&@n`f*bK;(J1&DD`y8H_|%qqwR1{_5=Q{BuE>@WrPJvHmlfa~PQ
z!d5QJkzcjGB(|>=PD5u_bFqxVKkw)nl)QObV)s3W`6>wxYBOa1aNXEEHCM{8u^=!Pl6ZHl;YqA}x@Mfc_b&iju
zt8PcOgvHyBxn%paj!AK~Sqeqc)m{&aaHb*p^Ys;ggx3ykn%%zDQq}UFu0cZWn7A?i
z{jpZ=ED>UE>Zls)sZdb^@oo82`07>$XP60w^bB9_tk`Smld|?V#N9&6b~>KGX^MES
z0lX0+)aTgg<(zhK?u3^OAN?1LA59;jxg7hw#uEdjB$V_99ZU&(ou&U|NH*Xn5r4&T
zV{YuT4sA?~BmKSLT-ZCO!!2&kYHJ8xLi7%2`|~;{Y%qEHKbZ)618DT$`J$?)h>vnEjx}8pcrQ)SA5E$};%3|Ce}KBU!VS4f
z;a{`YZp#U-8&alKWGmmyT;X(cc^LH>A&w&hTFMxj{O-u|ulSFhJCnJwl}yNuY3C4n
z7QQ!{4@TaI^WlxhVD=w?c1l?Ql_ES~kDQ!spvfzoLOY
zTeiL$L_OrVUb%>@Ii#5tT~nQ=<3KJwNpWd8;4c&%zKJ3A%#_V?LpLu9w>@$KJb%Kk
zj!yV<<@r>3K8_Ody>0{6kP0KzzvDkMs`1wHZc~8Uy(VH1XlEXjvs9I)0joIAK17J5
zAn6#1>HmPR-DJGuY2UoCClyWjQB)@$uH3q+L9-$Mx%dii+_|vljU5WY%5rz`*6w_`
zh6Bezk8eBzgB)4_4@Vv-x{>@%aRDXdEHO~kP6LKDml9XBrqPG?AM=e=1f~ybB2_R-
zxGvI?mX(#BHiaCb%FOC|XPj~=D;9U3!X0!x_3F}NT?JPnhZ2Vx4x_^A$i>
zE_>XSF1&`509mg@*5M6cw6msZROQ%U906;&6k}qp+f=ZgXUGs%Wa1wL8nTZRq`D=Q
zsg3RrHhkg5THnyl-jAmA_zM)pl|dgwKU&4y4m3B5Q)c@0CVjvnO!&PM+)apJFL@tV;bgRRgq766bcgE+z3DBS=63NP6C>?jgv3=}
zW3g8PSxwB7PHCCph(&X~a49JdI8g_b2~&-gMG!8pEl>hB}H
zE{nu=YdLQ{EmtYSNaf=?u9!84ua_X7wcfeie%m
zfIY~m(_4WWLUg?{4z9d3kWy4`wzarvZKuCLI=mr@0V=J*z$Z`jTnbv`JVmLcIbTI6
zyoAdS-Tuv|;1cwS{H$yyAsxSTwg@%ZV-`8I#KRFL!>nvVs$-2NVSH0!jM*=5(^JlAcf`hgZrAu(s5HX>ij(IyzEOkYS`UgWAIltIKn$bQvhr0`JL
znpXgcjY8J`h?-wH0Z7p
z$D8--eQ`ZSP>?P0)~_Nz-~{1PHl-3!?AfG9BD+W7x=+P50+Xib#%TGXhI^LmaG)|h!}9!*Eo
zu=L)#+HC|iA!675vY|kwov9NJ^Ep)pgCjT7J|RB`wjczQd(rjx&(%IgJY=627>p8m
z$x?4ltxw_-jIuNxoBq6A#Mih8fss6A>`#Q2eyQ~jv>d8vwxh~`njvLD8C>bW?hc3;
zk8CrM2UdWhM*vUwOX43-K&QU2KZVlC%s%Z$LM9BHl-3mIoHhMQp$gCcZeD+^0=BV8
zhL7ePv#Q(bWFa>bG8^zwQ`qncySG?Tyl}dE%~58!N*7qAty+suqAId0OD1!aqkofY
znlJl^ouKaE;@pQnh9OVOpfo;{eI?&#MKFRhk*bZD&C`8@uH&wOgo~S4W-DuO8rtb1
zdOJy2LAa89p@iMeMsEu>oJf}rhsaCVu)%Z)u!23sdKA`(2JFSfyfTu>{8;^*`zXTZ
zI42`4`v9h7DCNE5XoUjk4hOCMX^j69;(
zuW|J7k-;IKn1*6}F^8tOFGA)E`_DoxO5nZv80u*@-@@+pT7G7cr)wZQ;{xbqn2;GHU^|Mbb3u<+
zO1Ek8@23HE@>bUZWy^BCdRHAO?s;xhl!aACj}0>WA#ABufHniQA2)Ckg^<%
zt6X$yysPZHozfG{MXwufjG;EuRNAS`uI^+rNYkgnlMJdVs9f6Z6~iGeW^k47XjmvW
z-Tn1EoLcge7JYLUKsYAWx>El!xvP&l<~+s-u+x-T`uuB9B>x0`;8lWfoQC$iyf2w}
zE69Zz@I_KmugoXU=JBU9sMRj>>a?EqDI+`iFKcHex@(PVD@I#lLNjuUyQOcth5SPS
zDvltb8A^Xfs@qETA~qSpPo+f+Hea+U=*A#bwE1@=$TK1Y@F*lA?)%>#jIwv>BXZfwf9*NH;o*2_io
z@jle6W1J?LfcRM+f@94cwMtJ1^!fReMY;)jQ&*PL3^b@SWks5h&yfMyQ?#&mxnM(3
zj2ET9OL)zG5nNwPli7VKg3*$a-~9X*{iX@GfepFRPBk6p&f{2>M0YQRyS_`kkZU
zZ?tk`qsUdIz~__s=B*_JL{A~>_Rw{ykB_cnPL&U_wwV^LsSjUv%9(1?T{iWg4XQ#g
zCl}63M1|5h%uk5&wtUOixMyd3H@KBWD01y|IyWmYvB&{(T*q#WMdq4RT;+ceR3-on>iQd
zd%;0&f>>OpgsQot1IE#3^dOi9EEQ$&_jaq>(sh;jwCNqJ$~bfSWyl$mUO!TZd8S$62!EaS31K^QJ>
zm&ZrFFZ@dK?UuArU_Q9SWV42H75f{hX)}ukN#mYTH}90ip_lNh(>~Nk8VQY%Ic(oG
zTKfTA5!FRN=iFb^=(w9GIQr3foVCvpD0zP2maS=~e-s*LTrUdZqR#Wy9wVuMfR@_r
z!muOzZcy5JnsGFUgx^M^Xf0I)P9#pzvQ6
z+D&Eg`ksrI0w~Mq#_au0bX(qe9^S(xjdsY>znnu-KBiv{Sbc1SV4|&0rq4!wc&!Gk
z+VMd#aT`-n{)#y9rSf4pN(+ejkA=W8ht_qD6dY>v{8k9a=2D)Cy0La9+;vr@LWp|l
zl`b*FtR{^5Vc_=ny)2xsDjtjBHmts-(F`u+vFc=n#|F_unsHfNyknDVj{B0064oE)
zEhl{_G~G%*qN2`afZ%axD#TzDxu7?i=hwo}CQU*&+ayL+li9T}OUhbtxumh3;|BzLCE!}8sc
zxQ$#t5e(l7ZCOn;Iu}_v8Ew@N0QsXHJ&S^>4#jzujP>o$tXEeK)u#IfR6VbApKrt_
zb$>Ir6cEFyv-H@*X^cfZ=K7dRo5->5>^3!(-6htn`P=4XlDM3F2a{o?B7TTzFFwX{
z+bZQTT+#3t@vfVRg|EYIw=?BklxKlGO<*`DtPI;v2Jc9!K>IeOC~DuGYzxdfirch_WEFbr)fsFKe7&7hTbDzaB*#Qpd0oYxH#14B3V%%N^ohumWDA6r%j
zMHWX%pf3f1stZy8bXQ_w>sV{m=*!EBE{;Qn{k}FPiC1HAoReFJuP0rQpN-Gb(@E*N
zVfdTb!K1hc$d3_o|KLnByx==G(`vd7WuH$o`6xNm^^?3}WQQ?FkL0k4z^_0!!fQa+
zvW$lQ*k5IFhG
z_qsQ2jbDvlhTVFR|G28I-C4`1115>FdsK5NGQcZUfJeFNY@A{iTU(hTOElsr)Lp$^
zfl2>h7`aT@nd)OzTnE1Ufn}j3`~A;y#M9dAg%bX@lTlBZK_*SOvlYW4*1pu{_Q%&V
zdCX%uV2=0Qb7H-TpAa6pks@6-N@yT#Snt?tqUbZ~7?0q?xgsl%qT-@-$4C!Pu%G-G1k^!*R|&sL)t(GX7iCJ%hh1liL~2uRa>*mdj4gAeDi6D;nrjL-wshyz2~&4`Dx+RxHoqKXxh!Zq09
z*}9x|Yi
zXE>+bXM}*&02Xx!`+i04$pLzSqv}eSrys%NQn>_YJWVmq2>o+r=y}=|4-f<+OPCyn
zNGIBEO208>056=?ZaNY^5~Q~xHk}=@c4#SvtoN-X52?|apNUGp&>DHrH`$9;?caF$
zf`CZ6)D}tcj_=3Tk-K&q^-Wnq?Se|c{9{jhzGe0*WQ}r~&%TMWl0d`Hk}7G-t6dXD
z0ymJ6&Q1sC)V;1Z4X%m4i&DJt1OZU#zN0BVCH70Xo+YPB=OMQIhZP&#dAFa~Dsgy4
z@&S3W+^n}q^jybO@s8Ko(O{R`wvWtR0$UfrUWnj8o)R0mKD&qLpTQ?40cAYl8mGCs
zyoOPqDGrh^Es2!OWM9PM-&L#90m_esc1i54K6~3W3Q__jgn2
zTUnjtMrUtpiNLbiOZVQ;f6a<*vj}Qp>?R6h6W~g`KUIIb-LFHcaz)_P5Fjz_E`n_c
za`944vvY9r?J@ukE1pyIi!O?>s7g8(*^gk){Y#%JZc$N5a_d&+PoCTOp0Cp<{J5wj
z42maf{uq|FNw!tF(5?9vioGWr5DTR7>efDM#Z^oo*0e~_#aO8N7GjYh7Ki*qQc~3O
zOa_{aUTv$c)}`)zydSP!4<&TAn1oy_20g~oRY!ATT}F1x7WL#}4A3cEB%hgdXt&k9
zE;ltEQ1I_LuZPq#<(u$Mw){MBczT5)`6afeMzmz|DlEjrJ3E^Kar{O@ZhBO=kMTG=
z$e#TIf>4?~{a4S8{XcqcjO+{y|7*4}5iqebGym`D|HmkpSvi^h&pkKgf4c2Y{J6wL
zDoKcAcgI-m)(owAPL}A>=@OfRVy4YT=tmwauhe$3NIFI3a0*f7VnxT8$Ri}ul*vC4
zL@D;pg2%7XTxB0QyLr8u$pPxTH{CeFcH!L&>u%7{P48i)497@>!LM2Ryc$fxI
zZ#2?ZT=CS1459#lW~P>r^NUUh0W!hx!m%gE2(%9VA^k*rX<=iY=|=Spat)SV
zIC}%epnZY|0DOJHYyQx&k-*yf1WV)KF!~U8z#cMy;QpKdxKBVBEDKy%dtEaz`*O@f
z1u+qf(9{RaX?%-OO4;)T>XApW^z{>(vQX+DdB5W};qfZFHg+HBym$tg;b&P3ar
zIU;?NGXMzwHTe=#X))h`n~o@$cknDt$;5%cWnchcJRpc0d-sBZf*iFX(R=!c=;SCe
z`$dm$h6E63pa5re5GejiLlOjRQse``W!PhM;Q4nN1SDW4qH=f63k)7P25XKLLAi?Pf(6S=w;UX|^a_@8e@iqLg2p~8lLytNSkWdQA=0i-l|
zx~%V#t&MF1r(zxu@u-4B0KzB{g7&@P!hiIIhNo%>{KHA~-OpMBZfRi*K>JP_5tQ%G
zR`g3dk_h2o)a5(>kmw2PHRSPKC1
z7X&)HqCI8ZIhat31+dcmN#$U2hZUf6;42KFQbiR2By!k)eWC&(eBlAu^@+}S46uvb
z3{J#u|XGU-nW~fB0;sUhy+Jt!+xi%pc(Q$t*ULMFd0~fZEf>
zdGzlrlqU?L9_Jv+N)n(8aelqu&9`lUZokyMEK4shbu3FL{b`zc;!G`dU}_rK!QLt(
zS$-vr`Ou-oSLJ=Z1oW~)?
z!i9311hd);tR8z#*8%R)Led(Rf+sWod4@fG9bL|r*=h^NeTl@{*^C8#X4Jq2El4$G
zQ6N-{tOnKL#1?K;$FZaIiyw{?7sJkZ-d$wI3s}0wkX0)5=ZRsHO`k2Rxh&HSdi>GH
zb$*Rz%tLn)>P4&1BpTiZ@6&jreI}8e_vxEsaX7YhxGfyVj)D{X%p-cp!Mx|QYK@iM
zz4$baf;tQpH!7vxK84G~IB_cG?=DBmf)?s$-_>QLYmg95?E4KtPVn;px0Nozx`jW4tQO}>Bu
z54chaMw}-7YtviTqCgX8#glJ03gjVf^Uzx?%*kiR8YxcXWSyzz8r9Kd+^lXVA9Ihc
ziO}HKvKSxpk>@8@Nweg~)`K+RLCXi^L(I#m+w|SyzU_%D7F6TviD0JXG}0XUpT%AC
z+07g$_ReCqSVy*)ON&VSK?C06najQ0WQ|_W!L_~zVQxF4r|3D4li1X1G4DfL<*Z5B
zor|7&AyNC_vOT_Efy3+abn;!8H@o>(e>b&g*(Eo;+3RK4VQ!ba8rw^jo!VohdmVe&
z+j-*}PZvj9nVSuF#>jEwJ9)~r6Vr_#Xtz-pFgh447gag)7ANXwg<*DmCrYMdWd`1G
zk42w;O0{P0uk5X%ali8I1gLD7F!)ExKTx?jHVdB`{^XK;HvX2*C_%v?=N~;xggv|D=vu_=B>#%%QSy4zp5op&JiQ4Q?})`D37
zfF-ANlONy%VXmXaKX8sqGDZH0Oxi8a@%|*`yE+#-j^Lb0Dl|-KBs7xba>5~FAMKd;
z+ITpMMc3GIPngNTA2w5WXuZmO-fu2ne>(*3T>F5|kg`kBA}lp$dvhR~;RMWk&}L-cUq3wU_#FKVRb#BBhG_-O#$!uQyG-1yUU8qB6JoDs
z*6Tdim|cKe%d$jkJMs8SX4?OcL5hBvSA*xL+FEA7we-W^O&}aM;P3ak{nu(3U2=C(
zA|$TAGBIDq-@X|@k~w8IInHHnT?sC+RT?K$#Z8g=?C%1er=J8uZBlg9e8T3p^Sl*<
ztncly^@p)~?xNcGe5Vop8HZ6cT&s|-8AaJF|6#lJ#cEOR%~Kb1u{!v`YuNqf3N+f(
zLd?;%M%#>g-SJZCn37)QjQD9#6x`p}$+7W3!2>_LVz_s*N8a9e3dC|`*f3QhYw(gA
ziYb5WivL#@N{^fZ7`?Ade7!}%X}M`JYE{6eq@3kBF^d
zgjOY~iD5K1Gn(7m;at4P`Ih^enOv&42-9_T1wu{N`WOiPzzU^iK`hZU>#e-w80458
z##+|PrRa2PlC>EZ8mt4x6W?51LE11khikWDC|rup&wjIu^~%@E4vD%`SKAJ;Ku@mr
z+IoG}Vo@Ci{yev8u*R#Axb}sM-;_V2Dmww2^Fh~%du89iIR3Ak-m>x_qX@^!jV8wp
zx4WK8n&jUdq2{!sc4YCSF7>stor^@3R?E2fL#I=n>=<&KLiwX!!fB+g89s-ouTl(M
zHnE)s>(!P_@&15EJ}>JW@%7@K!ykbH;kL(mOL+dq=zKm4W@7glDDy5$`Hc~WE11!!
zNorNP?I2~fn(N1V=9IQUu#|6|@%r~>xs#Zy&KY{eMfsS|%$|mxLg4mc3ek=k-8w2i
zlVUNmX1xhww`cL}0k#WFFCW+<(?oRjUOD?yv4B|HkeX-eqz2Hej(+o^fe|L{&i87z
z_PmBpxT%g5$J$Oj=xsf7lzyM`gz?pTc?$+kLRPYy?W)biz5ga%%bEnd|O4
z#p#|6p0p_5I{V8>H2iMkN1i?}n?E77T40BPV)jhaZ-N(?2J(ZxxdKX!G>eiw$+0qG
zBHiT{2bvjLSaAJum&-Lrep;fI@*4M3t?8+4Mt8vRPA_fZzY2~9v$y$b+$?v2xYvBm##6JbbmHq4l;G*+7GKJ
z8y&98kY5|FTX%A{iN9D&WmrRxFFU$a;Ask>hEz}k-9D-7%AB}gi_2AdyFrevjLFYv
zCGvH)>aD-KL{m=1sW{vDJx+PicU?aZ*8tdO)R>1MWo$6`eJ<=zvH0a-Me!Si-p`5N
zs7Ck<=B$7z`mZmeX=+V_zOPT1&rH7pYyZ|$91_E`*y(QA0u_U}mxre_!K
z)2G?L(@Se5aUEvvTzbSr{5If~H&iojTg8sOzklZ^qfYs&s3Xu(V7hWIsnZ#{OcHzu
z$Z~n#P6xm@J--*f<$#xsD(vaheqJB13-^4tUZf48u0=8*r<83U7}{DtcU!VAg$oBp
z3lrK((VfhFEQW`+Lbb`cJ8Q#El0_I0i$P0Q$01dRaXadI)c#rdG+wfPmj^DPMg_og
zRl{r=DL1>z+HOKg{uMqo3~2;^5};NVcPtt=W){qv#_)DSsS)3cegVHdY0UpCe{%j;
z{``kl{V#Dc5wNha|0id~%E8R?f6kvO9z4l5o3;rw3C4PMpGeGL8p5fBoAEll22)#8
z45lMfYl%AJNE!_?Aq~dzLM;%a>!L5g);1KPk>0|$k4uZMzJ9*F@6(Rimk#)=9=gPY
zh8DDSQDkj_p&&e>cwm3#0H&s<6HuZVK_dnWm5*x#-0*P{LHYR!>
zz$Bo`0sE9GF#51a6b0zj;z8zwmNG~Hh!Nlu&3#$+umIA7P>N*%5Q701#&q;036zd>
z5{xskhdPb!^I>R3^8{cB47M4QbFicy9=KCn8=eXLA(4(H?})t&0ZH&+(*>po0Rs2L
znGFFgfbi@=86~o@1?+*X1;G1Ouxw3WF*G#RG&Ls$%tLB?!bO12`L&;FY^-CB0DyIZ
zVQ~Pij{u0Ztg>?Cu@U4CZ|T4Ztj&LN_?r51oSoWYYGaB>alp1c{zc8VKoTa`vTOV;
z5{oH+y(RMsTav(4TVMe!;Q|EKd^UQ1%n778hnSB6Lb(C=OAa2Rf;5<3{%;YAL%$x^!#!Q6doXoboeE>|5-(|2M
zlG~F^4b4nIT?v|8po0W=6x#T};2xnXwrF2cy)o3gnJ9w&;6Wem%TW&V;DMXXNe6_Ha9eXs
z7pBe2-NOk;CYL+Ja9#s;x03v7pA4{qsuG~S6jk3$_Q1WlPhc8)3~PpkI3F;LEICvF
z3c&a6<8=i2+dl9wd>J-coa-hha1%VDWz?wWr~e4puVg_+y6^gyAXGkp{|>-e+<<@^
z*6qx6EM|*QDI^k#1PKBlJ4q}In8CKd2D%af<~cMFN+LW4%3J5*Za8EqRDL}Ll+BTa
z)nJp>HpZ;64h0KrZw;4t;Dq)jNVnGc}rvE4?fcabuWF&!tc@g3FZ=-|X6Vp@w
zK7s_j!bk7LhZ(~HbQJX4?!I5}dJuPY2_i@$I?`(xyfoX;N07NR4QR+a<-PpptW<})
z8xF+SU@{?;6jdZYZwq%;Xi*HywdTN$wlz!dKpx*lQz6eiK?V#3yA$IC9c1v^@-IO3do7k9q9HmE?N#25e>=!QE59zPt&fH`v<*tgp$xWCXH
z2ox^V6XoPew!p$dTI54@rhNWN
zZ+#z%dCA{iK~#W?d+g$W4G-8#V1<7I^HKnT`(8m!vjh5aiN}xI0lzBKcBg_X+wcK;
z%wMO)8$&Coh&{*+qxq)aj9nq~g~Y!st7<{{eM2X0+Mh$EwO|Q%!
zPvfg&`j7g)s%^GrsklRIUmfof=6{OHr?8W%H)GwB^YrSZtC4}MxpyzB*ja3}|6Hx}
zJy5DV)fIL`Xn}t{fEmDZNN~9RBC_>_eG&V>zbdw{b08hW({>a!0MoV7JGf-dA!lX9
z&X}#xTFxQ_deNi_?aD#6zbW<&-=uF9b@0H3k=GEI9l*19Y`!Yv3;`CpgfT@;vhOLvJUXW@OHFNb~jEZ%BIu_i6-7zJ)>zTq}*k
zOh?4GuJ!s5MmB~GbykOoaXt)JnIH-L`f?WzEQotTA*_+wS{`24xrJ5@6fN*K)0w<$
z4k-1S8#3`bTD*V89vQXka$lddZF=3z(w@#L%e{Mk86PpL#BQ>Q&M+1t@qp%Y#&pjY
z5~vDsbvaR}WSCWN_w6vv+c@2tsQW0OWdhZy#(ZlHdX2@cR#|j}^qxY?TOV%wlUpteZuLF|X%~_ZMz1C6%ERK|r?FXA{4ig#v5XW1T}FmXv20H|qkE_Ea3{8O;HcM0(jZbB-z}4&C9`kL*UfDA%z}vowOo~HSh#xo(CGencXS=eTo~UPI3>j37klSX
zMWE_e3eCTr2BW=3rmrNoZ$Zp}ZxPI~E_-}!c~K?)zrUYihHC+bBekt2_}#-WJE{Zb$?B4Z53+^HVWmbmlx--+efbNjV_OOd)>AqWXC;$^Hpn8xkkC#%Tq1
zI$V@}#pqA=K9%S$C1z2w9LBkpbw#-{4F#GT74edUgCs-tM}shES*=Z_mU~&mp`%65
zfFreRibtfuEIgr=rcJ%M$=L@-ADf{c$eI9B6bPj(02
zSf_3j3NyKRGTO!teVWn2{^H2krCa@>jHTSL4dZ}-98q>*{0HHl5Xr7_p5vEPKu^za
zd$HLr-j)=P^HcILvE1pwV{C0%U-=sLqReJLA@MnAkoX)-<4KDW`Wj#Toj!nbzM
zqo(U(mEqhf7DDm`6j2n@RTM#hjXJz=%=ZG27c0)szC^f$MQi~!XTK(M*E$z{UY~tj_pU{;xJGdc$0<0jqUxi!S*p%EZ-Y&hg(Vr|&K)FH9c+
z4WM2ga-`gN*o4HFl2z$Hf3=-Q`V){)}X$8$end-i?iFO>&e}z)E
zwu2YVN;fz-33F7V5!^K!10KF&auPdO@b`##A}YlpDS9cAbFXsQB};HAnSMc?=dl>y
zUtab0{9Yrjf`PTVyYxtEV{-mPE?Fc?8y_B45|oy$nub=yS?TJ?>@)GNlQv*AV3`rP
zlKWvHW~m*zS;xX|46+@ZVt6yRb5TovE9*=irGF#}pU0T^ikcNeM3t4XN$m?18ac5n
z5Y{lXG1aslEaDjo1Mykwc|^_AzKQ>nbzDTd?hD84*Xb6f%E@PuTLtHZqBk9!I^^Xo
zrMdV^*V{&uedO)$S1eQ`UZl-x-c_Dt$HPsD(f!R)l#fD!ymG;Ua$ihZ%IFoe82at
z*@}o?nJ^`mblo1b864r_kKCJk+`j4CF
zn{T&4t|+5DI(fXjiMGm-d~->Sm|f&|P5@No-lH5|zSGgEAmUmR+bfxJ3%A3;R!Fp|
zdR1v#jEMi|a?s^X_BY_+dIFWn4z;rvQ)Br?o`?y4UJ++Fu$p@{Uig&N&=daKPii{US}qRJNl&6c!~z+mwx8(xaNS`W}#?zew{-M}NOS4Gm}JmiMT&eRpCr>l$pzn2Gh
zTb9N|hj3i*7;oyVU#udUu>bNx;G$}IgX6cK3(plTgm>gi1DpCmQw!!$(0Rz;XObH9
z))TUHLXMYS=x~BNS)i-Y$?g9Ub`C+JFkO&s+qP}nwr$&dZQHhO+qP}ncK3WSo7wyk
z6H&X`SKP{*d6IR_{H)>2mq$u6vR}hs`0={LF!;flw|HG7cn6nNU8Yqn+xHxJ63^Ry
z%fyD2Ot+wy8N)0m(Kw^NC$@J87D9~^dh2g|iCuR!}G`(%IcxFxJ0F$@u=
zF%9i$mr-^5RQ>)gN3V~oa=Oi1=SWL{CB*Xp$4(*zaPM7wr6Re1dx&=q>G~I(wSWo7qULTWR?Er=Vm=D*fj3
zKlU*PxA&gKA<<4m$G9M(VMBZ{e%LGZoh)U7{%i((-ab-X7fnKxgL;-S&qRfu4a>TV
zBe%ZH49hC6>S&|cg6AIm
z0}tY@LT}@V3@tW=ogURzVzD(pfA3HA*{eliL6UBkg^(^?P^#W&1MdXnPqlvGTe&=o
zOon1m5)Uwl=X|2OhX_3mJTw_fj2
z65+_TsbPYsDiLZ7J{t~a4{0xwR^AM^#w_01Gm_H#xqwu5-}O$p(B487Oq#J{VZ-gH
z8gDC;cV>Y47soiEQVt
z$uj;`s?iDWJjkD7=effKI-e{yF87%H=pMhi89CNzt^;3cjYHR1%a1kPQ;A~tmyEQG@
zQk-T*6@|bQis7r|AIO3v)tk|IbrkQAiM3qFrD+qhxsj8R`P`o
z7xpXF#2kfepJs;auCveTR{Oc2N~Ll7W;Z^3Eh`ho9QjFF73pP29{)4O{s(U!mEn_x
zE6Y;LDcw0;lsuu>g`p6kmer0Ayr`%)w9KY#v
ziMcZ=z|V<&JIOsTEwGUAL;ev;>fuWsZ$l{UxKVyh*7{WQkJj;_HuUKlqF{N|lw#iY
zWQ<~k(`0B2@H{~?Em-}MNb5YOuU>s}Dek(#H5ZE=JGC^BpQsJc~dj6c(Lj@A#&R3pl`XzRO7A`v@pT|DG9{^$|%?
zH{Yetyg(I!nmPe27u(qI*PObM!NrYES6}AkKSY1Ky4*
zCSI>eA~NHR(U;41N8?klKApKq%&8IoAX(b~o>APME+!86)y(oJj8wYt+F|e|*t-9E
zpid34M=KE<4gMErQ8WH&wev4bUTb5|F!khtz^dIUiJQF2}oZhwSQ
zDA^9Ppm&>a)5~KKd-ovCAVt-^+E32z{h?{N0oLL>%o5sHHKH;ssXN&5?C0fl6Ils1Sy*U!QDlFRbBG-0UuFj&6MS@YZ)wO|#lR-nl=6*DY+>
zj1~Ai^hFCBCm5EMeoq?LXV3OfV9jl(dK0pUoM~9x3@M8Yd{0W)68dzgv~zDt2ip&?
zx!r^AVkC1)_Mfx!OpuA{`#kBli3?((M$1xymyZC-G$(cf@9|^Bx{=X884m_iz`Axd
zWG&=8`vOKxDP5F@0F{eIhhU#=x2TyUg**Ht589^nG48qj*ZaZ2sFu_88G~P89mdI~
zJg+7k;6bPLD6J}HE87Xv|5F}9=>qI
z8JBRhYkYsHkH`^u0|s1QAf*ClubRq+-?p9SqZBX5N}IM)Kb`t~SDCQ#`JV$}?tQl=
zC$6~P1gu{_KX`pgFdb<|e8I^z?VPx{t}Pvlj<)RVQSVUdeH;#q;LJJfX6q)-;ATSVW0{NEyv>fkF_h%|h;WdGQ_n5Zo)i6-sapOlwiyFg4CAX|3aL`BlDa
z*W0g^h*5Tj3tu$ynX58=S_RAG{r}7A<0yIrP$I3eIfwWDlh9N8kDFzhm38A2o-eu#
zK(AahW)@=4{z0NYIUmnF9irI&gmF=YLtUHsg3+DnIuxV5Q*UjV+Gzp?M*Z|gr6&B(
zAfo`=`Nc^(&&X&nU`F2asSFPCd76c;aZRD_B%D7}hUTezo6+n#ZJ|Sr1Y3Cut$=m0
zKUeRyx|N#vMN94DELq^c{=T5wKVTXJ_vHVfpffZ64+WixgMsb;Dd^11jGX`5!o|YQ
z$o_vd^#4Bzx4F|)0Y|5mHM$F#N8sqzZeV98XLM&LNZ7$0nhr3q3&b5dczT4FN!_uYxz*gzOaV}8TA;t(ZK9fFdsky=q6y6paipj7@7dy
zg;njz38WFAJmwEj3XZN8hHR9MfOUR%a1u`f*a^|q;n}IRvEJneHz+3Hz`b~Hmna$l
z(^wfB93CB;+P|R(Qy^3^7cqZ-9}r0kupIXqkW4_@Slih@h*JQz98^IaRb3HOqLQML
zx?+(b5U8q?lZ%VvPoB20(&{qt8Gz9cl+_VHAqbD4qPFt&s|;%%Po)o(S3nT9_%8?C
z;$zjk6yV;yaEdb(HS!eUg%L0`NMz=2|!4
zb@a8AKv>;aogLkOfuCufeLnh99~20%&~A3N&w2c>Zt`Q#^0&oQ%GjfR?bfa+ClZi^mJh
z0~~?-HM%yp_D|z4mYjxc40_+}6r}#X9ync4n99zc?9#v-ymAxtfZfMPjUZPCH@heO
zcWz1Mw+v@5tx|4y2Q_>hbhV7G
zU=iFV9mJnVS}VId!2P#0Oy#kk>K`-6@}G?=);u4pbA8(jP&(iYVmbJ32oAr*N|nedx(a!2SQ3ZJ+6fAq|i|g4zLO3jGk+
z{UxvA4*?k>{;(YxfzwHT3D5wNj`4?~3=%(s)&XP|{SdhQB~RgjOw+!C1U63o2o&5x
z_#u*YANInrlXWlPflib^f&@NP{=497;fGl6`ossBtbG6vXte$n%)i$96XPMsdB6u*
z%=i|}-<8c4Y=UuV0nP0CO$ef5`(Zx>_(BfY-vBXx80J8WbK+GP5I4kQBp
z69G{Dn(P17OxW`zXE)yRlU^6-2}<2vUB9iPAaI-h6HrsoJXP~6Z2Sqm^;@j34~S&=
zLry$u>k7(x(gzfzL-P|^m(2dHp0hs&bNM)vd2gG!6OglmtL^`bo*M)@{BKy(
z6e#|QmAj}CbLi@t`FPe7-Z>n8=mcx+^SZS{sSrqef*0KqI&wU_Aj8D_b)L2Ue8Yu;cwS35?J&RjO&kj#V3toT8uSfH@gHW*S{qJxO9t;5C9Fj%!
zz4zQ;65T52m9=_XzuTe{C^InZp&jZgC{+e%`EB|F4oIirn^q~`z)?Zk0L`=^C>y1&
z*H0?mrvgosBK78KZuR)G!kXE%z}DE$!w(3>@t}bRR_tY;=HsOQKND~!a7MS#&D?d9
zrY)qc2*68&u;PCd$S{r73HIkH6{ViYR(#NTJklqcnY(OTMxL~Vdv3_oGVTH0J-l6?
zNXA>_Upa1b6^u|`GT@{2t(XpAJY}MH9NC_v5wAE(xu5-KXVD)3HTK04Fl)kHj?R&ldF9yutJVHBu;LJiVs+6o2F%@nG3y4w>%{
zR(rmgu$=axImK;KYU!d~22EuEh2Z*7H5>$dRr?aBQ$;*)B4|G`TR)>yX+F5zH;ln+
zwC!J|cVZ5T6C$sg;ycj@#-fJwWim_U_g0}QyGZXW5!DJPDDVPUn?z_X3VWzYKgY1a*D3Ws}t&H7X-8=3AH;QrBNRc`UY06A_>;`Jk*f@AE2NfOCsv2N8z>n
z`|s&$ZBV|g7IP`&l;;il=+jdKIyDK-gkJ^t(|;3m6b&6%gr7xHo?T~(kU5$G4TyRW
z3FzVk?x?4RRuNS~Jb!Z)JXZgUw3%0|EWlcNJ}5z&wj>tn9}aAWcLL1B7CZ|>+U&{1
zR3KEbR25oKIZV!Sle_yQ`-u*>KZQ1e9O0m+)m%3_Hv&%iL-lZ4Aw
zc_79PLKA-8T6vxhi^%~AKdKRZ{7zirX0Wj31Xj`bzy_9hfod(toeIH@%_NZ?jq3!l
zZUw(2n=Qc2fW$;cq<@yxz2LFtaqAjn7TkA>I>YKt#(zL>Wk*1dxp|6eIKupw6}$AE
z%LF?b*A|m|cKL+b5QGas!Ji9iPa@`=_3Qw$0&>o~p=>`3vt^o<=igS|v~*it2)04u(7_3}15)?mDk*tV{dX
zTN8#7^~x(9edj|RbJ&PeVwVV$ZT2Gf&8?r4T7;Mup@A2nYoa=%yLZAJI}Ur_RD-6f
zzPDKM&qU2B6TrB*u!^3nLwoj$a#}Q0_)bORs|QU*8ALK2G?FZPG6iQ!U$jcxyynCp
zd!H9kh&Va7@tuIr*Gs2_&dC#Ac;xJbg&Zp<=IhE+zvB)m@ol9Rqlq9JKNR)wjP+5>
zAxXoy26?3*=e8`%8~wz^Pc?E#BQvgr&*Kj6!YAcmCMYxK9JwMqs8Bz*==PwAdzEaL
z4VLX`vuD1rxrT^!*%JXKL;7|=ePn69-J#>(2VaR;A2OCyxu1*EUb;-N5M0>`s9g#~
z#A1moh{ld)vZsEF?L0Xu{i(;0Au}ESh7~o!dZDZCuLAVz%rQ*BgNVrk()n8Cpx^?a
zx8U>3o1^YMnQ@Emt`V_o9}#SZkZU8C>V?aji9OnTI)~7v5Usg&X(9ERgUStS342^#
zRU<72@d3v%KopL1oB1)Zqx8jc(b`>bN|?=d_rynh5k{*b*Q+Oj59bfxed6Ks&dL;d
z+=bO~E9D?!;JcnTH-T6l{K8z?B|JXO`bEBmxKf^9$Gsz>4rYV%x!ui8k6A*!eV*&f
zX7cjD7w4E)d-)86qaobsy_>BlKawp7=h_=?x?-xAhs3P?&8?$@XaVm-pVq~Pndk6o
z-9iPXUK`m66P|zxEfD1x6e8gS1?(kgnDUhFJNMHdbY5>l7Q&kubS6X;h%3tjx;8i_
zy~C?)(9u%eI34W)wXr?XrCi5?vzh%pYde12>5(
za{}^t3hRa5sGAWt@hbehd8w;q#X?k`wbTkaNZmfX7c*P;?7;Cgm(wc9hDenJeO#!_dEjgI!J%;3ULw2DR?v+y>Req2i3M(
zOkA-gmFys<6?9tnVndF7YPrxuDL)AtVulRV%Y|1VRt*?_Qk8Gh)E*x(>E^L`R+4WxNgns
zR9B;k=3+Fd@Cz&?Xu{`-cNh9t#0p7R<0-VRwDe@DvWG75^T(P*VDLJgo1O!QMTLM;
z3^~k|2io;TGXo+6}D(W=tkV(l@~*y$Ug=J#j%saR@6g#PH7
z1AXne87|m{u|B9Bwb)e?C<^$YL47>_N6o)aV~m2jAQ~>`j=-k#Y;u7
zB9Iz`)-{znkT17R+f(9C5`r=Gcd&-F3azvJs~g>qb_&N;jD+ACNZI_4{W^txPn7W*
zk&GXc%c{xrcuuSYcuh>G3JlL^u2YfMU;~TAk0^eET5Dr5K`xhgx{oc+j$)W66Q|KQ
z_QC$wYZd3@efOFA{@S@2i(}tWt6?4no%g5$ixrXkfkr
zc3p3L6*I}z9SYgh<;SxN=%KkoUnxc6uAv7wM3H+2e&0fBw-lvihsG$np#ozpG3mv+
zqz)1GfY{%Zdd1mi!NqZgEblg84;Ec$0aZ&^5!r3V_Z1P)5g>3T-f_%tia~@>{c8^rXP4v!6En*LvSY#B!z6
z1WjB^F*mjV8Xq`+Bwi*@qPwFX#wLoD5osn}n0p%>7xBV=-%
zrEyaclF^+FMS}GzmIYT2TH$O_-&82#@A&L|F$fTR>n5w%n^%e#`YB5WqHLlC*^qYA
zgyT^ei+6r}vhNJz1;&*@@IrWrIpzLyFqTD3t;QqMB~Cw()6mQ1rLYW7Ya8?HDh28G
z9c~#cXYL?#Kc`@N7lHeY5_%B!Ivj|_8iXGd#cZ54;b5j;q<98Vt@-4g=3taGry2Xb
znaDi$_z32Yw|(efs00@OCj8<%bX%7yKI?izQvE7m@E?poPP<6E(VKp8eJ9vw^|kAuy+
z%Lnvx3!=%hO$+@v1HGv&kNc*a%N!1d*9dTp<_D+cPmx$uKgBr%?mi8*4^-Pzac_suSJ4vfu^aK_(FJKjaDzo3Yd
zJW`+)#@7o1o+=zo)H_xWmd_^Y8p#mPy~U`0E)}=yd;4n*k2u{Gb3{PKuA9yPspH5Z
z+LzYWb6D;wtlTsHXIL*@O{R6aOD&>(ul&tP``irMIfjn(b%v~ZN)Yq|5tUTqCe(0q
zB8TbX)ODTBq)%_v8)S}ModCka1yr>aga-IW(*qz)wbM@aQfQYhNp=fiUNV2u8Tg_j
z%9zXTu|!{1;x;8sAI(*(KEyEm0+u8ucK5>1PvILTndN<=}^cI|V$(D&0q
z*S=r$l7x}AX3b?v=FS;=BOzh~tX$jc3S{c
zE_K*kntQ9ha>ml?bcmRAUAA4Zj{0+5#RGl(!j`APU$YtE~s4ndatzgjCpl58hm
zWGQ%F^$bl^&8C%fS!RKbb|xB@anA-8Qc2Seq*&WIR1MyRvgS$Lai|#6$>xV|RnzLi
zmnjz$wVw-)y|XprL*SV#)s6M&&|TFi)?`m=Fv0-!oGYc`e|i{JT3M3XW4$-1p~l11
zcX7KuNY(l=wkXVjpZ_}LYb_^LF4l1Yo5&+>N(82QMhiM(Na9K$&84A3*!?|_#czWdlDIbO#&c|m8dfl
zE~60^l6kaJdro9h8|W+x1L?h}wlN9sG$WEEo)6vSSUESCqb4;LMH|-uYL6Nkc_r(6
z*f!o$c&;Y%VtezA`-+|)(3(*Eci`Sh^#!s5CkxbahuoMoK4x{gZfB}>`6;tJhI6{a
zU!*RBSZhQ^XAf!v)eaGqd^mXze6hPusX2u@--1dD?9QsU#BYR4>vo*
z(#JBrbv)faJ(RmwI|@H&;@8~=&R0<1Pbi%j2SG!I>9~p7OFRZiL%#0ks_hR6SgA)8P?s+%IeD~d-zL_Jr+JN
zdyMC(n;?e?K6E(phT=17y22gdXZkA^PL#XK+40E=VCn48+hu{}0s@%!>VVNopWIUD
zMU-5MUhbpa^5*r;98toNYgUVdpkJ{NZf5?ABkKqo^SQC;(^A?stD+Is<4iyDYAepb
zBq2O71>GFv!Ix!dAsT^Ty>YS8)EydcoW*AWFibiQB4YV8?>vW!SP`$DR&)6;PA#ES
z@F#s3cB`_r6(5x+H|Okh@I4)?+RmRs6rP!#Rp#antR#v!%|^M#vNKpL1>hPbEiJ*j
z##NwA)n98DZ|#9~3wB!`C61GbzrXMa6cz{4qtU2x;FNT6#pY}p#O>UywXt(zhc#?$
zSG&6g_5CaC56dTHG3K58XyX=*S_jH3FDl*Wp#eUM*N`GA%Q;xM`h*_NNT;)_&;@in
z%Ynagl>zH=7QGPdZoGLx#crnIxaMy`Dr%Y9R&y{M_yo!GI3cqs=NL&^r}VTleb{!&
zhV;&d9qBZdhk?Ca0I@8mI(SU&ZGqQ5zwu;v=fX?@=4yze?ya>+2_1~#@h?Cfy!DmX
zHjp5HVur<$h2*sv)h=v`HSC>d`D?R8?kBVUua)^v_!G(DY_J!0;raau13`VCEM5ecSIak6h
zkeb_nWZH8qkSOl={=gfd+XJmaWw4zcLPKP>f=xMo&MOKy6vAC6a-LpQ#-$Fc_G*fF
z9A{dPyR!cxypTk*`XsvZLnKAa0n8jA-l-(seEi|vIev3kY(uj-5f2>Ac(>926y2=0
zGIj$0`T0{s=^dlt1sXWOmO73c@mr0-9)Hf+vj+*DPneLz;hG4UlB4*nCOH$~OtY<`
z_0V*OV)+1LC8u65&X)+y&7${)i+g*|pq;2Ub*}1Ti1Ff~HzyIM@#;I$L3qcf>wR*J
zz5S#Sm_4mNoq;X5<;o>y7VKp=Iq&h7kZ%{~
zr##zlN1dD@VZ?I<%{2eWA%kNd+7)->>RXEwniRFZ6b+UqKQFsaqdT15;ol-nDiS0M
z)Q<6Fh9^fW;C_M=j?1+*M`+Xas%y38aKtGf1eS0xq|3aVgDIG&$vZ-QlQmPzkFyec
zq!f_(_fJ67{!G_3gLv$AZR+8Ln9!`gsU22TKX#qXVn+oSWs*ZBZl~xkWD)vJKop{
zBe;fiE4iju{a}Ftx|T%Glv6o%IGK472N*pPTm$_AF<+{-Wsz-3!yebTB1SGHnlvln
z1McXwY8R#yy!My$*n*fbvSzuohzPo&k>h~KW(6?YK;DPe4(OfJgA0}l-O*B4QW5PC
zZEiJRd|x;|b3hz*2smkDhVh&z3v@wy*7B5`D>?0sbk@L>r8ex$xiuc{ri~b556L?R
z&Zvn;anp0im#Mce4M~%aMC3i!UVW|J;plvCK@q<6?oPz
zbn~C-l2yISia;Q(eKzIPwM$M3^=|u-w;hHM#B{bH7YUnKkS=$+;Atm1qJ#USg)Ixfgp_3p2cb
zhvSy-AOd7BhPD#!w#XTl69xnx+}yJx#WVzdy3MoEERTcbT)3TIN>ZcVY>kK*Q3l=0
zVd}G12_a!L2=rk`GQ*v7-9L^H4o`YW`hca_M}%`O%V=F!GpL53`t(Di1O}Fyh6X&1
zrQdp{C7~p!Jz^Sd`UR(QyVZ>P19Dy&s!_*3$^z`A1bq^
zmjNl^vM=FX#-fc5Y=cS+=c8T!aYLQ~U$Sjaid5a9o^IQTwny@joXfTP
zBJUu$cDAFo_@l(}th+O;^bsB_-0B=el4BarkmCN7L@XAah&RGsMhmgk*@OS^@%yRDndD8r$@LOV|AD&?aeXwrx$l(8UD<33IHvEh=Wk1Nzzf?1!jFBV%$*ceLTQmUBkPzg4)<7U|A6(vh9
z%7cU+jWzHdyn=tc&S^-Lruu5rdyZz-;#sMnijptO!nuD3fM0-7;u$gk`Skspd+225
zpEHT*UKicsE`_-dB2fnQIE&hNL+wvzgEaeJAOvO}hesHVM)lt7W{DdgbNF?eMQ3pG
z%fJUN(<0#`?fiW>jBf7b+hwL+S1B|Ms4Szy`VcLR_?DJUVY|5xS!kP+dLhx{ekDfR
zMpRx@ofd^zDA4HL_x@PvO*&gvrnz~SRz89+^-jIp|LC2osEx>_|5-{gutQ(5t>;pj
zYKF}-%;Jy*54k;6dt2)Ay0^;RV4+_KlJr7V2`6N5aAt4Yvq={gz^^igldUK9feiL`$&_&=V9Q0;dr?7>y2!Eqr59
zBi=CU+hs1Vhf1Z`8_bzQ1r*8$RQNoy!TEGdwH4h%yl_J_ahgK(-nR~dg}CI7gDJO>
z>a=s~@=`*jLt@P4LSA2<2U7p$um`oe#JAbHLaRO=2P6!3?bYTvadVLWFWAqxLojSzGr8s+=!b^8VTd&fJyz33vi9Zv$19G#)%=W)2=rd4|H
zD4n}GJlpN%_}Yi3Xl+`qoO>tBFolp*BGBgyo=
zzmFV*c(*W|oty_GR6ASSra+!UrfG(2W)7!@D7$L$%#R6vn84~lN4&p;uc!u|6Y%e;
z3-a>qSuo}T=o9xnt{qql;741Wl5(!@e(;nNh=Tj!u{YBTVnOZm+f;p%Mkl7(xE;HA
zc5?NV8L5|14u?p<%NCKd(bE6!itcZX^*z})$co0-T-WFd+R?*|aCX$SbLiNB8Okg3
z1fXxDcN1GaWX5v~DTUjaoPtt{^4Eyt4B`+88`+97@
zVFKF=wX%xKDTJv+-1J;K7t)*>Y2dW8W^Z;@Rm8J<5
zBF^VN<5U2_$3#-v4wpQO--O&AC`Ju7%>v*I=~
zpP&t5)dVM_M}FLbMLT_9G!)PXvm&fNm&~+&-{jqlsm*H&zFRu3T(yW+8b=;gXUvC$
zhl4oF%UzqMbaR}7p~c6-PKm3nGE|)KJ?qlm$9^)t#1)mUrfU|j3#M?xwD30lp(b5L
zzDEi#t9QmK!&xI$LLI{Z_B{)!E8B)DD
zqXxJci}i2q~9BD^mS{5#fQeL7|l
zvc?8;vPnu-?*&m8St@yWKtU0EWQr6beO65AownXYy8cRD#C!4HM7$jj?J-d{fzA*#
zMe$|=4EH|PIzh9c0s9zi;H$bDC()n(GDBhkdj9YRr_xwpoB-8
zV82WBc_@FgtOQn=zgj|g`&caw3=b1YJyl4cI-|HL&xg~P)
zfH~2j3!xa1WU?$R8FRXn{9cPCIc|D0veEWkeQ~=s8J^o1Qb;J+iK??igVE$&O?e4q
zGwqG|a<^OhqzLS&uiOYA#oVLfp=1cguJLYvNIj$hNQT9T{tx3Mh_tmw45-ujTq^TA
z#e|DkNGBPkATPH(;0Tpgq8sVGl=J7B>w)wXF}wg&Nj*QUX^NBPjl
zdl&l^&$zD_&slr32u)r~XTuRDZ-$lM>VwPq^rUlbVFvVs&u2Q78aU`=R(AO=}jR0
zie|01AG!Mli|0fKLjl0iL-@7V*G0hN$?~mHV2I;FfM-QCYa~MD=BAaN(SEuoerO?|
z9Ed6_9U!F@_Uyfqba30U2@kGsTg2$;$Y1E?;`S`}Yc6h0YRL^LR8UXnUP4M@R<%gSvJbdjxdDr{>bCG}%M!?DuK66x1TLC67F
z5Rdqd?-=6PHIyzCploZ$H>T+(2ES){4nR)lxV$#Mi9JtoHs8j=nx
zRk*evk4*9lU{TJT7MOI}CW%rh8&s@-8dJfm+s#
zFlD_3Oc8vMUq{z0tg_23JA7ebav4c^+O5VV4129fHN-KRRF|$*7l8s%O#Ksv(Ow@1
z{#UrZ(;X9}wy=Ip_$bIN-wu_&1;a3aGhPGJC~$&kOsW_-BWP8MEQrg`AeG)5dHd4A
z-_?eY6?tAhn2)kyvDN>uux(~G>Xm;|rt#(7<}4A>43G*uT*BadCBRH7Zn>&*q}j^1
zz`Rk~8>#gRaCuZpZN=#re(W;bGux$E!R9+)`u0qHtG)KlnHf)zwJQC`h~ylHYVrI2
zTSAsY%=qa{*(@zP#e+NFg;g|SSFn0fubqlFgpv@^AFFGZH>50=Xv!qT#)Dq*A>TnA
z3n>N7MxcTO2RZuj9k2?TxfP1_||X8M6nOCDio2nFzgcwppANT5c+qj
z2sm4(gYM`e*@H6#vWpNo<)5dYgQPU7fDAOXng6BhB7lXS5bp=$`~->P4KNPt!0h+x
zdRPm0K$DWAL%fDj%}RQF!E2gFKV%$Qsaq^%G#$%u{zw-VNroace>@;
zpL85W^LAXNYD>ib$08*)>3z&GGX#-zMjC}v#%<`OIGk%$ew-RQ@a>pQ*$_U7B@A@;
zp&7%a3p!7IkMd|t9Cza9G#OXpZKTc0Fqai}fyW~ZZkOaih`Syr(2gAW*YR%LrZSOk
z=upUP`^ozVLag=HN||xV!EO&h(Ix(c+^k9cv;;yqIX-Fhts6eWqfADwByO618K}vk
zxCzbx*(hF05f#R^-ga%nwZ!$7i!+tE?g)~^>ygW`vi*IcPZr@t7;HWK*+Ax^BTFCn
z6fPXZt%-|Kdu1C&m#aky0A=z4*h-OpU~R=dlv96)x_IO$2Z=ZPg70#kF-)F=>9jXC
z33QG`IYQ&^$i<39ntmnS71g27=QS@x8CQEofw5|&#(8BQX$^W=);-{tCG?o`D4Fn^
z{kiVe<1CtKd)P&j!UCcVp-&fVG_7r&W#2&rfToSKHT7Gk1GSYb!}^X7_?X^B^+
zc=HT+w|}$zUWz0BTAYK_aL+YC`
zg9*OX2L%v8Fz|+I3P_PfCZ)ftF(V9yRZsu20w)uonPV~GU(g;GX0~G(t{fDJ<@;jC
z*}mXqs9W-@QL`wlMJr9d)8G*5AX6YJReG^XB#ZQ23i5Jl8eNBCl-;&{AqP{6i)}Gb
z{^TOqu&~ZBFrk}5YW$=gT`~}TGou?8YmBgrdZTT{a(}otY&PKK+4HKF_B3)|nq6-MsW}wMsTaz97$Q{Lr8wjTwT9&u+?(!8Up?x=KQ=X7v;?Yc*~>j+
zsr64GIjtW&dGf%hh}JGD{ar9H749^DGF{?7#{wE^Cw%eFQOqPr
zy`sD{!OoJxf^WGcDU>ZRH=4`cd=s8LZ=6;l$!W_jY&W(-*8M#COhlmCTsh^>DtQ~{
zGBWhlSS*h?LiC|HvmtHUgypS~b|qIu#~lvKOX1>acK8P5SmF8`kfUo>s1?*x2uc$~
zt$Drh5Uyz-LRfRQV9%s#XI%KOn|6gToeKz6Rtd<`q}O=*OcpQ~dg9%|Ztcm^JOLt4
zx&~K>LP%^Dk77nV_(&EA*R;;HTY1>1F<2{y#)I}HL|fSJWESw7)Uj)?u9_}WiFmX>
zcDGmii`1LDmoS^@q8Dg(Ne_fhVbX<$;%D6dlfs?*%gE1$kyIbk-2{>JLyjm)GVh18eSPL8B?vmcd<%_TqV)
zEic!8J*W&SBH{-`;8@>}F&Z2BDdn}`$yWmN$Y`aAR2HkdEoj%k*@|FHd@%bl(n#ff
z-Xgj0y`<;eKESc@1Ue6Wf8@hp%mD$`i5~_|*^eqRYw@Okb7>xS$Mwr40<VMGJ7<(Oi;z4DWRgYaZtml95ynl5~YL6>5Z{MVd?|9Mkuz|+3x`)nE|tWRMA
z4i&<$fB`R9Qwi~v%W|rH-CM|R>|8MZdEF??pfTxWcy}vJ(*ydezHCR8fV7B@+7^pv
z_Y0l;;e=ODym%h9q~gE?I7}rJ^bU#L0;fQ?+&VIZXw@eKAg{NGkP4&RW||8F{J089
zHfpa8cs^j^Thzlm^YU@WX;saoq&o8@knp4tQ6l+s^24cSMDNX=ny92j8MCf#;o{rc
z@{1Bs(cu;;JhY@GQ2LwTVV$Vw>y6r(Zd+AdUMbLVd$_CLQy>`{E8sIc|NIQ+-hPJJ
zP>t+9W#Qb_i*~C+JYMOPk10E=)=3!B>q70;+tZlw>Nn~X7wj_uF^{|t}<%hMg*SkwkvZnzD{YV
zp~bYx)&xYpS~AjpIR;5FCD?!VWo#m8oENyzk$bM$tcAh&%iZ&tkb4=iZ5i*__-)
zHV+JC%n{f_-i#c%|RVJwzHaJWWXE^s^Q6#qw13A|JzW)*<2=
zYA)1-ds40wdvvMBu5)*EaN9V+|545dSICVR7t+jFa>w3t(&oo%htaMYV4E}8I5wZQ
zE%x|`7~k|STpK2f0#DJNqQ3#2jT!}J@0R19jeP~Npw~-W(^Bs>a{H0ia4W?mC0xJ
z6|{i2c8C^H!YczvXTxWQZXxhV@S4sG@j!vWxXS{zhzam5RI#n$x+k2b?nyAKg38_sn08F
z8N8l4!)~Y*s0H+$1>f;X6vV~Ru5x6TsZLmfX5~7JqB&RlJ__;mF*u|@+l{7IZhh_5
zRC#rDg;ywjvP4W=UZ4*{r7=8pMk*z9o4`QC%(p9JZzGa$JKT@PkmPQ0E@&140|OR6
zS1-k?d$XCGMYO0)l>Glzk^hJwm#{V(UfPcfB^ATcr2pPsKPCplV&fLAV!Ng)sMnM`
z0LuqEC97sY$0Ma~{Kx4#RUd_^ixyBKL?6t<=08M#gaV-
zDdJF5X=;B^WbRcIX8T2fK9EKnUQ@1cFNT|&DARWXjrBmmNc!gd+l&|m$6n3$s|A<1
zTW?d`0xxLBhA0x*8z{wx@-?E9U4Q4m=+!_^j63k9k4&wwRc=FWM?Mf>-R)=rLa$I`
z8`sY5!6EVh^`3^3S{eTj09HV$zj-k>{){U)zBiXTm44!0Xe4-9TMUuOh;zER!SqB&
zHnW#5-#?%jzNbsvMmLxQ
z<7c6E^#ts1^M;Vwoc(H)vZ4wGWN5+EB`l_D?~8F}B`#x5*EH3Q$b#;uNVAnaemI~A3(HoI(*W#A>A+qIC
znEslyGZlA>Kj6FGD#Q;WBp7k41|X)Ot9(+}kMC0P70v!6`4bdAL$f
z^2MK2b%M|7Qvb)FyWE!a@}oD;oN8cM@AFQGD7*AqO1i{$42Djuro|7Ww+zuPuVd`s
zK0g@RyrFd6I5@6)#2K+}le4?GxJrmvRh4vyvdW`|WlD{dMat9k2~daQVm#thN`#MK
z&I1TCWu={}8`GN=KZTv@ldR6k_zE()=`gXv&p)g`SzXfDPJAy
z#C;?Sf}+dJg_&6`oI3eRqZZ1`9!lJOuR3VU}nVyO#!5Xk|zG
zQT;Q~6d}g{@uq##euaxL*TjbO=D2ZfhbyBLzB089-oc%7`wr(sOE@FUlgCLH7Q1_+
z(bRk>UPU!haC*2MiE6L4+?(SNUpqrpf|T!jSl1XbOUfgX@nsvk$Se>a{<7e6Y?Rqf
z#tziCRZZVl-6w*Lo+=}#Fs{oEj0~6857GeaM?IB;L_{8*H4k`4??u%S`*zAazA}`~
zV`^l#?C1n`bUZQG-QG^qCo4$?PkLaUr-=pe$!X23@X7a`z`3g2>lU`}%9CGVdC59M
z$bu7ntID7z0#mWOL`6!Vd|~%2BX{}6FH*^BeZ;w~apzTjB;iSTx|p*$_a}RQ*yh7c
zxqsj=2@R}4Tz3Ya?++Ji#Eo6&o-$6Zg|u&ghf-!CWQ4@p+x#F#%2O86@(g^*IU9;W
zqT{NwBD(<0u1W)b$`O;*TPF)ugWa}ifKPs%_G%N|jKZI_>J$#Sh{f(I`AJuB_i;tq
z^%FPl3kLV$8ox@@&^@>^fbh;s-qrV%REB6;?eWN0IR9gaS|b%q&BXVb&w7Y8=gSY*
zLh_~qOai9*BX|ZS!A)^oOaqtmSKfS-JW+%?v-gT9m^@%nd{CLiYs93C^z(1&KLWf)
zBFB5eM$2C&*A)ktVZPRve^(1H%IN<^VE;-=
z9T{Dg$hv{zsn<)A2H0+s(MO2%HBWQTA4Al$KLk%K)rWWq?%~k$NxesZBkWBtD7S`@
z^=gn&DVU&;t@%4L-+=#s-}&%q#30-pl3z2p&skEb88X2f!BROM-Y_`OSa80h*LD7B
zk#KLVW#D?x)|h6L>ZH1fXhi?s=<{fl;IK|?9r8#1VWC#0-r3k)(UQxwdjJ>?jP
z0f;ax&}m710o09ZS(Z#@IQj@v6A4+kEHqCuP($t13)bH&;yi>*cU-Q)sC<-Ol9O9!
zzfZ$6C&Kq?^r<*RU3_FpG>6{GjPCUMYHKWXm8O}TBhR@i0UTefi~J#o$a6Sc;}yf(H}uOK~Y
z*KE~~13~qoxNIsr@ER=nB?2b|bq(Gslk51aiEK>*4Veo+&h4Iau4!TQQ2p)FO&i5#
z@i736OU2_WbMLrcN=d%=9w`L5_EoU@wHjeq&JKH==sflgv|W6Eddl;AWIW9B-*?yq
zmksjC&fVr(97H-F(!>3^xma-mW>E-Xz2S=J>iX;BWvSmwc+yc6vbl081QlH5ICQxT
zO9F&ttE7hZwmq7cZ&S<S1v0=6fMDK>VT&UY445TM&}+`RM>|EQDTOvG%2qx%(0WOR028*N9Hrf
zjtW=$pyD;3o3zQMGOqL|-MR2kqVoZmYZ!^tDF2L;G_}L83|%s%6zAd&g$f}<(+Ju&
z1{^a4WwQ}R{o_u%1+R0Baqd4;R@Vo8eH2KE
zI3w`om~E6YaqM1BokLprb?c$S=0OEx-@}upE6`FAcl&;i@&9Y!pm6;kIxe$*cJ)JujoLna4xi30P
zBBGaMRt=pRthqRG4NK=wiM*w(_Q;a)<7$3t>0hI?Ak8b&2U>H`TF2GnP23sSb{KnQDsdX
zyJbcTTeDkeD*ZxSzF}Kh*%Zx=XNdW~gjwRgsmT|)MD}R^e{9^tVkiKDMA6u`Z6`0b
zZQHhO+qP}nwr$(y%q)6V{TH>kbuLg#MfNC75bUWbl5s+bi`(~YdOwA}^CQ={kzH+L
zz+9ZrT8(JRLikRlfdZ$S+3M$-9OZ^XaM#r1TbfObYT4?c%WK|#^(LHvOV$ZGC{xu(
zOA1qZC8L*upM3AboXFodiy(5Fh6l_vq>#`C%S0{Z6aWeys9yo;&N@OKnZrA&_VYd2
zjM`B935mBSd{Q%LLKL`rl!4fPE-L|fYO6C{(8GSB=E#k`&^t}qf&1phe#r~C&DIQ8
z`m=0%DBt(n)-D7YO1Bi1So1@U2l>9;Y>?japx5*RIm4i7C!OPHRkF9-O-godtILMG
zC~ppGY~Vjjxuv4QJC>_9`lluMEvqC)W-zS$_&}LPX%dA~NPEKaEF8Ydt-5gJ;531b
z^z?oSdFg@NQd!6rGJcS|;I2^!p?$8m=FPHWHy2jO32s_OP!N6#Q(FvYp9UCKz}v47
z)Z8Z!R7h%Xf4QMoghk5iSN1_5EBIE(HJDg7*!L<{9NX*GfcYKyq4CGOY%A#s9&JYe
z;uMJ#__hqTdulbnozTH{rMf+Qy6q0%7vXl-4X)9kXB-EQki;K6B1YYL{sRfQ`r=mbxt
zLPJY0$>X>Wb41x%P8f+xoLyy3|GL0E=KL5we{!`|W@sFj)gC-;jrG;7-b8K;PnQEDUbc_w@!|
zC8E)94SOp*+B_n#!~~wseKNz7#9I{60j%a?L6pkfy^dFFCWdHIPj6qAc!#>g50~Jbb~hIwS?|*%1}8j0mrvy1m|y6VE`!2d
zpRB5?Yq0LIRZD%SZI=$^a)#6~
zNa2C9mL^MYp~1C2-rNdP%aKHw7E)%xjM0heUCGpp7Kl#8i~?Vy^0{Rf5J?|r9#-pI
zGnBhYcX&raw!K@q*|Bp%h>9V_jho(prrwTbh78**<4Ma|BOFhY_3^zK8E4hR^K;j=>EFAz_{?0$?q{;&SnuS)zUFi!F$FjYxU|e2)hSt?%UBg{4
zFa^=ztZc5bVi&!3oC-)X7;NsW1_O|SusSOV;l<3Tyz6xsB%=b^5
z+M;`3rs@Gnlwm)z9J72dF}dTjMRi0dDz=nlK$HWf&Zk@_@GZCmHy>`$F~KiyBiJU~
z1y{K}eSzxH%VZlA?=kD`K0Tn2UNUUab?mc1=Z>;WW1(cJ_6ec(B9ecV`!x>WC}WdM
zctSM&KOJT{1_~7?5-iBgt6Z>ygfpx1F|(j9!8g0GuU!GJuu$HA6jKmGBE;Yc&jNYf
zk9auN(LJ)f|A9?cOnz%wD*)kJ9jIf7?5Q*HD2jnC1D&8
z=g){c)BTNdEt>N2{2SH*HgTG;g5e2#LEnJ%2;XZk$4%f5Ryb<_zPY628sSZMW31U~n-@VU73pflZ&5<}AlDmh9S|$4}
z$)TZ`jn~1ruReT1jXl;|d3*xq_N3@E+2kbpjlD4vRY`m9yDc4~bXA{JntLeY#Fs&s
zcwaXp;Z-&`S}Gy%oE6YK1wqQFoQ*a?G4ZrIo=XG4~*GK@Fx9
zXyL^0@T2;jpJ*6rg{N(@s5I)68%}rkc@);?;fiN=1+Ilq
zFn)WaY(JZYf4`fRM1pq(sx=atsLW$+9nCNYb4U_6P({S%azh^F
zk7mtq>!)Cjsp7(a8@zDfcRm#%;#2U>5pkldmJqm`W*t{`TXiYMmTVy8?D&GGybp2X
zkFB1fKMwy4I6fRNZ&uT)G{-YZo}USKgwr1Kp%ZHS%$C))4QGhK9)Lf;31zn$G!PuP
zgGOu-GIx@iP?cH1}l-de^Yqx_l8HjE(eWO2O2?5#b$t7+1wiG{A2O
zo6Kh+g_%Sm^w6@L!^Xn4C;9XapL0?TV#Ru5I8q?g2)4|$Hd1X>zL0F+{J}sNGci+}
z-$865ILCQBPp1irylh4^oD>ddMuu&!Fu3~96PbY8(~e9u((hT_8+LY7IknG4ZqgRR
z}z@!6s(VaO^4QW
z8~D!T!Kywhs0mXAiwQYQ%g&s|k~n({0AQv}7880m=E#D5PA?W!`M5RSi+17~HV&O^
zA>df^3E8g>a3WRN`ufuh`mKl$tg*X^(>y)R>$vBhX#V)t^sEvRt#kEDYjN?WFQJ0D
zSH;06LAc41f&ms2iBLJ{g+1`uvomgFSYeYZrj|L&ZSn*Vpsy!D3*D$BO8K2(s
zg|Y5e*?BpidVRhb!}|cge8~%>-15=p9Bj&!<~xEQkMtiE9?Y$B{DZo+TP`VZ={Tb#
zTYGESu|#f8Pu0H%_me_UV8dWDCc30le$RssU>+8Dndq?S+_Jw#xt{=nNlH$Qnz}@y
zgJaOL>OJx(zGD=yP|mStgZ(d~HU6JUkEc~FI?dbPa^0^2&s%XWk`&UcoNoVDuP5_^
zc<0pH04DN)xhA1$1tRD)7IvR5R2KMP$K)F6H@`ZH#pjwCszo}03m#4FmvRGNR-Gde
z;rgh(LI>M&YgPdiIA+Kkz8wW|YO#>ZQYR+}$)J+_N~vsw8Vh4f0GF7TRfv|*0TLJ7
zCjRMb+6^bWc8yb$g&1QaZUd^X9Kyq6iokM$R+Qeh$jAWJns-Bm?LorXv7wehr0
znq|`Zw3i+ILq%q?2Obs$1b0
zq~?K;`@)K-6XK3+jO|pU_|Et?$MH0hQWCa3j
zFvX#UP&Hd~!BXP(AUlyDrXx7voBBak%r6ed3@+TVj
zYTRw7I`@SP>(AT~qpT)0AdYOX5DMLBx)r*KQ6`&Y4GB;EIujU3iC4;hNR8(@nwyBE
zEth@SFoLar!bGaUE$}0z30<7HMAmd5u;f;=mk$lST6yz}ZY%V|-?PRTe`CYoLyT6Y
zCoVpF&6Z;6<2Hk-eh0MDE6sl3QajS7f2=rCBU41y>-x3McFsaib0JpM!0O9DB{%2*6OQg
z5l(%&O~oAxx?~u|A`yj}R&|7S%*IZ+1_~~$1KClqHJjsZi1obtJ4qM{F#|ZcgZtro
zwVh!)-tbh37@*j{`uLlI8Jrg?=OEf4fmp?_U-?WKN_*KJ_kuc5JVnrF0qv6t0{Wmr
z0{y6yqf2Lt@s$TQ>|8lXe{Ph}rx5^)<{2D$ANX=i-%qF{aeQ=PAgUYB)EhdF#_3rX
zRC*4@wJs;@P07zG9D-rC5szcj7L{akk7m3k9!^G!io_5AfDPo-&J!}T2|ZJ9Ymymd
za;~DS40PsVXx=M+ZNksYoj0NunKezBm%=;SwtwE;=JSLH?DsqM*bi7Z7;eZ^;0yV#9miYPiNs9uHZ?Dkq4%QD_iMyXeQ1<
z
zZP-J4E+y;THtU|BBSaXR
z0M;w4Ix>J{DpoDAfdtMu>dWMWBQ6DS)TlZlg1%G~ZE$(DanGbYpVgX^ygUn8>4{*g
zkrE=@anep7VfsvNMdu6T1?HV>QPkQ_(ngotgUMd*x<wN}M)*SV9oIW1_3I)TPl5
z;~2ie(RGAQyD~PiOZImq&2Mthm0EVZHoZI7ClvOco{X0--RPZw%^Ic^uY*7N
z`b1X*)&W6Zv%dKth9kQnJto&41KxuGZ-l%$Z$Ag67`~>$8Imn}Vn(dYWp8M7|1>R?
z5r<2AHm$R~i{-HwXsJGEjjD}3T+6+GLqD2GFKh&iQH|}IM2T97(sTrq3>(9!^U13X
zX6yi{ZdPYAaJIR@M
zMaFnmsemHtl>YoV!mX|8I0tuA^!1FBjSZ#%@xu=((pQKSo#IcUcfUe6q?x7nh6kcF
zR;C`F^JgZ(ux9xYK`Vb>Qsa`avG+96{2_tpS12|bk)MkX9-^~8{!CCPU;SPle!h+n
zBpc_D@AY{Dm(_Ggk8uc+r)k+N#O1F~r2Q9Xec6!;MCeLzGTVX&xqs^HKqR%PDy_#g
zwNy`SI7T<5alf6YFs&l$`~lV}UW!sPBX@hy$C#JZ?+j!gO5|U{5+RGP2@s~MZ=q^t
zFmar?EpAN1Y67}Hr^TmSL@f9oWp1!J(j&*@$$$aR>FK2J%>oSV7W_)kB_uHZ6^0<<
z@->d0pCPN}W9OX^0G_oX*F}5!P465%5$U*+$b;~FG&cvC&2N>z4PE^Nt(ST-mK;JJ
zMytltm+<69%~5N5hqzq24;)K`@Kv-f(?}S3?_JEV*q#?BvGG)gA^GB>zk^)E;D|N~
zcOf?hNdNu6OjK}`wK52Mh}Og!*c#8vid(5Qz+?*&?r2MEXGW}8ps`pR)>>5QeT3w`
zBSved>(7xfO#iTu0?nCC2gKslmX5NHW4P7l%af$2)xvy^fZsD%XV}kJ#T(MV7qN<`
zJ)jfRR>l|vAK@^S4@!8Dxc71BPEgAzymH$vQRW5#dZXqwJ?Gk-Q
zk5|3_s5{3{GyyyW;#`;qWsUQqF@k#M+Ir~!Bp}EciQI6uRz)&>|HcAuHqPrFRN|T;x-GX$9EVWr-8>BAD1JJrcJ)lv@_*=
    cQ3TK&r>n2Yo
zQ2QY?dc4r?8C&qe9cC3GqF
z!6l<1eET_SPI!*U-jh+b+>sxjEhr!6GQA3cc{~j$R!iM-;KgZHO}ehG1cavlZwj6J
zy|aF`{JgKvmrw!ke%})*K!rlr6=5kpyxlY%=UNvD?AcxqKKi>8rd767yI~vsr>~bPuf4l!
zCmEBC%PbwC;Mkb+<49Xd6He6Sl>zlH)ILnARk9Nmop@7`0lf6zb9cn9#
zlsr5((JU7l;-XNgTM0u&SL7%ZCqQz%B~MhH%D<#GWcZq3UqXxHOqhrRb8Q2VP*ab7
z*8rjKRNM!~woX@B?Lm0HhvfdriYQIPR4Y%bl2f^=Q%$xLSPKlqFcO^ps)z}Xf0v{<
zt1efJR;2qe+E5FC+cp+JFeS=V5c!WV_sb}ac!{{o6e$Yb&(1jLS=&r9wT=aDy-k*M
z={xSHtI<xm
zn>A?X0nKVyOouBYaTI+&^4jRvrhvno@r8#|_w)-o^zO@KB4c@+Jiu5)U86uprW^sw
zV;(Lp6HAya452%Das_q0&c8X7w`)^PACIj3yb>OZMJ9Ie!LIT!7%~W0nlNC1E^)@v
z<@H(GylH$EcRtiiBYd7(aIAo=ozmG$evM`pHHGLwebnYZL@;Ym&r(hF*|^n5SCG{e
zQX&523`(qO>Mx+or?@5nfW7|3k<|FCYmj9Yh_P$*u;61gl4}l{PW<-+j^enQ2*bqK
z`HNg5w&Ato^bM-}3Xglkhi^Hp8fjNc5Qq1oQh8LbMCZ%#99K6{-u9NLmq{<5?f~ex
zu3%;d|7`ZO-J(k)Shps~8)}BIIy-0flPG(P*J{Wq*V~1NtYVL&(_^5E$wZm;0%Sk~
z))aADXOLdTk#;fC!NsxO+EFf7UiOu6)a3NFf%*7v9D&~~HG!aY&9BI)!mmk3!>(#b
zxc;~y83rcYkM{EL=6}%x$!V*7M<1=!B}GX_Fp#EYCQ5Xpp~@oon5U|VtR62IP*tY<
z(vH`womPGv5?vRy@XgAcS9ghT9Hw^G$>&j?BQHhyZ)N82sj
zZ3j(`kod*66jboE)tX8UjPE_YFb#Os0ARZ@qZOcn_E*liry#e*-D!IhSR|fVY~0yZ
zJ?%^pyN}Xq4_#aSSS2YEiB6YEai8%(F=~VKOh>*n70zCOLxv0XFPvelE8A1y_6gjc
zzXj%A2&N}o*r!;;sx$@;bE+X)RAUFgjJ^`2o)km?^A2Uemc!9U4o4H5IkfXCiR5J^
z09}-GZOrFJZbywG?*|-3?
z6Go)fNzVH^Sdl=LOrY8Xxuz~}>Tar5_#1mWkDo+UZhbmYpcZEBu!upH-gpOksfM!$ZC!s4Xi5kXzx6PD*{@LCz;N?K+
zRtPh|c}Uf{Yp6tf1VT_!TJTSvzdO)jUpqRc;ARhB_t8$!*XTIZzyAx~_}JtT^!X*UQRl>T?BIdvKp-3&v{pu0)4ChK
z74Me?i%eJkr%)dWX)hy4)GD;w$zYln%J8LPPXyTW1PAL#RK-D6=uoO8M&AbDs?^zJJrU@p8=MMPCMhoq-Lp4}1zVvC*BG~UqS%|s#qu2JK
zv+OObd97W|VJRup=x^fim;aBP>-J&}d^nvTLaWhS
z(Vko?tOzruO>Lw_Bl+tHAcMj_=zE8wL0{lTFl8c7P3FZh@bHL5{7%O(e-D@ec@WyI
zKt-5?xxNbrUCAWiUK4GbFK*%C2VFx1{Am{$6g++^lU-*>aeo2n+#tQyLtMlo
zoUwv=+pY-|W+f|tHX83DPN+tGu$KO05RHY903KI!s3}M+ATzgN@}w4OYZLk4W&z9q
z=Dl+A)x1sQ(LA-^;s|}t9O9*FPnWYv!io~Ir=(+}a-mmD_bPGMuqavN-l_c~fF$43
znI0L4kf@J_DX2Tn6uXDR=~5dCiWue-A@kZbq}IB^9idT26&jnFUnazsj2eB`GCr{@
zK(@lOPZtf2gs4%bUKZWW2_w43FD?LZ5TFlc-g6znhl=0Wn$7<9&$+M6PSX@%7HUS}VXORlR3z}A&j~4<
z!$=UE+&pJb^*qG$bc}buwa!x~Iek;~uTc7#+@%BqEjKAOQ+=A-ktW(XKK|WpMwrHunWDsDe?)Lm-%BJNQ`OB27}{{}OJR5$49LWIG
zzihayT_AnDM5p*D+;Cv)Jod)`Mdu)RP6mdTs2a$dAh;!J=EfG0)v0TKlyH&={snk`
zQLBNBd7c87Z
zB{U>$rL>F~k(sH~d*h|5w{PZ&j~~FdJ-dC@&g}CC*m4enhi?7{q5Y}1ti=$kk7v+5
z!YD~#Z&jmAY#n<&JT#5N3efS(X8kQC`oR<0kl>Pa;9e%mz%aT#-Z{W3(w8M5s+)kM
zkdzT>eLvkzS+P$G%I-9||3zucro`m&@jo7PWQJcoBKsAyLndd3r@eR-a3=&-LPDs#
z4O&4&lTvC96CmQQ5Qw~%43PL2tPumB>EQNrzc{6~rb(Dfh*Dro;L;LkOZyfERL@hq
zkpIaER*N4Bv>Lq`8TN>*F(t_|(lBB4>9CJH$cUWX`C}cX_9|3fG5+~MDsu$ZqM%P+@wAh0^|Xg?
zkHKbVzUVn*s{h__&n#THh{v?#knK456&2`QrCb`qGwLpsUfL%duev;0-egJGfVE#&
z(FN|}2XtrXq6eR{ZBTJp(egisITVp_kB?u{Gcuz4^N1#RpjH7>HOu-H=*r
z3_N!5;dZT6mKB^VQQ|BPZWH^95BMfB{>xK+&M_me|HD3kS^KrS2U&q!*Iz(;k7mMv
z#j>{J;+&+@i+Q~kg1|BdWDgA-E`|CJq94;!^L`9K&jc6^dJ`vwiw*l0a1+a9#TT}b
za%nacSA4EeSEz@VN;XjVUy_u!x>dT>YTi<=aVfyPy($b;Q`Qf0B!ot%8c4=ur9hs%
zA<>U@tCU!7Dfe{hvu#*O-9E|*sM3aAf}ZxdBW>;5a`xtiKG>7v_cZ2^+&x9~K3$DD
zN1W08n{;y~(1|LDMK)!4&por_T|?Q~_q-2b6b>89myd!sv*hn+94n(k0uyp6PBz!Z
zKg%obUN_jgjH@K(DSh&_H??>*e3SVCbLr&n)cmJxN?)sTU!|JZ2({pjHDQwZVNJka
zhnx?vf|0D*LmU~--EP3R%N;$}R|0rJfBBjZTaD#(ns2QXvq^Q=Ht#E$h68JP>wDn9
zSU;skSxECTl|nI%4x)03VUi9}w@W&f+Ih;7J09-5SyHRB-6WIibHqtgEeGw+{L6?s
zKg;{-dAtBbdT3rZz~zq@|c~S?qs|@~q-}d+yTlo9NFi3LhdfeNF^y
zDi@5WjS(}TEK^r%?V+iydpQ}zJMcQx0S$>4USG0A=5lOWI6JB|NbmR|R0Sw(pw~=fQK-aB8TVekaEvdihx?eHol5#0fxfl_9IrmXEx6IVQlWQq
z5B&#uLKZb-&?UKGyqJ>nGWl8|A<58DfLk0;HYljuB?iG{g2x<&GtdOw_e0XMP&V2E
zV0`#)ABG=l0~z1ESJ*Ul(Y%bg5mL&X3&DZuJ(-;8BM>-*`$;C46{!S`F0ebi6!{g1xiZ@X=|4KDi)e~C?@i&2;60Zc{op;2
zau81`;ohf~R(4z1Sf_ev;i_ELa7!B>)$ismbAv1#XVYSsaPk=jHIG!hxQsouPI2YJ%DLzm3P9O9&t<
zp&)Co{}qF8_{k33)1{aa2Rx}t9aqOfzg35<@dHD;0@iKHAp+M_rEPGOwY
z#AxpC-UsgPEl`wJR}#qw?7MR-6Lg4HKvPnQXh~IC@6k49XFrl=TjUr{r`zE~M#g{DfMDfa3S*eiT^a#;*J_2{wZ(V+gRbDwFL%HRPIUv;Gt^WzJY_4))tej)dE~
zZ>VK`d_K1X5+j(fA3#I+{_wvP!;+KtX3|<_f3?e}z4)!SM1^-?o)*)I{;
ztaOTz-|N4tRghs{G#+|_Gs5wfV+@2+}oIj`DYkwnxeXTyT~H8)PunV%f`
z!$xcJg#l?qg#1oP@z00lj&N@M`w7lRceTg@#Vgg$#-NApQhA1X@c(jj`o=;<01Km3s
z)xo{k`uQCP(;$^XFQtxEzV-f!%H54wdFo24cSLvc7&{HZ#*DBsFZ`rz|I0akq$6NM
zPXIpSrj42p;fSTbSw3dxyPixnyC0f7$|G;(s~-d!wq3Y-KQ89bq3Whl?OLoI!q$}=
z!}58_+&S;9pI+QwhzH-(a1gEP_k-f$uzO#DJS}{%YkRZ7SieLMNHMYx-Y|<*|2
zvf}_q?)Z4KzF@Hb;ym(bx0tEwX}-)hqe}E=g47!REy9K~)jNd7h#(Zm9A?(sSLdZW
zvYgPb{Z+P?Z9qUous)K~n%H}QrFuQ3^UWp@@UZS_MtD@r6<$;P4lozKhEH*7q!zYQ
zNsPqC;jq2oZ-uNFYDb`gXZueWuB#HF>H3mvI-B92{XjM^6(j0IbEc$(
zy!H5{6iIKF;LAwZO`7~20bpkS-o=S^A^Xyc-onV9oS+S5gH=^gx&bB>2-C?
zu@Fi%=Mf+sjcmVP+u~q9{a!*&0Ff6zV{mcbpL305d-t%XbF|P9H@Ohu&9Z$Y5B16Z
zP|o5~6O{Z)?lL9uN0}aD@3WOrKRAF&72os+(GkrEO(a_2{&~qDB>wHjidq9&&jYTW
zTYngE2C%(Ef>m*6legqjrUw7C`htCLvr9Xwg+7H6sO+4bPi6eQ^R40=$IGH+VI5kklH=Jf{0s>=V<&?(o$8TZCt4ZQa!2DK
zo+i>iiFC{(>Q9o&Co{D=4{c&pD5lO;?12&?oX^@qyAE5^smg3O&n^4c>^t0h#_aj|
zy3!N6n1t4~XJ?y2c(<+Ii7PZ7uv2{0I?TO4%!_*Z)5-X2|mzat*QxkwCi-WO{$=2Q&HdM0`3K1s~P
z_*#p^1p+w2a#$~hU`?87<(KjB)6pWZfd~rI1+B(k;auNbI^5-rWjl}2`tr3ioOcy4
z=`=T4w%r?r2=9y2w+OTO5^l_CWdc*MUCqb(5o#>UKxBd>U(T21mEkOQ;hNtN{a43!
zBy}k_q@secR5M?9zfm=`U(JtycORYI8WA?*kF@5+Mn6wzzE=K;nZfE{+)@D)zj(+V4au;^jb6c1EMB2M=k8TYZ_FL3PNk(}@B
zLjO#n3Mm9u1!#X>fc_c$O0~)(odv?Ofom;cp(NA@F&i#I(aKg>vUNYhsAbL&_yI+k
z6{&+!_@E@m$U)PnH2AIn2;9nY#{s9PUE}+8H;f`+o(h?K^ieDIhApf;Md_-5U}sL4
z{P|)IIPVij4@|dyRfBGnj&dr2snXTHHNwsxzzanOA*nP6=_1bz*w3mbG;>kLV^W8WcV^jA2WSn4t)
zq%sH9>%}v0^udi?V3$A$#o=UMP=p~+^;aoN8SiMObZdDGMqfd1g=pUVaJcOfSI-Ac
zEXo9(f-z_X-@d!VjGp5O&4Yr);Ro$}c4x23?tgGFG^T*}RHJD~-X=B|S
zp{Y>`0JJ8cl))mVgDM0A?`idQbKjY_?8tkC3|h31)PY922)nu0+{#0iENsaF!#k>(
zQ05!PQd>O<+MdpJRgKUZ=xHgv@tubAOe|?>ymfVuzE-%37(^E_qEz-CFtPT8Z4edt
z>e9!<{?21R@{amE5n4H}1Bf8$KNBK(#p~*bOb|p=5(ArUYRv4n_!~n*5Kd!S(;9YE
z`!loiYIqLmv(C_-{mCsbEKcP0DFT!F1wbPQCOks*@>61$#{!hMzQRZc_%ZKK`q5>R
zkGE9X#MFM_UKV4U_4xOo(JMNWsXe?e`?ZI_xPd>Q_IWa*QM5z43b{=8&&6`jz8ccU
zfF){O-Bf+;l5X70gqK!D5MuM36bm7a~#O>*BD4QU1z57ru)15Riap`thU=p
z^)`%}y;Qv^GaZTj08qLSH-Gqc=Z+aJvWSXfyrz4<1WzY8T>talZI942#=0*;trCW4{d;*>qZh>3Cu*bSa$Lc)5M@CB*><-Ol<
zz?hm0m%Zl=YgLqRYfm$Pt9mS)FhFJ9N!v#>m@>fdlvebcr0Bpb)wLj>xJ-Thuc>y*
z*r_dDLk-}-&PaCR4%&K>kHVk>VtXzZP)Y=Kyqh}q@_oUd!#oSqrIUtKinoAnXY2BA
zDnv{)Y?w44Eg+3pDsfIK0nV4=<)WAyv_`WzwKSn%HC^7Sv*
zwL!xSySd4SAwn%y9v}R+mp+o#vTdLfo`ju9)uf;EUx4HNdB+ri>(Z
zn36#xmtl)R0wlfO9?SmOJj-OIaV27GLD<~BUV0J9A8bCNj(tJ~8D3&ijyiqqjuzMF
zeD5dVU8bJ?T9L`)xZ8r$od>}+_wPMv!j|0CGB?!^CWzvqOj_lOnktTeo5eu@HmPXd
z6Fw22c%n}x0j-MJlMXh((!FmD1#FH*D0;0g?~+8V
z*V<-^YdiDqQB+CN-5J#>JDs66ccc`=3z&gP>tW|X#xAKGGB}QOlTH#UncW+|3CM7u
z6s6zYJmPl8pt6OL_Vux$=N63lQ@}%eDD<0D(;2%Hqk;R`qCFM_m4j4fS=EsMNIDUi
zq{^t(33Ps>XJ0DN*Kqa-%D1N#<-T~(I`)X)1bi-lXG?|NCAnU8vLkibBNs^sB_Gx0
zr~-F+(w@B}kWA6M#0#7|5CGRmG4wTyS4!zv;#we5I~9}bYb
z+vnolb=;i?n3^jj(U~5BGg8e6-@y0PX=I(l$Uj|b=QStUtv^=u#7-UuX;+odnlx7s
zBMgnIui+xha7M*g58$s7Mb-N3(E-iLR{2Qqff+@MHk&Es$_-33-t;0T(=s{
zA;RSMd_4@Rfqq!M-;+tl#WfPoy)grvfn#eKa!z~#KFzNcywqVpCJ%89N>H2kNByKt
zR8eF$8{l!-&cPnJ4D;yxi85Y2K=)<$VpH@CmyIe<3nYDu(G`3wa&!-YBL5tH<*Upu
z&&0J~HpVY@KG2t$qppla-k&y{UmDrcM>d$2w>+JQx2g`53VFJ6FIx6%l#d0qdO2+g
zN(KZjw9T~0Pt@}vt8p9c0}X!8VRGQN(V78vTNGuh_hh7lMd_W`)E=jbF!50LXeVIM
ze94Bf?#=3HOkE)gbiJ@Z9KD6&it3g^_R#xo`3){cHs|sB)R=>&^z`VcHK5u&WC$t0ySt$IbL@_NBF{bF
zlJ~93=Nx;z(3JTnVIMS`L%mmQ3Y_vc-`Wy~#vYkt7og*G=`xPYkH+-KXd0kDc=d!s
zhhks%rG#eyX+W00ft+&E0UPyNwsIAI$g?ajZ|X`*j4+d#I+Gj1`GC04cp+9ce24A*
zePp8z9n}a-W2~JplhQxwJG+d|1U;&qwYAaz(Bme+1cso;Zva;=MHzcd=2-ZNd7hYF
zspMp7vajey6r{hZAvfG~m3^En7(0PqW215nTO!ipCHHF+4xzay>4?^ixJYam%9zk#7VPiC70Yc*bqfS##X+r9v2U`@JyN00O^xk43
zV0Ri^X;z;WLH@};2ERcF99iFzgejdSW=N|1RQ@X|L$G!)D+7ww
z>Xj{18+h%WtS-pypO{w|$+NxdD%gVB_B?Dn@(<5%Fdt?lL2Ntc`n<9xz>7Cp^;3Yg
ztW^5AJisl;?EDz5DhXA}gfnLb-LPi~lcCO(lEY99Vng#R(Of*h0H9gqn3&GE`WXti
zKov&!DcY26WF4F}eFpb#q50=aag9*viOHRKxf3L;_0k3{=*0852}ayz4%08OvYiY;u)64Yq(p)(OfA4x;lX
z7awhc!?oN8o+_e9q{bg;iwU%u?jFt?D|j&NFFCEObVKa?2z{3t@2I=nVHSTGatL}s
zA|lwg7@Q+Pt(UhiDg)OVEZ&n+&Is)w!H-kEcB~PUHu+JGp(%ytpJNn|7WROIUgh9*
z6klJ@gTL`lZO&z{`3}d9=v-lM4;i)p^N;Ir7i~Ar?B~#$-V`eJP|}7G3m1{Q-NT|j
z3v4CZ9q&WGB*rC_TlBXPK;X*;Hz{nUSsUEFJAWB<=d}iL8k9nZ>((E`aH<*;KRW!a
z!FS%H2Axy2cKZQq^hl5{4sGBtxfjF-|DRxJbTqbq*u6Jo0i*W1VO>a(7e0^C^>;Q2
ztbWlWvFj_{M72WV1-ys@kBq)|&_6W}Ya4%`_k1J4&?L$ix(3eKb
z?VgAq>Hu*sQpqQyIf9<$OAI>0iA@{20;AGG$TJQ%bH{C5JRB?dv_>b9%Af|j{*_bE
zy;Sw}Sx_HE>vZiH-_2%TyIBCqZ?eq|ncGwQo_+GDCEf#DAgy)4J9Yc)Tyv5I;2DIu
zZZEIUeDasJpS>SxsE99L>{H?)4c{3l4<3h!+?)TM8HTuw}mcjO<#0Z(v8?A2
z+LqH?)20Q2OqufY_-|=e1HLTPkR@`IE4qU?bVb#hckw5v>Klk
zZ?m`xhQ>r-NZ(Z*^uFmtMj$fdD@o-_y9(r+WOGBcZ#?-ak2pL3kmvs#T3xizitKqZ
zWqMp1@d5Nx@RD?a`Vnu#+t4Q866vySZm~j|>an-93&m*ha~>Z1<69qe$uS(s_<5c(>n+v^_F~yISXsB
ze87)u5S@{b-DqeKmFWzeh*`*2JXB}0Ll-_+zIvUml*B8q&sCARaEMYzGW7Ud)t2;T
zm}O-u{1XLL4v(e?@6EI0ISWP2@rr##mTq-6S9@K+u$u#agW`W>kK3#7!*6_nJEsQ_
zM}?gVVgurI}QVRWUYoV*AK=@k_QI66(nsbEQ>TisIUV(
zVCz}lzI#V@^Gy3zmB1~ZYf0QEL*6YrX~D$>48!B6hhOhPcrjM!O~W(75+^Pgr`Vu}
z>X(H|m^Sp?WW8t6cQP+UE%N{a#2$+B0F%q?mK={~NBP3%jluUqX2FeSF1b5$vg~y$
zztGx-XR|3tD7|YV1I>0-Y&x{Rd7l9-$K>X~4h8ZBzl6QfG#h8p>=kC6+wXZy@&oq03ZKeObE|oBmIZ>o-9eslgXp`@Ww0-(tZg;82fB)D
zuGqXPg>0by4lzM!qs@ZYr@9~%=jkFIk~pjwN+Gt#nbah>k^O;GwFh_ycf!xxg~~JC
zu1lMe_n&+w%|LQ5UJP6>U^rwZ|5ynHPK0fI@2}b9
zkh4lNbX(qGR0$$LY-QK;{UD(t6A2Ozb#$IN2C?Np<4eQd+mPuUR?QE_Ft;go^g|$F
zrem{!KMCi>&G`%xlF+Ea(X4Cx-4`#tRp^0pPm&<%%$Hh4qjX~
zoS0`oQ@g8c8j&B{0#fM{&x4LVPUA9W@zR44-=+-z4wInYEWhgSNVzR#(6|k!9izY
zP6ll8>ShyJ@o#KiHYyB1WHvzN=HJ9xIpCE#1)w|CIJI8l(uTej!UMGdqIE5qFZEwZ
zg)eT{ml)Xx%GMF>(g5Yy4g29MKIW0#M-|w(F#SsQ1`2#apXxz8=7Urbs
zzF)D>AO|%JdVu_B@Nr=D+2N<tib&zx$i0s
z=Lok4vNVhj!j&PTuVH(k(%}{ZsAekzuJ#u!qxtxB{h4$&eDes5)}rLmIzpv_7~}VP
zC70KdgRmAZ>uM86o-re(vp!R3&Tn4blSMn%PBL%kT>X*5|kOK57wGb$)p3?
zT%-IkLe2M0t-&_98g3ELFo+w$=6-Zv6nNiAWeYV2LxsB84TnWEOFqKbTOMDy%0btb
z=71_`iM9E!0{%D|Uw-G;Qw5Y0Y}RTg8oTtW^mWxRKp_G~KgE`*q@AC*e$5*qCmvGQvVV2>!M1gIYz5`1qaLn$CL`!m0SRQ0M`b!5H=?9&A4ZTA|s3G6QC>z9k5z0`(ZEQ~Ru*N4ut&O_%
z>-s69bs1{7Ty9(*-Bc3xlc}+QHe)p}iCxaS;|~0<@;Gp+olW1LhLp>z-RO_MFk`LQ
zJSKqcOlj|aXMqGgweXd=>d9bo9khO%p=UraI4`re2~w|J&Ebgz?zU`IU`frF&{lC_
zST`imqCDC>CNg)q*#!X=JSyW~(R4sFDJ;MQo&Wa1%+CQF^P`kK-A#Oi86y9iAkV){
zeEwwdvl^XcmnwM1{~M3%oA~Q7X+LY0U{P98*G{ah?h}z*T%_au;utO#bOJPIf4uhz
zHj5be5Plg<@ug*XD%wz{`h+1Rk1a9r6a5%zVYXTCl7jN+Ax7-AEu=lGi#ZaQC6{k_
zHXq2uW0j1m_k8j2Ee2tf_g5MNnpk~j-7iR?=uha_19$-gkCK=YP^&g@w=>Xqh34f%
zG;T=kFdFUXc%=8Rc2_+$5@b|-n={N`0JEZN42TSRo_PLsx~WlJ8d)O#sx&B{R
zOl59obZ9alF*Y|cH3~0GWo~D5Xfq%%3NK7$ZfA68ATc*MFgYL~ARr(LFGgu>bY*fN
zFGg%(bY(O<&|g!jcpBqVzxF;w2X8N
zTmWeqMO$kF8&+CHpsBN!fdhbvj)8#%mXuW30chZ4ZfhfA-~{9XusfLn00v~dLfO%k*Hm(%)_%Fm=vyVL&x_DO`(-
zf&bRV%)s%VTxofEX@IqXxs4Oh#=yqtGtkMv$=MO0|BvnS4KybI7eF9D*xAA1Zx5OO
zb~*f?HvbJ>$o6w)+E$+42Co0PVg@$OjvoJun}46Sk*$rRxucWgzbXO&CgxVazvvzR
zu9>;bKQ3Cc
z#=^GN)}N>yVgJ%6V*bgKldXe0{r{zFOB-8P8_)mSZeng@Z1R_RV`n>h6&rJVXP|`0
ze?@ft4fB+uQ{Bd4ctG
zG;jd|oE)5i-k$&4@$Vy8Ms|R)xslVS&_6c|>_62dY)otcoc}U^BKdFEe@TGipN&fS
zxl4_0ZLHh@#y}HTdRbeiPeD-p|Fh5jqnDVom6fc4HIU*zDgDnl18Z|D_kWB0j|es3
zU#t|ewhqr*1#b
z9`nCP7l4+X`9EBriZrsc0XjMYm^l7%0Y6#z54@k$|3wR+R~8pmms6(xza;Zdn5d1B
zt+Ba{DS(NU4PfBlVBikR@TmhPR#t!~7#2
zOcqW6y{!%KukoKTD+7Su$<_9s@xSAd|GORkQ5wVl0Q|3W>YuTalY^}#P|e)<^Njvq
zAulm9*byI%o7pc~K#c4^+$h&#w4xhbU7M>Kz`41$!C?z94fPhBz&
zd?{_J0~bpqXPXGfU*O;3mrhn-EB&2I$LFWNO_4fj5W3a-dTV0ka~Bj1MlN
zsNiU}DxFF|y3C?qSr-xIkz{tBMls7wdRtr@K0sykSeUb;`zeZP;sfP6iAcMwZ8aCd
zwQ!cP8O_1!tJ@+nPABf-R41s@#}iu24}(6zC8|IAc_A31l8|=2-Ke{uKMMI@BPn*%
zxgJqPDW2@tmsO6Sw89-b;l6?3`I3AqGZhDrCuqZBxTf{XVEzOPn?wYSD{$;`_Ro`;
z5+=_IJjQNBnO=P6WWMpz5KvwTNESU^={roeB;iW67WYt04>4`$Z+7E_bLIX?B?^d5
z&pnZ>dzRrCPp7$ae%ySkJJht+WwR7DeXJcqad*b!Scf@oRP2=*m)Wcd>GGQmaz?~+
zh>*vxw_YY<0;Mb8wUtxuw4f!d?*
zObuij|GOEC7_f4L))oet-%Nu6ZkJ?lpVl&02RD1+x?mUnS7uaz2Ol8G&Juxc5dZ5g
zm+&3jFu!P_T*&QKN-n6&Rk?pktC&7ygIN{Tq%e!S=%YU_58JxiT
zYgGzwVv9(z3*NT`&jKsDuUwu}TT#04BJjbAEf(J~kdG#QlmVdZM#{=pB~cUUIp~}_
z1muT#!HpDQYO^!1Z9rkn(eU+QE6po)M
zrg(p1y8)d~xB$30K5-~Adt3!qhof$iP%3lJUc4T|KKIs)xeSjN7F>o-t=tA)+yTV6
zu%5P%=-QgR?GIbI;SAZ)z3KJBC(5pe+U@y;xu7bi&UOC!X`ND61xjs)m+
z#4u4H$krr~Ver9i>)@av!(iP*Ozv5iw3h19+aDQk>=Mo6EU;DKW3XWg<9izwS0lRr
z@+tzp-jEioov^L^X-QNWNq?t5JR<{eQM2;!q_Ex=YLXEA7+l%GDnlHOn+bOXyj|pkIzb3(_R|7=sN`6{;iobtw(b}MN
zkBl2gqc
z&dT3=BE6|P?FRG7*yJW#uUvYS@4K1e%)RYJQ1YL8mR&-K=52!%?8Y-fY=DvR^$liL
zbacx_ni0z5mgKH(?v`0$a~&rmiB93azI0;2Xg2no(i1UEWe)`HIzXf-$<*JZg^<5?m=Jsq)2WIz+_<{mt0kJAiS%t-z0KX`1&VjjRY_X
zk#qJ8MoO-QbSob>hSIp3X~~ZToEEFUGZoXGZ{dk`rQq0o8i`gg34yLa;k|#u>hW*X`
zd?^Hr)qwbu4kiSxZ-u|mTL6M##J1Gc)uqKw!%Xe~7Qu9h#`*%24^D+t{wG1{FHGf-
z$JBB+V4OM9dqYhZ?EBg1@=fU@>{GxKke~LVh)uH{Ji%a}^;(0a7S91ON6
zKk*2#3f@SPUm!HJhrmL$nX2>EZQp;noyt+3zTK}S+os=VUz(M8pWDw2xy!cnwxDyMp|=7oz!
z>Ar|ios@s0;LfZj1+@oJpxxWyqS$%>6c6xui*=6?ym+V{{l1cVPJzaM<%w1A7N&xVwoO(c7G#
z(Y|w_lLLmL?S#>nM&6~hnh0uK@|K}{-IRw1z)ZaR8j
znb+gr_Os?KX^gK?g}3>b2gV{|4N@(j)MM2@G%nxSt6k1hY{s-q;+Vh^;&RRqJui@9
z+XZm+YxBHwI=eKxW+iRm;@+`C*5&|<1(dI=1ukbBTe?P!u#uZuo&ju&)M_8l^Rzg*
z(iIdTj~miOUMQy!o>e(PGhkcE_X8)8aEaK38D<`O2Oxw~H26tn>oN=zilYbx#AHm2WsW|;(}~gI&twT_{JPW--tSyG=P!
zaVk=qpXVyFyE(X-jT;fv^^bCmE#gZB4F4a8m9~W=6l3&+pR0#nAte}`yJpTrl+zw%
zjYGFrdmeMlji<@V0=S8N-`+-?PZne4WbwFG3vv?TpwTVtEk=)fVV2>buDuJI;d}Ns
zk}KLm{uJQYp0IR~T(-fBDUu4|JHbVt{>>3tRm6@JUwi3(@jQ?0Yy0uFy&04ocZF&y
z2x-EG0xzTTrYlF@PNv1MP&N*0=HVgoGHBdF)PM%Ot)!$)un8$yFh)SZ;INq!8QASx
zO-<9RSY5Ux1o$f5v34$q>B>)XTPej#}+%;a>beLEp(4M07{H
zFng8SzUZ_jLNz~YM^l3XA`JLSR9K+f&621~xDq_L!*l_wA@2As+(<m#s)r
zR?Cl%*)C4^kde(AwZ{l=k5^jMfwN4cn-1rvIimx`hIql2>UaFNw0(eVB0FrlyL;Xnj
z7jivi)|<6B+%3F%{~nwSaAQ@55t7Hzi&g)LcYL2)WKbsmtR_q!OEiW#xuSDy!Tf{~
z%w3Hgw)%F4J@gTB$7do*cQgmYcsFTXJJi#h5HyA^{7=oX=`6OW>~+~GER|;KX#eeSVF2hGz}*W$lB8f6n_JUA23Io$y5*y4iDp(;>t0H6-=~G
z0-slTWR9N*@&T`39@@o(k;zv~xF*vwZ}Iv<26l9e2;nNH{I28A~e
zNeap1Bg(H?H8=eJ+=b=x#FHm|9OH5{VDix3ZYzG1d>79+RlbmSw#z5(8R{HRE3%i<
zQTDyS0R&etc#e##ufQSFq>RkN*}Jk``#n%%P|scJry`N@0h4ehmFK1uV
zg~^d%)TE+k2Ali^LjV@7+6&ft6|uy%?*fg_IJh0WF-R(PKVt!QT8hs$TGgx7cJ8W+
zkSvo{SOqGra$zZS<8?kgM-q~NSBz%&d?8VCSE$F^cWw_l(YetYWB`bj3#TPPOGOp(
z_PIX|Tkm#5Flknf^R;S7yaz0*wIW5hp&Q8=LT>jlC2hkJOhZhep5sf>_a?LN$A=mB
zW>cHF8{W{@gYodmwdztnAPUGn2X3{9eS?IB{5@%g*Ckm&NmA5i`n2&!7nk7R8#_*S
z14v%3W<{PMczS(dz8I`!ljMGRf5;{#BkkaidIS~i6(lEfGpRG_==4fDLlIY|UUb3H|BL;
zhd|v@GLSNuNRgfbq8Ys@pUPw%X|&N*^JZCs{4^z|iyRakuQF(?QlD&4gO;D(74uw6qbCsRN
zq`xr9kbzN1TbF=c$YMRp7DNb>a$TnG+B@Eu*+3DJ`lZ;#WF`bOv(Oo{wRHjaxq
z4;N&7*b=+*viJzg2r_5o6Bq4?fAr?dh?#MxkE6b*S2+jqoqFmN=?$VVP~456_kCy<
zx52{8%usVQ(dE%sY)!jQh{}805r+ncpN;TGrx^FMM+Y@
z9vULeNZ+l=_u8JoUe6vqPpWeKn=>o<>eR0
zBZH6;#Uv^g(j+=zwnZ7K>q@7ZaWrR)CIq=cVI5hRfJC)v^{QkFMU4~;r0{!L!$
z*FCA8r&%Y9XWPNqQ*b_x#`K>Ylw^WIpUe+cBdk8sb7%5n=OhQYiz`8yQ#^%u
zu7_pM!d-haU$Y2_+6@Mh!|>sDkMK7Vu&WB#xlp2`2iTAZ)P+KaQ&&{!u{SqP-Sv<=2Eb=Ey7~C#R+!G~(&RIUpfk7s%eVJ>A3gcEjSwXYnXnHi6#!Uk$+jJP*{|M@%8C2d+?y2Y@W
z{@KS<)pDflOdL*@Mn7)F&Yc>si_{-!i?p#Xyi7?AKX(ycY<}S0E*~cLn`wE6!P|A^937_P?vD{Ocij5peodt8
zizxINQRJd>N$MZ_-rXlYc-1SfU^JsOyfDTaQKS!@;|Bp!E6=e5SVGaqzK8O(5eAJ&
zVvVg|>2H~2&)H#1v(ILj;qq>NENA6Z;O}~9+aX&VY`EC*w^ounCR6!2Ka{F6m49go
zPd%kL*#Z|baHAXyA-|{L1qPYz{U&3aqt5Azvu{A89@sN&ST)K^hYkrbEO8@8pzqr8
z$YWqe@W1Fa*_dNyPT6#b9W!TklDBg={{dCpT0&UmRUHgJl{V52!Bh)61F$JhfGK|T
zB=A#g7&i2~4_Yx^G6&U4zs#bm^A4K`d@lGc2swtS==ZY4q9~l5ujZiQVOu{hCWf0Q
z35MzbJHZqh96{=!R16tU|=0G~hk;jK8D!EyZ4TK%rb$@y>q;@^`g3XWsTj
zHhB(c+6}Cnv%&!pX)m>Jb?ageafoQ2(gEsLS?Bbm55#i%aq$71G`x1bpU-pm>eD0J
zv}Oq;cn51G?rlxKdGYK4plvKX4+T8QxNayFKP!Ac{a(
zmj7A?3=725IU#OY^YDZP$)@_xADeYRE#eU?rinUP4t_R#x`?ved0cuJioY>U?R~%8
zBj$6oKUSFrKg~4JmOiqX*J!@lsrENl;$-G|#XW}mfsu<|9Uad?5!qHDb=o=~D;$+@
zdk<|DJa($N;C8`qcb%c-4JwSCq%%&A83{nNUme^>kN&j-?Joj3+=OQ=|Q
zgE?~-a=C$%b8mJH1ypHK&QrBX9VJEwebFn5GcB<;KN1|SG$1WBWGS_>_`|RIo2@-E
zC?MV0ThwTyU1HCeY-UIwq+ATK1d5<9$b+?SZ>G@6lCoPZpQSrY72S{VrZiI(o1>
z<20ZNi*m_xMQ^ZCR&ZaZw_YzR3r5sRfk2=_hPtig*dSE`jsq95#o33dOU1g)0p9R=
zFX|V~AjspzEs07k3bkkzIqwy+1^1xRhSZ6baB)$i7I)$~;>kyE*wCfEyyo^?KQWI)
z6%7MH8Nauju!;PR#YiVWRK4jC%HXQbnG=`KEPgQUfYu18)w-6FgzxP9^Br}?mHk-LK1eIG^EV$+VA=QP-Nn^~Lm~TlxB&Ae(MA;)(U6u;>
zAZbLHO#USCFO(*i~QrjeAfRaYSC<*S{7QX$GOu%sA7sgG-Br|s@T0l4kO1b?Nxl9
z*ae4gk%OWvGbdX58Q0Hxo*^#o=&(qy3*yo50<}gz*Gl-Y~}SRMVX
zrphr`E>l^dlxAQeVeFQlP~yl)M^LZq)=?G?uW
zAyCS+Q$7tk%cB(-fR24ZVi-UlmmjLAM)oTQ&6U461vzDRb`|{2c%dmyCQnNNXl9|Y
zgG88{ks(qFersSA>ztdtU%gdA&9Wl+mX4tw8nG&Nw
zRtA4N4RqkBwuUA|_MkYra7H~)9_r^;V)t5g+z$H;YBarT1#JW=s_UhGa37sj#?i=g
z2@k|Mg!=0T=gDWWVrecIDsL?ZrxkThKKwBIfKdL05L>+<^phhxg@X@B$$_D+sx)D{
z(J>2d{lnzET-d*N1m|OD_sW064lg92^Q8_nJc)Uvv{T`J3H0dK;2!9?dau2*Nr$LU
zK{VcCEqF*2n&&zoC80i@gC@cTk&PpjBCp&s)AOEJC225#Uo*QTu>
zz7>0o6qX2{Nb;83lN%*NfMwm!fsPTmI5XBeGk&m1FAu3kq@kD5bIxh9jJuz4n+f79
z0?$M)CwjONG%fM6R34E+H8i9n?{xJ?l=NHyFV%z{S*5s6r|6j_ION)nv=!S2UP2!P
zF`M_1IzrC43$(f$l{>&OLcaU5vbBj(a4(kmZhufmv{Qp@iP^PO8B!l
zcMml))X3044#VvTM%K^=35sYd7#Oz>s^Q*jV+D3Bitz+dMg=8r`*T0=ax7Qo;ggyp
zr{{|K&^=CDZ|*V>9L>cXrOjJ6|Ic4|637Yl2qM>-ltJrBi{q~*VZJwmMp&wb=i8{K
zA@+ng?)?l5HA1B_6a8#F-U-(92B27Zh)|4oC
z@BW+xh}!C1qOFzsu0Vq)L!*mF>`!aCHZ3bR_m58nu!?mFQ?~j%OTOQqi1^fdFeE&7
zgq|u7Yz|OEYZ)7;!YvCjegp+LRV+X6cvk=#r0nl!O9vF_9Y5B&hqJ^ely2IpR8q*_mf0g
z7VC?&Hcghl0Nn2FIEXdM7!8>aSP&dFLvo@kXNECjjLXPh9NeE_lJrxY<(PoJ^@HWw
z#aRWTJ_c;!rTkSxs;sPGHWZXpLPS7Yrj*IhTrRF;dk6`L!brD~sf3BP9ThMCONS9S
zLu**=+7?nt*}E{bjGzG|wLtyGXcD-sBf&{~yOY@dlX{@;Zt)%wD4lsbf^cpSjo^}!
z?gfNAw49)-eHL#9n^`e_(M}vwFRbx$<U@9UUh9~t2r?E$
zK{T4t*OU$07*^8)eMCcf{;VnNhTh4&?Jq`H
zunM7Ic+4aTkM_v60c+Z~zps}ZcUOh7Z;bpCO>2lZyYXUT*V*ws=R9kCJ!-Dp0*E|`
zaX3;Cg|%+8NwQb}J*1jg8wQFjduBu@biB};6m;S0WNg=nh#1H!CLeamyB&r*F(T{6
zUN}z0b-iXSD_{B?aqqSB9>68bpT!{N;TyAqfyiYxSa)@tYJ2eQe)>qLg;B{USzn%xzz%XYGY72?YRhZ;hF$(zBNz+LeL~ni
zT_}?bts1^JbmbR+sOMBrtFgx_cPVt+2p2uYlv;Y2FZx-$SiO?7b3|V?lw#oX$+&as
zXmpiufJ7f0^!Ei$9Y3dyYbZ~`oJ~D=5QWzX)P&c{Fn;F)zvBg5O%cSZCp~&>_Xmg+
zG>OArZOF-X#i9Ejb020&FXv0ZE~d#HzDc}hgrg1Q(2kO)WI-vG$dW-OHL0Pm8Rg0y
zz;*i^E1>;-peZaVfvjz#oHDE2bqukQd-dtn#RIEWk`g>Coo*9B$9jAn=TN;IZqBi9
zx5*z~FnDe5Y}6{Oz&HX&fLJv=J^}I=0$0ot#>(L%#AAIZ>n{4hYq;t}_z^;^Q9Sd6
zL1|kjNkL{aOozKL*+wR|56z7xtZIzev_#kzOvl|`+uSaFuBTIJIuKjGf6WuhX|1A?
zXFrOGyA-__Ax+r#xe-z>gor3!YbI3k6+cSSinQYUwdwb20(AqSIz`FoVl#j8=s{75
zn^JsCfcpjE16Rjayr>3F!VduuVDef;-LSfOpUu5shV<8iyhPME|5%eQwjnCy;pQ>6
z#_OC=5JUr!?OSUcIkGRmV~jlnaXr*bm~f*E#Ti#
z6!>v^uI1T;$!^rQzcxHP)uZbWI#y6XG+C&}MM|WAf!}CW^uF*uc6vc8Sf4
z$lg?xj{$Z{ntG2(wUY8U7Nba2mWp-q)veLlRA~=NXzZRFJZzr_L_NKrttJdb=Jsol~4`{AtomjM1%~hr!{S^|thkchsC(mek
z$=1;firZOrAWJ=pzCT@tM`T9mn}<%i{L5>`QAqo0+HBEK7m|9Q^G@eP*Si7Dn*)E|L)19qAu1W>+yuj#@;=eB9+6c2k58FhSo
z>tgF57l~6dQ%TJ%6YPZj_I;>L)j0=zn;fdad>~XE^2v}rh~PSdigkO(5Mjf_@^r_s
z04W)En%rhl9fhs4XlNSEWf7DwU~EX(g3p7!3LQ0$ALb^V3FO@%Mho>G>!Cedw9^wZ
z6MudQ@o3+uHF;0eYRUH#=$w^doua2TT{1%L$FFnOUHm9bhw!R;>X5%Yx43F$k~fdY
zXCJ%UN)k&@671x_9fQFZN+R~Q|3Da{Qufk@o^RSy>zf$;mqf`J9W^s%cS
zU68soH#%8lWbmp-ytCxXjsWSbrrEwd_of6fHShJg?5zh!y*%OFi3s(ro%jcqC0P~T
z7*!!OZk&eYlu#m=LHcB0I&%6phyE|Sxf3nKT?I|0O#va)U!{^B1J{^4q4wG_|0T+y<<)-EFyh`jpR9R*4=B2^uaMv$ysuH{k))GR=_`_)?GYK{)(o#2o?{rX
zw8o{)p93%Io?YwG`Q5Qkl`fe+FW=T{Yc(q%E2yRW8fXAasu-gLlRJa$V)&7Aiqj@0aR-
z+=cl;w3j2#>5SSNrMIW*(aD>taHkMB0vDSi?u*KE?jp3>^!3ZNLy*i9OOOkRHO)LSNsFfD}9myC#?DVh-W`yPt@*&S1F$5ri$;i^&Wi9GL?-kX`T
zholyi+R4a&G^7-XWevHMHCNNo5~kHg0uQJK!e(7FuNH&qEle*-3I}Fk2n}}bufM*h
zihHc`o0BIa^K_n#hYbBzhJ65MZ-)S4e+bhY%QIqK=G}QSAB7L)p)FwIRy$~=kW1rT
z>G^NWG$U3)P&p=H2I2uKpDGT`b4j0sgH10<-R3E5Pnq{+fSbRNCmNb9Ux@Ahf~^@sS{0
z8mVX%@BLH{i9`d9jL99CONaJP+)uhC5Sa|dOfQr=BZF#ouJrKl)j1Q9h^12YWk`C7
z#A5JE#4V_GOV|tkic8ma4bxm!nuL>N6$KMbL>H3Gr}!Lh6x2p}Eou<6q-5QLDp~lr
zWMF7Rm!i#2i!gl2<>P8Vpz)c6zwq`=MS%(=A=U^_yg*4dR=nk9IRCv%6M>z-MXzfY
zDILq{tvV$Psf*=iU{mV~<^iPEZe&o@aO-$Y)eJBx^{ni>#~^9#7Z6m
ze9tKoQ-C1~5_6>N&H_jD&VJo`7Nac01^tXN4_+usZuEzpoi%V#LqT5HrPVu^rVql~
z2zgprfc_MQ0C%@h;bk?EoZ$jNTqb~*##=8+hyCh>fT+NrP;5gsh!jH~GtSIHx
zAXmkv8c5|azBC^uIQB51z5b*k3oMFVBc#eOfFHsURhR(hNfh)k%;SL;RaZK>CU)l(
z4Tl%ybPKhWZ&DMRnh2*)z8=+{-{u-8JX8Dix)^-LfbhW8{oOg$nD{-?L^y8@<1#j9
ztnf68kEZ4ts8t&~a&dy-eDR)qECE7eg7w^h*}XW0iW0Dw81@+7Pj2|5*-@$?`TI8$
z%3|PXYrSQkL6e2!j%Wpq2O(Y=Y?cNz5`aL;jf~fSV}zu)K#jt9@D@O}p%iooYbxDU
zj@zn2;+v0>{e*95@hJstNqX=q;kwYQU_-c|9Y;zXIPl`wkXcTyt`X*-VzbXOH9e$f
zs1741AMDM|yh9^+U+>=+pAvH!``t=Lz}poiwa7;QicCQu$s8r{q2wtzNi_MR6!g{T
zq=ThFW6GyrHCEqiLhQ|fQ_~43SDz;PX#L|Rr{~$3lwJ7vbeAv%g^J5IYgYmJP
z1znns6;k!NkQoxG6<)N;Y&!MTw^ltgkS?+=F6R<4%GjdNexU#&)
zrxZ4IJygTP;orE!{GwKmcq-_QVy&9c{up&<`~+ZzAY)NC)Q2h#@HyL%ImLf8ux_RN
zO^ZPkn^8!Hd|tl3Yw{6$Yz6PSZFq)tr+Jt@re|FU(B|!ktMZ4A2bMwW#z+qiv+^|R
z2M8^J47GeeSl`kCCIOC|>l&n`e~7#4R1JTauuZECwnOjKP!F|Jw7|Ts)Uz*+HZ^tr
zP|z=#tsz)jmP*>vmV7uPhs~WiC0mjHru-mZlw;ESnwQ`Mtu^1JaPMK#5D`a#T0G$M)U?iFmwf#PF51Ijd_cSxBu4
z{0+gMv!k7ACn8Ki4cqiGeRHQ^p0=8Xo#|e=nv?rkNs4U+ET7&|FkZw&^+k4#L+%lg
zSop*v!d{()CZ46!LBy321Mh>5Z%fD4M!g8)8@pyrWmXfifQGJNH;{l5gHF?Tnepg4
z<#SXdi-8_C
zm=_{wpzVklq}zr`OI*AEJkF&GM4ycuuAPl)g{(e<+;$q{X7@<^*eoXg+)xW{roB6`
zp|Y}AH&uM(J$7s;)Jl|EV4bSVVTd0aV3h<10|7{lW+*G*O?!yJJs*RF=FisW4o&9Z
zy(xt_G^85x=sw2rpKLL1raNM=!zo$+PMQQ)g2sLD8Fk+7e|dT(xnv{01Pr0v?i1-9Apv)Nss
zB*u36;C4tZn}Ond`C0k-E4qwsbrICQkm@{#Dahcm8Pcyw1I~@uznT8mENvqe%+5D_
zt;i{(#t#P9Ic|G&ebs9>6nC>Q0j?oOpnYc?@Ye=4M0&?)JrVbjbJ5DjvyvaNP^*zp
z+t|!%8xT;M*{Z2GIbfm2O3+O*yY}=m^L1%LI049C
z&Bb(RZR*pI(sm>rc{)G9>hF}6CU^84n4t|MZz
z6(flX5=HOeGh%rbNrnX8d)EAt=2`3l*byn!zSWqJzqkEXqAZ#I!>(EA(VvsHK%Rnc
zr1(4|j+lw;a0+n)32UB}U&j|-4F2NLqan)1mmn$35>!|}XOM8o+PZiX_SS;SVi*@`;7g4SV%J1%oqsdo_+xM%)S`3+Mc8~|Ir|rN6N|em%rbPv
zK6FN_itKU_r2$_^>EO#wH#;XcR;#Jf_vVWJXdUeAnl>ctOC2IiWEr(BvgWLLjgSoS
zzxoTa$`ktGPEv@C8wsb*UET*XQBb)q+yRWVbSil(Qg9<#GT!#nto0$X%^Xv5c}upF
zR+p@G2Y#z0!lep8m$iexH^veF%dz}pS>oi4ar+~}5I7pQ|LNdD;SV3lY1C{cL9?4*CreDcgS?oAaf1zJ|2ZWD9XXodljj
z^B3$Wqy21tGZ>~X*Lf{LSBX5KJ-Vz~rg`}@VGfEt#rN20FcTj8a+ooBsJD3LBR)DX
z-^@$1O}4^HW^N}d8PEO=3Zlb(r}=
ziMw?hGnsR?!&%M6aaZu&OZ8oB1f-Pc@+kY*JNvhl8fbfx;+1x#*iU|DjujFpi{gma
zvlxWDagTOhxOSP_Sfv{=C?UL@Or0I(Bew9RY6R-DG<%rCoUXy$Gi3A;q+BRMcF-o~
z0heWrA11R0y^XZE9~oPDESOIWxujt?%2yuevG6c;!LC3IyS|~gxJ*Ye;
zvud5$z#G*A0x_O{Bd3{B88j<~#`g$4dSU3qgqh!R&h`1nib%0^4b|T(ui(RYpvqP%
z`of&Ru9wOsXUFUMfSFHY&vn>>I2nnJ!%^F~9$nu|_oOeYNiii(w_Au#xH$c-s%c0k
z*k#7KfMVI0g%m8&`H+CshzLS@3a2hLs)jV=@_o|^${_*YEMW^Yl0&DQ_AM#Ku6{&t
zVWV)*Xeo2xI!9(}_d?6xnIB>`o{JV$5tEQCMzBRr)1$P3WnlQqamDOe(jKR^Wvi4K
zk(#9bd?&}9Bd2;53{&lVnfiqKKmmw*-w#&lMul%w9P06bnW@l8)e0PlYdN`Dp?-qO
zWXf>f|CQ_UB8evQee-*~7ujX98dpgcmGgJB%e4~EzT_4DyLnidGqL=P8vIIpR~>Wg
zU=`^%o-`fZqNj%`TXc$31IrdDwQ4l*oj;Z7y1N*=;y?4aPaEKq^APtC_DawlN_fjx
zqjZDASwW)PJbLxdzyl&UDik{j{S0(qIlWjN$T~DPc$s;!Q>H4_MOY52;u5Dt|Z&k|JTsb
z@Cf@Y7Z;2s#f77W{X!d|f+EwKX42CkH^%zLg_;T&9iplmb^JopBS*t6L3jCcaT`;2
zY0G`PP7s@{`%MA#rhOs0wviF?ek^tX(!B%6ue+(8U`)cHEBwMmggL~5S6N9!Aw7ayksblrI)
znq@oix2yy$GafxibU5_4w01`#?5t+gg3vb<>2sbT6@6GOQe~h@Ex{%hzAXj&V3;VL
z{bM8OD8S@xoz4-syid5@R$Ku}73cXZRi0wX3N-xq=#e_9o1x94C`qiO&*tnCv&v^s
zFvv^tT5Z?Ilc(K2FW-S`os>0~u3@J>vw!{p{ZIPb$;lGO%cCpYgsGA50a(#_Uwm6t
z7Xx#gV#*1}lSNRNyNZ7jg1VTXHMJwN9~}wm@0bOG(-{7;unew!pa}^hs}UKBi&dk+
z;_g_XQN^!5dzsw#{%^fvP{)!zTU3U)mQOdarYxAShf(&)=rcN0O#lw_U%BcxwxZ!b
z?X>pty@HcRGtDacIqy8)B!cdIWweOk=SEP(8fH|VRl7p10VWOsp{ytc8-5jZj|ran
z(D9C#wI*@0
zPQ_@23u+sbkC2XQ7>r5O+BkC2AGBUY-sALedJ&b8@wqMCA2WS0@p8i`Rq0M1W^5q7
zF2kM!+(|Edz30Es8X#q#7C6RTUnt1RL1Kx-Xql((#&G|IcwxEn8*y4ygIXT033KYS
zrY9hBD|-(e#94baCAmdaL#&GOqsD@MSBnYrCPJBv-)~>0;=DBuHoEuWj#zaYh(yUx^9ZcbW)CvbcLvK<2k4Ty
z^_kb|kjy>-pm8F>;sUD9>+Cf!(i2_dm+wtUIxSh?0+LCs&8e8=tP&<)l)3AyNodLh
zSWXOMw$p*27nzXbHpepkZRG{EllAsitrGS9iKu`;k_PW_I2EmDj{i#-hwGq~7RaP?)9X()tvd08}9PE7zj
zy%U&Iaf5Gfp1-2YdN_A$OfGGRO~_3jDV;V8ti>&QZKbMeLBU3A4Ld?%XEy6u>^2$`
z*AF1FqYspV{n;^ivLonvWg*Y$Q}=|vGK5rqS6>1Z7PXJ{6DPp?m%a4lQd>zS3lP2$
z>`SNyoBuT+54v};=0jwP3pdF+oXx`;EK+~Nh)y}Kvn!VUc6ECao+R(xjs_W_K5PoH
zzP~PoTzQxT2vFQ)#)0e~I|{GYw3>Yl)31~qBFFKfcprGI7x=`UHy}**i<7b@9iWLo
zOPB#I9qU2jlGKTIw)%7W1zG$#rB?sehTZv*wQLHKYG-r8Qo(51g%_ZQ-Z3YHPqy>8
zX>2~(fDDo59RvSy;e|b)C^M8MuF!=Ne*C%V%yRKq7jfM+HT)rX%b?J~uaLn^8FTb`
zF_b+=@0EwC*k(j9)`fn@C@lmk4M4R#u0BCoW`QdL&o=Ch28y`j2}S^1t~Wu>2GJH+
z=k2=#Do5@3x`TEhUjykVp5B~d9c1yq_HQX!r?)?`WW;rn~O8C_ab0sj-kV_t><
zi^ESC)j7tfvv8z_F@H%KFK_YuY_^M$Mf>9XHr84DQyjJJKvWi9GamW{yBP30aVgv?
z7(*UDc|1ScwDgatE?$k6aMd&j=I~zJ7z&Hj7h$^fJkJxB>0T9&n>L-8OkbGnM=g}O
zryJ~LMNj2Zsu|lsFvP-^Yjr(5%cc5w$`~ZLXmvyNlz+~hAi5Lsr8Ga`1Oi91c&O;+
zg&^LTYoDiOtu;WRA7L|u`J-};M*x*cV`CUtc9xQ*{|x;PVoVrH5)%;;Hc7j08`$oS
zylcT1uGytBcG!%XCzPy>5sOi(yPj;r6y3lKbOh00a_5hK5VZY>Kek=I0FH_p
zI5wY-Vu5ARs~A!uuydq&RcMeRoGgBkoTgudoJVv_)Ax^8Njw-5II#JVoQ9L0vc}4g
zx%6LCgF_8phe~>uOJmlXArK0
z7CLom&w>=9V<@%55=+b!8kSJGyXqGFfIQ`FxlH_tWm-a=?0;l!u5MMH~o~^R9oA;9RE~C;O_43vUhtR
zkap=uPwySGo8PZ-Sft+PJ}0@>ziUP1V=2pjM5g-41SA>8G9qI$QsD{e@}XLp!F&yr
zK(+p^VQ~YkWoT-0B36_S;RKG%s2L>uZD3A-@4L2vRA6%TllF}^Fg+G629f|26c}YM
z!|WRZ&IfQ|Q+urkZv?c6{iRFkmz$9x8%B2yL!{t7+P9lEye&mAd0#`0>~#L7LXE_{hlr3LU;<>4MNN>
z|2g?xJf*#p?+6O%i0FuG1i_xnGY6;#)Cr7}Q}f^W0mgpDX6o}7_LgRK2ig3#1;}r4
zacey_Hgt4$Hg0urbTR|u)?(P){24B@vpfcI0OsBR*y+IyZ0+kE`WrB{bLP?t@>A89
zCkr^HodF~$*T*lJwCB%xy;JTZ?bXEn%Wq&W!tq;S`B~=>fQ;~lZ%Y%K^K&vmO+5*G
z6;mTHBapy2_C!0kzP3ILY4EiA@&cAQ{{>BeKyY$!_Er`?(5oBwTQ}WX5FNCedzD<1
z9^bIZ?>>cbba3>}|Mz`%lnpdEyE;2N=SLM0f}yPu^hfaa$$Yk2z~2^%D2gcw$?8Jp
z!F~XkkuqR+4Umz$llwRJM_FD(Cl|hN@B^@KWD?TA7r6ptOa{ir{55O?@*dvzK+T?q
zGZ>dw<0rk=R=|KgVfD8)OFM9C>X*U(#AvnsbJw+Iepiy)Zqwff`}&>I~@V{v$8$SBJ3C
z5qPRvCTBO~hm9t}ziKjIsvls-KVvuHYx7wjDgT?h+{Iqm3Jh%B3t%eXB!VW&;FMi(
z?&p`;{P&FX+{n7h!Unwfo4n*VhPkz=@%b%&%wNO;@YkS36@+^ObMt4={Fub@7>-4`
z)rqa;3%%jDagqbO*HRzE#1PtNGo_a-mjBS?vvjkqY4l=t9@OAq|6kB5n^?+5ke`ek
z5PV$+{Wdbdf7he<3r-s#Dz2z3t(~>+FL`4B2V@Mz#tgI(px)UR0E3$g%bTFF=R6=M
z8hUT|h0KyU|5LvJK;w^~!D$;xKzDQ#d>_tL#G5%Yrzaqd0m5DU{cUD6YTv;X*iYw=
zJkNyx$bH>{#xM4se~z;Ms?|)K0d{^`>iQ<;H@^@{Ob%|%*0aX0$-s-5zxVfN`j0LPHX-kao!iC(h4g_lznRN>gN{qy
zpL~;n1w=Pg`Z-$Z`L&X$%KWk_AZ{@}soszZr6OXEEtsr9%qVC+(_NDmaVJiWE>WCk
z%$i5qk^opn;)XmnJ8z8F$n_cNl*jl}TWU}HJ5XnuSPi*1j5_FxdT$AIa&H1${JKYf
zaBR1z8-@7IQ}Cm>$)L@57eib04GkhGQ=S=ugI|J5kmn}}1`IYhgIcU;YIA|3W4T2D
zx`$`3fAstzStAbNNWoQ4gf)~P`JrXlTB*mQ|H#cD`0{Y?I3P*v|LJmxHD#2lDZ!gS
zo%M9aztm8(G$E)CumWLkXxugdmJqj@TS9!2D~-1X4rHT_0|if3%)ok%5wudSKA=Jg
z>+KwuzF5)N;Jdz{p>n+?GINP%h@Adx8jys)JPPPfW9?^X0IDS|ZBsuZaM!?`9#svR
zeAdk=j2lN?iWj48t6I6^Xkw;O{OPEiD_J;*nbM+!Mh0+gNMH>aaze&XLE?wgNe_u8
z5&f{TSpqZrjhd4HZUO5^vuM*rzJE;fv`$kKgxIIlHR?W-ZpPm+WIOMLw&yWQ2W5p9
zUeTO&dC!1>nwwI1Sf#lOvTJm@PA0vo9EPTG>`3V6-aNqysVj(`6Oc10{FpV|t_i-F&R{r6X
z$(LTXfjmjv>oBKshhN42S6N3%a^{D8>?rN%ie8pN!pzwMxKe#rhQ!n3lNN`(3gWFq
zKQH}qh_nrPk?mL@AUjQal<`?AVhG_vvb)nvlv9DG6{wi#EnA!Rx)8{dJw=hH(7M+d
zSQAuc%*lw*Qo;djle7x+o`Gx=c$;s)70RGny{T3GsJA(Y>rNFb)c4{Kv`Ugs6?D36
zLJQ}_z9O2<4VHFsli^v{6lkR>4W;B8c@HpUhJ;huk0Diq3(PAUjGp(1Z9LzbaH5l`
zp8p;zTg`eBhvlS`gGctcrwK#rrR~lrRCj5|?p<4v$M>|Rpr+oUPO*z=KCdRM^DtHvXb^B5%v+jRZ34sHv1(=xp;
z>7E-jYWkZ0n%26@&SBVESAkah1L-b?uM7d9CuKVl^Sv4POAp#O>9o=^=RUgF-|oX&l#
zP{<|J^zyD~_LJ~F1q)}@ER?dYsO(#W1-M#(y9^)r8bG+mVH*ay@%vs7wn8J7hvGflan)dBP
z26}ylj3TLY70E`C;I2$hww!Qe#`U_=)p9hIuOO`z7etJIRHMIKHY2gGcrp&rdFbNd
zV8|4sj3xyC$iiRbEiBc7dk6V&^kYp$(WzKTYm=hun@zsV@^X@lk6MX|R%qj)?P7VP
z9*>+ALI6ojTz&;S)OXF#9p?30J~4Y$1q3ZURgKj&xcxT|bMDd2jpOK*by{d=e*JO_
z_<`)MFUqyD#b2de(81Co51HEOiio96rF42-(wIj-_j7>80dw!)!vwj|^<@nk^%Oy1
zjiaokWl^FTdi#kV3m?%iKRT4r5fg3`v}Y8dnN_dV6EITYiKqA7Y(sW?3w*LpUjoGw
zQ2*I=zh+M(VWfgv491W_ObhjUe8WKG+~&Bh%7+@*v+Zj^cwB>R0eO2SQj^i8Qc^l?
zQgWpLyJ_sA(|VeCk349@@2)CeOk6vnQ{G1=j+a_5=~?Zal1^+{Av3jh6VHx}lIzSQ
z^Dgp?U0;9GT-rK9^_2=Ow({^`B{MmSpgtX}|)(
zsu(3|X>@6Nq;b!|zJA(FW5*7v_I%|iE6GE8``dcE-f{uf?;Qk{4w-r#llLAwo9>&e
z;OD0HITEhq;O)Pw!Q2C!{Q2)&ngb|aF5wG$_KTs#Aku#n`(W4{h7c_n!mKSVpPiRs
zqVH6EUfeef+z5XXnAm-+qH{noKN&nH=x)ALgLe*TJpyrV=at2xyd?_Cny7TA{A{)|
zch3k|`QaUTIBPA*>R~NyRBs+}tX3ZiRDxPgeFRGhe*cD`-#E+lB--k?F=zgneHP8R
zcjF{kIx-4&4F+YGd*bX(T>t>`LBrP`Hem(jLJ|B2d_M0)jn?II)iux_;h?Fm{Vw|{
z@lY1v!n~T#PqUwIxI-$$%}Yj079an2SQ4=d%xyj^y~s;>D!7^#WZo_6+TkA)R;_2b
zt!=JO9P%nOI&l7^R7;Di;*e0}7I$h%a=2B~=Hh;=%GZ@*O+HnIuZYL@6sCS>%PSO|
zOl^P^9BRAKF|#BL9afR*CNd0=WT=5|Rod=mr0Lp+b0G>`d%>HwXV+&rh|EG5!d}p1
zRAZJ&GPps9;v06nmA~ZcibkH6=BwAe;@aV@i#8AMLgQ4Rw&NK$so2qg*H9Ba3f$KL
z)l}b|#Sxf*D$U?u68H%dXABzlAI|HvIQjiDwPZl6hH_mRp|IOrO@&R`=Pf7#*Q*U`sErRlsws@Eky!O%cS`aI%VuIb
z+zFV?rL*dRA2Zze<9*F~sC57Z@lscU8<{Bx-m%2ni#bCj`W1wOJN$g}ls%Q`h22Gb
zQEC?kE98EjoT|>))RFheor(@cZZsjfBP-D|m}^4p{!YFC$MNp5m+DGToAM@!O%Q<@
z*nLCSd5bA-`aw*^vYAqOA^OsAvAERIu!;GWfsO#R+)msi`6c=r(n)q&519<;j+oE#;Al^BrWoFa1Rc;r+8~raW%$r;DtUgf=6Ls^?RXK
zTgHl*`7T*nu$pWCIqbDGA?V0$g0Am?5=*xB2Z7nlxE3?+seGJEglIA4b-rZ~_UkWo
zG*Vb`*BmbO#$^#odSU{2)K)1Ygp;$3>NLxrWu%P=EiD~fxn`AFUKk(_T(nSwsFM_V_9CS
zmZp}qHdXZH#98s%)eWT13M+;20%9f%%WaCyOn)pp!$X1R1X1h1zBCv*whbhdj<3%Q
znONOnR#XAZqPI^&bM@c}U|zo?>1QGxRkct_Ju8~k#k?$S7f!ewhnc{k&`y7F_PS0h
zK+siH^}yA@j*bm9#G-_i!CbU`dv?G6E#sbBtu*xd-J|a)Z!8QZM29O_lmX7!PnfCE
zgG~;LaFr@oQgI0jsf51jzXaFNzm7ODzCD0(rL6J1PId=0{^YAb
zT1bF{*xXTw^=g;r-cuie%ULfE7K2KGEwyA}$P{bp^rdqwXR`rH5!9{+W`E-Wr$ZM8
zHPvJQ3-&WG&KG>#Qs;~hvPkCd2HqMc;h#l$YX^`;UQgB5+N1x
z)^*H|&<>dKG+2~ooh#2i@Xp-*Ox3eQfu3|W;e^*POqg8Y#9s}P7mq{7-MTzHfmTC=
zZ-*>Xu`iKxF(;PWRD8K#Q4JDV6wE4OdJOvgV^$TE;yth^#P
zQr^{&4xBu_`_?tP;?w!D5qcoy37F+OxB@T_q#FqCG@Tm%V4DzxSl2#O(1onE}PG|{Rump43Ipx!scFA0}6
zcD!Zyyy;DwuQoKqyA)S1x;M||B7{zqa50gDN-@r3X8}5jgc?e1an_&<$#!UBZP+11
zN1(ibQbUgcd!X=0K{tD-WFYZYJym8EE)yX=W?QW40NKHB8z{-cVmN2k-g3H(>lAv~
zC8(Dec&*`%KK7MP8q*Q+7%vEkZRje~!)TovsJMC;5_`)F_8Y*ORpQ8084Fz5aL@?+
z+)UI!v22{M^-LjlSyFk|2xgVmJkpy{n_|7{2L87ihkl4>00-#
ztLSDdw$C6VKOUlQ>s7H?PYM>_foDkdjlMX!?ulCDH|2D(UVq~>27|4ALVH3X+XEH`
z^k8UPnRR_#h*Vp72DK%jwXsvQ2kmmEP#*^-LuHW54cNxCHFlH%)rrv<%!$UF^kZwN
z-sh&ZaZ$$2wn+jp_WCB2bvvffVu9GvT{ij)@d=L26|rl_`@?Utd~>YE0Y72SE)0pU
z1*@Dh#^=wXD+cG%>XTG7r){Bhe<4{iMno*Zij#RFLwJ?oA
z@Esa>%g3dN>d$uHXJ550_K=P9BV|i@I~*k@0uBV5^_kwYPH0ljJLDS(e+M+TA3jWY
z9#5%u;pZ60_Boo=cRfOQ30j>m^g8EbNtPJiuP~vguFC69e=Z4WYQgMb_Se=6@O4d%
zPfs2#r}`Pe=0o<$(7TTmlVqk$(K2~samv~0|00-wYa*IoIa;QPXe|nWY(jHIYkVT=
zC#9MWJW!k6c)LFxigrqzfLsB}cswr?=|4lL?9|xY7I5}4zLpb-1)(m?
zu-(Iort0NQ29lyyC;wQro^NTVCC{#z_L&IBC(D$CU@z;B8D`PNIm78ZooUi^pZGad
z%RB?gY6PKMIh(p&cLd?f*sG1XJ39dWv#tIzDLrg{uWCYs&bGm3M*U9Gk%wfS>`i7KAtqPuR};R9TD9LF`0S2!yn2>8-x*%$c%hi$2>0QpOUfQUX2FQh8wdsgY0g
zHJRdF$S)e5H|$%2Bnm!iN`3z3578c6fk_j)eA>f4y+|zN)cC_cv&K1Vj7<0AVl=%B
zr(w515?@T}h
zBmM-zKLrF6WYgsfVcWOqp3GBng2gHq>_Cn@FBB7oPB=`XXcE29`gr>-Zy-4UGeLxw
z3PsOoEAERdThlE*-mJagNEXuKTK+0`0Thueius~8r;uZl8L*Sy9*Po#;cU7|ci&tH
z86T29H7wwnsbb*PWPQW2Y8F0v%R@SiPV|+B0wb*dV3~G%bDjr5p)cH#fVTAAfj~}V
za5c&2yQ$;D?@6evyhSq96=|9yc|REg6Xx^akib@WtH=*Fx!=6LXNWE(<1jOV(SL;F
zN=}IyGe%_F8y3UnW)q$TmV94CrF$D})21{wiZT@#5OGKi*8m?VrI4X+lcyG5!nyxI
z6=6>}sElhGwj29z6UlzOo2vWqVRVLj#{DFCXOso0FsH#NO?4#)(GllRyi+R5cbQ2$
z0?DY)y>%MxW3UJ*zSc!<=NehOHptlpQMmXn*@wP#vXgJ75HIF={kW7PDArY9lo{RkX38HNuNDG&1jW*)JS{U-s6;Mjq>>Zn=
z+5yeOI+w%-D?VQj2}gFHg2z+KIDf=E^4!O{VY088W|DDU6<_|z0c#3|hFhiPq3D`!
ziz<>KyP~e08m;fTxMso`Pf6gBQr;Z~NgFvUJD<GQ1)iDzT=W+TPtZ}){H|UQ%rW3KeT?@)l`&U5v&%wIDF>2R
z*~xSa;DUq_ZC2L(al?7vODls-XTiesTD8wLqtoxTN7zH2LH*~)x+;#4soH4b?^_G%
z=R=d5Tb%|;?QFJqW>N81k6pX<3I=zv4*n$4B`NN-hMm|GO>OeUtlCKmwJ=h&+x}8^
z%$%$fz=;>)vo2q29%$-HGJ#Of89S=y1mx*Dk%oep$lvQX4OESi0PRaoGbVkm6=>)Y
zlVcqyCgKyS))uB2_#0QQ
z1>JlPf2X9|h{Mnu$bAlHkY%}HZHcf-AH}JgKin$fgo!HME4JU=U{`$d$*6`JnN_F8
z0JvST8A;Gg3cS>kh@2xH&N^;pQ;@dv!MzsU_eHdMT&WV7*ac+`g&PNUZmu@X&Ne=O(r99P@K7Gya|s_og}6Mv$o7R*pE
z*}eMzdZbcPOS`W}GtphDDF<|FfGn)2W}Af{H$g2q(S#27J-^8K%%Z-|mVWFEp)?rT
zads0bWKIrIi)olyMUCNzm^u&0Ze6f|P^?2QwlXvZ7w>muuPk1}4%B@BX$T*Q2{D({
zOfL+9Jx*6T{Hu$k!7O@O0hf=v9I|5kF~?v_a|+#!aI_52Ev}N_y`3PwYCT3!Y*zD=
zubE8l*?ON|iDBw3d**_zOFQ7^l&0M^rOv3B$NtRl4BYtRB-tlvXM6<{*a{;nVe?uAEj(88n+?Og!2wd=Ck=Cf>-k5g
zc`bhL!9#9z*~?Y_cxOi>lHveixD+LX7mb5lbY!pt1+1jjK6w$DPvnnL5Os3I>Vure
zkyUS)M>-3J>c?grqEa*Ma_mze<9sb3*9QY6Lr+CAvM+qB5V4FmyVD+!c8IE9ubnpj
zK9(T=nLa*Q!Ql!?oJ3-^v#jFC$GhYke3n+rn5ZODMo_?x<3YAMvfe{pm%jmmEzU4}
zgEoBk3)*=ml3Cs#ub{w>&Ue?FfZPK(kim^a&|I!_Qyc#o%d=GTk;i_zmO*igYO^vm>g7D5dW(xOtxdz-*
zA}nf|^_=XBT*!)g={c39Z6Sq=MZ6{u2}vh}Gs?N3$i
zLaZTdo-z$5I1o_UdFDI7gY=zugy*Zt0?#RS9FxjZZVH}F)^v1kii$6Wv{|Ws@U972
zwp+u?)f|F>S|KnBY~mY1A#ti+hG$3k{q`ZDR8}aSJiu?v(+2jujcZVQD)UvAc&Bq)
zVN*Y(&+`hv6%^a=t1O#K;b9{RUjjm!#heFQSb}zM#1;Ce22Pio8H)+YgdmO{>!qH~
zeo=wN&Y1AuD1YRS954%Ur>^@8zpevvA4PT+6bZFp0iJfto}yO1nZ4S;LDy3Ku{}@N
zShRLh*+A2nEmi&wvbhJ#e<2RpD4IW4IYFWHWFu3g+$kznW!%@I!`PBCN7p&IA~Iy>
zIUZU5*@WXf=t)ah#b2der8pdXXQa<7OLpJoh_eZ5eL%`mxEG_
z$&&?w@<=cMez5VQ{8jYXS`m{b4K0bd?G=L0Eu7$o5k8qe?La0sM?Z#iahi)XINu~@RWBJcAnwq?e($>O#L)$A{np
zuenr}fSOH{-D5sdlK%k|QS*AC(LzBb#l8&IZJowtlsZgvVPx5fbbkh@;43?{3=FJs
zs5RF4offT4@$T^me3S>lDVCA7X4jJ?YhD})BitbeLT4bcWjc6@
z%)!V*HZNRQg*)=aU7b_q{B0H3-h@X?Jp?wYr6Pdch|TRVFq5{u)Szt{M^MR`8TeH#}avM&lS|$ZLaI$0!r9^G2PV}2l?J;
zVZVQUS~DIH$9J~BX-Iph97U8=gORbS?bPYr__qh2SV+X}wY~d7BcR*+r3HahXl>p!
z!V2FeC(n!18x*as7OJYHv<64BIZCuna|Gp{FYMPyC7q5erNgkYh@$$Wtb;lqAv_H_
z?m4ga7S(PFcvj&(k9s#vQPw8q%8Q3T9BJIoWl1$|h%nv{-We8oVmFCmz#GtA1coo}hxP(u~u&
zHk93Q9j2aqUd-tRr+n=U_XwC=gIPJfLkFcFGP7^2J&S*7W5@q%dU+$3X=e9{A)*cn
zBiY0WS#uJB%jZ^1-7$vax+hg<=<~Cvy5HfjM!F93o0k`&UGP#o!|;#kIDZ~|1gni=
z&ntWh<08T6Bje_DU
zk%s_Rsr6VBi`D#;p=03qV!aS%w(C?JAILB7QtZyuVaT&e#n#daZ>-
z{eT#3j0=+X9j#a&=s;6l<+nguOmRAEIRzb8`8Knle1axoeGU-*&gZ$eU8_@g=Az9}
z7&LW!ssJ0mM3<-xDg<@d>=PxJ`&xR8P--Thp*G~ILco0q5)=N>*uVK#{@N1s$lRuPusZc%aHsrIbmt>3%9D$K
zEkz`vuHYANe4lZ_;0BQ&BT=q)d>E-Dv*Q7yjH4v3}kIM2U@F$;2}y9de$q<8E_Zsb&5i
zib@3>Kyi@8*;W?&ClJWll=|U%B#~wc1-E67#`{2gN~oO`!58Y);K>^12Y!w$|Duc2
zSHF9b9^u1B!@81+iN#7m`^b2*@Q2z&#R_M=Q^fTORD~wpbwTIa$C}4QSE|Z2lcAm=
zy@P?4o@Z<6vjxrCoK7~@m|4zjaHM9|gh{@)(|&kjr*~8YA*+dG8taiF5ve6FrRz0A
zum@9)C@oUHwRelV&3Py+3>C-%Kg2pOx4Q}T#AXcNhw@30j`zJL(-D3*heWaM63$aI
z=1?8&O8M_JP#`E-(R)CX0QUMi*#-1FvS(7uNh70QS@CHnloMz!=iQH-5w!;n8zg_}
zZY%@P$(2X87WA=l0`4h&DDsRn-+UkKeL#H&Q#%?&Cdh@&70*E7fLmU&Ed-avy8|_J
zI!9?%Wcc+%^F!fmYB+>`nxA;?K@DykbYxp-bJI2%IpcZfwTc(coo@Y_Qv*uwBm+lmoj?emdTRs<2HZPmMoYU2U{DyC
zV5^3*$9>)cUG=bh(ZDD~#gOCMPWn`>8rfq9_&CC{w&9fJCVY5TmGqjGA@^=*<^c79
ztj~E^;t&tVkBUkC0GnMvt1Bijh~D!Tej9~gZ5IT2gvfbnGtaG5yPtqHoS^O_4ly(6
zx!DW?938B8?Esqd&svO~Ntdxt2n&BNhQfg++?U^-F*j0=%OiEh+KIANoEE<`L83L)
zs&RR1V!6@%vxpacBKik9HXwMrc}7-?v)bljF}|c%2ik7B{2l*ijcvLb*p|@T;=e+L
zgvsPnN29O_pE;jjGL7q)tj!FlaTWWt*tTmlK`c!24TBKlEHRQovr?q(`xQq7=bQ
z5)--D2dI7Qta_rdkT0aS{MY5ZHLij!D>7(eVB$r9unyu+L!*^e>&*aN2dKZ*%m
zFGvMadOPj3?5UDTTP0fU_uP^!Kb(K?mc<+Ch)fPVU&K}IBse?*kgejRk|7ybXT6;Q
zHJ6~Sp@q^NUa6^@jLPR~Urt_uSEvRjPhXcJ%h%v{gLPDtGlq3GLi-4NdLd(K6#sxu
z#S9NctPN1+y?INr_*E!aP@g#wc7?`oal4w%QrD
z@jZGg+^k>Eh}1bN8pPw}0o%`6C#W>-IgcIUN6qOM9+NjmNsHC)Ga|JmE;mP;KKbXj
zm>W2h3dd8v!rtg|qezV;Pbb)87HE6Lqgv$n5IGe|9BZg-i(+;yN5m`Z#|J}axKL?2
zM46pYC2pb
zuC5*^xH5=92kIsQlRi*6rx4G^bLj&1MCSP^`xKAZ1xtNK_{-uYs~K8ex3_v68GPVQ
z7Mm;}*MZCRtpBY_8CfkUH5u!$IGU6wVL7Z|KORkSPy=?WPaA9H$F?lWY-2R+0jVIl
z2MUHtKsl{PHiq#!^1J|N7a6r$CYlH8mQUAsU=tmt9*>hFs69^@6vN8=G%6I6oNYd2
zgQ5bFDx=9ZE5k=|Y&wtaE7iyD@X@ql7xVDk@f_whT)B2;a=^Ku=T2BQRKn@*D%Y#H
zR73|3%YxIT#eD<2!1IDQ*H(jgTuhzx_We!TQ)#4)?W`|?9&Z-e4a~EM_t8M*OvVSF
z_H#2w+?s(B?bw=sA6`}?-r}ALHcj1YYxc(2f8$1B1g3dhgu}Hz
zPDq3Lvq%M7C^rcV8HAau7nlA%4EcgD#JIcu7@B8V35OV_bl&R=nqv7ugxYRw)KK;=
zFS{(Z8~2Ul^(HZx`yY&*L#!}bkgbnx+qP}%Kelb#wr$(CZQHhOygPYG4|>qUJ+16i
zQmg7)%PBn!2sgLqH%A9CNb)Xybp>r=@9Js9NV(scw^LtBXaJWacP%dZZglG)EGR)<
zg0^t?K4JsbRHICcj=XL&4z=e?W!dJUJ`rO9GbXI)cEv?_Dp!Od5H#Kg;m^wuKf#2r
zej7^Mo&s5kbY4F6vjE($j!dj`)Upv`Yo-tEXVM~dypPvedRB=s6Lp<2jz(P*^&z&&
z=H*R-W~3y*TyD%|8N6_>-*05HLxdJsjfz|9h%Uen3^YHpl?tkD)K{oki4(MC4uMK5
zHZ~-AubL-AbnYTS0G!LH*>kr2=5bD4Xy)^!G%-wfwbC`Q^7eDxTRa2e+dAWd2eMAv
zB0X&Nr|AG{dEB_={iJ%L@b}(D&%BV!5^cZ*c!>mb4}9u6{ohb3QX?gCVuq^Q;&Vl5
z>Xst%k;lu{fYW~tQ;cXswG_Bj*Pg{CIiOy!{a26aKWWDG6YmGl#zrs54pE;x3vz9w
zCyonv9D6akesytK<2#|P9(Ku~6$Qg|QijgC^GTGsNY3OX1yLu6CD;7C#OvbM
zF>`+5pCtzWq@%6K55nq?j%Kf|70NxKoxGj>^lWTWS$(?LW2PK>twl}lV-OZUKY~VM
zLLRq=XyCc%rh#ayX#*rrFqW{$t%s+KCIXsq>nw{UTw<
z>{^d1|6TeTSS!*KY^*@K&ViQXJnaYc$!m3+6Q|aaSvU>TMqK57Px5DN;rB?s_D=Tc
zu@JkGF;{H`QVxZXSOJS$@E#AHsppC&=gV{(b9Oy0mOAEZjW`P#+?6~sw>oKa?qFxw
z#!^{&h5$W@9T;grDL*Tzf;2l}!)QNc5G%bstRuJu-=|SH`(I(IF4*cP&w5N5DIypi
zi2dtv^(RvaZbwvo(aeY3=6ft?-|I`dCmv}bq|Z7;sSmc~7!Nf%)}~mIsRv$UbV}+*
zNIC`*B1~&c6QOY<%r4-aG460T^V9%uq7c~T?|gfT$RyY|V|`O>KZ&+?aYm?1uB444
zis%rx%j|pq+09Et$_xPP;EEzc9n;~mCf8_5?hG6xzRFAJ(*r-7uIu{Xts3)5WuB#N
zcGg4+90c%@Y`T7G@5++hJ^MluIvgBy_TGr#s$njvgw6wtQ=F!P8?XqqkM|>lm~U?C
zGERuiuud`g7{C-lolvI4<7
zEQTXb?5+IZu}Q{~*O(J%=d1z7LQc+Au^iJHmSwderEtmMWEvA4L{92qk_K1f2_Igk
z9!rvtmF;*A+mKpDSsKn69m)X+VL()q9jyg-dk=9n2E%4tbWR^+lk(i?4H3H_z|h21
zTnx#TlU$fIRka}|-3%wEiO(+rCmRtE=FTQ!XlX{s(Emsyw@
zU~%fvDRLk^U29{e
zMM|_w!)AA1jgsEQQvAKcw$-nVDM#K|Di1hV)*GT+!wK5$H6)`Dk}44j;YFZYU{Buc}PF4+tAFg4r|~9nQ7u
zE+m2a5_7IrVt|iT>45L1MQH%0iVXMttqCm+nWM(&x)x)3miXk#TmB|iWol{7=PAF=
zpmU1U2Mx~I*AU)6kUs}@U$5EEd9uQJoy(pFUlQc#Z0|dhlpg6Na-YC3ep$jWXUs)G
z66(Z;?w#X-Sq3ADc}VQ0kn1ETPQ|aGzdDlcc3^?Xr%v&}y(1ER`hu~)f-f-r3~$DY
zR_~yU4M2zcjuIG*nR*P1TA-Ay!Hs&*SNB4|cXo$vX&~2+=Lt@oERZPM#|xid$I?C#
zy0;;aDB~I3=2sjW0LP8@$CRD8QI4&5U%#;6DSE|Sv_uwjF;#ftUPqeBF6v#%anIl`o+-$+wr0JJ^L7R`Zp}*2-c{8w$gI
zsHvB^?Vd@;8fb|Sie_NfyT&FoPTG?g`mC<20yH%z25x9=oY0bwm30~jxRMLikKecr
zv_rm~SR}XFvSeuv%`)VxSR12JYJ!N9>MAoonhruwWd6;z@$cV{$AfJi;Lu7-KS4XH
z=&5BBO>?0)OE%78s_VC&aRhWG?QQwRvSw0csq~MslaVf(LvI90xK)YjnYBGoywU9F
zED^x5rVvNr^=Q-B!6nj2i#6ZF!g^duhhSVL+<3CXimuSX6=_8vTQn%0W5w$Re*UCz
zKY2`HiLza-PX@~@wYp8rS8G;}N~|ilcnKh^GDw;#siZi?GZGD6@#YtJdjaVA
z&z~})?!YU(tPpoydAa+MYQ9GC9XijU2Rlb{|CAiIyy-o6b2i5J$dz3(=NxqsE$RRQ
zJkCuJ5tla{+)}#mvC{0h9a+qo<-Vj64Jo;GCFW|H_m@V2=pUKY_OiH$p;%URwp=kg
z!`AKywzV=u?aYJe<0xA3%XN9loYTn3&jfP@>H>d=y6AWk{;~9b;umFs;$^+B&0!{U
z{naI{p9EI!BTe&>n)`%`s5*mt$*mhDBsj?eoL7V{Ns2eGX9}wxen^ekguO{e)Dksu
z^H9O;!9*VHWn%qOoF^yVQj_C7@tlfjP`4afI4#8Co@@TbWgGznIq{n>|K%c01N)9o
z7meY7CT-hSvP5EGLi5BehXtJrY)L-TL2vTKocA(YCWotC#N8Xff>(WaiGOSxp3IdX
zO4#Jh30;CwqSBbPJL+gU4iTC?Bd?9$tDuAIjE&V#+o+guD^{pplmqFa+n4!l-pqu`
zS>VY2PHe_3BPobYL`U&^+UN6v;~QpB#V`XGlgHGWUJ-Exd9(hi6h(w|C#Uq>3mvGv
zZM4xaR7hzoSSRb$kg#)XMNS5OShfa@7(T<%Rgkbpe7OS46`gQJB{n>*xE9&#{s!p5
z(gLBX7)VOaz}b)~OBYu+N)5gD+EE_=r>YGYy0m$d)S`^rlOWUjDs#)&uaI9PY-8gR
zAHENkk2y4u2~tPcOA>Hj6bT%5m;hRztl@~+fefaQ`lJK#d&c7mybp7?wr0FhmX3@D
zxdLt}?d%XEgOmEnPUhQCK_48}4(ZF&77a4di^jlGNz(a>;|~e!$2pb_j_<7F#P5c<
zoS47ZFCoSS_}RIB5c%}vf&sMkBh1(-1@GR88@K4mV%RwLT~3++pJ>4~{4SJs0py3l
z+I{0fd=1w7gF9kdt?p7W!cqbDwd-!kI7ysK%3|8EFq(N~V!&6@{-@q)C0L$=EeNsG
z^y>D}axmK=*sff)k7gyVS34-K{Wan
zx+R)1mlG;2bOkHhv^DAHc&gMdBe!p
z!Li{54?w8ou7{!d!tYN3-};3C_DSK_t<5Jeyo>FkMzZh#5IC^scHGH|SEDgI1tT;$1A$!8IvhXkMNn(plX(l#p`
zKAOMZ7!UORA&kgo33Q;tpm>u+tRKK)@bSO2B*HFJG>ga
z+0tUWC@4ac!E+pv!Ni5%8y?e&uzTdBo=%MoS(?}w4=^B;kXs(1kKUhq3r99nyN5zv
zD-dn>!gL53b-lmGYWgMwoMk`SkE?}v!;hl+4K@@WE}PyZzH98$WAQcgRE>Vi+x~JP
z7OU)Pm2_q9{%NY#JkaNzrq{6oL-KrTx7FV~W1iEYw`4AGH%j>qrR&s|#`4cucncZ4
zW>h4yEyK1up0BT&Pj4R#h9xpc8mCNkqn9jlwq>%kTuSC;RYf#{^5J1E7u`+umT8UY
zcn;DuSze(|Hm4MQd&~+B)3zM>Mp{#u?yY@h`Z?r%ISl4WB#c81fzG6`!2VLtz3ol5
zu3D)SYD$gtSX(VmNZFyv$0ZFt`(STua
z6pGyos%dO#PnCkYIXb&`l0siMPSfq&R|2ngPn32NnAb4}nMzirf;TAQ%nH!1WaFZY
ztki%joOc?a1Q~-GSQb;k1~xPY1g;EH5w&rTzZA9=1u|veppDhgqcEIxo2I#~B>8fR
z;yAs!E#+cJ)HM2l?c3LaI?bcDRee*TkK`2c~nh
zf;f=A00zF2wdT1Jz72Bt<@c(6JNw@_Q}+LjGi7IBgQAx)u{CoxXCh$aWMX6f|Ih!)
znKCgjb29uNIn(A&7kTZOPul3NuK%Ezft{V59Q~l7|6tSGTe<;40ed&-Tibwrgk$2n
zj|twhUBAA}%$~9}qpCdJoffwhd2dC<5_C3DOkh$0*V*T37#SV_P7l=hngm?(9;o^4
zkLimIZ!b()c9io`!^_FidkZTMO{w4+uoijsHN?Of8PTnE@%eKlY_P
zb3`!z=!6K?@tKKDBzCheVAs3Mn+)&X#h(BTK~?<_DCuL
zMNtJE$&}Co;)cqo9^UbXzV`6S$_i;;z;HZ~B%vd1mqW8-7rL<^O*3
zD^GN^t*tH0DV^*r41u^4m{*qHd&Pg4Ea0|4?5clU7x6sUyPT2swG98}#eZ3=YJq)G
zK=Dm1fSa6ve_)dSzb4H;<|pFw;J4rW!+cl=-|S^GecOPt0{eX_4GfN-`
zn92K>7l%KD&+)TB5Av=TT}=r9-um|8$^L9U=^^F+niMSfTq`KYXGekcp;(0Vl>s>Y
zAOb)C@_b(LTS&C6tx=+_1D5|b7XO}^8(W$m-r48=?&m>%Po!!99Gh7iKj5=Ul7dUP
z=anZGH--=OVt&x+ZmgdRdD^
z;m(gi^i96TK)%g~{1AT6fAl&5#28=VI;<9c^fEuj{%y~K7+YGH!RZ^E0bp=)VR;ZR
z_&@}N2?K<$v)@0WvVH2W0aL0ld=9@&Pc%6A$jK7@q+#Wcpk4sr&Z-5#US%
zh~)=GCCvO0SZ@P}#pB-ljb6(Dh?%j^|84R+x&z_&nT`aMrP_ceXiE#2fb
zR5lThW0hN2d^$A#;nO{e&BmtQcFN!%8vge4^Zoru|LsA^f8X`K^B0PW2;AyETWePX
zc%s>&0WfoqjfG@{%$)5R{`+;A|0ceUd;NQRzoP>O2)GqM4o=A$Q{e*Qr1gWoVt*GWBwt
zd9JBwoX<_D>!v4+Y`=HU1q<@jPyKH6JBgcPA!qUk*)*|s(%3`pO#)aRlN0UW?#eS%
zFWp_JQ{ul;@ztaEcl~dWu^n-29C;zRUiQXMr|Jd3dI;jzfzdHd2)`xk^tb34D)R~{*7
z8!vP$qu02Ur55Yx;=S)WXcl>FmSY=G!|<=}wM@Tv7hvR$m8xIBI<`eljh&3+M}gvT-gUZ+wv52P1s}typXG
zsNrIoLUn1(+llrg>m3vjp+anjeBippBOfvwh1**__O
z{4}IYNHK7Umbf+eZ1#fO;Vn$HcDH{0o?M^kkJs?@J^v{z-p^{p{tE!(h617OVU_j%u
z3exHMrBNs0(d2xpd|w>s0-lHK;`ZZggtDtTZ>$|rl#Uh5jm4As$YJAp)?M_
zR>+{qL$~F=Hyh}u;U&k{Y#gkvszgZ7ZZ*59XcDGE4+YjD-as8iz2fB~eNuf^i_9xe
zqz_&^ue$r%7sjS1`>(JKagVCn`W{E|K8H+=Z+1Af-vvXf#gfdN2p5MadT$){T`QJK
zUt3}}w~BCD4d{{2io0%*r@z?#1oJyh9G(~F&&
zY=U;nR@IoR2b7fIE~;(ySx#?tTo>B5DEJ}!CR@iP?m8H?m1;qqGS*`?Gu&Im&I5
zORS0!Orebh+exT%fnzFB9XITzNt}s3pMf$Wh1*b2kxcqDssI7nIPU?ZckUJM?0U~9
zmN%L&RbiY%3V`=8n?++8f=S*%wQlLq&q0u|Z{dpiAUgJ{7bC_4(bbYeu62OnTZwL
z$A=UgYCGt-)!5MnK{TTsXg@nRK4tOVO>C2~L&iw$?<*qCZfPP3s4o^9j%S~u_pJx1
zQb+TUgr~s*)ecqm%b|k|?`KrKx!L9k0tT<}o3H1XWIw2{Cskc~H1SoI=cA|1qhh+A
ze5umi-{yQ!9&aPgK=9p>4?y6CRX61j-A1{&vjEl
zW$^ptZe^B2w{4$~I-|J~JQ=;d$?}JqKgbuOID&@o2Sjfd49@yx`3X|C1mZco*sZBO
zO?w!fi(MAfyEB2uzZ76)6~ygK`SSg;1X4xy(9
z$hky)&aY@5uR%S|2qO+UctQO7155`-yM!LR%!Wc8ZzoP1VFjlvGYc~-D>6a{@U~;I
z`VAA^-cn7uCtaifw|BZLHi5~FX0+dp*hlDT#s<<9u-B4}n4b)(*%2u|rNN@ZtPRW!IkD_V5l`KpzG-c_bv6_*~02fGub$igoD~XFh6HkH&erWm!
zw{Kl+^FSeY#66?o_pM^Dd9CjDA?0J(P`KSoy9zf4!DEuYG6+7@jXq77P_um$NvL#$
z(TApbu3*Sd^HMEKr4ZKB6+klft_A
zEIJ65p70(-X_^4~HVX>fQaKPgL-v2>KmR849V&em$#cCb5v}hhI1mGAdEJu)e}1rT
z&yb${;}U}8x0(fa{H;()1ntKw-HP6q=$liAy(mu?OGQFTvqEHU=Fd{XK(W&k>d&i<
zHM{v)82ET;De#y!!{h;3lIGPqPd6i1{5PMMLdK(9V+V(e1!-^la`!WuCclU|c8kl~eRBaI6w^n_^6d)8z6->Zh-B2=8~6dhfLltz3&a|PA@
z_{f_jF?^l@Nh6$A?PgQs;M(X`3#gc34)4!zW}A=T4Ky!uSfp}z7=EPLQ^a7(u}jc7
zc0eE~j6$TRaGkDY(c$qD3N7eYm*KT)qBSW?aw=yV4rTmy~r^z%!irMnVfiwql!8nNTBW$sKU!q{syjzK9CG|m;IneW%ZXsITy3T`iv^7vY|Kq+uUuHNu@v
z+u3*rMlnj2oFhyw?vQKXr^@WS6O-+EYhbJ1Q>L1so
zCxtzwfArT2r}AX|x}_++68GV$3EXjI_>Ff4deOge%V}dLXXD8{PTSe6U0~Nn4>|Gc
z0;YJyBa-P=3L{Im8bf^FpiwYP=&Up!{}yb|a%|JhX9nNY=#lyUcGSf-4`^}xLvW|)
zmn6+|F(v*0av;!;bTys3m1pK;XU~v`f?r<6dooXv!i;!ibdEscdi_bdHggl^F<|*m
z*pfU!h`lj8iXfW!QdmmXPOIga
zb$Eyo1e+SDm5jV)t}2j569?D5F0PU9B*-#qpiq7LKwZZ&*nIJaTO5jr3OP3d#XvR$ZIowXmD
zw?Q}JJnMTbS_@1^e6t;3f*v9tbSi-%IoOb6n7p6Ekf2|a$)C3hZMCu=I8VQQL3^H@
zyhUd;hR(ti3XQHiv2ncXQIKo1u@fikpv2odj>L~2$etRKSK$;F=yxC3*+Nu42Mj
zh@>)Exp|;lYxpTq7Q!W!;tp9HB^x$f{@KHx{Sw}bbomI-(#o_+&3NEW)VL{oy59O_
z9{;$y04b6N%B=(kqA87jy9MTLRAAs`Q7n1j=&`8O2Lol^J``cmzP7H?Ea&Vmr9lycjK$#SSiur99wPR!f8=
z{3LTH&3aTuE9!A>p1uj;F9qWQYW|ke_Zu)IR)oi77kC9PfG{)fQnc+8N77r4S8QO_
z*RJ!ht$8#Hg`uou#f8V<5FPTuAihX!(?ARvNRRWsBanW*RTX_UucA@WOh_;^;$EM|&mK_SXgzcl#Dmu#`XUb{$LKU&P4W
zSNsJdfs^KVxuB62YO1lP^STAli2uye#x9YwP5Z2jDt;ydLkXOdJxT4cQZjAoez9@|h-r41?`*t>7C
z?<$baUnmv)pyXZ^x+d$q>Bv6kaYPAyV)v?dZ{~6Q4k|R~4&5igh~xcyjH|?`RF)RqD#Er#R?#4VVv(Rnouk5<|PN^4)rpF=q3TwbTEu$*27+_^R
zSN+D_bu-YmPX4z{;M)!r$iLWYSdBC_Zv%ISj;*+b9AmbE(mpU2gM_GjWm&oM|_VKx{sed
zO4$G2aBdM!(}g8eZUd<`gC03M(&4T+;yO~>$Nwk%c<9c;s5y+AgRW&R7T1d2csSJ7
zQS(r1QBerVXROae#mhdMC6cF^H_JWdsv#yHw5`_PD&tf@nn=NCmihOj&ol@iH&ME<
z3Ai34(`C~LSd0Q0Ex#42>3<-o93>8{PrznI7!sx}coxq7gV-LWiXK+;Yu3W@!WkvkUu>=Kg5p}|v7LmhBq;}|goDvf96Hw>DGnzq2vuhR
zOjsX}i|Oo>)~G>*sI#$#RTY%yyfZk#P412Saf9uYZggz7Bi%I{?qG*=OPm`jz4B^&oiK_!j)6tib{t^q<(ILczoN%Y}>w%
zPt}5$rk-BTYOwGf9I45k?wTE<&}@U)|Ev6GL{hIEaw&?S*TIW(IkP
zHb0)#87ZI(KoWAEZB_XslZQ0g4~mfaKM5!IiD(#?Imk6$ZQM;=Vgt{<99hy=uSKgG
z=L3tc;Q%$1RRTN0lGtXv7aSj4AtRY;@PH)NTRRMxj@wft!($T=To7|wxCVNE+=u6?
z+rm^n`$7FmeP@>c&Y+tS7f^CCTn9uXAV-tM!doxESBIV{l|ZV$~K#nxfvYUJ$|znESNv(8`M`XfLBU3GIpQ
zw~w9}9geV_ASX7~v>z%VKPdc7=W*AoeB*eq
z#Fi}5VlDbuVhD8$`oDQ=8?k;tR_AVS@lG{qe6)53jlqxia!D8iPPqFm+oosM$X$&=
zf2V2QG4qJ}jzd^B#o$gamSjXokyAk%W@hrAjzgSMgwZFeXtvB&A)n6wmf*_}&AL4v
z-7>{FLdXZ#EB2r4HGy7629qk=h@>zc19*0iWQ!Kr3ZeeA84x=XgHxe4&SIj>H%bA1
zMSkrz{=`vLEez%8RTDe2`E0gOx8rMA-I?l^R~BWO$Jd&tdup
zCz-SC3V8~xL8pM4Z!f?D_I!*k;u@i4mt86jP>*6fz#gD;!=(X*wY}o|2i3xIi(v2e
z5EkyUh^X3@uA$gYN%2~&{_o>BJKr8kQLgzLo28?83Kf2SJ3=dmd;4|$diO}PXWZ#o
zWWzD8Vk7;^T0t*C1igHt&-e);dXl0(lAWMAnx2ubhvy|QQou$2@Z{ckVHFHidg3Q6
zDZ2q%ga*XC3{$C$2}O}ZB(fPsxhs@9iP^Bat=g{oYXj`@p70x}s{XreK_%uk-0Nt$
zfdukws6KxFg|*r9s`Oc1!H=nj&diZEj~D&53RQsP
zFwJvd-$u`HHKX{Tg010rFWfjAp3WSWHin!8(T70HM?~oCWWF(E=6kl?RlUovB(cpw
z7FX!BB}Cp*R-kz7-G5HvrJlRmEEPo`NMp{q;Mr{{d-OB5Z$_s&w_%1YJdB1>$hTCL
zrJU2?A7r35+1@IoafHp;zKwaZy^QQ{
zrINl66!`tf!rv|0iC5sUu?D$g?>4cHD$D65f|Wm@wEdLxitojapj3Kx|JD*;1EAQ{
zdHWOEN>(&u;{pWPN!r1*S300&#)0uu6|)rdTqm=w0(am`P6GJtM@uzX_Rh=u#n9^(XPV@;{rTLQIaDCjj;r81jM0FC(A6DQnoIb2a1DsFkXoTyB
zN`yk0Puv2FHpYz|rX=7@4r;g5V~}N|QXB+o>R*Smu>DF0dK%=&DLqv!lOD5gSP+}oO6Ll9<)7O?(RUEPuW@5dG4h(>}iQ2IwnHxw(@7XVlP~dQlkO`XVhgXWpOs?1T
zQy#wpxn}ClmE2kk-u90^PF!9AN-rAfkBs@DIpbozcjS$(_H<*hts9N)s&P(!f=i0E
z-b0>@N4NL`OIdt
ze^4`%h0a~6kG=uNnUMtBN+Y6_0;YaTF6D}yu`S-w*-4(raf_V27{rSgUYPQVoA5N?
zeg1lmn_PTX6VL{ZRi5ZMMbF0wXy$roj4r3&o@o3T|T0si|<5>jUwdoB7m
z9M;NgCI^?{MXqb~0sdVvmg`Nhww!IYBdaM|=_Zmt^B#L$IOstf^IRvm_xsV6`hEb&
zqdlV>=zXi0j9lyh0tG*~GLxccx&!DSx){Rej*4bOQEi+p3(TMOJKcZ(nD1=Mfh6m3
zsb=yA9wc3aA4?>r(Wh9{;UpN289tBW+(!(^x3+a`!O87umtyq{6$Y)xFMWiM-%
zXlGxE@@de@f=h-vZK+e7RdW>^6LQFLw#Y#wR6YVK<}Ts0M8W=KFH_x0o`o@zTI$hd
zCvbb>u;OwaeeVs|OleSFV4u
z%?a=}x&aJwXREJYOsqXN5ZL+rnJ@c2ZREN=h*}sC%Z5HMJbE*c$Vs4)6$lfiYNpMu
zoZgKRiVtnF);$i7(B=!3oT(=p7)H
z@o|GQW5|_^noh6f|4nolS#0Mr>&yM#
zYg+a&zr0>MjlG6BRa6NuJVDG$l{vzf06)lZpl%2K^8Buu8tLxb0XrsQ4MD!{xvf$f
zBr7N5A|xo~92(8Fp79#7@rG@tpTJyd!hYoD(_(lZwT+y(^&*vzn3>@ORl4$qj
zekZEgH(bay?v{Dn%1v^TH_*2u%L@WWQZH{=eB}JrLtEkk?TXE2i|!WVs|6s;UF=2l
z5m~oqG#zGitbX5r1uMs&gbg8!r+r4X^M|9VOx;Sfp2;tV(zGf1qtg1k!eXUHm;0UX
zPixv<1GZJE_a~OSX>xF{-mh~bA%A`NWtj-;y?RF0$?Hjva|(+DCsSvgcC{cP=?
z=V)G51?O_2dxfDJ;B_=lsr(r}S*L_dY9X2Gr>{xbJ9b?uy+~vY@90Peu7*i5*JXiM
zuAbr(5kXk2h|Ub8>!bI(@J0QOY?~wi=xqLk=xMA`#6di1#-G)C;k;B@MKUKwd
za|mtjKS*F~8rvutwY!05;zV~FA_?fMc{(n7$HG$(Wr^vvG~r5PQCETT*MkDCx{2{Q
zf^FG+QkKN5s{!Tc?3nFu^A*Gcy%>r}O)AlED#cL+-COOID7{eq#@?p0B8wV~mbA^L
z(|g7?c+&##*@@#gQVYZT?=0rI#*m$RC7&)L|78}Do;+}!3zrgyRSu5hfOZ;b)%5=R
zUZL>ajjC+>J^Jvp+!T`l`+L&7FIIqV2U;>+-DN9A_W5u}JXg
zF3|J!6xdyFsFQX~Mg^GR675-oy3No&4u`-xw<>SMtpTk{U#aJ`-1C(B=$K4}!Ar;O
zDi@Q0{VLvaSauqp#7tCQA++_WmwVkQq5pF!dj03Yc%*^j@tt=cg8d>A($oL!3@;ck
zw=!+t1vtU{lQhrU1e0HwPUPxdnEwRvm38Kx=9|ed*@!YMR;JM%^8x^hPN*~gVtp1e
zZ=qYJqv+Bbu*2=ytZw4?QTa2X1!;kPj%uDl&AVLErR^RpjqlLX-&5x_x9J>2{pmOP
ziS6s<#GS1+8k`PLCFDyXmbQy@kH7w>ia-KdFJKKvAD*I}!Qi`jrf>IE7|FQ}J=c9<
zn2__39n4?QMZ3^5BzUKD
zJfW$UmO1&>u@tR;+ffImuq~V9I^I+JAq9N@jc`h7HCX692
zF$B_;Dg-{ZEsIChcgxQki6#Dukd;7$`+x8d9P!LDIki7rI~_bPeg~Yz9NT
zx2yl7d`0(q7Mp7AX$%9;ju3cm@QW<@@A<&BKh|#6&v!v&)ssiv{?iTK&-ePB1yc!$
z;m|_yB=jp)c*SR0s$Il#*PvZ&&S%;qEGq(Z(x^1|7M-kYshB?Rn
z&50>_Y(uVw5*S6TOcV{&;$uKnJCOzfEQJ7J|6T4R%VL*Wg!B7$lW3A{@fgl4q(p2V
zhD`RQ_kBj(T&^`hEczsU3HONa$5o8{%1c9S#kN;|irRmJvF&AW-d~iY>jfyV{`ANh
zl=YD^C3=IyEBxK-tDgOPfX%p_W=9~y9PvOnT(%vV`Uo#|_8Uw^VVCxi?tBxs
zQV9ftFjk^`)!jL95B(kX3qt#Vt#ueUSKH0A%J{5A@$mOKB$+o;1WTZ`Q?x>Fd>}?>
zjxMoz5IPM*6u9$UHNE`9xoh`BfJ*;mW^(o~aQ=sm;-M7pbw`INM@IpvTsPmsg(-qbHj&8=>Q`W33T_?vp*h9-lx`$7g*FF*k^ml+kV;Dhc9om9lo
z>3#vz&M2s~o&tol$@Fks6Wntg99Q
z3IS+DdN!QDT$4_>n3*^^uUG$-IAu4rW2LN*C!e6vVxkz?j>Z?gt$#z+n5nR&C-wD;
z#85qL-=1(2}f-kXMWXJ_41%
z+1>5Y{8Y2|)EUR#_dutw!^?zi{f@UKZ6Fn4?-E_XlobnC{cJb-VCoRX(`Y;bu&m51
zPG?qXc@vFTUyE)R}l07F=)>Jas#iKr7?&v+Ab{!Gs(1MzQHBoZ%E^
zx^h#%==g5ZwYixY-hoDuv@7ME0%j8#JEr=Q9E;-Z0v5oop>9K0Yw@hyr_sUEYFjQQ
zb1Dj?D2wEwP}+9>2v
z^BxCcd=-Se)`Xc%?kuCBzMiLNwseeF>J>BH6@TF2yQ6=(F^DW1XM{Y
z;Na>*j6w`>{FTxhn-dP*S7|V&U>OWww}B)YXX&^#d?t9Hn-W2Gpd$rgBpE7Hu)oY?
zRxnkvx}aeQ@%eJIE0Kqhc8k^#D?E
z-9Cx~P9eusn)+S4pK9j&+~NHJI+}=cbZlsq+an*x9a9$Lpj+siT`)CfiJVcGQ2dEFTf^MGdanzfCvQV7gK9T{hmjNa|GhZE3W6ppDgw
zUj|fMZM^eEOlkPbkv$R~=&`Qr)#ypaWF1`1(6mt}KbUid16drM5-m%#PXz42Hm@<&Y
z76Gyi!lM&VD>g6{7;Emr%~sV{J{sbY-G0Yp16lq(x=5Xkwq_q|TR^L$uJ&xie6SFi
z;vrx~k^&J2tFQ;z=5r^09%gl~cseFYrKyyd_=NJpt>_1zlL%1(hAwxa$~;VuB3wVy
zux8S+TN?IB$*{1>g!0W_5U8
z*rXIk`LYImA;ih2aB^G)=a?90()z3Tzqx0PG+n|h+LN-K$YSD2*fH9UK>L|OxTwC!Y7)sgB~g^q0h%MKL;qT050(@vqV(D&-7#992nfeH#!`&tm%kkdr4
zYCO%ESeSaB^icF^d^j+r#!l=!`$1&>6s26?O3|RHGW@bq18Kz$#Nd9m4%MDidr9^O
z?BH*rcNGv#VZ@q1$H{>R+6w4Lv3cmA*~~JwsFXaV>VG)3gvB_6Q@#9a?WQ;!H=5FO
zGgcC}Z=7F{luu6_ptQW5Kemf=G{U;n{%EgM8{OTnF)jBMf~)&Ye$m&}y?fp%A1s?0
zDojDAGE3P+>VLr5p5@ZVpNR*j31yAvuY62whtT{i{~9*r5j#hmt}OwqFCmF9@WG#|
ze1#NL+LMAywTIKC^w_iw8p6go#%^9(A(t{=1pmuYirJ;cS2@V%$EBA#|L(76fP>#V
z9#Kr4m{y2%UYDL*)p0ge?4nntU>Z3;iyO$NpjGvDTSGa+Y_e#921a{5DCsmbcD3eH
z#h%qc3J{>-e5DYI)=l-O_xe=D_{OPi#>JQOI{JhM
zm@EPEv^pmEHN&>n+ClOi&CDER`25{5NQ<!3TUw4XsF9fqlMzPLn6)+!?O%g({
z^4rn%%NS6+x#zbDHL+yZ%8dSu=Mn+_OqY@yqd5z+*7gACyChvbGl*1z>(d6#OcP4l
zS|X}EUJNdqS?kTPFFd7^FN;buqVt{~#<;&_j!vUMM7-D^1S}jnuXeB>=M#lRs>{q2
z{dJ(9z#AfeNp=T~LM@kAndqzZsfAD4lh-!ZyQbY!LyaJ_run6St};x@(=k=)FG9Ys
z5CcN53Z#QPQuE}k9e%xR=!HW0=R8bHDX$GVB)qVbWUN_Dpc7S~ul91!&2kJmb+fpG
zNTAEK2ct_={hDB#j=+YS&aM&9;AWca7^pvyaZ+hv#X5)o0+8D4;)42T#%vuQ8SGf4
zOoWeVSKCBcd&zm^&ueP5yb5=LKdNT}QXf7$K0UBzCdMsxrpUd>mdx
zfws~5&N`m9ntg_tRwyd6l(8Hi{qnnexdyGY65zSCkr=J{vM?-&_Vl*Qygt{ZcXc)4
zrCDaXGhKYM49C3e5~zLKZ84+0UA|&YV^<&aSB3d!BWo
z-)2sse|d&T1;hKrvn9nLUXRvI<BBdHf?B3bUjI&0NM@mPnRObIh*
zL9f%1J751*SQ>zmb|LFZ6Ll-@)g83hct(mL`HXkh9y
z@8wL-DbT)FzDu4nyvWBw`wch-LPGE3Y9UlNFZzLdVcco(F!Dx5t^TBG`BJZ!CX>VBb<6
zu*X1+91s!%c|ws-7OF!ByW7yUOOlqdXP(*%B@cP%0gdhG+#s&H(J7>#thnD1TD$McQ>j9+H^MdGkQN
zaw^v`qT3jApNN<)H*)6#t`l-`(^y~dW3@j;jkOtB#{Mw6h{LefT7G
z!Bw*(YkHToF6Zc#E{pq4#3Wbc?lZ#Zh1M&ie)Zx!x;X|
zCtznJ`vqft+JhP()ZAm6vuV}8v`|AC_meQdJbXgTb(|ZIr0()a*6syk6?J2;e4n9-
za67>zIiHaNMQA`Ry&9$e7!yE-<+}}r{ET;wf7E8WvNomZTN!8MqqDtp#9^+Zhj6k{
z2Eq9cofocL4*W?9#nL=qQmJHX7E9FW)aC9GapevGua$HfiR
z@$FFHSzX%Y0Aw|P_ni~WaH~fI`jn9S#0xen;RM8_ld3g#T<=rmmu4>zUiUnrnIZ>t
zJpN}45(_Wtq^4zh1t$s$a^^hrj*T1v*_+ysELFF-D%wn)6cgq>CKRJ_b+j~}kOt%4
zNl12ULlannx2{&!Jjw~xV=`Iw9EBA%hIo*T(PYhN_R+qD56ISM?`90nZ5N;JXLSLt
zRl&Oq#|#9LB%#Q6*s&VQ&9yl-D>gH3ettxFPZ1UGlN~Q>c5C_x>tA;qEq1g+_4|nu
z?UY_-{LdVu_CxHMigWNhcj8l*X~x|vtnzR;(r|vgd^un+y>&s3D}	`S)}#Nh2Pl
z?k~J@VYMkkq5F!5(&DU4}>*tLZAGx%pY5OHz
zzR|?gnrd?-O~|4PG|bpH<6EXlguj;VKmq)FMrX6s_cE4gswXzruZ1AyG&{K08yrGx
z87h4=t6h&?6Qo3`TGj@6g|mArq7N&(K#A&V^I2n6Kom@y!jRllutNFUEZ>2Aw%!c@
zbVFf_wYm0cc;w^Ay*mxXsTfxKM}xQ7(6T=!X?4`Mk`Zxe(MP6YnUM$MEigxx8Iu@H
zJ_hRLd}@Zj@+ID{ukd*DwpsHSfqP>(+51qe*T$14&%~-$$~Hm{yG24I@oW6bE>Q(L
z9!u1c7VORs&71}?X)Wgg%9G3sZoa^#ICQG`9-z8^VTu}x~uu>+pWVpnc!ajZ_JBLBr
zEq)>---+b3mF^_uL)K%}gZ5(K)ys7$ynX`WywebLj1kli8co%oKA)w;9_h4dr+FR(
zya+r4i)jnXFd<}p8d_X{8=$|
zIoYAl{9bUR`N*625yWFSpG~|IQ$bC3jm^dNO#`zqQHe%ULufm#a1pscNdXebPK#pN
z?x?0?G;Zq+4#?=?q
z*6{h}S`TpibY2d~fQj5qKQD#yz@cr}Yw075&*XWXah~sTL=_?4g~@GQwv%#yO$#^6
zAl_wrcOw%7cwu??BAi;*PW1-iz<9_fvh4g>b+~08u?z4Q1ay0{{TaW%3>RP
z)4!?2@c1UFJ)}`Ch#P$VG3K%rU4TVLScKLnO6V*c9D%EHQu?IE&hDr7ypMHA`)9#XdoFw8Y<8XBspkK4x9={ifUjh(%olEx?bT-VHUNgx!P!y)h6bXO;mh*
zi}q@|`#u3Y!>{NOI(0(XemM}(p_SVq>iP&4q>zJmP5;-LLPKw8{Ptv9YkMKB%z%!EH1+IfSl-q|0ZByo7W=YBRbLx
zsXXD1dnO4?k0_iSq!*OIR1RKPdI{W#5`63|;nwaSGWQ$DN*GfNf8Zb3zN1DfPv
zz=mE=2a)3De6Sv1!Pb$+hgdWddxZa9)7>y9b=>{V;Seq>e4(-Lz1~OAt6lw$qmBRk
zP!p1M>dH;?U_j!CTztW(#Yt_&vWYuW)hub?5t|;-zH{;4;5=2sXLON89h7dv$8p2b
zC|pB0xtV*Z)V90=RuA`vYW(z@>bC1SyJ|QLr*DQ4?pc9yBGZ~0aS8oIVR@>c)#0AX
z8z(m9%|>QqtaVcy(A8dyd(HMrLbPFn0aU`bhARrfz}N<{J^OXU5qBVE!Vz^cBqxiO
z%jl$z%Y3X`f5KJSAh8yi9uA<}a2K+A$4QhtgE?kON+9;y82Pu)XsW$6PmDk=&BiA@
zb1a#FkoKX6Vgl)+ln}KrgpJ;>^oEjnU?Q#Kk8zkRGA3eib=7Qg3^sOrjsTg`<}0FQ
zU^cV0r|*U?)H!4qcv?ElVTFba#2{x>1q`_UcsZbO^X=E=Fz)hEyfA8L1%n8Zxd?2M
zs9wi|`~=`CaPTS1CJyPxd2>XR&H1Ond;J>>r+mne^|uHqA1T7U?&Knu(fyYaYZ!-~
z7K>H_YkZi*-7zbHmPAgQC+2xX?m-|STP{^;S4!2p@ffOGNbusiwyi1ouxs0hg00?d
zgU7i0aO93!@!m)I)j7lX+HN^NtP9}nhj+OK4Przn2FIJA;<4BijBjAhC{D~+5fQh10>|s%
zCbmryDr@`9sQo10ImbUpvZvq)I%9JqMZKLm
zU8lvfySrVby0@BhbK4|~zL1E4geO>E?r%_A;>(Na1C`~t0ZR~0jd2L#-jw;856XZ#
z^I$KJCP`A5B}QEQfY?(*WLc*i)MV%}-bt&*^wZlxvzC!RfjyL>9jz+!RXa564R`8_
zZ;x}<0@HhLycSu?&Kr|iwxD^=9Z{bpQt8E;$^A}1=`1DzfZ{J=##O72TR2EQm~Iyp
zH~ymLrYze=b`jg!vVXwZBxyE+3Tw9PP}O9U8n_M^Ta2_X!K99hij(V-ockG8kEAk1
zdYp0~|0$0f{r<8}INl0A2|l-dBG#a4S|O1ZRoK);y9^MPLR$Kw$F>I7s8
z`x{GtaiDXATsDA;^a}|PZ+Rt{sG>fwMD=I#l?9_SvA4lTa4;ZHD>?X26Pi*?8t1>I
z8|Wm=H;yF&H(O4jb=3LG7tjPQ{oVnGAbBDyJgsO~RZ^e11Ts?#=Dag0e%-`bNg@&9
zthOftHT2)tH8!Yw>94aM{m8`YrxXjnHCijV&Zf46HfF;h9^_8?ko|rmiczhOU6Vp!K*D%pVMJwOk7Z0pxOma2E?Q86(gO-s6wqz=G~-s#sJlKJY6vcY0c
zdIy$5Jo;jUoav929rWYv(wCEtrtJ^ngB;YZ>blPzkhWwJUola9^Yh~Di<(w~t
zdNMS%CwM`;BOB9P)%=_me4GNGNm+L^z+5H%571oN4m@R`89JB2sB({3dsx`Huk$Z2
z%Im3os5*vp438;zGbod4JOvfx=_BY3cR*AR-M!UOo5*~H@QbRRu}K2$GY3Bcn{K8~
z5PH!7nS#IU3XcRvxdBUPP~hYM2qXg3Ao!4rQf7E^
z?Sjn~w1{wfQeop9sYNG?kV&E+eZcffAc$v=3{VkK;>{mT?(6j?OF}J1FncrxLo*JA
zn%~`b>_%sJtVvZ|&#u^2fs#Bg^sj;ZrApNiJyymDQq&``)2TP!93j~zHbc2P^&YIV
zbkZDffA`&egVo3`+R5^iu>!%)_Q~3lr9%l6Cn}KON|gG{jv975HqbRkZrIhF>0cn1Fi#`Xlp~@}t#<(|)P>fSO<{`FtZ(2|Dfj_>iGGp2wHzRaR
zB(OL`YtPhVCZrV*-=p|KpmoC0aAd7Q9+?p-GsQ`;sx_&4A5w@l&S=_VD`jR}*0tbP
zw`&?erGRLJlr+@N5}HWbiR!b+QOE39(5Qmi2JAc^?^r3oeoDNJR}3+z?z~nMFcec2
zo-now`Vx9VpIKco31mkalj}!{Eu)8a#E}omQ7))wO@n*Fc=Iow#RP*(E0Wd`@$A_F`in
zl2`AA!(nTGAU0Vde5@db?HHTLMB=7H!Fr7*rAA@*8BQ^YE+i+rROnLFIT0
z^0h);f}{+hd?XKWw9@LhOnrP9#@+mQ5#A5YV$VM?n;&#@W7>#43yLiHW+8w1sXk$d
zt`H8D{a##EvrazUJ%#V~G0g=oJ2;wDJ@Y)MP2{OqA>W;b*prF;mlMzU!*~=;-lH
z>f6h)&h33aB8Vr_oNq8X*I}$m=(GxLMrBOH}F%FhWQ%KYkHqXk|4ie1#=
zqTfJ|9>#{v^+8}?)fLaC0abYW-1n6UweKB4$g#Fdmp7N4e^6CDR>Dttk7=;zm#o1v
zbNmR_Y&8XRp|0BegPxaCCsNa#H1z|g|4a-ZpzG=PgA)KQg0cE9>D7+EgXqd_#5`$y
zz=M&75kO}(r?`d~Q9=5Z9;&eA7NL1B>%f8@H@d2L%4+}kiPQNQXTz`UWSg`Xw{yWh
zQQ5N_pDUj;+S;>u$(tLPnwrlVub)~JEy9Knz*emEACF
zb#vOVav-_Q9|4TZLknPa|>c#BDv;vAI#5h~s
z;V`s9!uctyr@8}MIq)@W%{r;iiR=Bnd<2_fLEo?Euz2h#%Pdw~_0Qo5fZ8n%{N;z$*yxgkiAdW4TvXDRE{a
zJ{+aCN8R&1lX4JhB8y4!KTR1El^znqa&C}ce=+@GBoPe$$qYU=o;gBp@|LA2CQkmL
zM8v41L`JUs8|zB*C{Us9a1B$pTTWHmQ0nxbwD+dOLhVt}qj|bCvjY!c(|;b?E&z?i
zfIb{R+3Oq;fccqg?fRDV2@t1u6mN)j@IEvC`QJC}AjDzZWvat+o_t*dkWmImyERU9
zcLDVX(O*nKG_ExTE86cb5+tl1%md5xPyhOaNck~6H10+Lbd7=0pf@nx(-co+Lavhg
zhO#02TvFI2CB>^s-%L1}NZh%X8v_Qo5s4geMAY$i(-vT=hRVh1MCiB^Qm@&J+hmyK
za}A@REiP<;y}edVYyJjtl_03`3}TGfkXFDoi;Gn@HO|`&IJe(dF{-RGVn$aTMW@_B
zEXlt0WNVhhg}!>(0{;6Aq3busjt}xs#cDb5GXT@4e%^%c8{*cMhW0FO2M%MGq4VtS
za7eJ9CkFCAxXC{GL*d`yGCumCarx;rc@4^^mv+GK-^Hs2m#IeYygbobU)RhqHl-zq
z1#hhwU|e}%$K6afF*~D)F&c_I=fJV`nqGk=A~W2$
z+I=GEIv);hXG1rHkP|!kAttA14PuHC?{b4GwY36LSXk^5FZO*(xuifOB5je?`2Y%>
zZOz)30do*^+D~&eqmwn%OqU9g8aEgFbK#A`JLQp%LU3>7HG4-~1kyvX=v6i={&}Cf
zl5RBS%qHY9_S!5{nPfxBuJ4^dbC!Kw=}CO#zh6ozH6XaJd%eomhp`XM_UAmW9gGqL
zn!4SEXn1(7@kEG2H42A0bqf7HS3&tWsqzIjXQwMX@+ayeqn{iq{PeKXW2Yl4jH>1Y
z3dp)ZTfX#897&I3nh1MJdC-NxkW}&J$5+kEzt5hp@tB_y%XbOXlv3a6t`d&>RJ355
zI{1m4A)YiTs|&Z|SZ)bK0Gd_P*T`~3&!ca|TbXu8zIVK?+a#gPNUC7%#w2W~upuA!
zPD!I0SvaY-=Q~?|U#!IKr&l#lMJ=e2gXuVp6-o?-Rc9AvJxZ5!bu?0{NG+44yqo78d#3Pi-!_tP~u_?ZWY_F-zE%Rcd4BSH2(6rc{4tC2JCRL~F=
z896&jMcoT9dyY1Cxq`CXFQ?x2?%)=hH`TJ00jZur(Q-r3p6Y7%_Ba!9EXaDvV2hy=
z%xCO|IZGS_;}ymY?*odj1cG#BpZQDACA+rwC|bem5kVn?KP@F@9W
z6Vb{>enO4|qd}`33XDKyqNH@^Gjd<)EWLL^|pJr>{xspW=>M)>FDFYuEip!GIKX8s=IK4diH<1SBL
z8)X#%>E5j}Qp2N>d<=fhJcg253g7>Y2ju*}@PJGVoUH%x0GSAwSQ*$D|MU7kctAF0
zMz;UoJfN`>_C^P5)F1oR|5|eXJt*WHAg>31o7y@%r6_OUhP5L!?O$VW9A~#bvtEC`
z!H7v7=i75wS0%xMvV7g6-C0OP>9u*==o
zfYSk$M12AOqXTwg{~sODNACa80n7hO2P|F4wfg_k0e7NP0I1?{`!LM(jSRs6(gBzJ
zGyt5H?NOEeiXsX+5~x@QJ@u8*Jw4+Nf9(*Jl@(Kqfd8WdIsr&HD>w7^F8jIMeog1~tmVg~+_x#nG-LNxkR1qMhCZfxX*Coas)#13qa
zOk@Tgo6qXsIidHlKueJYV2>+g-TPigGz0#!WT0~?xjl4h0r;ux%as8fQ_ldn&hqn3
zAQAX=*yxFU;62~I`RX0o0pI^BEt&2e_%9vMx9PuhKx)*x31~JMVh|899Cr(3Zvl
zFtud`P~sXNv6&tbU}e3l&fWiunVWX}nUy6NgBSH1TT^{4VB@0mTf6x9|DgjS
zf7=B7ZvIalFxSPm{5UrKJ@(xjMz`xq@CAUG7@K?#U5a_{6S*rjF`gP>@9GTb>)=^#
z)R)N3MAomoIQ+45ZWkW%>J5_~;n?RT5`kSg&#Yv16$di}X=_L$h&%Sq3#xZ}@CB*|<`DGTn2B=*AiB?A;!7wC
zPRi1|
z9bn4pR!GmpME~~zzn$#29-$RIyufov1PZrGtk+&=*HXWHi+1(bl35|1}m*te62_JM#VbM+Y?GMT?t
zCONnz+Uv+0B5pWXtdSLC33%Zc{djm*JIgp<}|9hZj*+#69wLk~lxqmcpH
zP8POu8Si_ZV;x7V&d2g#3%QO>kv@i9*B~oU4Xb2yC!qsP0?~pn{T$&=VL(u4w9QK@
zYDFaI1u<}*oVT<-)kunn6Rh?#s~Z-#6`biU_ms2FDN1fJ_E==n16~paY*jA?i|(jV
zbmL5v
z9B~Fuf2G+8hg>rwB76@|doMMbsUr%*XLsn4txcc~`-B7PXutJN<1_l@8*e_>4vIu3
z_z1OWGv|d47d6%I
z)h=TGbd8VopT!}qeXq4dogU80ua@mcO+k)zBgg%yz+8E#U-C-lketFFp@^VoJI$4!&<+50gxrxV`gG2)Aa&
zUx`A>uj#jmfx%NN8+mGH3pZlbkIAlXLoB7rMqKrxgfAoN&(6cOyqVj~qGO$7zyJiP
zlv4ob*aWa=8kwX-DZ=(Cpso9}F)_sqx86?d*UaR1)Q0W=3L
zw_%45*;L=fdpo!aoUttvIDyNGr)9`tgrmvoDVbL4voUI(J<6u+^b0A7I5$!{
z82Tduorg9QYBsthq>s(e)b%oeqa#F$?ky`0^bo;
zi2)x_xw&kNYI5a{_$+1V6A{eKAZ7<0dxHo%LvMqZLqpwmvO>CaP_^U;nAR++v;>GO
zT0(?xCC5YY&3CPldm`q&Yq!|cfuflupwMzEksu?sG1vNID+
zCF%e&|4da{K^z;I;oYB($Ld~!$2OCT8L;Z8>|+%oRQ;;F%M%h+=GCUIyfq_^ymn&t
z#o-uVH*1nN-Mv844$k8XmgKBC&rSjxTTL4@Eu8UiU{y#&TV3i)z5=l)C0J?81E|Dw
zV|J6NWfi(f8!NTqBa>Hg8#ZB&=S&CT*82%&+
zW~sgN&{*6KV9#xX50_*b^r+zK(~owzWk^ixRyHPV0%ZU1I)953Q-g`3*$K*(b>Bh-
z6=@|Y?&s}pOMh+->wXRuF9?(R0P~AdTv;$8wBOoeH%{Im#=jyc&Jg3dKGV;Zlw*xI
zZOkHd&I)CTL<#ppka-XY`(gcK+!L0v@X74ilgOqYGp0^0GoTJ&>q6uQ3VbNlag*>>
z1}W7XeuHID8&VT!J(8^i2BRq%z`f#=z;W-42QX4!Ea=wN`t(hbrjHsBoe2FYKP^h)
zBR#p!l~lL@!j<}x?q&)2)9duE2L>2Ga(_FQO9{-HQ-k)+;iWHnv~*jIw0WlfRG=WZ
zsG)oYrm_B+MczTV#kSTgKAsdfvw?h9pF4c{X3|Xq2NJ|KM{10-Tq1mN+MDMsO$!8AtHhQuLd9_eVibsVmXZ=Yp`xYkDw$AOTdgK
zxo1;E)ZI2cEz&FmAh5a;P|
z<>NK5re5#^P)EzK7c#irBbr@4LFLh*WoeFV;&oFUX=%yHN`m%d4ejsDYyv`nwJ+!$
zv3m`SzNv2kO@n7PL@R40tyKazvL>8@nvNoD0L})5;A@dbT?}h}&o;CIiO@1tkA8>R
z5R~=8=7mQ2X*Ig9zCM5VeF!YEcL$kyFJINi65c$@`;f+z&RP9OKpORL+)L^S`;sLbE3qX*K0e$237I<&-V*^WSQ(p@}e5C^%klb@E_K`_G6%o@|py0
z9!>553ajJr73>!yLzuG7-}DI`a2ccyo6juq*3kY~`!0Gh5|#sZJ2(pnh*KB`rb_r9
z8eXh4rwd|p0qwH!;_IE
zt5!)hKJLfni0R-rvU)un7wNg`(Hx$aKayDn9;Blsm<=YX^)J|jwNwsGQn#l_Gr3PM
zwfZ*UHu>5KMQ!gLhl2h=ZhOrAsc5~@|D?4edkeK<q&Uq_1_-7QUj-DHFN4j7W*a%1UAhK#+Kv=+7jZe9i+^*HM!JSvUgK)jjW||m
zYQ8!IVF#;dQ>kX{>S)4c##Vbb%B_hmJZTEIr_$jvqw+|Mtn=TTWFGf(As{l5xg09|
zi`Bmu91JfCB^XP$me8nsJfedn6?AAR$$VH|dDAA0x@gFAcvB66oM3s|>s8zPbYRy_
za?Vjy1LongBtt!HMbSx}VdT$ZPz2#qW-9TiZH9McpYCxNspD4RmsRmDjHmm03PDIt
z5W~{Ze9#i6C|0Y!q8LFkg`Hay$`VQ$i^z{!W_E9Y0qNTL*ikivnpSux!$Z#O47%%}
zor&-InR)P(A|S*|T_q>BZ<;a%c=sU4or9_bcDYlI2D&6)i$%L~$;VCiA|m1HAT{P)
zgZ>5f&Q)uvwlLnrq*ewrtz$bOgf|n2QQb`Z44yP?d$$=Rz?vcUi;5DUCoD{SmK>~N
z9qzLEvU!`7W$mXKv(W%so}fa{s~?fdv_mUB{;=(ak*$s-&#eC<$j}$q6jSil6UO5L
z^2qP^hD>%hswnEq?sgx~jlq>~S6;0ITMIxnl$Nnk3eYz7{D~`ze21ilFIeOS@F-je
zDud8Ib8r>ppy$Yu*l13e)%BE-G&(w|UOH@&YziD=bg|gg1oV~hT>L_P-7licXD;yz
z&(s&A$LQp>>@kIWSH_qqv2FxpK|hGs{5`x3d&
zBI!mWu!4N9po>S+kNA!F=T#ljoA-Q;wLU5>c<4*7qY(shGp~#JAr2jPyfn&y=YNb6dN=DWbPa24(37
zBog@g__@spGj0AIPOO=2!a>)kXQ5^x8Q(?YtMzAiU4AoSZe4ce3_$!#sfv?mcP9b>
z$XHPvg8~vJxaQ}YvjMCt&NBhLar<-$Xg4yfXb5*r_mds<6T*eYL#Wn4CBR#i>z#P;
z%g+lgdZ@%JoH?A8W=}3kNRd%p#vTMTb<&`*r`G~T2tNTog(TfhGc6P&J;T-9kP_S@
z-PxGhOGu)G!H1oq^l3f*9Us0mcla5Jg
z*s+MNqk0D5H#j_q0?S)rkqsSISHDa-rdIJCN4$`LTnICaC@O{yVh0Tqs@T9>#a|Z7
zfts#am!sk@W9vs+q{}K_EITgC4~!bMs;-&PW{2Oe8tSoJFCJ5`vlg1T@I4@{_zJCC
zArzo-)=Dt0b_Wz7&S4Z~f7O+)wm5y2<7P$0VSITm)jn9Ld<@&YOftYGIboW4gH(84W0rTkS@ds=kTF|5*bsF`ae`A6;n5S8J_TkTCvp};qxtg3h
zSZ>o+645Vp6|-H!wY;Y-X;YX14gb_X!%7#@RkUc9gO8pMdQ5%)=$^@;Es~nqi}i{n
zmQ)e&{|kCX8k88apaMbZ3+hH!rioRDcbog1CW1VH_I}9P8-LX{DT0&%TIil!Rjf35OkajWTzEn)7^-VB*OAPE
z1j&qfcvna-URL8%uV)ZUBgOAQK(=PK07cfP)r1%EUdu>L`idI(%pDMxFFI}eZ889Uxt$Lt7eO(x
zuI3F&-V%Ho{X{=)44Y83?uU{twBDWl@l#R4*td
z?`Mhn+U6hHgjN@@?jkaQ$$>%(FOGwNkuFqT8nFc!6U{};Mj
zEd8#@1Zki(+dr>VmUv$&&yLpHsvzH*sK`srA_14Nv!WW=PD&fju3FHw}xvI{DIYejw^P88vLGbF5i>IUv
zmV90&9hJ?p$(^B*N$;Io_pHBVTbP}
zydY}F{O$I`=@=wa=c={13BLv!nmmk-n49;O`N`5h@%Lfnojr9;&)!2>oifCaotwld
z8`;Pmq>YubREDI|Mi|Va{mTYd|kEF;IidoUZl
zfE0l|sMkzc5N-J5i%`14+}$b}WzEC^lUXEt)X)E*AFK&YQd<+jcdAf>Ty0t&o##v8
zXX_MEn5IG``LqZb>tFF>Q|zZ+Z4pNI7szyw1L6iPJ^^4$|9fKss1(w*r1Z$86sRY=
zt~aKSTpnb9S9g}_uU=nz)5!hJgEl;&)~pONK<*wP;x;|4K#h-m90-N_G{dluFlCA3VNKq3e|du1oTB{6
zR&m76n;FNkm;#NGD|^7hDG?*QOVP(*{-3jbj9_RE5FFp3Wbif9*?oc1C)|k(kg>Z(
zd;&IHTl`a)kv+#MaV8U(!mgkdVCM%oriO-Er_X*)lei**b;_Ax5WH4~0CbMQ8XX-R
zt(ZX^!5d=pxI4=ekePF!7!#PSt#1ZX#6i{YJ!L|JksHTi#=le{f8)@2mX1+Mi_Tv6
zDLh?xcdlpT#wwERND9wz`4Z*U#e#i^+j8Cq6EYeLHn=}K-p9*#lNPT%r{jqb2Ed#Y
z3L1q=yvff(oUl{bb6JyR4Eb0B_Xto$RO2c#Wz)*GUtY?qGB>TGe}N`y>8KInIr4pJ`AJn0Sh9
zfL=?(+g(*>s9gbNw?#)_u&Xa>e`=PhXsws1dLwW2vd4h=S)c}%t{1r!qbhVI)JD@7
zMle$F0h&wlVrFLTfi^Y{=J|+s|5y|_wWUo(a6giRnP;5FT;>nL_7{-o&Yf~xodan?
z!PxblO!ZjK0Yjb@o;BB1v3n3Rz;giwVOttI;%_=9z>CQD;vS6grzD2+jI1iO5A~-0
zYopnMX050(6-`Fhby0GL`p1oTPn1cP2f1WK=vs8_TN2@)J1e*_o{r#SL
z;@I!*rHXZ|*m;)Q%@Raam~}xF|7ss*Io7EZ8pQFt3G@j5f_k}R)=FQH57ni8=HY_l
zbAm=!6`wwlW6oXULak2qk|>`&s%Tzf;(%f~CT%mYDejS3>pmVGqol1y`
znm?`1L#=I~3DN46QPJG*@4n>Zl3^tXpW2CbY$5*ugX9;bEsGY;
z)=(X6K<)V0A{FL`-Ae*UUwA^YoSm%m_2in=3OOm!;B=jQNM~%B%a)Iu8#jD^x(L+q
z@O*s#f%-!EhxP~lbAWET6ez2#7-7L-E@!e-Y-vgtq|uwb8F7mbI)pTiy8-igk}!e5
zq%i@MrQfX3*J{fK=|9MFXjH9oQlmzXxfM68F+4Q{Eq>Z|JoS
zYtPCCvJRd(sZpZ@MOPNy4pJpyU~74OUJR^C4!$<8b3C>;`Zj1C7k=ew(?&L^CsW_kEsu4IiK{
zDryZft!KImu~8J^6t+eT1rK(N(RKJsdNL(xCQ6PP0D+F19RKy}8nfvDmMF;^v|z
zrPI7)N}+314(xjo_=i}z!{arJF1?F1OcGi10m1gU2<#;bi)huoXnIV!Y~F^*$yBzL?$;C*d^3Hl7_2QTk|8^_0!
zEmjNM<5DA-%bS?S(yR34DAfv9-xPhSOP$k2q6bWBPku(iUTlv25m=?>#Mi%2U;Xlp
zuyAs0{nJP|yxZrFVuA=wAmWZ%hkyW(GJXmFkfaQ~3m7G+B|gYjai)~RnGQV<3?bw#z=MROeOYm!Qi9()K(QD08#BhrWofNT2pQRnuTv_N)%
z)i&*g&NP2`#jkEb6W$RaYK5babbZhimJEP*lxR{3)VHgPYTo7T0mDI-$V>{e(dpAXOp!IbR+wKJ
z=Bk=aMNQJhS@+e=yr;jS`F{a7K*+zvD53?S`QyfwalC>;K2vjnzuGLS6n!p$k|Pf>
zY{O+GYIzqX@dYm1DhxBdt2}fTEU(x7tyEUQ@svtVth6Vgt~M%-L1_t2!wLXaB~w_>
z-Tn23f~oP-hd@ONTZiTH^Mh5=)}N`0qUxXG1(FF_aiEm)p5o?jKm=R-X)~evbeGvf
zA=W4X7oT5+Cbbr_ql9CndDvm4CORb(klOD-fAbY#8HAB
zRZZY+cTMT7=Xv=%j(eTXmrb-fi(`lbm0T}W{Jw>=*zXJW0Bq+7@wqe?8hw@!t(WBf
zxQM#iwXmHeij0`IJZ~LkW&Rtjl7`g5CalOCTT#eFsWkm1{pRg$=oLX}KFYJIxtxl*
z_`s8>?d{AZ#3f8u)FW
zAAy3phAz_)6F$A@7Ahp?kb5jJ)8mM)9dMH^A9A_Lrj*v4PT{-&u7=#
z`!a3%J~64coGqjXL<9^j3n^iB7bx<==%_5kv^Jrl^|`|}Vs}h_j)Lx9UOu??nY{!R
z`e{EJiFFJDBcV5G^IL-2KBj<*SGD5VaIt0)bkCivFH{7X8iIQ#geR><=iC%=b1PAG
ztmhDD>N#Bd6WFqe63=CZLw~}mSh-yhyrv=7*M|YWbEVGt)heoZxeV3B!VJ{H$=fY@
zFxds9eQO$hWkRG{xK2~myxfLC&7>JxRNZL@Ml&HaDQ~pZ!Vw6q-9{9B
z-n$UZ`iH(JfJ3=grdY$+HJPnxEE$^`J~z8=nz$Yhg@Z`bkMmcm@s7>UmB}S4_C(!;e(BS9i>|M9sWr7ePQJRCowR%ZvNG)gVRhKB
z)v5C_hy?ma$xIVYrAw>m$f6)dqrp6UO_kPb<$>;!otK6doQ@CB&7GZPG1j4nj@JC_
zc%6$_X`CLpBKW1f#0N}4`^OORaE!KWz<^3E_<38|%!Wa0gg5S1aifkyfIaM-f+2q>
z$w%f`b`t1DXtp7j2?urgO^qFqh=LIl_Y+^2>d2#ABo+z9e5+FwuB1G*+E1UA%-pgm
z;684y-amJy=n8}B=eQBZ&}`vtd!IK_PAP(Y!%)m|Al^Y3DkXC^?h?_9=9KFv9>p?5
zxhqH$hH9hE4ZZNT_s9qfgQY!gNLlRP8fC`tLJFA1nUZ8C@P_SnbMAQ(I7FEgn5jj?U9L`x{ny7
z0-J2CI(8LCxz8*Nly53uQtP}SbaA~s977Vdus4yaj;d6&
zE4B_^59C+hQkgDSqwaLwES?N+D-h%9yg+tlTqiP`e>xA71Ev|BLk~P}<{a8Xniis$
zQ*%5q_W!V!Z>RdYQT@_wH3l9s{!xyb=&)?nmf~WjrsmMYonr)nP)pbajDeJh%iUN8
z++bgov>3JKKvR__mMa{tD%l?VfTM_OHFO?3U
z`wR6s+Mb=X3=lOVRC_=2NrjiefP^U&;fFkmdoSJpB1Qdvl;_b(
zJ?{FEz01EYq}EFUz3C7&a`1gZme+vC;G}KXu
zt$?f;E|@F$D;Ls$=(%-)9>5@86HlD3=Yf~2XBwQPys9IrBuCzUH0@?CoU@mC4Z4;*e(fwMkQ#sL3DuGO;5ECb5@
zi$M2|ou6TNAQ(L&?@_K6p+~mo3IN%0qJvCce6PbnYVr^t2F9Kqmx5TB&!>B
zx@s$8bGcvdVHy`|8Gj|UV_Yli$OLe5yZIEy7DA41nYE`*5K_BBLB@;$Da`ZOcf-@^
zKF+i}p_{=5g@b(0tpoi)B}p){e)O<%I{p;n2ZE2D?CNdS)bERD32Zs=o7nuxLD^Zzhe1hEF-sW{g)3
zvPL8k@=o`qZtfoa!L@Odx1B37SNTmm^Ca9({NF*`yh3{=EFGjQLyyRp{c(uNwOX=1
zV9KaI`|h>KJ^i0B4Re(9h2)c1<*TnsWMy*p!!(BtxF0S+7@la7E3j{LSUPUXLj2?TJLUTb45DC@{_G>o~;s=!9)$UdyCST`vTp)
z$(YuufVBDvA7Xs=H|Dze>P+SXQx}#T=Y8G6t@Dp5&RUBrD`UU)Nxh#|=cQ_vs-{&5aKj
zhphKOk0e7&vTKnMIm85g$>bOM2GU|`Lg^z;H?IOs)xb|G>hAk;27KRjQg6C77nTIC
zWDav&=uq~F74{YRFE>ItLD`7&Ne_w%h0e44l%S-8*i+dU14n}BZIY-sZPk9ge(1P44QjYqJ9MbevXIPIJ$i7*I
zn~u@K{TiqjJTTx*PtGO(%`2;(&||C5(3ylQ1puZ#dtWWV!NkADSyCqXp*#1Us*bcn
zZ`yfV1e5clxNwuHK3Y
zIH`z_4rL&?!dF3K((*1a^mRS0mUy-;=PjFbCW5lla0t$JD`tP|J%D>5sa7m=6#Bt9
z0%MH0mWQQpm@{k^$Gih|wY76U(a^vR}cz%aX1Wk_6t}92-U7qX``*V@W4v0%t
zu(6iF8n8C2ZqsME_|L)~Zx=Aus1DbxIsy|(=z^&6?Kwm!=F>7{
z!uD7$JH#X}Ivwf9r;|ig>_O2eQM2YwSoSt2lJ$P8%Nr}WWv(p1(b&v$(AM6ab7fYi
z$%3tEMZ1;(yV0{&ofbmiDmVnTv1PfHTpw2S!`C|45k!ltJwGcvnng%%p!1Y(CUPeI
zaH~BE6+1e-{itNes#
zuDw+AKt80RU3oGA%FN0XBp;XIvm*vQ9wdo99tMjsI}ycS1$UqLABx%)_N@(h;jR>E
z&}PQpLd4HIh{%Ox?=dhuax^JSg9ZYa{mm(oTh(;sf>+Aq9(*}pZTKTu?Db?_uZUx-7+X)9P^VIX!NY3R5=$^~4cYFoGl$jfr*~f67Pb%T^RU8$xh}8e8{z!W
z5PNZ2wEo1u)s@AE%~j?^EmJr=@}HGMeaIOi3ryIcVn)o^I*V?;OFW6W9$eDtSAlhp
z=)>36_8k(kZ@Bv0zRHKd62NqVHXC}{2hduWF`+FrEESDfpRQ~QfkNC@oy&nmu4
zGzBe5Y+jZJVlDuj_IU6RGsyGAMb2M!d!kqD+&ADs#t90je#Tt9Y+f^cJA2L^QFzP{
z$b6kxX|jtej@V~1S`fO#lDh&3=rU5H7QVkR+I$phVelM3{E~DOQ`KsGSjQ@v^pZUB
zZ*lq+e=Q1sRj}P&lvrUw+IP*+u^?(%Xf0#^i4pmpQ7l;del&P3^+U1Dd{Rq>m0pmb
z&?^i#yPK)@5ZNQ6)NqkVxW($1X8B{=Vy3uDVyHV@u*?>%PEmpjSo04kI^cu#dPBh<
z<$A{qa%E->f?)7@(72AQU?NSOH1EYb*k->ZCuganVi_l+
zb$Ia(F!~}gCR*o~)EsoX&`UcF(+#^;!z_$8ntAm*d0OD2tKm8ZNRM0kVr4>q1KP0x
zztrke*CFw~Pz-3*e}HbZnONt%axj1TnKzbpl~%IJx|Q4QYaRC3HMtm&o+=tr?cz+D
zriXx`>PdVj2-@djVWYi_t(!hNS+
zm{0Z=7W`vzFs(G}$)|e!r2;dxxVyuhAcQ&pOXm&pfYpNhxI){kwJ#48M@{Y{%E}546e==A3$VnnIv=Xyxu9%^A)s@D*K$>T=jS
zi(Mhjg~`z&gHRy7ge5Lp>U0t}YedpeGRM_ubf-l1KizR_IaH!yMLy^AK~=ouzt)ky
z!U>MZG}tg>9Ozv^wx?tWE<#5<2&C*;=N|}mfr(!F5;-;5V;4fpCouJ(WIcQz^(DlS
zWk^UARQt#qG|Lfy?u3CG@b>!q@Kr?}a)Z3i3xSJwUW-yls;}X>@ua~RS95!zpHe2a
zy9K>X?Fy@oTAG*r7LUj6FhzpiO6FqW_stHm%89Qvk5O(!&ViMi!ib4EDKdDFL2$b%
z;P%0y?zW1AHJ@>^_qdJkmnFkn2f1ROk0a=@ijD>gh7Z_iVr|%Vj#1a@BYEU~#a~NH
zT9R<@tcMC=Mac!m94$V&*@|z6`19jJY`{z%dxJ76r@+OkRC`1YzU>Di>haXK7rw=>
zxCZt;15DrrSo6u>h7en~lv<#T%WdhE$mW-+BD40wUGEk-CaattRZ@%C_TQXi93?)qRHX&_nNdvM01sImOPfH55BHPBTb>j?c|GzS1lG7*+o9!!%!(7B(X*q1<2U
zCPtMsddXVy24gdP(9Eb0P0mr8qp-nbomRf(TN$-FRUXQ}ktLfXA`j0xaAyP^^5Yc-
z3KoW?3fe{orocI%2cKC!Vbu0-t?PaaecT{THukt^rZUh!USGa{GKM}#^RkSFP(i0<
zz1D}AL4`RvuI**LfT}W^5Oesn@mW+_<71l_y1EyS9LDb={3r?n9%m-c3a&;(WOWK7
z3VR~#61>yLqrqP`8MCR^{A35YRjjA0Vq_(7%vZ=RPZifjIM@QP+sr=wmSy5p8eEeI
zbBMIda1}oFVo;soP_Gd`^)ZcO9t3{=Zn7g0&E&saxD7}IxriRdHj|9Lpr?i_lV+{w
zMu(^fovlOQQ|^=1tKuL@%+|4u=v;^}+{jX5S9;G&rm<~)*U)Srerc}$n8nX`s9&>T
z4)skh&uVb8d^!M;QNhdg@_a5KRuKG6h(q*Z9_DouCUz=oQE``{)WjKrWsIZvt(KAQ
zTYla|X+)5RFQi|r)GrG0!}O`Nf|E&b@L8E(3-DS?V7zUSlY~WcV$6;4
zydwGW!#ejCA&H5YPZmK7uK>f9*2awb@8ErEgoM@jL1~(uC@vyz7}SCIWYe^
z8+?qNJ}kHX_t)?IO0`aVNR0h(?X_!KApoBfnag$Pjj17ZImQt(ciNbsX|~_JDamy*r8z0}>FH5?9(y(8kI~Ter?&
z@qghZ15h^^ntVR|6si%ICV|58hBI*0py>|MC+1
zPH<^RW19OII|hZ!;i1G}@MC;vPEXZ9rh@8t8w#(7SFImERbq?;{@Q$T-l
zgZLQ{wM)5{-@a@{?eI$iS}SJn#%o+y0$*+)qVFIg=(u0DsKtN}jvlx7C7YyxvbG8q
z5zajU?MbBL>g&_@q>>#bdk|m2Ap}(bg*=`4!-vU;568ICB_q|5_}@z63A2Rmzo#`E
z_yCT`*1o}2v@ie9QP78{la
zhIoPlVeNT99u;$&8k*^V&D1@
zm**_zwU}MOU+)>P*6{R2Lb_URKc{fdw`BW@Vt_C05Z&FbA^KV{(Cu802j53!54K{0d!|kH-kYlWg6B*{-UHcR)d>e1dn8U+cHR6Q@k!exvw2JsoV1!
zlZU`3RIA@y{Je(Gzc5AmjQD#kW|AGXkx6n2L(6g_-ph>vSW>g-*UH!vH7e&1R2H=v
zu-h6`l8`Yq2SXW_5vo-5)U;TQAYc#T(4foW4%*#xG4fe0y6B;?B{;DZcny;;-^}`z
zL|Jl#DB6|Qah-EkZ_hKZ_&ctA)JQx4_c!Hap64#gmsREyMDU-hQg0%5Y0Yn)O+~&@
z%2vRs8L64gOe*GLjcQj|Uu6cC$^-5Y#ZyJ#a)1f-lI3K>4g>N3X;9*6Hfl
z#bU|KI6Lo~W~YGy!lWC;=%tt?5xdiz;%GW@moS^rPsmOqV7_9w+U)s04Uey-7fL1F
zAuegJhxpeloaY!MOtH%3MEF~1#~HS`*uy!RgcM=F-wZLj1m*<}XBrMXRjygpF2O^%
z8m20$qn))s)FuYJvIzKOU#aK2vVfN!eQx(=F}r~$J%cW~?>{VGJ{wEA5#~hcThb_1
z(-)LC5=zPo%5xu{N+@I5?vIZ0_2bn(tE5Mef9&cHm-3Uy(!-PuAP=Ey?9}kfA5uJ~H^i0n
zdG$nI0T{^nb^w9JYj7F<%j$s#bsEm`-V<4e}9#pK(DG^VtLXn>KFF%YS+r>1a9V31@T`jOi09x`bVreRFrvg_y@W#)iA$dCa6%($tk
z=DP!yR#&6S@E(pPg+x^L1Y&)I838ZGV
zvA2xGI*Y~K>JO$6+{RYSwmJngQOLx%10vCT*J&R+WDy5L
zGS4Aj?)Ui5c=pZdsVfpRSjV2mljG#^z9s4kHe((liiqpvvxis!WzP!d6!Jjt~U-ywWssA
z_GL14dEY7@!pY>FlM{sut23gXVDYCk%FETqthV`Nc|}_-a*FakY1b@+m=ihLwhV!-
z93DeB-$(Je0)I=zUCvg0jAp`$R1){~)~Or$iXof%=oi$zVJwpawAu`@U(Rm}-7is3
zefE%vpS8P<+PPiJY|K|~(afULTTS6mncN?jkQB2rRe^r}lmMoy?^UHGf)!wLkuLmu
zP>)N2m1y~zzZG{pNpgmV66wX5qW`3~7z|p|F+;oom#!i+9_r0~k0cddz$KBO9GXuT
z=6ue=U=p2jZY+B6m&ks8w2qC
z&p8FLD?)7ww@DYCgOAqbmb^)RCV99!;84&b`PJ$XWSjOjZ_ao2%ovgMt5YSGTL-3w
z!l%lD3TnKSYOy^mTQbC#eZ+9?U{gJ)
zly#eZ)15R>ZP%?(S*bANqZt?QLZhT%K|bu3*^wFNXh8nmKJ$~dJ;(6yTf@>3)X`h>
z=4+r;w(O>|;_Ga@>GS~()_&Ce#%f`(NCM*3i|t842;t#x{#c_L$~zM~n28Z49?P^_
zQ;gpZ*?!8d(*@IYy>!xYy}w1(u{u>=#93fT1O+gj1M@-jA!C(2p?1>p|NcubH2;sl
zfP0aei;0wIC@(UBipyh?8^cLBAuc6l=BkDWA|+nXmUr6!*Ty+?*8wPMG`4NqYHTz-
zIkDN;wr$(Cb)qJX(Z;rI8~0A$T5oWF!#CN3y}xwcS}Si4?E$OOrG9pM3mUSyU3KBg
zI-E)|{m3uUPTI>w)Ep*|!LTEjLSZ3io(+>;NR~>mc!zX-RIK(7B?1H)G@E-24t@%jfWq)e5tQ#kI#$|ha+HeuMSDwv-70~`lIe)*#hX3MY
zRye#>$?H|iRDg}N`k=NGEJPGi0>n2Mo@wia+sE;|>
z#R=42W|~}MM1KfgF<+(OBDk_m0&&$1p^V+D=p(}GYY6Y@CPv7}I=#+UBvXZazP&l(
zUp2=*skjxCoy<-;MToGtqPRI5BjvRCj`ggSjV`D!7c(PfRTQ(a#!K)=kozE3kzk$h
zBv?8@UYxtRdXUvHm!~Sc_`u*84(u?jXQo1!UfkGs1!4C5cOD_-V$0{Ycwm=r@*4d)jwB9E+lBPz
zuAI4*coPcti|R!9+5pkL>NE-G(*<9M4;gUNiswNuJa3M1hc(Aoq>6Dt@cU_|USfpD
z6m=6TBh&)AMjVv@nM1kDPdFF~*Gl6@w%9<9val7(V80g}9Fnns7h)AOREojIsZw}il7v9rbe_t_U1?pc=N!17N5vH6
z`otJTjW*7O=Bs`3WoT|jE6dZCjg5wNZ!(9svc8f|`NK%)Hm5>xguZO`CG+wC
zbjRVYGmuf@c*v)atdtlo>eaG$IVXK;8w&gYkv$XH~3JWZwG#xL*Tdf*n
z9g|)7$)uF=p2OoEUap|
zWpOjpM%{8OK6HmPqsFm%J)%8o#I2Q~O~w>=lpox$c1OHQN#2Yv)hO(ABP1cCNb?@x
z8T4&=5*Hrn6OnUZ4eGL&hs5@NU|lDy8_6%-WB43uF~BPDZ`2w1{S`biM`mI9`o8=E`9>`MFdYx)GUW(1VKF9^Ex+;sCvA}K{A|xT!X~Kgp$3miLK3QV54+*{=aP@5975%Qn)LrPmrMi+^
z()R-t>sdwFrX`Tw6Z7Lz{d^s*ff+Wj(Y|}@0hHZF@?fo9^lZY46
z1bic@4fbE(Ece8)<1aLR=f-M}fzBqJID(AB1B@!N;8~kXB(--aQUq%Y`$J})>K%SQ
z*EOf>8wbt61M7p;_uzt?4HIEBZ}7e;JZ$~kH{}~-{i*6h0O9IySt6JLJHl+rA-H?@jIVTLzQe`e?&W>9(oMSVUG0re=_N
zWzO3ZE(3234RvO1xB2y7o(r~RZ}&P#0iV=Paa^dh+qK*hSi7VSpVk-RFkJ+YM0}5O
z^M4JTv_{3u7lO_gR6hAjaM
zMPdV;BEQb;_Tgjnms97vlz-b&Ifma`C)6!Ly9mQXSOs3_p?g#Cwm}qSWv8`ulrm>h
zhM9d#%K1Mcp$dUxMBXJ7{Nx*f#0)Qhj-=Tp(ZcJAAWCvzl9rqN8!00YM>Ij#23`>C
zOYc%X8d#Ggg6iXgFrzLQG9fD4F%TP`qvo(L)pkn%4}5#%pkDMBWGPe9w~SD`Cj$cw
zxw5Dxwu3oM5(`NS*EJTrd{u&T2oXjZL%r!a9f+*#oNFRpS*B~ia>`pFcf*9j&JwhZ
zB4$xkIij&L6{?N{pU0EhjDPYYj?929M3b>-^LonXf4b0NOebqw`@Nl6c9Ja<}h?9J-^06HQX9^}*vBVSQm
z@g$X_O)rPl=yyH6K>cW$-UXh+TinFB`!c2=vz43$h5i28?>j^W$t1!KtCttXe6d3Z
zu6N-%C^IxkV`?>9@ESBu|ZGjsu-7pf{wpn=cM6;oC1!qX^{H
zk)IB+UoBF|2@Kgxkm%K5f8!(AyKp8FhEt!aHdO|DPShC|Pv-x~OjzvDlGmE6vMHJR
zU$y72ZSXrM*D^K%RiHo@U?w$L(>nUt(KF(~gl2ALLE-P%Ww@p}MyYupLN=5Ecfyc4
z@kvIy^IO2U_1NA|Lh!C|r;e2i?V(&aj@_1ImZXd{E2WvLBH|d}+4h?S#|Mz_`)yyk
z;iL^tB0LmW`gPyCD!7spdJDiS9#Oqk*Yh*PGGV4fosrlK&g_g?<7eVg4zJzzZjk+s<{SNR%GK(LJ#1T@spm;EXMhpuxmDGKL4{
zohW1%@D5p>-FT-ff|fZ0zT{K0Qk&1Ux`4XSWN!8M9_mQrNqZzU77vc^{<$#lF1D(2
ziu7yfQ?oIOw2Q^9M_(V;I68ZC3uMq)?Lv&->hH_CZ&dHK_Af2W2KZI_H~br5Q#+ftWj`esF5GGU0}YA}_Pgx}xT_w^
zLIcIMO@1Os0Musek*_s9OY$!3tg4ox7>OB&5lo?PCa-PfX*D}n_)s2rYQT(nFXa@s
z_0*=wYo);|Q~cAq5tz}p+W6_*^{`$6P9oW?61cNy*yixC$5pZ5Z!#X>i?_rgS@?q}
z$U;3rs?>uqOuoEI?SFkGtaUOk<5p{+jj^7BUfGnl%)=V{Zj|^mf)dyR@8h0XEgYnA
z$sZF2^Gp;4%~f%)PPQJ8#So)xT8@+}`>=c6(v4lj
z)KJ_2H=)OBfM|E@$JJh|+?h9%f|jGUW##f4_cRD2%721L1xPg;mdnjT`>*|2&wEFT
zD$Bs~-^ORwA4-;b5&*^9<|}i!5#%@Gf^~t!>w3wvZ&^2gWirvl)#ZNCH2V0BHy!X$
z*BA7zDkvK7a#Vl}S$Y9}PJ=CXnY@%=!dDYbWE_MV#z%=x5$l$Kz$3MycHc<{ncE0L
zeHv7@Ycb|DJmQ$wD-YmXk;AX}1q%Yv*{{V=XysnmZyBKF+n^8Yyq4p%m{hr>Brl0@
zCzFzq2zP-LU5(`IF}`7op40h!R{>ZEo|iJxg^h&EDL>S6Hv$%`RTx|TbW
zb}w#K*c>RKU;g|rifKiSSk(NxBc}LDae>OV5L=H!$!qtG0Phr%y_^8Vsj;UBevrkb
zZqdX?-&L^$MoZQJ7WG7FQZTPfx-!$R#o{-QBdfpk11&8aHozae~NmnlX^D)-Zl
zk9+-jX{S$O(k>}B)`A0G#%~9s^fIVYRsdV@0=1#~@)yf(qAkCC?X(#}H>SqD&F|OP
zs5M!BYbsW|yx$&wZLs^H!s9v^=(Ko_=)+$yw7;T4=dZy}BC!%?h4n79zq9EL4B#s;
z@P?2OKxsxT^zBzg@~@1P`s;twUzGullt!jZ9%^~ov(;v(&od@Ldb
z-g`g9o`9wS^7ynC9I5_WL!M%N=5`6xe??puCz*`gKspsVXBEb$ie#esB{pRhJ*{3uY(*d%2sVV{C0Xg;$S&epdyRkynZnC`miOD
z9?1axR-3ot-|13e-XEgyCy-n*bAr08E6^dC5Mu!;su&?7v<3nsOr+Ya=Si|H($u{-
zX!AFZut<_T25E3}A(qR&Lau!7?q2N3oz0udikG8%KL$6|M_5a&6MObG2(Xi2tt=9T
zSLwHg9jRZKt{T4IjmB2@M4a~@!4Z0*^ti&ygSX#9AbYNrafp2c3U5x*%&r&tGif2h
zIs^}hnlfIPYgV2qrB70)BL$ayHwa?5cJ%0Y(?!Iv=?J*rRyG`_mxRDn%bdaqXq$@~
zKEEe8=-Pg<0t?s|Ka^O4Rk!WcNsQ{SBR;mg^K%|l>9~&nx1f&Q76jCahx?BRWm_z
zD?;^UZ$Ah=vV2E%C6lVO1>jX3xL|>NM>?tJU0qo&9h$iEtf04
z9Pw%Zm%~UA-7gLAgWhI_Q
zD+@)Jj9eJV|0d>aK;bjFUPolG>$#TMI1&+&v0BUX_&f+
zJuUb{dGmG6s@kHeQ|9H=WO0;IJ!>ZRIE0ays)$2QfH4(xA^Onk(1y*u&#Xr&hF|!^STZTroPS6jkq_MQ&N*~M!nA1FL}9o$@h1gFe_>w~;aiMwh)T01
z1SA{D3cD*D(Q5^j6`Lb(7QXu||M2^e0F=A_bTkPZ1)K|>9MXrfStAHQX=du~r6j0-
z3SXX@?@pTf<2@|5e~lUN(7>+O5n~h*HP(LcG;pKlgWYXgRNtjFIcS(?lM$DaR?2-B
z=<@R+IfiyA+9P{5tXna(xo{s}pROpKlvi18nV?Nk$lHRwJUrp^Hyn}
zG!qqUPgVH^Z_KUef*3Wq#tyLJ?n9xd%Imk$wEL`m{X6dmIBz7x(`h+D=qgv$q_&~y
zKfcj&S`^WsXjS?;_`dVCp29XSemFv4Kf&@`Y}!WbxhrT7Tz^T*;!uc3@PYA|zvBxr
zLcM)Vw2Y_F9#%eXhUrxy%iuOCiGHtwJ!Vob6P&~?$0Xie7JNqD>JZkuk6e9t^|MLH
zNG;tA78X97X5&jX-?g549FONu>OQ}Z+c+0H`0Bo6+3WI}vISKU7p+>(kKJV)U!b8!
zbibyfXf##h3AQBep@wHWt&Y8NoQQg0?!}qQp&8YWy{bo6S_j^SdIwI3dt?fr_UQ(J)!TIiSVp96MIj&3vJDWX2U5HSoGdkF#PMWmBrI0>l<|9k%u&=>jkvH!-$Owx`a1FmRki
zh!=0+BGusN_M^!?c~R6c;#Azc$6{e=s*?BzB58C9gmq*ptk`C9izSrz=rP*qrp;QHcI^;$-wuPF
zLBRbv8!FwOIo3Dd6%Uu!@`hQa^m7m053Hi+ZOICaNkFhtD8tMyUpOA$-1(O6Sonit
zgQOk&iFI&_GjFnI#=(3mhWZZrjL_-YKU!Pf(c~ufEvM6h49PNSx9tjAJcJn(B~^nM
zlAC%)W1C7#@6#A3e8%@xPRAk{xytG%qJ7;rBR_bTd_fdNBw7Ecwd2lI^658^!|>Lg
z?emf-8zmd+CDf?~sCjsMB3tiCw>~v=XUvH}^yC?+%9
z==3J~SImikFz8~SDz({sFbF*M%^D#u$T?}#$pLcSH2Jc
z?%%i_&StH{#C$nt_9P;W*?*wq%N`h23FsG8MMC3S^M_6
z)2UIuB2h+8eV<_N6-Lc(xFu$SB86JI@Rx0c%5`?YgJrKc(J=ejLCYkS%edDp5qkS@
zNMxVnjO^)U`A=wq7kkJ#rc{L|^Y5@N$8qTRSu?u1cqOTx%QU_eYBFt4nFwK;&3`m7
zpAsjY`N|}?q7BFYhQ#JXWMhk!LObqX?u=!E$HFlBofDC;8FiP0*|N@XGh}GQ?U+=B
z`k;`^8l;bX6A3!|ABFZko@=%qL7s{gz7=iA&4CAf0k^N*4C
zZlDDueZl>TD9E1SXKp2MHsmv4uq-WS2y7$%xJREE{>47|*BzDD5QlrdwGNRq=-Lx)(oFj<2HWrJbvreQmJKD|G;8PP@+77QnJWs
zF~OJl&}o0h&5g&4L0CHZ7cICI{!1If6OO8Q3oGnrP;uwxiYkbTiS%#v>d~eb3%ypu
zrIN1ItOVr%s>i;y%90LgwbyW`Id}sPul(_ZN=cr#*8g!+3^vt6Ip)yq#|O&j);c9*
z+y$z3R58?0Q%5_rMHe)w8D=Yy{3{wcg+OiVI3eCXZHxbx!e^GEpjQkfkb|Nt$248$
zs&8b|>&iSik^$YDS}Hqy9)18*430%1Lu=y?jG}Tw3Nij{dgBO&BIf3-c&~+^@FGs)
zn4cuCjyWuFD|u!FmvhnEFNPkx%cO4k*LV|PZ`{-1{6aTYb7}RtPDSk8{1KPYRSz&|{^p!8dJph1)r(MO%zBTrE*7Sl
zuHHnpF*D0tog3N_tk)2z7Z?#B+~FU7o&QW(LV%2Qp_@V07P}UKYs~;XBK>nLc6?TQ
z!ima`U#CG2C=jBpv3Al>hB|l*z>;-*vPL1YC27!~U6_pBQ;aCl+9u$(ZM%E7ZQHhO+qP}n
zw%xtkwr$%z`^?O_nB>2hn@TEG7b{iCmsP9Q_dc~h$--&I3{p4yVzr#MB=h@GA_5OH
zw{(UD_uM
zFgDuf00QL|*)cy^6zbD)PGf;ik;1XmHJ^)
zfzXEmmr2@@`LJJ(9=LJdX_AH*vdzQEppQ6~L=89kN|5pZuv*gvMntvJA*>0F!a2;Y
zwVUmPUb+7MIOM}f+{pocZet}ox)67Gr9JV?MbrXhXOsBw#-0y8s|u$?*o^K1hL$>-
z*wS!Ne;zi{Y}QLyXb{d9AWF=xGMt`^PhiY>S9pU8^y3wGUa77c5M$xRaPItDX)7FT
ziZdm(rv|gv1{!SO0>#{{r^XA2^~&v8C=72&(k%)4WFwd6Gk&A4IL0SCbxkE@8|$XP
z`wk?ooKY1;dCNdJK2vg>t46!WjZ`|x$(q^~<(jh*5gu%}eV}lrKfvUhEGp3=9fzkL
zZb`Li^ik5xTk*j~va3n{A(fB4XdTR&V4yfcyPQk*Ly?foz}f?A*OlK|jJoOihi=ah
zJfA;sk51B+tcWjmf-=F{XG4DAjxuZ=@*B-}Uh+G*SlEy`ivyJPmgRE{+s!@kbgnd=Z
z!c38w`CKu5NM}+u2MFylpLSr}TOAt_*7f>|O0S&+3?GEWf`VnT8uQV~zbGKlUaXYD
zoT`6s^uT%b6{DNP0zSZM3T9ABn68sz9nrmg?vt^fX-7uS=t++*v|B<02l54SHtiW6
zthsOPGY%6?TbyYT0l@|;4N2uXE?44oBg+GrNJr`hqqJy&_RJ{kkF|11GUbeKVPF$u
zw&E8o54fcjhiN6;5XbfdYIRAQd6($e27zoR=;}48+j8bfv2g1j6?iW@-;u{*xyjI|
zuV>T}{4)2|;KtS5$+gEbbLa3CEZTH;@uP|6Fvq@2xu(3`8Gd9Im_@`kYd;XH7IrKe
z>DQsK#UCWzwB`yK}mA
zyK84mVs}#X=(D0tMGOgA(TEiko)H8ozc8>95)hJ*flg4Df1z(?Jv)>S)okCw;tE#J
z$jH<*q$D5C0UV=KjdL{;qvPKVK#MNrpI9;Jb>RiV#01X`$`54@@9gWA2`J4C2*rHNXGR4rmBz=ihJo>;{ENv5Sv~%`W&M@XS=#@ed9v>DiP`@38i~Hv5zzfjpZ42ntq9^7
zgE|8GP!l2O-qVb!e_M5sjx_fqY
zwg*Ya-VF5bRXyo{h3;k>TZ4tLcW?#$@cdo8&x6Pffz~%Qxp<&`dYuOPRdd1gasmgK
z?VGxl{#pGGqVr=_w9tL4$OP`CCFhShhm$KYKX9yru=M+T)%1HodTwb|Wo`{#{BxV|
z5oL2KDpX`RGVe}P$up0o=mwrTVw^@E-+;^=0c{)2*nP2}o{JYfDSM-Ye>c$NH=!M=|28vtfPdGMeayGVMOEe3??%-657Lpo`U$1J^|}59-2LSI0`7fv
z`2zR0JNklu+ktq3f8#Ys^Xq?gX#WB3U9|tiZtP`y_=5f^vrhiS$@0sAnDV6XR2>%
z0>n`FPzR*G!nyLZ&E0(EbCmgYY5sLd?uP#HyU%k%1px5`klFW}=1PAEQ7>Z&OZk*Q
z`Kk6V-80~>7(`o2mGS>bd2dPrP~*0Xf+1H%S;LAK(KsTYo36S|Lo(MZltBzM`RrUt
z)#mXqDkbW+xhmTe^QR_cOsyI(M@%lM$B>_w5p^QZ(#ltw;LTaios)%`!{9+6N>H
zWjsv6=q`v@zR6P0e^Q2_B729Zi@dRPfDFIc5*J(l;V7RuJnpzFkO|9S7n+ltW~rJk
z5+)xN1`x=l2UP>I=cM}6VEvs0^eT+@U2{1!D)l|^hbxESZIxy%tW;05USWKAbd9dI
zowo?ZnYI^A
zM!~i^v5ru+n_!`}XxvizHq{K57ruqR*zEe#1~n4`hf|j(ymCUZ0%?mlQJQ_TrHp>`
zZPNFFpA^vFUJ-Y=BpQp%>SQ8j`T+0EEdXQvRr~C0fqE
zY55QZ;c3Bpxtoqw#ly5&1E493WV!We_{xOSmc#eFkJ_%M^$nR`xAAx2`@96lZQQNN
z&YM5;;^-}(J%CV!{Ysg@WMa$vVJ&*Qlw-v)bcBPji8+PC2Izdf%F#YKE=-Eg{WDz4
zt~%Q2)ecBUzIfX5!Z}LQ-O!ljx0YZmC4P|`P=6d^1i}|Jk}-!~J-_o6mrrg>kP*)F
zqmoj+AX)Av;XeQ#*d6eAJ8k(`8u*S{KnCo%^t7DD%1p00fQ9oWxRDg-5G2Z0{l;kDkGFi^dNj!3f7@2OQRI_^D6i$OcgyF`o5KRd7bBV>Mx!iHEG*?YW=
zA4g5XHehLV{}`*5XP4#Kr16z9TL|;Kv;bn
zy#G*?lMDt;1YOGg@82;hcKk(kQO953_3~Bgm<=bd^E`30bJsdGIbxIT=Uh<=Hq3_%
zO#ue$I)9Cxva!;L{$%2*Oj7_4OVyYTq*WeWzd(2Z#45kiuEQy@qa?
zTKU&%X$42x)U&BD5p@{dC>g1|fLV>E18&9VRU1(>E6j3(Sq26{I2cljMI(fOyLmwg5b
zVpLG{Rs|flE1b6KuBzRv*ACfXWk`ynwZtbkw8HEaO?pVlI6&ZPU}#mHv$gFOi8|u}
zXA!Dk2JS(fb4zsG&hW5-8CNzYH5)!u#A&!#c#3#f~TZ^4iWH$2@Om@*2;2q85$Aso;i-fccGf+5=C
zp}uMaA0EIK-6&(;T)n}8ykv(RApzi;)>p0D^m-y>5gEsZTQn5zdie>^zlv8n9V(;j
zM6KrRd3iEXK?CN6^LCqAC~nq^OCYQfvu}=V06;=Ro&GP^9@;bCA#`5OM(Fp5KtZ3o
zmN4Yb`-ZS(ha@G4@xI5Bo}Eg-8z`Z>STHSo&+o6X%XFT7{0Ds<1$9`U8^+oMglNk@
zSPYdUz2+q;h={6|-HH6Uip7JBI5daNRj}Y;MdBH#qL=Xj$aPy$1)yh?1h0#LdVk!8$#$Oic9E=fBf9pDw5~qxMbiCSt|%#lLC-Zh3~FsSq?7
zhF#D_<+Zl{VT~fod+X;b-Q8MhO+PfVc0-VM08oO2Xpf$0Zf!i&Fv0nnkUk0VWfp3|7_@V{MQSgC&ciOH(oaGq5^sXEjl`9Rf(n@RUvLHWi?QijuC4h4%8Ye4b#
z#^9`9sB)0&?i~V9#hXXdd(v0Jw}Tt&8dZo0=U^z-8y6%sW6qGAG`KeZY8S{fh%2K&
zL?;{712tk8E*&piU&!lT8lT#Ri!Pv`a4Br`;SIe=c?Qux48*KQ1Wl`s0wLF#sOv;Zla$*?{A2D(tubFmQ
ztvT71Cwv9LkN|N1v3V*)Rcyp{*mE#S$4^15fd&Ax+pik~UVmAO&Hhk(fy76=<7^M{
zN4B(=H(|~~OUZ-z{?kvoPj`CrkNwCm!!#RmQWaZLcph3r+S6gz3M$d_20Ot6P9HyC
zqoEn>6ua(Pzn^wvr4#4%n-}Z^3`#+m;XefRdt2|f
zTXKbOXs%Wq$Js3^cQQ-Gq#lqy1Udh>9Bd9av}VKrU7UJy9vy1AH5K>d9Ei>~U4AT7)u6>Q%%y(h^Bw;Jy-!I#N0}Rw8~ojMGUunXn=%bhjURc0lpdLm`HEM19{}C
zxF(@?nOyHPz)UTQWC>;iAglStE{(IQw5v$rY5{UiAzfD85h}N4pGyXVF5`Jd7tvkl
zp`1EALd6d82+8-SX;c4t?uxnL%(uR@v|ALk2m=iU6hH%$(F%i^(W@K7>|yo%yjv~}
z?NNhNTeM6!ADU;Fhp8MpZ3{T(p1NL7h@lsHMO3`?jb`T5p?S+#f{@~7ORb`ECtJ(q
zNP-9btT-Kph*my<&9S9Aun224hmJAOg$xBS&iv)LhHE3HXcJMhK>~e5JEw3$!!?gS
zb3{vM>Z$L9vh`Q|$j<}A$Ncjrd;JtyZWiUmGgS2>-CqA6#{_U15i(m3yRRjsh8=HS
z8#%3g)yw>xodHN8Aa+4N&yW=g9z~FUJwO!AyJ6Wtf1Xay{dIvfFA77W=s^JL^~Fl?
z>15tRbS*A8)Au%pA3bAV*k-ehbc%r9`x;j@o^
zdV}IrV>rt|YFI9(kP@Db>aCIS3Em6QWsJ}Ah?gNiAOu^9%K9~7#Zy&lIirT1xn)<1A&
zF`3t+DEWYxINu?{Ou4OE_BjK~v@cV6UnNAq2BA85dWqIT;D}7-rMftgj_y5#7(!qQ
zD&*eL+*F_G-hC#1`7eoCgvh!l>U{cBAB`HJ9reMqo=YcLutYm|SW&-c6nF-~cZuTc
zbp5?57OZp0{2c)#2{JlRP5VuqPs)nw5*&pPc
z1Z=?nQU*qM6
zOTDT*<#(4RMEKp9B4>S2eUja{r=2&Ypag+`S0!P&nQ)Ccdl|y=A!cl}Xi9TauluP0
zoO2B-3ptxBd10lw2-p0e6RQ2`9Z%_w>|HX5lY6D#&P%M!Id6O5AazFQC&ICp655=q
z!^>`{&)>{3uXbQFmU3D2QV^)JCs%2}R=`68!{|t;WZjF1IXFzbuOyr6M2tYfJsCyo
z@Qrgv)03!9QmbA(PDm=53)K;kngSMGYy8IKZ2)&Z=~n*(k`l}Yb+*Tfw}@2w)W2W`
z#xG8xa80rnUIUtqB(QZYJ+xv7XYCkIC5+pvvrFcx&mOLQz&8>dOyg~O73hUhSga&Z
zTh`!7kE>RUe=%iDNrVyR`~G^yf~%SRm#J0^pr#Jn!n*YyLOPMepFbuzm(Tv9S$cn`N2_U
zg$_+vi03V*xjdkU2i4F2`qyf{oLyIvm>RtEE;EQY?ks4U^0w@hLy&fEl4XAnWDAM$v
zZP1IV*TnwPrG)KO#K^PQ%34#WHkD^jE*@WaI;ac>yy*%Nh0A0bRV*!3i4X)Uz4zWh
z0DE5=Z7nIZ_~`-FFyb`_Drt1u09O8uraF?z^EoB%CGwnD{u!eyt~*co#Zxy&*JQD5
zSDBNw&2su&?ZooN0SkbB%?WSGi|BQKlgx^*$-L>u{GHn%I&iS3tr`A-tJwnRCp&iL
zE6KV@lS0YcdSA+dp3L}1WQ>>W6`?;#bS(%$T=`2W5IS27qFO?~2suGZ8;kcs@ZLg4
z@7#%?o_0yKkARS!0yJZoj$Sn+8LZlnD~uESm&P?seQHt&FB2m$964z%9$;PIBcVfh
z{_ck1&eTmuBc8h0Npmes!*axF`9dE-sCIJdgqc8JRoPmP`-$VMC?Y$wIIVA-d7SMr
ze1b&Qz*ERbwXaN0-U?Gj)+n`%aT&-D<;$oSs1oK%BBm*5~qR70y{aI;QCv-Dc%5p0rR!
z5m2v(js{FnRF+FKYE1KlHvo=`nrMIAfV@uv{pE7JtGIuFI*U8;7tqz9B7*;neh
zfS(k9!nBOxo112zHdn=fgmBv!ZYWxU!uNvgUp!EBcKL*2+NwRVt`xG%+u{WvuoCrW
zeUX_gGCM4w>3E^qB@f0o>tOOa>l01ISRJ;0eXKhQo{EKDS)*gvM_P-Wt2|%H(1_g~
z{C#Ron64ntChM2s?!HV&ybd!(FKRV+nrk%GsODUI6`_$YKXd>v>k*B63P6h)gRDXn
z5?E9MIZ!zkz)tLh6miI+B2_F;#Z5B*Y0>P(=k>Z!9>>9R$D3YC$=z$2hx`*3M=tpz
zXy;YKfn82&EM=iPzFFc04aE1Xx_lh6J#nlOc9?{N$7LN4Vq^m3XgP5Xs4@xVBK)eP
zeVTC&P+~$r!7noEGQkFimjiWc{QBNCEhQI&HnJRiKxKhKGHbYRPM2y`^5N-bh|S98
zd(mUOYu1kY8QDlwvc9nknX}XRr!19W%PB(VHs4;eanl%#+yIX$QmTdtd3)wF_6HfMP$y)!CsWZHsy_V0YrVpe3$PunyJ_oW@uft}(k3b%0$y62z&4N_N
zF0u&+sUQsJ>jCV-GpKRm<}o^kfz#C^26$Fvd0hMl8&3g4%Lr3dOBO+S5p}6hl?W%;
zA%*v;8@vS|WD=l4=Z6!U=s;y9V{ZjW8nbu3A;<2$THHUAp1;ZK|t}X|2{+qI_Id
zG#6HzG(uonEt_8=i>G~hy${Z9Qa#RG5J|1KD0bkW8UX`gQ%8LRAhZgoW?tOww8(`K
zxu0PA7{NM=vBk=Du=-d*$%dV#GjW+E7)NliQ=({J{FIMY
zk>OIW(^x+tB%g{m`R!}Hz}#%heQ<_LtvUju093j{yoJY+%$FZol5EjE@|ML08n1Jq
z(AWTf4_GXiDQdMLoTUu0HZRiRoPJ=BRvf_YCFr*Er!}6dbH;$51WyjqqYhVL%LT^gZ}AIliJM%>WAIwUwOqW`_4q+*vMgP$Q0;ruNLUTE-uz~FE-4-_6`5Cf
zqoqWJjFmn-ErB#b&2~DLX1y5^A}y2kg?_RtomwvK||vX_bod@QKdiS4^xz7dx1
zn)s&OePHJ1EiL>AY(GMY*FgrhrwO3ml!ZxOj5!npb`k+e%ty{baN&Yuv_T1Coy8OSSh}W|76c=*JNHb9WwD{f`|1}*u%W!Kl-QDY30N-9iTTBsN7VLFSb?1%
zOKhOa@!sWsu(kh1_*}x@g{9$U@-3<7Ap$(mZFf4rXvO$?*qo2Ze&xrNC`I^DnI;o5
z)kj_}^l|j`A41jq^6POM{e&VB0=+i8&pwC1vNj9EC@;
z;L%R6^oC$pITb+(CEYBUMpY=mN}ALu~)oSQfaWY
z(%S-*p1mY$24zZsWAmwaK02R_tpDC;D6Lz3%*S$A_vtQ2A6Y}8w&nRI6LQ>Nyf3g{
z5(`g`b~PBvAdlkgyaiX`Hd(}xbah^}AKFTGS6{KEt#$;hPI(Ku(!(4Ofe{jLMqj>kYhGMjWe
zvBRiaO+^|LB^Ax|>I#F2D>$7yO5fd=U_RrB1#SyQDxJ*Nql1ltwIApR0X$9^-zvRX
z)810BgNKgH;TN!692=gd_j-kpC@!^V*tS)rRyJBAPv-KFZRSA9fAi_cawXqt*SncM
z_$@U?0e#A7<~|oA;ACnQBrECmS1f!FnUTH_M*taZ6L@~}6%TB(bYjk@eYHXYvFb_5
zhc&xqx;BX8Bp5x^Wna!kaz($Tg)dHf`IGxzJlX8So*z$9ap6i{suzLDV}dpnwLm{5
zjcT6R>@Rw8ezj>BZEb|FxQqvB5Wv$n717XO-8t%Eg6x-fNZ$ttL&L(7$p?UXAew-m
zEfymAsOyhvUqwcaUxtk=y+y%ca#xrwmr;x+hs4cUMpbqM@zzI9DVC2Ftbd20QSw25
z*t^<;fOpFKM61oy2HrVAT!T8K?b6gUd}&N%dekhGMff8+`Q1Kg(p!b?ch4!T_YSjs
z7wA20`83lZNr0c)wXACb+75#)d<7TtDM~^_Yv3I!T`v`DpI&8>Qzsw)bhBj~gzLuw
zIhUVWXPQzrLBSZ&Qk(Jq!@{pBi2SAK-@moIYS7uqD*BtT
znhA2+liPG_Dwd7lC3rrt+iEDKZhao>4TgK)dR6r65wnv!owR3cJ0Y>*6&y>0tu$2+aFhTt1KK$EwZ
z#v>7wa}6|CjSMlST1ic#K$8~Bd7KujBvVZe6*|atdGrmcgo7+l=6oI3kt-liWg^w8
zOl4-jm*lnBzjv!%=bT!$FhSmsn?%&^wG+Nf!vhgP?v`F(pG2MEgSCVMA&H&S`izBc
zpxf6&PSE?XZM(M_fyRs~_02ar05pf9G!{SrNpxc#OZHSiw8oh8+A$N$h^Ls7D8(kB
z90}(Rjm4`u8X5ysWEJw|Cs6h(hRjH0Y6_+!WNj0v
zQO=cE?Dus%+a&PhwV!JzvR1orY31c0B8xX&WGnBTuMy6LdGrkbn(-6A@6`(WTKlmA
zYENbJs#)95Wswk{6Wvdwd~tPjk|duMCKQdinevfj+9~fE1zlOc!8`1+6g=NQqS?%T
z{9u9%%On78<=lfcFBIw<3SjmJj?Ia!)QQ;9Ek10>#W&Sg=P$ZE&y)AeS(6XZ24d_~
zz$fGfM>fkgoXJ4~Jp&D0D6Sx%M-@G2O5;C5{J_=`M2W%|c;oufc7513YHvX;L#R3Z
z=yS@%(zxeG-c|r>FBoSUjs2xZ(Gp|c=E(xIF@%#v2b&h#tO!s`8a=MXAO7Ho>Np7B
zZ-|=$4#&X(VTt%}L+}DyJ6-(obPaKz$uI)Kt=?2G&1r_pFRw5Y`?V++9!H7lu4O~1
zX=@!e0P%h+AG`+T#Y(8qt9~BI%m&f${5e(bhfU{8yZ)c$)!|&BhF^vffloR^X+IghA0BzUf
zC~~sHo$Q(%sWUnG!HCbLy1f2P{gSAvze+2{V-cYQwwmHZ1>=_$GHHrb=9e2T=XUnk
zBQ);n2Zb#%P8CAD73*veeYH=n)9<2Za8biJbkUOTMtw(U80#;7P6CLJJw4X+JC%lf
z^0IXc2CAzpHDiPsuKB_B8I|bv`0mSWvigYqI3W#G>d(beR{1(|fwP#1ZaY?S-YU#G
zb2vYW99xcbX}$fp=IaYRLCJBNI&UYp5yvns*QC1)qF*#=ABI~S&?e-XIS`xOg^#3qfTEafc(8PeoNuoRiGi?yq}0bz#$#d{RbRLp
z$Asv7Kt8>`e`iWG(0{wEKZH#bRXI2N)~QF5t%~zyFA?5O+-t;vX}r^WRAY6ew7z$b
zTaYL5Mf&x<84Ouufm@ANi4vYgBfwtT>f9u*yRaf*9|<{B^&1J_oR~&Ox40ZWexZ@_
z@E69nc!PO$=3w#k>|;hgs1JoaKGSR!vlg&Q?M!}|6dWB~ES)$xqJa>EjHHVUo-*YS
zPBg1;D+I;z<>!pLcvZkP2*(iMCqgsT>tsC0%x(O2fsfAEY8^K9q;-!Invid^;Q$&9
zVj`cPCx6QY5jO?lNL|GKCphmQW&ch8toCNLCqdy5#R+Q$j7CYfoa~+ZxWKt{BqQ&r
ztN}?eE?W7`xm)dp)#K;>J7&M)%7qT1+8e7`1>r1jd04L+H%gN&W-Xy-^^wLR3nakA
zmj=|}KYQl_HuV%i?;-1ivAUSy9Q;ffFI1R%Xh?^~L5)DVeE4}1nw>JQp<@E9s{F8Y
z!IJxL^A0w0D5!=*lS%OoRl)9&F8Re|+^#Lm3`XlHCB1pk$PA_|Gx#Y3?n>ZBF6ZRr
z5OoONUyS?q#O~a`qWug8SxpgKERj^U1Sfvc6Ij>sm(q{o`>6L__QJkw?u`2lPB~3}
z2VC?t2v#BU*Nyjd%iUwnK^OJOMGP(l3@^N_CeqlTvNCQ?k@#bd{RAT-TGsD6pI~X{
zx!8`=W^pjc7f0XSres4g6)4}%+yzo`2`C=I(L(1Me$f=Ym+g;JB(9ex?ISVv^&|=(
zgS8*nr%1V^YKkIz`U0ZMBaPixSj;BPlCTx}n=pngVRoUjS9qP?QuC9`wR}=4D+Eq~
z{!&So0cnj;Df2N>JAeXAfM9O+lD)bXP(cIn0+{NCsA
zO~w1VsgntFXWt8$ca2Q8mY!4)A^!2jD*E7wfZ%_vGGD*KCFN!z)Dg?nA5+t+!$k
z5<>=O9~puv=!AfeHlc4bdD*zkB#j)bS?Q-?3)$`JiQ-BC(yaS$UGkhmBKdT%%?E3r^@PP3|x)I>A0>x6)y
zJztXdkx8uhZC_aq_O$tYg4DYBQHTV)AgnjzoTOzays4}kl;3!j_1=>4Nx?c^vB-9UEq4t|=P&}auEyF@9vx<0Je@aT24E8|lw5H#*L!3K*rAMMy2n?x~YDz~%B8v0;
zzI_yWE6;tetLQQ#srJe+u}o~0?ZUTYlc2gxqTh>D7hI`vbQ|fDZu!Jh#Z!Q{9%m26
z_G5~nw=aJ>0ok8g&=j
zYsIMPczr7xed|!^;LDk$$_&H0kgA|nghZdg7P&^(M!-42L4x-Uo?|g$wimV-%94`^
z(2U7{5#Ptgqtzy=8{=f^Zqx5<9j&6IWlc}a!iqIm(m*@izZ3F+Y~Mb?e8KGf|n{5}!!q5}F@P$e5V_m5+gOQ|O&^batvIf+=rvCm6IRxT~~d`&|kX@W2ZmyW@-}nfcyjc9FtJt
zL4O+-=b&x*&^qvVJs-R<&1T4{WTX(b%R#Ix;udlKIRHYeIG
zFl77HIg>9fv>db2(>42?kAQTVXPebEyWEJXsB0X--4JqQ$BEdkio1$DKif@v$I;h3
zg)Enz8WLAIG4kiTa$mt2vL%joqlw=oApS=84zueF>zM8R{pK#}>|rlz*9}nWmjSde
zesJNoqrR)PLi&g^k%*^OwqQgIxD48;rafkB>E)
z((%oVzPwgmpVh)?XH_M5E?p4cGzTnNZ9U=Y5_28KhLMMECs$UPpVo{nr>4h*nrRH@
z%-4A>Q&pF<<8cIJix+5YgyPD*Y|WupkWnGg9Go9i1^r-#l@#j^H(yh>O2wU316)xq
zF17VHNU=Z9*8pSK)KY;7Rya2{yaCR6(_QcLi#N@h%5ECV(*(1KgFdhx3enCuTgfG<
z@J8x}1!sSc*5x)^f>UP
z54TcRR6r(DS<@jOY#`;FccljU`uUJS1&vm4$Fse*Qur$
zN;47t!H55pt$o+R67mLrJT**y_dcwdc1TQ6#*x8GaqfXd+qOi_lUVB>o_$=Y>dg05
zTWCU1)}+W4PCLEb>lw~vI7QyzB$41?8b3d3_%jduk7Gb_ZG&M8++WsT1Xl)S5Tnov
zXff3z^Xo1)^_r$tSBI8GkGZ1;DH)uykc0eO|K2Lp
z2l$u3mMQ(vISi!%b08Cx=iW-*?@^7Rg(9xb_c{Wc`(~4
zU>D)caX=K<4TMaBhr&(|=)!Gz??z!Hm*V=~8W$yGVodY8
z^Xf^N$MnvkC%#}KWu1m+;h$+2KV&N=JY%F-0cG_ey5{5L-mAfadDjQ#Z_TeRH+38E
zW|6Ipn6rWy3b8pLH|$GNu0Urr?yyjrvK5TMLd?k$jE>`S0&@|K@u_L!nx%frMm7Se
zP!`VO4`tLi-^??X(WUMwm|qEa9Rg996kWmGRZ)SQ)&UI}SD%3_8wTR{gwkGMl|zC0jq$@p
zq~Eom$&dvXHFGI_|*COpj=W7WU
zTjdh{(4B=?6{=HRm?>tuGAn3Ddp?EF$yZh!O^oIBn8z~=rA%q!e_J9+?Z#&zW7?gq
zd4x&(v>K$(dPMWETxVinnI)!_^NXb~IIOi?V#l0D?hWG5>{)a|V6RHJ?8neV64gww
zDb?#JW83)_0x`0(S=Ya)SfaF+R8`5Dooq6Do@;Rot7FU}wn0Jq^YeS~{f;l=6?FOj
z&4WLQpVCdRzFvBdzI+`wN_MYWK1(6F#`L;^5syVxXBvfvz}tKpsq<};4m6+(@Cb1Z
zME;VQ6J&pT$?BW031d^P+}W4PN$f;8(MgF%%k(&-Gkq&Q7Mw$z9PXXkt<#R2#;-dJ
zg{|jk9*3hjX7wiCSGYT-
zjL{B)R))u8*S<%7z;*K>Q+t27ZkdM-MI+G3!Z?On+1k&RRF;w_(?H)$tC#k(S9UG-CeyhyZ4cePL_AYRPna%4lC=!katjy?Dhb5oTfV2%qXnh7lZ;AJfmGkY=yH(_ElzUAix&R
z*+tqGewu~hfV{IvldMjo2M!gS_wJ{uOl7!S2Ql`k1U#lFB$dON4eC6jfa43-vdJss
z@-uGULDx3rS>R0)9eeI$5e}nDxez`Z^vZFsUx75YLXGsjZjRA&dH-P9%k3yWx0V?b
z#MJx^sVwbtY@hMzOSuB&)YAN%|NCerZn*3kLf|*EEMf0+bcB_-M&mIxNUp*UBq`~~
zXj!T-{hnm^`@YJAQXlI>cAIbv4#%EyxRO3Aeb8%T7bNN^0PE^n7kp+nn7rLdp8|{;
zhzjL^o2&)VB5ph%snWO>NbX4ZWe4fQlYsUlSpoUaG%@WawIQa()-B!y`STB>W+WC9
z{WvPOC1fwg#yhuhQ6x!zC%RdoW}u@e$iXdWw8sofSbkQx!}GQb53g>cw^f_@*Lm7-
zt=x$fsAvFpvSBTFKq5&DaK3{GZ;Zn?TTum1f7QdI90FRGna#i><`u9ngIw~ePU(LYGR%5u!vU$EQ>ma4ctUa;N
zqaTpsg!9zl4|;3SzQUIU%RBMsRh+E;qDeEmT+4-6R>Uvhi6orq%I5oz+f2O&|GhJ@
zw+)^iJ>~bO%3k_BprU|*e){<{JzQ|*tt7JUKTe~+jfn9ZGr%QU`Fk8k9ewW
z%H<=OZaUKd%$UQtHW#~DuVpRW*>K{b@?222F6*jP6K_T2_UsPNqPfx)p+>Up7lV@v?_
zh!qi;E%&YRqn(G&6=l3i9R!vX#^B*Mx^e*~akRY08iQQc_D7BFYOAGts`SKd@XMA>
z>?k|_jIZZ0EdKm{ozkt;B=^$|fz#I@ni
zBnX90uFfDQ=CeLXp5-Z0bH43EcN6*Em32i=zE=#dHx|s`Wy;^@!oTXkMjGuhj|5^k
zC0u1ejMM%>CTvuq6Em(t@Q6VbXI5&|YGLRV4tCKWjf|!PXB+SGaRCwgFI&y1_G=@@
zqiTF=vMgMIN!L6)>JQ_uD(D~RH*}LwE9$-IJ&pFxsxg|lO}DNcr*!I@5Bs|ZL9z^V
z@Y^={4pX+v#vO;c?ISw`y5_;bZ5<8_t$7|?Ul_jV>-zX(L_b*zF5kZtEZRf`6tDuq
z?lI$Vd@CqSg>1FC8OQ&khT3}u&4p;u2KXldZx1Cc1eEH^;lg2v)_2i@hAVp8vdQf`dnB@eV|P0aRi~%93YtP6$t`m5UerTI$X@&hV_*5tyBHo
zaG62~P`Kb7?m)XPx0!WhzT)OlEAbR2!G!Kfl%e;NjeFpSingX6|7=A4cJ-2EH(6mT
zwqK>``e`@5wbN>CKtMl)5?+vW6e*u1#Mljr9#O$iH6CtZMVB(SP$7Y?9t9|3>xQ8*
zi6qNOYt0#fXAdBQ`z>A|@*7=5cNo$2culOh>cJQTUmGJQ$HcWG*jBGCBCNea9|xQK
z8ZNP3JgrnMz=K+OUI}{wX-GV+TXv49QI=5UL`Bq{UFKpxOp$&b$zQsT?l!NcsSCFm
zY`OhkjGaStE=<&Blib+0ZQHhO-IzDFZQHhO+qRuIw)1~I_}A(|54uKm=4Vi~&U5x&
z1zW^F;xM{T(-ucWqThbD&Eu+t6WJz(4ob
z8YK$Gp0tdzTA^?_BJgTKR{*xUiIqg{iRsjqSf1EGUHj(mb@nsny}U_Wvvi_RPfsPO
z)%+80fHT(sjp2jHD)(8ws*zwKn!nr1(2PHG)W@)GBy_6!0_SzM~w~x?uew_HUps
zo6pqMX@aB=%MarSR+?J-C^x
zLV;~w-A22&{F&HdO4{?5M3Fvgg&=Zh7QW_e1Z?}M2s*%Yby3z8KtnB#X=5VRW^VQ|
z1C8|{05YXi2#RL|L@0p7ZT{W1KPu!eM=Pq5;cP~Knl`3z4=XZvjdk+M(mg?ywN|*U
z2j{=0(mXC=0tkD3jh7X`Aez?3CM$;IbW~P!3UR+cE9Rh!3^k{E4vy^fL1!&z5&X6t
zW_kIHU{!E}$qG2qhZ(2XS1S$Q7#>p{PkWf@bUOa=-~apvD?=5x3ya|5z+B_KA~HXN
zYj;o6jqhmc@r2`P`)%}uu181txq6zNF65)))B)}e2N9lA^XlbB^Lo7^eEEmJk12b(
zxtPQ8Q)8Jn*awb3^`dth8?jt(k!n4_8A~UB=ZOQmC(!+x#RHeeqV?|I7k>N|&B0VR
z+e77CXmB_Q_0>xa(v+l@0vr{gB<_^`#q^Q2vg*2%LxSk<4@M`qA#q&NRyC&%SVBi0X)daek?Bv)((U%W(
zH!9?bT0le@~N
zCk@XQ(|<=23v?e>uRv%Ut+yE<;kc$UauMh?K)ok8L?O`&kqb%(;7+W|PgPIlRO2Hw%h;kBQg&+bn8e6sc}IDQ<^h
zo4i%%rJ;3=WQ^cHSUUvd1r5CP^m&hzJ@d)|s)Cb=S|~I{D44HTzgi>JqZsN=M+f_2
z3quPxINH>oI*QsJT)w#4oD@tF@~|tJTv)hXN_0>1PBws;z({eKbblJjlG9daWnG1;0tB
zA36z9$@S#$OgBS?+3J0Dr`2|(!}f!=!_X;Yl#okbWLcN=gcknf{y2WJQxE
z=)c#YRWA@(F2=M1CLus2o@e%A)RY`n-qgcq>W4{nx_bci;ReKaNX?9HC71PoqwilK`Q|LcA&^f
zxKHb@pfWPoH`FQ)Uc)m^76gN>lBLVK>qNv}l?62IW#w2a{E
zurALS5o|F#hFTv>0J{dsBeOU|cp=T_*=K{AX*f=~H__C%2SOA(@igCsMdA5?_V2FU
z837_%iZa>YRMu`|d(8s#IKu1PLZlmGm$w|`VTFDa=_Wuz_nu_+RNaDK^l`Mhg+0Q3%YAsoMl
z38pD)Hz62mF+vSS9qKPe{7q&(W3viqcj@Bn(j#&W!qY9^cXheZ2J42`K~k|@YMQ%$
zdE>(N6KTsaMUT>t^eB5yz{_L6li@3{P1mqt@
zqgzBChU_*~o;FBMSDa@={w0;Si<^Gyi*FxVuN7}!HidOC;fcp(8Y^e0P+mr{l{s?rI0u+J@6+2s_ae0s`}yj0Rqho
z#Q8<&helZZr(l`BvCsQTbH=Ic2Vjh0y`erTe9$stFFmmgYTmdCkSre(@snR=SL*?n
zd=36MwpOZqr`$l6N6nQ=K-x~b!u_ZlB>W@mzz;<=<16`~Avxg7Y5TkcZVTRP=zUF7e6pjofn-jw?5Cx1pYM>qv
zoU{h=XzYtkgOy2$B
zSHcKM3%2fBLeP;m(dNtfc$tQj)fa#9TTuvQc($YTHnSeYRQ_vYLo+BlH5N;+^MQHm
zJ;^|oK-UTn9B*LSDFJ|@Fpx^g5j5cnGX4%IP3W=-%2#sdPmSjW@mbqmzF8_d!;*1&%(vwm%RruRT=+?*lo4+(<2#&dzXIkO3A
z@eWo&+e!8xx!_K*XuO(z9?bO;Ns!kn=mT%H>HBgKN2dzWCZyj_b3(3;KR^XW-YA=d
zbaClsVR4ZQ6@n+4bVH*H!mu!a-1FR8f=8*FlCk~)NkV~*|JDp+o?`GN2c-jE9nC$l
z=o12<8Xt^E-F=C}I;qguRKb{cpS2+R!Ps?nrT7d~78z%4&%mwuFzC#F{h87$W^2je
z`)%jmMdux_|Jo}wAJKZ@6Fvudf~D|iDU0mg!I`}Utmha$T!4gcH4_6mv(GE`2q059
zwN$$$JWiW0GMERcV*Syng+C*eh#Wv||0(U#aI11TqYBvUVs%!ZPDxe%^3EuK$Ajr*<_@-6sZZn;iQ?&_2i+Crn<<_c@#Y2OKpH7aPh%Ua+pKcdEA-9c)-o2sC
zk}V!M+A!lSDfSx6K4(a5_4Z5oIjMggw$=3RI^$`uJ5nv!hIxl|PR368mS?qa-YK&4Mh#-TQK|n4pDU`VUnk{xD{y@Y?YdRD-n$oSp8;JQAl};=0Fwob#Q~OEm_pH=RkU0QPb&^E#%khM}nC*
zN?NIgdufPl95h+p59$$<<4rOD;?^R{TA>+aL^7APFI}uz|@0iTX%+T~Dd?YJ2G@1NM-=05K!C
zEla(g7|(A$?LYUb%H>=fAM#i{5_w5qbV98UVlz!DuB%p?tR85c*;Eh=`N(Zyl-$BJ
zE(dhK8->&KiGR7`TFEvyk{vsAERRJ}<6D2aS#{-mJGcF&Q)t1vq!rP78uMy)rQW>W
zx@g|ukt^m14Fd%FUH+V{8KsH!!8cJ1%e>J)LU%~tdksHT1U!26V7|dR3ZG2bxXB0P
zRpkC##pLSQ8EkL+aMfntv;;j!fqbElt02Sj36-HZyUAp)`<^2GWckP5V;&|yS9Hwg
zXi3>rDphy3NOt1UE}8xx_pZzqc??Jqy%!0i
zn%g)O{ZUC~w}z)Ss?vPK?k?os9}f4tI7AR~-yCUxCOcElOj!=zbr(h%D{xPP9a1UE
zs|c6BJ9R{I>QD{8PTMJ=O#5RUwF0CFaz6HprL5!MNDg{lJ^9WMsBu5kKTjX1A^T&A
zuAQ?V;MJv`e%_r+lJDna5$*(K7rArGBt#Wg1qGXZY{JFT1fiPvpI(`VRCql%UX7ri
zx+J8XYsT5?7X&Ly8eD7*g}gUapAK2zovJp+-&f2Ztnlu-$A!g>5yI;~z8=ictglGE
zhw;x_O*B-jLSkSEB}+$Y5!A6Qj=P%FAml9llmC{R&AjsA6$t;TdX<
zUw+gyjTm=9g$C8!JM31asb&oGMD)HhhBv}dXcjE5S|fJ;&M0prJ!RhPAa$#&nUHB~
z5odg4+?M6VsS@<_asw=&1qB_n_!i{r5YSr*(ug9lJ&ibFyYMF>1>~p<>h=oOpMtdT
z?SJD>XurR+!*$w$i%|CKHn{d!V;NX(Iz-s&De3ezIg%K}o?QWsXD2*7>Q@QC)pGap
zXv4EwpBfDReQjYw*1$m^qlI5%22JMG1Ifs48fsQ&%S%%dN^Arv_^0oZ{)-dTQiRH-
z?suWP(4AHwu^SLd{zSZlg0Vdb>M==to5UaQ3@DlsM0MEG*78hv>}H-Cju#KAb7wI5
zKI0*KiMe&Spj@SMj8V7#YzFWPFgCKxnM)@R&ofEcD5^0OEk~F?ht>Zw`z-%exhA_W
zTT)K3%W$^LS{7h!zKWbD@vr0r4I=E?h%Qxp%-XGj&Q6MK1oIv!Q&kqyIRZ($BB(vg
zt~E3EgFlMUM`}t}{qjL3@Q#5`3w|pEjMy1iJWCPxF8Y`+8$e;{A6#_Z$kaFe!LVW{
zdh=p70m)t0OabTS2*@Xx69g#CINOTQWqI#p%=9{KXZlg&Kf{`%0@NG_4wgGE)%{R>y2
zF05socjU9K@owCa;vr9gGqQR69kidA+a^L!PP8Wa)@W>wj3Y-adeXB!xk?0=USdVX
zrcmL#e-|fni5PzLG_K6_kvfBkXyn_K7e?`h5=65q+!;&f5yFvWGz+s&)0xXb
zEXkCtmF?r@=F7L)X=Syp9S
zz0AT6%F`}A?<~IhOLPbhYO`xgMn-5u7LR~bCiZS)DdLjZjQ;8e(tZ949LKDR+6yG{
z?NvqlP5Z(db`J*u=#Yu$a_>Vvr^Ks~f>%Em#)!Hd_l0(pdS@i!nDX&{Y0rl;zuuQ-
zh^Vm!YW43mPM3KMJ(&mYRgOB%X-ry|aD(3FJ|X2{P3#+MwSuE?4CViTsu$)3_Z=Q+Ca
zT@ZJr^`i@!Ffko<0W;Wo!x6^%i8%APaxA@)yX<`Pvu#)3t=zJdE8|nN+X?wESCYV^
zD_HtbAs-%XF`Wl>CYC_0`f3Bu>_HuyV~vnOThU`!eu^TItHC;2qQ}76sJ}-N+yyxb
zSxHC_KKGoRvCb5nBH2m)EsJ}qmgw=ia%WtORx8;jQ|{@1Jh^DiS|L+m8DaRsBAICO
zI?8MnG4Cg@)pmJpoyCdO`R?xfScY6ss|dWZVUoZV$ithR^sIMm$GAk+?D)qVu6LJe
z38=WkJ2EE6?TUipx9ANB_7+oZ;J<;R3258SZuxX;u?#aHagI<+%5
z2t;$8DOwY~t$_^vdB}<{BrnnF8IdHJv(#R(kLN#qp)moL>#xsI4zEpuz{v1TFM*nJ
zL!a<)0!uh!ZPS6objgZ3Xcyurdc2|}OJ&PEevKc;cm=!b?VRe`q?V|9ZRtH{hYS%#slZ#SVF;_B@p8E
zeQwIdsnaRRJ8ha3Z?;QywL8QsD31RHm6|%>$sYLJtbYq)Nx^o;{|NgV`sBD3_z-eN
z^33-P21?&k@6ww8+Kst=Dg3hhx(6?6JmycQ?KLvkdVpXAI-yu62elb{v8>vxY%VVH
z)4xC!;4|KO{%x&K8=x{qY?cf
zWj|tVj%{$cyF87-ONhCx_xU0$B)C09ARjbMQu7qAG%5<{HYMs=`P4H{GPIdfZh{n6
zijM)vX!_&8FCH&nA#bz~6H~{`-V$V~%eN-3&W`R)Hhp>LJ6R7-1lj|c6L(j1$oU4cnn^S|7m!S
zRdC5#cxrx-V0R6(gfxKKIVUqC80{W0CixaNqc(@(;!ctrcqPg<3iMkl|8XB~rXony
z-q(}^8lN6+lRIkO<<3{#T<$Z!Tl4&PkiS2EYpuL`?5lC&)vgBZP(kR1$%0q$`=qwj
zdCrUU)lm)me{;Awg|;4gKKbs|QDRc7-nMq3jU7&cjtTcHSCi0vM%x#ZhI>S=O|kBM
zJb_oGjat9VM49C-E)8NUEM%&5jLx4RWe2p5AMj+-l*)}^7c|9GuUC0BWRb!
z6q003#~h!?O`wa}SbsO>``?fW)PCuVNuDOPhufjmb<@4Cd^c=VSj)p4zhyd*43aSL
z*6IRxemP*q8olIE9J8Fp0mcexGsKI92zGe;#z}J
zJ#%8qxF_}S1Re&Y`bcY3IRoioO$Wc(M9PdmDxVqf8(Ge~W6
zWrrTHIgw|!m*(o=ee?fCgfqRIWFQfYb%_z|(0PgWJTjpnIJ%j5)6A;{gyE_$rYBS}
zZS~aTtN8zm&e2SA&}dpz^8XjJleKk?AXO2hmQO*5D;Hg&rdmq=0YmlaCR2WOOu}Nt
z5bZ@WR}vWNUa@zU9MQ8zzgaznlYh*)qT1UFvtrfXJav;%`a*%&jAlNh-9Izm_oreE
ziJc?MrdYxOF(->I+jin`xLv1`*DeeKG@r!(cjeAI&Q1`tjwGOK>ZU*X^*)V=D
zvvA!IdMZ!$dyBe_^OVH>T2p|1t_@=4tX;jai5xvPQNN43k*^>1S*6_v%jG0BMd&bv
z+Og;H{;W>0nX}vFK&pxoNPp#U=Wc
zuwJ*mg>8z?b#hpfFpcbUQ{#aLSyq>cvS3xay_AWTaFHhRWUwbGbyb3tmi-Vnha^fR
zH{YAAvd~+G2^zck7cM=BLiK}!piM`lRDRv*aID{LTbk0!lhVs26blkBU1e7QjrCZ~
zb=3fIy)1iopDC0^ba53a%)84J2fi*6a8KlZ#A6@(TVU-bD5Y5=ZTCPx$x$myn{W*^
z_9>pt&0KaT-@<1WCxp1{N537U;M
zICktRNXhnkJjGwOfi7CkFkHy3ogTG|OZjH3iYh3w%)8Jtq-ogP@L-@qw%MDJudn%Q
zC!zExjCkUc&KT0JA?GznE5Q*QGA
zba;)$;!QBZUf?Rpbi~dSp`y5nc(E!dF~j}^J39@Uy11=_9kBoN+*V3MZ7mUliuO=7
z;_)~_Ra8k0`e9SfC*o|rG_;w{Wx6Sr!01l5JiL1?_20uO3rPtK0>rqZvJLqmH0>_L
z1uh4ZN9~pZ=hVmS`rv@KVN)A0W0?MS)%Z2^p>CSu^+XJYwMKz>SBu`C<9~Vd>})Ue
zF!Ku0Vfmov`uQ&h08^?}``Oqmxdtvg1lX{<&?w@DPvf#SV8{N-rAkH4<$~IK_l5^f
zQ}TO4aAD`k2A;buG0C3VxmVqp9!?j)m1IWsgMc}v`zAicBq4|Q*?!Rbd4D^8awT^x
z;d-|I{4R%^^moV`n~~K7;Ut&On40aE>KeoSa$+-H5MEY@d*w+T)qWc0aRqY{J@-0M+vqJP%DRN=*MW#-S?j$3Z)DlaQ{s1Cu}?!hgh4mpKOtfi0xPymW-6Zw6hUIG3L{pH9XjTHZHali20$I;sAXBo1U{1-;
z#~kYsF4D8EXRFl=C)!`*cKv?cb;*sZ$>>X3rI54k_a;jLNW$Sea?u`zstLDs{HEe{
zX+#Jh$9qR!X8rtbtT=HZ13{Gjha	bCQ%Qr%h_Jh_-h2A)L9wKtNO`Qx3vV@hpzo
zc(|x|nFtMHsPZe|4n=qOrl}KTJY*CA%~o21`8AP?RW3*T$l8#4yA#dv()|%F#h24G
z>r?dhwQKBBL7=et-FbQ1$ULH#jO+;|aqw#}-BKN%e&9p5=qxE9QzG*GPdF-5*}NSv
zy~|^{+US3lpQ(Ic4$A81s_~)D>mt)ulsb_-d!rA;d!V(Idf9lc-kfZ)%
z=KnPjHqN&vanpO%As*@dc2h{+#6=%_^G(KC!P}CI-pKR
zkVL~UbNFpuuFhFwVLhAQlQqx>g!>*~zAp|`6*TPC`UR|Dz7UB_|p
zpz1%$W~*e*B=+$B+8iv_sr%#Zb`Js>zSezRH&Xa$}~dpzK5Uq
zi9fEd1Sf+5_!|DT6l~Jooxb-%Lp}3gJ=q5cGtO#oBi4rlmrpo8YN1?1t+xSwn@ql~
z7>NiQN+GfcG?eY8Wp?R}X&m|lNIPehjQVU@Rgm=iXLg^P0$1^RJNXXu;x>kym)3*U
z>ue$6)4+eAh)D@=63zN`x>$KCm*L2n`fjNe!97wDXMA-~pe8*jjM~Dz
zDx8vCk4LXH>cYI3y;vvkkjqrh9AXCcs(`{xl!@dnq&^O}y!0`vBB!)r$LV{tY=0q7
zDhH-xw)suM0_0Cv&IB7=e+t0od)?~+Vlj+qCyC!e1uXm7)!rp$?2W76o)%;&#F4)p
z+sW8-8nPdFTE&Kj(vOsCq1c6k{T-r$d-9XrU|>rJC5r4U_%`;uiz-Gk24icjk-r>XKV|^oG-UeEX1KMSD7D3nAns
z1PaSQjDhSF6IVT8A5Wb$CwsaDg&rkDU~&E9+wEVk>!IxJDE0`&l}kX80n3`wC?u5V
z&0~V}8o5C={RXs$^t#fGBfs@G0N}XAUIsfTyWE=VM
zH1zS1V{$}g_e2$8iOz%k;`Rnc*4h0okU$_I(#BE^axvGv8ZEq_QyU;5bGpSePA>Nq
z0o%p~fe@+{P~i^IjMa)iD~08Mhs(ERfL@XRl=!lMA@;X12vwDg5)qXV6gNKfWcvik
zov^>C4S$+bFa7;>Ypw&%@OLveaqr4J%l9}$?(QH!6?4;Me{|$mtbQA*O62cszCHr3
z%I|mF{gud`F#rB98#{4n@8-OI+&)o9r-{HW0o;)NDIsX?WG2AsG;fs+92Uz+ZP4%B
z>4WOx6fB;V$1ktiu2O@X2Z_btz;IYH3P9A1of}+7875UK=35j*G`D
zN3ed^QXTzP7VfZ;mTCeiQ$ie3VsT=cz&DRXOKigjR5hvCA#EJkmA<(&jtlx`SB;KQ
zhnY%JjVkd80+yfIjIcdB!b)N*-rj6Q?iuPE_3B&Llu4h>im5Q{G+Hp>;;xRP9xJQ8
zwZV=MwQdWb&}QPYEHY|ukEg#*C=U(59*nbYOTegkbuLc10A`ch4`SXvr*u9`#q@SnX8oWR%!{Dj%XR%D!ptZp74^xG5F7za&8d1TS^6D?M26
zt@xj6k(Mj526a9Y)2cVZ;$TP#3bkL6a1m&e3bjj@%d`TkfwAg=CN*IhrbkuQv_P>2
zfO{oJ@+}7s<01b$+^9>O9j@`p&1*lnlOj8%
zDW}}^JO4~naGJ9R{lpIESe#$jFrjGwp!cy5eRVl^T9_M?P3qKlh}Qmaa*uqAM`E|o
zI@D1+8HHNB#4ypjv+-tk+8Z8ECV6a)*xT_fm2T02igI~1sE8TAd3qv5>A$c#fOFWe
zg+!u~%1y2~Lr71JuNQu)T!GEz%pKa*70TMgRZT9j00ucB5x-VT(t5yMnRykYV2n%!
zXIT6-`kGTjQJlt3)8v|qtD9-Z{7}Resv9B3`NVEnt>bj%PVr-R9~4MuC5xIPdN3s3pyjHFn;u;d0|^w*td4pzZxxz
zvCg)vd)u-W@~Z*ftWYu6qF=MFG?SthAei$mK_Q9h{gFzlG6ug{d)N|T6fBL7)#Fov
zC*(Y3-+BxmS$bj5;QVYWp%FUvqjjCuDlrSm>yK&cCtr|BVbW0;ci0y`H2|_{3LjT<
z55{uqZqYc-m$p(=2VW8-?~@T4UanX%c<8_###EFLC%}t=cMYBTAGh358-KZ
zkl5Squt69addBgAGU2_He+$=V=A_>o!rzz-cZmyDPsm3n|C*ficwxL}XXbOVi3J7R
ztne@+l#Yoy!OWPKJs>qeo^Vd41@QLWun?&vTSyDV1I$fMN=aXFic+|FRys1A+kCJu
zj6Vg3jf{bo!ef%??SSgtA!2a6PbOXuR>hP%|0z1bCp9LHiW8bcUcmW-XRt$9-$8~c`khG(-QeU_&)W9y
zLnSCB_eY^bV1pPI2rEB=ZFps26B!6J3&aHT7z+p*aZRI$%pV#A95aa4pWO^797_ul|hpJwfwKll^SH30GsU(JOpQ;;P1jH$oZQ_
zZFXWJZa=yOlq*O;fML>_AVq#s6G{I$!xjL=MEadW2vKl(W8gpm;oYwT_or>9uRKan
zU!&Cg=mO69`xfs{U=AMNPr-~I{T3*QYqPVn<5w3kq@kJRpO2y4`?*}Lz~9CR7#c~L
z@T$s%sUCRzbbb7XO8E4FiJ6PKbAjjJkz;otU0GZpGdp&mB|v;+kenD%YinIKp3=`a
zh1q*$63XyE&i(In3mfJk^vhSDD~n8h#RkENT+TJ-%0TMNETLt?Z-g&f!Pk}~5|rBF}#^m{Fu-TT5_HACY@!&`y6KdH7jE`#C@E
zACv+{ClFb>|N05y`?)^=VfoK5m-+9C%;3hTIIsmw;fKA{FAuzhwdK)IUcxV0A=HORsxcry2@S(lwZ|m=l45p3UACP&LCML!P$NH~prC+CN8~wA_XZXf1r;zV^r?0eI
z7go9I0YV80Ael
z5cxdNO8~+8|AuS483yKCsK+mB!$6o={fK32sQx)kNgVyKhyD-d8Xa7~(f^_2`;a*N
z7modqSWr;>7oG^>vHBrDAWR?_wf_^koNoT$F!!AX{sJ2~5&inp0*QC|39`$3mJogKabWEsD827Wt~eg`Y%;SsIlOG_>QZ4&~*#{u&2
zsdk;Rb$@2Qto(ew07*Z7tc1U>T3-^a;gG{KNx3mOFoJV~-<7;H7-M-S4PL(zwG(@3
zTLxhlE>*uFzuUAxh~sc3fsLFHOb2j{(zOUw_egVjlpz#nW}MVxDD#OCAv&^I7DW9R
zQyau!Q0Q+QZzU?Ig2%<~Pd>=s17h08LhbBx)-h0MM;fJ*!drp_9t44Hhl4{K*szzt
z7Y{JLCAP)OpUfK`nAv3sakPz|fME?Gcvg`R0xezWz)$@$}Gfa)fPBU2-6lhCW)
z?(WYK)J);=s)Zr20q|3s7@(NEG(EX_SiIcZNyl+KGK=OePt=>7&1Zr>3d4g0qMWKK+W(;CTzIAV7_e(*2A?YMhtkEuGZra9w~00E%yIst>5a@C?z+Uz*f{vyVx1sie%*N7wn?W
zSj?l1;%ck&_uuQ%xYbXvu9|*_kIS1g<9(KaVPb?gC9cqIJ>U>?E9*TW1dC+@pbq$C
z@O>=m&044Hm(*UFiHFXy+fHrE@oWqTRPs_rTkh-y^oWFW@-T~wNL%SgCKIxt$sZUm
zB%0eVyz!&CS!}SS(IW>FDOEu!KO09i18Ko&P03Q7%sJD||E+S8K&GB;xQ;O<=HhH!
zxPSjJcnBBT6h%5oH#o@yOOSJ4NJA8rYG!+}jBGng#O|>r7I-t6T8Rz;4>l81+r
zErAhOlcxSbr(}}8I&zz4WeFLAH$pY2wOJPIU$=hGHZ-f`3nl$$qhj~s823ZJa35qk
zu4C65ecGToUDpu(J>OOQ_zF_MXtF(jhNG))n)b>fE^`T%
zt9uDdE~SCVyvSATTW~K2Ac7!dDHtsqf1hwVwXCjnps;@l+h^m_N@)YMtHp6un)y!l
z?Z48E6AOJU7~+{Ec)7Pp%)fbEpgqFLox
zDXy#OM{|zj$5Qbm&$;UL!ARAATV`A9R=hr#we*LST@EG&7ryxF`NaoIX?*DnO<}`5
zv_n0q-Sx+zZ8|Vf?trB4M+=%BxgBmsA(8$6+H>!9l3LCCcEemq64WH
zgQrlYP}Q`A9pj7@f+Bei-pJJ
z271?Fkh*KVAAT}@T%8YZPg<_{*)v!){`r%3R&rrF<8^hk{VbmPuU!nm)Fnj1^wu@8
z5%e_5KNYeF@Yp%{r+noqQ*vs3%RkULa`hcc`$8nSEAJbzDAd;sT>eW;N(y%SjLox_
zdG2!K8`)@}<+{4WP9(-#g-hB=r~vMRiG0>-%fsBRju5IXIU3XdJ)v(~(=~6}B*zK!
z#DXs>ZfY$A*numrSzSQW1cVxp8+&%0)pdDekwlB|H$HMBd}hf(W0G5>507)pg-}AB
ztmrj27kREFb8iXX!^uhY8fiiR-_15_+jD(^`%4WVKR+5oNZ&C=_0qD
zei$R%p8eoNU;haPcd%|0atMgp6;v5+SepZ`mj*G?=@Ps5%krKsJl{tUpgh2OO}~Gy
zeYRhtu~p`<#mR6jEMOlQy!kco;?SHYLtEfhYip2)sNul*Y|xvgWez?FNa~;Ue(i*L~|r#{iao3ap3$2OcnXrgik^LEl|2fnMy
zMwVV=FstY*aY^#xC5oQGFE=736{mla=q##>()4Ex03rII!U7vY3l33lW;BAzmPvaM
zz9Frk!uq357yDJC>v3X(OBdzc2Fu;|B5aai413PEWd@&W%*=$8A
ziNMR}r~HFCwt-MC;?$mVcM+4FH|FXYI9~)9qtZVx@SE3x)spFx6AQGhNGj4;5o$=>
zlN&Q@Zq*~P9;8_zD?$U%e?70{X1lQ||EwbN=;z48@
z*6LQre0)Uo$&9&Xu5Nl;B8j*DvI6SOFuAz7!v#FRiRnGqrBC6zK=WI!s)eb`!2|tb
zzEpW9skxMA|9pxLq79w?WB95$Kfu+f5@bCA8MVKx7o{vY3I
z?@M&>M@VVi$W6WX?cBcip#qG@o1_{$RTL*c?Tx(*X~qEUUd&)+aTQ)MtjZD>K5!+Q
z5}!CMa15P$=QZ`8(;%Y)-uY8Q!=9U37(wKbQ}=N}d~Na_3gdkIqiiCFBBxpjq^l1#
zs}00|nw`K=^3x`j?xeTQBTnbOy4}X2`PAr|Js5>@wXiZdUyHow*RTm}X)O(G`+4P4
z{|E{cpH^jE3=(b6kAk&Ie)KfW3rBPUyj=szuJBFjfOL#ldcI~GO0K16+_`WGdEOAu
z2jDd$evW;c>tgx~X^z6`GBR<>yGp!o7VAFgNermPf#)~%9eW}T`|yYAZbSX3GhcMiMZhP9&kC?({v?YZe=bn&kumpz%7?Y0=^)l{?U{^FDWYADfwOy5e
zNQRyu$^6P4WBZ@ev9F(1=*ZaZ_|FqOcX_M1(Jh5VX@i);*5fveacm!M$Gw&W=;0G}
z#C{5n@WVkW$wS@-fo3N1L#ZQ#8TA`?Xyib;Vh4)q!r*B+n}p628|bkcUJtr;*t@fa
zV&`FR&$1gGWdiG>AJ#F9te>z8&fp1a9zD#T844+zg_;)#3WhHC1-yUb!s#QMk1_M2
z`+XtjRX?z_Mj3-jDSuVQM5IeE3!>%I;1gFln&B7lnL(%!nT{e*!1N^R}-n_TmWCZN?(CeF%h+;Zy!FNDQA
z4qEqf@{-doKO~pHj5%|IVsRNJUpS9n($<&rE&1A(0?jq_$FqaXzO1fC9
z5k@xgP-v#h(CC5lB_DPeE<)LHtt29-1?6T(Opl1nN1+2ck~HQe_*5pTH^${{ERXq!
z=YDZgNaRIP(7C5b)G?{$ik4axVQ_j*+MVm;!9`0#y=t~^#VYO5jl$=O6lWhs5!x)G
zy`ZXiy^%`H3{ZsjV<3t)>QK~dheHnsH7)*&Rs;v$*UsAHuBXfU*n0ej-vTcUb@U_X
zFZ7Ethp%eA%|bY{jnTI&!Lrx`+cvEyZV45g;Sao$#7n~TCW+*TXbr;6V1P^kdx(P=
zpe(;hz;cc|RFto*iL*P#Z&!q1k(yGS>Ro?t2Ec&&kQK%>*LzK)iYeU9j^QBm4Fvo<
zH7%>`j*r0-fm%az!gMro^4SmPHVF}$rOVMx2mSBeSkH!ph@|C2s|dvif@;y&&uK!Z
zKmG}YV=&H=o8uNvrZ|%!GtZ1-^Hm!j4Dn3<=5~n-jZ@k7jbY$peWok|Gr(qozt@Y{
zqkW?pYhd!yn!4QJOc!M3oTsN%J{^&r^=r#XJN7`cQzCT$VLMsWgEu+)b7f^Sqd}e!
zIqBQVYKU2`-?DIof;{WGVe%I
zd~83T8$i(KLk@Y23}3_O^}**giN#zrkjN=b%kiGcu2VW&rcye4z_58RCcqcA`qI+v
z-Uel%4S)a`@IdZZSq_GP7c=U`mdZ(9$^kXkgMC9xbfGbcP`6;!0*qdsq&g-wCk}X`rTM>4<5wR_96&Cub8H_3
zCgGwb9IUE+sS1qd?X>V`?RZ})ib269zcHPIfjT{EG*qECzAceH4^8!cy<8o*IBdj3
zd26ws57j3OIMJ%mT?TWtnG0ddRuNR#p(|30w;Wb#^-c&ZBe54kZLOPb#LK{HVBh6Jc
zRQip47!&m>j!?xTXz1{8KcfBKY_WP_Zf4!FH|Ed{X&F6D&u)sz6Wh|o$D#T55Y%`$
zg4GbWoVmo{aoN#ILOYY|;;2KjiDFqRSZY^pEzJao4^4l4vx@WV8>NV6v2yJ=sO
z2wH!O{F^yjIOpM9Z97vIYM&6B2K9unEKO;sKqf*G7!ixGK6{#zf+K%sSptiqG;}f#
zH=gz;W&3oaws|)<%II#nEEM86fQvtS+a?C_pXXI`lJi{XyO)#s@IN5W3ZE{id19E4
zQUum*FaVg*Nq<83Gey^Q2KRtt^JdFYS~Fp8v5k8eqIX5NMmCIo&;ruzf|SYV^1(Vw
ze+cJYz|o0YS#>yai!-tby6p^X=VdA`iPNitMup^?ouG)Pk$zjX@b``m;HDaC?E{33svv{E(_DMDGS@U_
zZg+Ke2abe$stE9_baI7nTA+_DJ@^l4bn4{()ddA++bKpt`@Gsv8g5`U!+cSp>OG#W
zkCxN^-awdw;fLW!uQy6x4AvaR!b`HBy>UD}THIMxS~r6n)ufd+(OrMK5O;xEXn|XV
zjFb60B)^s$d}w}8_wn6^!K9#j%wOLn${P9i1{AWoJrct+Vd&`@p?1Frju##NxuFcl
zl^F*!k{O}}`qZb5aj%G5aH@`M48)9V_MZ)WFpKl`hlFny5VvHa}~C3Tl$j+
zZup}v^01W4n+u4gq-3KcJrDWRSpKh*pV7)FRVHsoqQ*!Y#8$$+rt2JIz=M(VzG{lW
zNVLXzn`pMtPgvT8ZZ!heYL5BF7vIDyZDH9vEvVL)Hw*6EBA8d&@jxSLe=cUN8K6R$Worhy`(AqvZ-I2KWnm3847{?7fBv`J0;uggLQpa<*C^c
zb#|m0)jxDW$9Z*iP(4OiWy$!vlc4lD#&(dsxkgVjg-AV+L=V-{)zMv)1ua2di{eN`
z)|FgLp;kZ?2Q_GL>=J19F=q7Sw6AxaoKmS6ex2(t?dSYX!r*h%9
z(&FBr&bQ@Wnvk$iTOE8Wn3KZHu_X@$%-O2J^?=d&e>k&!|H*D!>1n(5ZEnLIqgAU*
zGoo>EG@KDZ#h~~2a;yz2QLkuNoZAWFk>gF?K~5hH_Eq6r$D(S7n^SrSa~a_lAct{b
zCW)JE8$&M>gL7oLTO-x-6x*6`>+R0pa=KYOB_SFW87t8?)Dhiu1HJt(Iu@pft$Y(I>FxRakq`VwX}+|Pm?Zt?(}LU~QQ
zIDH-OO#ixPNwVaRC&en4<;%0VG4D>3<>3}Tp3*Z5GY)5A|IxQbuUI1u3d`MI=A?E!
zD)Mt)nxzb2C7&uAV5=UWU7}+6bHV+ko6PZ6WLyCe3JbqZv5mtkMO0?|t<^B7p>7ZP
zL4vOc_L@H-d&7+}8i$qB)8c1`V%J2t?g-xNPq*$LSXU@aTgo$5%?|$ptGJya3u`Iz
zy&?>HwHeLFnM?pg&vyG~>Izc%9&{oDI~4Yb*<0+T(2tB6<94RcqGA}jZ)h;I06v^ZScOi^)_d2Z9M`EftP%$ikH#WSP7Ri@iGFE(-r
zCG4X+h!7MqvRdU}-n84~)V2s_dc(mnCb6B8Le}e`*?v`b0XsVe4&A8}2aCElPd(HO
zQ*MT}3M0eCc8ZME<2||8fi}my*d?NJk&KW6>X?K7O||eXI(r=%tcAPX5zKGx0g}r5
zjgy@&(d2_{R5L9tTXsIxBRuHs@ejoX$?#neiBY>0%fhgQqJAF_*!z!<)A?UF&h`6f
z)Q9z%u}UG6dPh=hW9zRO#0DbAQss4GSp-mAqRYc7ONdNW-
zR6~-b=;o6~q+8v@F2J5g*VmMKf}nbU1w
zI5}aZo43jU{h^Od>PF&f>wt|L15o9D31rxW(Tv!SPVALH4rq8}Sn0x;yF9$P4dp(zKXi)2m^<<4?j_?)%El^y@wz>u*KQ-D}Ls%_iXXPC~engiZ(z~QV
z8r_iR;&e9~=Na?`{bLfV6@#_plXCL&%J9ls0JK5ul6ubd*Hd+
z=`E{!dSR|v(KD{Qi`U+mqF+b)Cx5^k61d3b@c>jaW|26olU1^}sTut7?5X7vveH~2
zegisPy2T(N<+|o?lw8iMZ#y(yz-^N{tCmU5m$o|l>C>tA^?-*L%%0I6Nwew8Yuu@e
zpky7w{C{%P8%{`n0hs(b9G2j}k|O3wBiJCI+>$w__@yguG3&w*IUGMXYZc6T^ydm4RwfO>g6#3sPr_?;cW)O
zwe`1)=IeKUoGEcgIpYDnc!NQ?ARc6{fX?kBHRa8>KB*&O*#7tyKPLX
zy_rswiIM%5gu3yhN+jO{Q@&AiiLr9Wmhn^4**sAC+2a;2CW3gG#1H(d@s(wCj?5b-
zpR*7eWUW$6-Dua)QEb^iq4$^NgjPqp(QBN@G=lgueH2#{5~Zp^=3Yzu%3h
z5x*yYLcA*xL7;Do?XYx-F=YzvS+bCkI*6eZRWqbqBCP&_@fP3T@@w(;+Sg9Y-@`c6
zO3>?qcq{(0rwYzEh`uAq5&gQW@A$0f{^{rAIIWfZ<%g$MJiIJcrJEDliZ8;h!$h3r
z5JEk|HM7CvcDRIL9-`c$C{6wOs7~f{J?q(iu4a&nBYMNgpL2V_zC-Zu9xU`&EVqjS
z(2N-zc#SQciD~+krl^K@@vfFnTN9`RE%7{%w7+yY8d2hqw~y9XBTf+LQ1{E4H}
zwXcSstsrHc6(4p3)f*)e*sX#dA>n3um#e*F2%hsX?S<$YMia)&b;I-vI>nI3Il50R
z;=9)laJAWlhZk7sXK=&cnd-}y#~H|YU}p~swJxGwtRM|J{SJCE`J3pwteaw{VLxc{
zIEf@}zk-2C`tL*}trLR|s?)LrJLWEuA7j+AECTY5S}K3Z{t*N@FJ6fKZTU2+m&XQT
zHQaAA?}$xEeDH_{%CI3wNY80%VR}D@*5x(;brP-$2WHJUbF)-Eqg^wSaxDV@7ObCrDwKFM0{IQOc#a2E|ntV~DSz
zSWTd+W?dYVz*3>mv0lXhk9GXY@OONkn)E}eLfEbEYXwLV0
zi7x;pX!bQN|MM-P&o3C(fWV9TbhBqTCxc&O8knDtoxTxtpNNX}&YOc$d!8=X3_l%>0Ewy;a3k7IC0Uq2n55*e*l9
z>Qtvhqat=Fp`4RD>sp4C;mFY{RrMk(z>Zbm#Y-UR#5bxnJQgNsU}A|3^J$V{Ocuu~E_sm{OaJeb+me<#TgOcd1O7PwCv9-Aoh}
z+7nnc|Brnw1QyNhscQw96(o)us_bcJk(N!Fzk!hHRcAHmR3E$ylO?yY3&L_O@5lG#
zGx+6YAY@&hnK;jTF57;}aLv}}FuJ1hyC^hAZEgxMtg76~Ik2SiG~szPAk!{krvr?M
ziQzgCcomBM2?nmeN;bO-)yPFf=cQ?OQeQ;{1)hFvW1thzB>A^hA~((m^8%vk8*wta
zj+d=F8I!Fej9yUpZVow!ACr&N>Z-{N%iR5~x|4z$h#HMF)V6=#b=zYQ-vFl_9#poYcYG9M3^37V?%a{>{nM4)p~t
zo)XT2w;v!IEoiC{0tZ1vxN#4~B4D`S3O8$E;7=9HC7b(8ap|)teL3Pad)E_!h?BN-
z+{80T4{3*P9Vlx$i>TW>J39C$;)%&^A`;1gvt;7HBP3QmeNGV_L_}s)XfWGLMfU0^
zil~|mEFu1RjD(rcuxy%YCX9U9PQvZ0{AbqJ4FAI|DIGT(6L8>4JANfl^CS>-RX-$=
zK(Q8Kleu@%U6bk@%&2-*sv*&yR-l&U1q0b2hV;g9VA%>qL;>cwyaeo~>ZhUpMP~|Y
zAmVASGc2{+$Fy*B=8XPjGLe<0;Z*|j6L-m3#J;V(&8t|583|SHb8u+*osA*go|46^
z*;^}T!u+srM`n$Sv7}U?CQKxM)BHzYy3X8--Ss>wwqci`4nxnhaHc>q4$i1)keNG}
z^4-oD1rBXp0qV=vSCXg^-C;d->6eVZGn;W$P
zE>3c;uyZ7s^3Wb_pp1aKyHwt^=fO*~?5TQz|IOuD!;GbGCM>mw%rNkKh9>Feb-Cuo
z3d9}(CE;oMKQ`P0OBt0ZIS_Pss5KlPiW2h1{_}{oc~48h;k_8<4t?~B|3J_EV-B=A
zil;P{*uDURLeK9@O8R}I(Nw601I6Nl$&{h>aIFoQs8iv}oi01G3>qSu;4V$ygs(mw
z0|hC`UKT1ixwUU3BV*?G!0@vDE@^wIeooR<^_>HNSVX?xS1(hb(H0!B&nx~2P$oTJ
z*Mx;{MTktrx-4s^h;kh7Dg9jg9n`Nq_ZjmIU`gUJE+$H-R*_1*ncGVRrBy5i9JMj2
z73)mxO#w^PR=c2_2OMaT%iE`$2EUiGkp2!-ba=G7i0H>oeKe^YXz1luC3DfkZa|ok
z5`(g}CuO~yDK9mDQ2B`jcEx4#TxY1UK1~ze>{@qtR1DpcFC(Zo!J|LpsR(C(xg$xn{lE$zydPdR
zX$`-FV`BN#>IC9n+u^gov`H(@jBGCP%x>za9Ix)#-cIzn`PTjowx!zT4WUnlNy8g#
zgwBI05BRgdC#W-7B;C$D=w6t5)32-bkPEU*I%Pn4V?HgF#SjG7}@v`2FX`?OPH$tlN-7~b0CV)YwRs9qt@n;LC<0YmL65Hq~YqwaB>#kfqZy3nWurw&)p
zY9>9Q4Ey5U?lUN|=oND%S{@qETCTo=0fk#{uCDE*{Hc8uc0F|E`}wQ-M08fMvmgIid7{>8Qq8PlMPHh#tl|(XA7#BsEtG*%Va=9!@F?!Rp6BFi`Yz#
zCG=omfl~GT>cm0zppGx4rw*$3S{kt-#HeeN&ag<#Ks^OrLVWy~HejB9*#p=9y?GJ6X4YK!%gP
zcCg+Yu5n37_BPfX+jQC3WQKVxf%%}Wx_+>Y-PlZbp7*r_Sr5oWGRBFJy^3D%kt6jn
z(@}SoibRvTsWrzWMlYIqn;Jnk!c>=RZv%mE
zycWYveJ~bFbAejs8L*P2LNbGWN%vuDaU!@##AoTf9cDV8+uK_uApF+Z@L}bYHl!;a
z*HKZJ_Jm!`3-w1b1&G%0hQsw`|7*~OYLbsVKTOP|VwhGogk|Em1wNUqwo#u6
z8}a7BSp-fMZl{1^V3q-zxD|WJBt=p~nhU!HwQTN4hiAnn*`v9ecmx$rCJcf!e{Uk6
z7LCu8VOESB4`ce>^lYKP)tIRx6s;sm82fks7<)!di`4q8n&Tl-f<+a$@QoI)he0I)
zgwwR{eB*o^#?E(}g8!~S(!JxF>D!gSi*gKEcerXh@UlXfYX0+c;)>XV^PW6&%>p0fpAr
z0bvKS{n{vUoby1*896GU5eTDM2Ys@!)6=Q?B!117!@4Z{C4Z0Om+2ETtP%zJG$g$L
z7fqC3>ypPuuJImd%50o`cX;xjYp$9R)?GzJA&A>0<6`P7+R>>H+MQ!kAx_7NyWPpA
z8gJK*9jR?W!myHM2?k#$P-`s<90@KqdcBo(qJ61!P)(HE*3~K`RLE+#bRheSlXM1V
zw71hQ5H#_Q=Rr~tQct-q+S9VT0bG(+#wfvOq`}x{Hu(hqOMCwFRkg&$WLBT4@Z9NG
zeJH|TUlB3RVsyXF&)vIzR$54@C$zgK`>Djgjl4;eizHM|CMsn#Gy`Q!<|hxke=x?z
z{NhN@!_~NBFb2=3&&RQAz&E;GssHpZQ^11ox6wUVLrlGgXMiGnnm{@Vc{(4Hs{L)&z@{SZ?3qN+`+0qlY
z-q@JNXp(8S|a-;;(%FtJL|7Q
zf#Q$=K$raIN4VCeNYT1iy#&OEW%RBr-+S(Gu07!b^&BW=C)6m
z?W5-|i-C>U-+)^1Ec1go^vTQ=9d_+jBjw_RcJ4%4
zF+2SHsD4&DtQ0qkr36pzsB={j_~e@1_da2d#EFpwWx&o9h6Sx1pI4&_ogV?BZ&?8<
z)&aSMmC*3Qx${1apsZ(T
zVG__aTJ}KVRgo9yXxlJWAU`rzVBCLKnW4^TyW3SMhoTuo&UwN0{xn6;N+Pk;#u*SL
z&&2=^T&Dp0T@c)6_FhTSAswY21O(p)2lt?<=jlFc@mm86V%*1+3!gFOqQ-Q&4}z1D
zY&$m;z`v5^@pO~{>CpIsD|NIcpF|c1E3XX_^a33sVEj7B#V-p(lsa-m$5yG!{ovW3BRLlDi0QeU8(74lXyR&)w2R!KpoP9|5SL+kz?MU%cOXGe
zXhS%v`hA%M`uO{`*T3^CoJ$pfgro?b93{!+bgq4j(L#8zDKGC7z
zCK(USZ+*Qz21e_aEL&?-#k)^MaMo}uuB`0sc01r+h~RERN;Ec1^w&8Gldl%+)7VX-
zM3_X3;JXwVr5cgzAaPit;4Xzzegu2ML2+xzM{c|5=Y1v`>K!{n{&z?e7YBR7JXY*=
z)>Bqcj&2=;!G2l8rF-Ax3;8D_1A{F)9mc&n#u+$WzIs@6ZLQbHjmw@YyLKi@Dmf$bAUmBUx(&CoFcTIX0>x*kyu_%zA!L2|wk1N8i!yEi8
z{YFK4ZG@pc%VNyCu5vzyb*BF5i|X3%kB~sPv;?uhUNGPBzmI>@P`xJ>I$vRW(7dWr
z^x8hu8#87LtSSWxwE^CalN+Rs*+;vZF~lQ2c8*Sa9#nZ#U_ustbI~;e5GJu!Zrs5j
zgm7-b@Bxigg4r=#JHh`*7BP+6tPHdQbhw<)+<+9dQ}CghqIF|a_e{ytkBn-jB+Td?G05o=oC
z$ZMrd=N88x$ybq5ph2EtvC+n}X7i=Z5uvDfHB@IlsN(4ySSBEo>0G;0H+Ux7R?~
ztQbYTiV;L!{rS-u*wP&Sw_yHyMk&AcMBN(%XBn}gkj=h96pWXn+{6X_d5XSgJN#Ng
z(y`!`4aKMWR)H(!fwGgqXxvIjWOv-P?d&b+CDnNwC@--&n$8Yhj6b=9oVnVq(kIbr
z6;Jh~8ROdW0o=J;s&j0X)9dg8{gRk0s?b0bm!OgfUEg>Iu@~cOd!Kmdb7K9Y>jwJe
zd}24M8|}77wbs)Q6T5K5?I0o+mbzK7UdCdnchqpT5SplR!<8=UiH0T5Jc#P|4ve2U$hw$0W%{z+kc(@Cws=i%=-V(GXf}j
zF-sd4Qzrs?F&je{QxQ{RdlOSAK0YXC7bjCgTPTmsR#i_9mCxVy5NSlZ)K+BUXiVk^
ziqGCTR^zE8M&~fYY0PE~Y(m=h4y6<|YfT{~3&M@7e^Q3h@gxgX2;F1vu-lu}r!N1W
zJWsP(AHBO9*WE2~Q*5+w7IMMB0$34aLPm-f#Qo!I83WLxq9GtcD&%0XKY@ZaRDpm7
z3S1T~5m8a3z$6V62s2UrK(PQZeAF2Ifi7qmXb?;Q8Y2oKpbemw1i)#@0sR*Jm?nt#
z3oZ!2707Dv8W9shU`QDa7$u;J76o87Mvx1|m?;GYXlvngNE0u^t!)B;a3Dq$C=)QH
z03cY94g#FT0O%A*V3Yw|BEUcc20*DW0ww_PUVyh8*6cV2SErY8BbEgQV;pR`aE!f<
zjl%~!wyglv&gk8&0A3b=ZMPgeZ|Lxg20wyY0nm79&PV-GI?m?S-JHAF*0Bqxy}$v0
zL;_&RgRq~-?5Y`<_qkX$Qf2~%)qnsm258ZCCAGMi|2h{VW=)C+#E>Fo_4k8dN*Mwq
zqY$9`WI&~XmEi_U5{}#&1BV&FcAfy_P>970g0*olE#$C)%8*!toELaT4#a<(9xt=F
z`auTp2SJcCs@vo=YLF4c3g$HkVg-l?qiBGlXF!0|82>^iEw4fZ1(0SB?2kJ9Xb=A-
zpWa)hCHOnzQ`v|kbUcSjJik7j;eT>EwniqzLKZqW=n?^!0I5PYUYH>Gmk?=+L?i$pdH{^G@fYpO18D#QfCMTLw4IlS
zi>P|{#?l0WK_}Rd%s;x9HRKN*R2~Jahm9mcEEa5nKvP5;^l;FY@b|CzrC*fkP??f0
zgkiylzsak)s}M#g*0(?ASIIe~Q)YFSKnw%W3pr+}rEnyWqy|(pMt9mN>vo=Kz}a99
zng}Q$i5Pt7>dKmCa)&c{98^17cOJzy=veC2FG+A$8io{bri=i54;9eC_Lm4g9Yi`$
z!Oi*kjalH@SG}LsN?REsAyOzB$ocg=I0OtBp%w@%d*o~!t^f_#tsfFb>!xS`TNGHM
z0dl~snIizxFv5WvRx>iJ1@gHI7z$`uGo(tLi)=wLxCQv2VlVI?hJFHs16X0RZ|L0z
zSUu}te+OP+jBsdhzcH72umCoj+1$#zD6ocLg4iuka@{#L21F5F>Q0a$
z(}%aXew*X3joUU(;;913J|WwU<}1&H=Um5p9Wo(Vr1z*Kl~n47+m)ufO(*#=V-pW^
zj!i^rX1AMq7HlKWalVb9@2M_&+$5x4=@zVc-nwntiuU*1#FiS$?xdsSzM@86C`_R811o*I7q-c8IzaBh82+YUM;_@Ar$XB9+r&3(R
zHtDyDagqb`1+WwPJX)4<6^@Ssqb&AVi_?Yey_=xVo{@gtp*nxN{B-zbq_WzXFaYaS
zVVEjmuVx2FYMLA~!A4x(9J|EZ%z?c6N&au7U};t1>nO7ew4c_uiJ&RjuglQ}3k$-?(uA2jy5p2IG%D>7enXs8@`8e)Pk?`@@S}+ol;8MjT8iux^Y_x4rdY
z;9@vOfM!QDNK+e*`b;Oq0t<~-upWhVR6CRN%ZQMEj~Jn~aE^9us^)tmI59jQ9QvA=
zev|?uZ{7WB#*9^TLzEUXcn?(>H-TxrTtb#xojH&=FR+}anSz^X8DVuCgmZO6{V~v>
z<4~NhJOS#rK1w_dEzXRYDQUww`(sF7PRitRc9-@t*}gU);Ys^X71Uqr;
za5?;J{1_!0MpUsK1~EEU5>b9p2cy*v-+eu5C5Da%Ytp43R3IYqX)4yfCBJEl==_Q{@AoYFHk`yrf&4%9g#C}Ba%x1G`+
zIOy9XW$dgN6%4&EjboqK4U|QjudT7;3f-+SP&+-oPGcAQva`BAEu6pigT=YUSJ!a2)j02~k5SulcesMpYnm+BlUQ
zzAA;aSE)0)v)*56!Ca`VMvYcz-u2khd{%ILwG7XQxBFrl`X>J=e)jW-Jge@;t+?;?
zBi#M7q~juk2u{l%54WCYl_^pR%EBV6sj0IoPV^b
zLxJD;oO#+vJ=(mM&Hw0jR8jn@%c+-j`9A~hl&Sb5t2c(rJ=n+b(&SHJ{iPte_+6Lk
zpq)4!Y0Q&ptEmVqH~-CJOe(Yv=9f*WCzYg_t}#)fm-V!|W;dqvoHpoikR7-x^fQKo
z#ij^oJtp7buVRSBV8KjB@XqgPq>_D}K$3E8eZ`!iNC}`gkmbR6hmA>XlS#gwm*!P+0YzyH(ZO%*K+w4?2zS;}It*Eu)
z^q>wE=dM*WweLggGhy?5E84R$h|@1c2-K(aU;cE0>%h^^b*OMByHV6iMYGdPDX+iz9q{?NH+Kh^;XbgJv>&cC5%V8-ALsyjgS4+
zv*YRE-&y9hdu+ItZ9hjde>~NOf8L$n@f5lJr>cjuiBfrk`AMEFg%2+VCF!(ULNn`t
zC3|;P5Y+$Yx$ezXCT)Ae^5Ngx$&0_=2S@LqF*#xZA*FF|yv3Z&NYj$|XVp!z=NGnI
zk5Z4#K4O@!i@3$qi3c7YCgs<(?OtLNk!NdTv@$xK>x96qE+dd4;p8_m`BxScP+3hH
zs5Q6u^{PHypE7NQm6#@0JJqeCNJ_ipsf2qoO)e`MwCA>iM-cLB&;laq0gv1-m|OHo
zw9Q_lZ0(_0M8iRpd-E=Kkz?Uu_xr#_>f5%MUFF05^^H^NI=`sP3TkIU$Am%Dn1s~Q
z9ul)+^LV5WHSoK~2^e8jPE+ixg&wZVaE2XUZ~poknc4GUQ(f<|ac6eaV<-l5XlmUo
z{YPCr8w^Y*1ldiL;vD3{QsT!5embFNBC4Jq0lPAs&iUl)Lv%emTG_4UK_)$=nx6sM
zFNM9LuTN8@siG}gyiQMYerIQUe?&{U!Ft+6#hYxAzO2vF*!eo#iXtUpaGICerQES9
z3?I3q6qodyNwUn(XFY`T$$@Ty!F^s1bzVxvghJH)zO&LYiLmOnV^MPO+%vJUw1f|9
zed5X&xKs`^B*)XA_*hNfMfg=p9MXUKpH%Ij@3vbT=2RuwM3=U%_iGMLBXyp;aq?aR
z*;Nulz0Xn$>Fwsn7r^4>Mf9~sOyN?sxlAWo$}+$3kkb?dUD{l|^W|fZ-WzEi)jO1S
zI1GA-mn)uYJnv8f)vgT}(^HMue7POZ>s$fa72L=^ux_n8_tVQeg^^6=N=L1tgp_e`
z(!%Vj+Urv*BZK!X=%p+(Yxab5=8e$3%DA;8-Zv3fgsV5($ahgj1?R+8*EWk;<3{!G
z!UyGxw2SL=?~6KMYL;YX%ArT!9bn+QCEFR+*0d>w*p2eS)E1v;ADj*Qs&f0Y5RMBs
zzZ7y^+eQDq010QOynl~)qde|NGYjkP16{$Rj*(l_!ae#AL_8MIVu!`Gg4H^wj-rOf
zUiuo{2aSM%aD$Kep6cmZs9K++;Fr+p{rqZPQC5)7(v{TZFbZISKDypI_TIX8qS9mc
zTN`a&4mf2Z>|5+4GS^x(__>{K8)T*52KjxCD*&G;cqs1j9epve_1%O
z<+DE9V0rB}>$1#zw26BXZIp3nu#KMGr1l5i*Eu~S`$SMdsPlXam0Yc#F+CqFzQguX
z(78=O^Wui1UE=Aj9~3Fq`<#X;@uk8_p=RGG$Xj_)FWbbsw}WWWA3OBSVjAJuTTPdv
zqI)J-DU$;dnBJhSSs6@+3xnfhslEo`d-T;YxyP)vd3gJ{U$0{)|@g61t-*t;zq2?TQCGU=V
z!Ag6p-|Y{35HDspdmPMub_c&n;yC$P8~Ek7MSnv}DK{^XH}JG~B7Vg5K-qkVsWo`*t$B9pdf0cbOYf
zln7C|EGH**@<_*Au;%QY+qND3lu~+D9`>4(!&pz%uBP;eh`9@(l6p8(Y{Wz?yfm_Z
zXAy@;ThIPSzQhkp;iDz>(WO4sJ}K^EwjmOb!Ssgi$q(8IT;wRcL~F~l#Eu2yk|oZ!P_B1+w>W@q_W(bGW7q-w<|
z*q6Xovw=K#?uk7Y*{`TzeSX2uNTtjEr}l;YKeaFH%#8olxBOH8aIpP%7xO>c7bZ>?
zw*M>s`QNlJD&9OQmiY0c*qtVdoy)~;+h?a(wwf{4VFNMiWiyP<4%5?|OVgM~NQFvC
zLQ?8Gl9g(e3DW0M?Xu{E=^{N!H(RPVRd3T3wLWgYKKy6QXVXugKXiL%Z$AF;6FI12
zulvRe4e()yCS0MXU_hBN5b@B`NTS9B>T|C4I%yS1YD2|hB!v#k%n9iK`GE^WFr#`2`K@FENFs;
z)l`|XLxD1CCQ2gSN&*1G$OcZ80DuUS;D9J(pdf}$gAoAKFrW(#I3zKE
z8ZZV!hylo_xL9C5!oi$14AxMI95Q5R#!2!z-Gl2m`rQ5Z5Juj>9NTbzod(yV*s;YR
ze;VBgB4qcYxI&7FaxP|G4aA$eFH}bG0B}kAktmBOxZ|=x2Vi^(#X?3U7!sQR8lr|A
z8?jTV(x)gJnJDE3NE%TZ6(-_w!zBW7k|6*dQu~E8HZU+J$u}d%28H_)IYdG2ACfl0;!m&m|n+G38f!8x7BP45p|^@cj%EsF2D2RI%^bWrr6ki9|_{
zF%1%cArAfn8b>k-D;7teAs*iz;Ea=jA#Vyz$i!U0OZOu?7PS-Sf)UTbaLzb@{{%BD
zoG)XacpDNS4roDyX<&e}0MLUv5Yfj~QPir$4a%rzM5dLLMKCf(1coSaphgBp#+yZI3i9v0Tsv#NRszJrD0&P&I!RGh#U4G9%0I3FjXhn938LPGn(d@r6Mp1K`Iklz
z5W-de&P~TwpAaKM1rs3Jyv)Vo-oZv97%uYvjepXHp!U!PHAF%r!3if+!Y4sIZAQpJ
zXJ0*L?Ba(*WIhNSg)$T<<5seOn|(Eg5NEU%5?UQz38A242P%&pgBW=Pp`b|x0v9I*
zLC(b-%j19pbr*Os6$7JxZvRGSq|EJ6#2Xc#*D$y{h9QUqky5-
zFMkM>7GTJQ05rnLgED4_C_7HO5Jf>D3KI#I03j-h0#YM-XF(Srp+^dR`eR?v!39tW
zMM$b-^`6Wq#k4Aww$Nk0yX;Y}8Fqjm_d_G{DV|cJ8#yD9I3mU-VB|I+&z<>4_fZOb
zFyZQpdV3i~QkwihCg|c?0NmM~!0VW9l>a76Vt0F>7?FMce9~fk0SH4j%scJj#<0Jg
z{9B}~X==IqX@{dk-l-He6?-ZUa$LQ$Y4++b~QQh79U!}`w3
z3{dVmKGoWGpj-Fd8n3KplQ+VvveYWB>$i{g-4ZBLk-uJNu4BIFS1bqm`*+r0h^u1a
zC{I>m&3B@)84bN@GK_2t+d%c
zy&5G!Z>+fmtY{U*entHPb{B7CB`;NBdjxO%?jOi0hHts-naBHC%%vFv
zCSGZqt9(zlmRpn6g`RE3J<6|Rl+ec_nqI=>F0V0hKI!W--d)VozglI&LSkM!+SHHy
zbV=VzLyghlU1oZ$rn!AJmHJ7&^p?(`yUNDDOD=Z7kCHB)%U#-bC^YN{d6g*V*{)R(
zV#wc7fBAi}5?cFbJU1;;xbW!~mtAE>o6*M2;%EBL>y9-m7&Oe;vF(Gsc7h^f`_@dX
zvmYaobIjlB!%qpp!i{|kCHFPIw{zD8t=dj^Y_BLS+yqqZLW5-4#g1-4U%|B*?Q5zy
z(W9yy8mAlBMEjG4t8Hz?&|F1>V4~`1{
zlnvz_`ceER`J34e0{vXh&YEiS&KvgDFrpoNE@Arj&2A_u>_s@Q(zc$umj0#$@4=}N
z0i4r?=bn%FcK*vI_4YZ6+nqb(jNE+F#IO4<4rHmVg;RXAi^H1s6=oQTXnegxi#F@<
zR1vpNc55{dk|?Y3mpN)Fwz~I=GyW}a)ttI)o0?pk=JS|3#r&hKso)=(W?BZ533pHEi*in>GA+83pYKxbYjl#tnbF9_XVnEqq}D2O1sMw^)B}kaJkbR
z|0NL8P0rPC^vLmbKK_pnw#o&wtZvQkji|RZ2jbJO7syNWQ$4?vnw2!^HG0<0{hUDy
z+cf)4WF%K`&8LZ0kpxz(&PHfqz;<=VFm
zr{MBz8r385HCZkj9b>137An5evXO-g1hzjN%}T9=Y!*NLcU^k)tBt+|Y|W(yk~iLO
z&v-vfCF;e8jdG=4u_s$6w$Q%PoZvai)^{|9JmW~MisJqkTtgh~lh1i4??cizakS;K
zN++)}Ya0jL;O{S$=Jm46-|VytYFEKXuFWuXhul!)o6gQ~_>*w#u9WvDZ+B_%;xxXP
zUMKG<&+5SLz5O%MNJ7&E}QtEbF($2pvXD;j?;-Rx`|J#XT9!{Uuw
zYUD3WIOW?^mge!W5}iNas@VJ1g}BzaOgE>Iz*en$%O-MBntI!fIo-=&V4Y{U)L6n#
z3ogrTa^##LGf|1e2gnlldcO&Z@8&kY$V>S7wvYqtgf72Pw#WClSk^Uf$nrn
zF0R(vF?FlvvlXg*3rIII+l-EeHBWvx%P;n1qxgRfjWgwsXn*(NZSNNMH14w%hkRMB
z(7yQi2c+`jtdFJrKv}D!Pt62*46N9cn#sn#hJ(X|fnj#LtcF6;0DU%FMse!o#Ko-7
z=y78+P3^nh^jEMqxC>p$*f*WY8Rk*+9~222F0m8$s)D^VX3Ui_Q1k)P#QvB9=QF&0XG%^Vc5W
zsikt}JcN)oqDj#|ufpqlx>6iDstCB{m@Hh5uxiJDe@(?XdAVh7}5>(uX+tPS~GPjn`LWniHb&Y
ztGCOVu|M;<`|)|-?ezQ#F
zGC#fn?}M$$O|xcN#>agt-Gp3bK2pcrW>wxot&mD9Is5#9uO(Ta{ZGlC^FJkjHqQT<
z{F&JQ`}F@Me`W^G|9$fJR6${?1VH<
z%H&`MB#PeGqAofNGqP^wqt-|*uLvNDPAhh;l}N?YizbLV40088^mY$YJ81FmO=iyQ
z?0f0j`}SKqfv$|9R!W>;$RWX~OCW~;VMKx)Eg6^jFi{ehKbQ@}Duyc#NI0WKaL~~a
zqKshVkPI<|fdLWXCj%WX2n_r%kpM(N>URW0Ob9@EtQKg3lpqM6kOiXvvCHrzB()67
zj#(Iog@lz8a8E&!!jeI$lQdk%O77HYL(0Y-1YXEck}g}&!Z8Se5_uxwL69yZB!=6T
z03sPektqi&!5TN@Q5YX?HYG56r3qCzofQr>6$=h5*)@broj^s@BalI%MhC_?10_OX
z;{2o!2N|q?0BUqPHvcRTVACZ^i#~QPH6qO40Y(T?kyzl^YQtw!4}tLs2vb@?sbpsd
zu%m=AhB$s&T3qzNBA&UIkQ!n>pZn;08%IxW#lcf+f~uG*&U82RFq44$<)-)@(t&4k;M-u0
z0{{j<1InnE-EXX#oivycT8pKam|{o(D;RTOnH^X~=;gp*Nrwg$#xPJH8H?|sil4))
zbR-EDi3blcs>c&8lR%ua5|iBN-W~6s&Imsp$AzhhB5@%r2(pIbOEXE}*!|{bUae7Y
zS1gSfd7Zz$4Zp^ZK6lJMvp;+vo(iC~Ytu+xFzWK-%Y`3ZD3Fd|no7}6^t@7l2{RKC
zv2a0j2;t`$H08V3ix0?U!%bl6%KmdA5b{sC;NvURC`k)sfhB)W(?w)lgyZ3YN8{d$
zI-ev$GUojfkU2MGC?rwk%mdh$Xv7pTAd3f}9Nt_(GlfX}rOL2NL)d?UND`#%H!x`m
z$SUEuW`zFs0o0FFmcmtOIva1Wlhfx>a?_c)MD_vrdyB_hZwe!7b%3K2YWEN?gkS+B
zpHD8Zh)Bf!(esEAvHEw?
z$HVL9kFfkQQx+aC)k&)cHPFNr?5+@d3}&NZv-7zt)H7CWgVs#8aCK_htoR;=9zi!a
z8Fb_L=Oh?-8p)%Iv{WdXGgTd?hMyH~qVyeoFPc@vw^N$92;V{;ea*V^`)2Q
zMcV588u$H;I$uL5&iBjo2k7r2lm4DA9w{qm|HE-2~N%XDHZ@-36A72W<
z6Bfj`_@>GtP;$P-T2mKT(okuD`!tl;Ys*8p)!{(IKP8ivkZe1?XjXcq*4S{iYEa>d
z*TiRa)6vSG#h|y$Szu(=lb0Obm4uBT!>Fg|eVT}J<^0Z8I{B5PAbPbHa#9wE$==%B
zI~}elN7y9eJqg|$kg^({gwz2f!7RD!Sh?wKND0if&5=Du@2Q@?x~_gLNL{*WXYWuxVMCErlLTsEUAsD;Aw
z^~w8W3OL$GC-o4NQzZ((!$lrs|{f$B8&l3@42
z-sCw-d|)NrB!yCw2>^C*irSaMM9pa75ErNbDK92EH_@Z1|*?u>p2*(FTU0m5!6;@(v4(+OdVgj2ER*a
zf^u}nqkdFc8Y-q|z4r6H%BRg}-pDsYGYZE&fI@jmE(0xdCoawETu>*SY2L`OP1Cj#
zU(|Rxgo@9dW{56TrN`S2;uT26)x`MOSJ_GxZ6BA!_w)?>$;44B4;0?s3%|@CmI^g*
zh0K{r3-3;|6gbVtly#4m9VH?R_p3OJSTEHG?d!Sk86AEJpF#{Wi6$CeGJIWR><(a@
zTvcq>VUS@kfkZtp?Yie%Pt!&B*QG7u*6
z%gxkg`@3;`&FgWgAJfk%yehwX3bLv++1Q9=V$x^!%#2)T>1WDYJTjaL^OT)d`6f6N032_oP}p!f^tmOC;;?i1seD>?K8`i%TeD>~5;H
zQZf0Xh4CsTFoIZrQd@!u8Ki>g@W{16ZH2@$-EJ0A!Kv9-Q3^BsLcL=QI
zRe_Wv+)wP4v99L?>L-+#Tsr3}SSF0#&jCW2QhD989M*C#Wy4x^-2n`T#zXGIMSM%e
zA<|lKY{I&c#CC?(!zB^99h@va7v*>iesIll;L6RpoxgTEZYh9|a1TT{OJrgjEEj!v
z-%dVVBxgg^TkK;Bux>OvgNm>Ibm|lVZ~4VkK~%Wgm&qBv
zFV`-hUHWy^IZ8uIk;-J2X75BA!sQ+jfy}6)+Pv%TDjfMNu7LjBO!+`s!z=@)yJ(yM
zarw;fV4`NO42Hhm5RtPb56#+nV;QZ}mH1s;f=I$2`EONzi!8Fbeh%!Uq!
z!lK^>%XIyjRzMZFlR3qO;E}_Qo)I;=N1q-nRdtHoV@f$>6_8=Gn3=F2~{u>qi`L#u};O_`TJkHWpEvY>-EL1<6!7=u6ATol%%vj>N%8XA7_Us%#|
zHdULo@O%H2e0SKw-s7?!UtxUb
za?y>3S+PT*=(>Mhx;0C6f&vL$$9^Q5UdAs&0m~qvL)nqT%=0}$Ep+nUiibBg;FqIx
z*OE-ifxfJ$o8#znN*TQ;g5xU136Xv96n_xPF3P|S7FA-N;&Y&WM520gksYZ5vUiYe
zW%{`#smuLOJwZ^K+^(8px@~q__l;V>XC-U}!{4fCa-EL=+NUtF3Ili^(!b?&WZmOmX=#GvIu^y$C}a)~Y4B*lF+90-
zJ7^6h+uG}|g#Za!XmOMfyS%mz87#JXuJ(`D34DfZ8hj5p#0DIPB!4ybAI*nE$Y*LI
z*KL;D9b*RxA*!h7lPUvm6q7$J9Jt#DiGy;+RECp=ZOIq-h({;G9u-#UX67+az8B1v
z4@x0h#%jyrBBlZ6c(2_|g>@9eVfBhkuallb@{O?g`eHY6xMbYY*;amIsA?-80i9OW
z&Q+~=2B`VLqg>xV~2LzR4jZ^;8^Lh)xt-I827
zWcO|{`K^rDjSppq3Xl6jwT_~fLiD-XG^Wm#lrG3K3v55O=*zaZZ58M|rhB1|*d&H8
z5!s}$^Lxf>bD&#$ioLKahAkOAE-+^hgCaR)9&I({Ksj9OX6W9&YWVB$&d@!BBm?Lt
zEn+NusK3o#A`DreRc}C7%b0(GUf$MEP*-&w)ec>sQIaHu3c{4!JH%9#-;?Eej9}3^
zm7O1G4n3o+td1rm0nY~AX8Z5v66d0qEM=u+E5qB8f?(^-nI}*9gNFI1ao%!He4oeY
zEZV^I>+==CdTD60ZJd6V;<#Y_NOK`^1yA1s?gAjuf>{+I2N*E$t0^@Q;#EMn}skOz0PP14ik&iwG-Sa$0Zv>P=e(lq5z_$-)
zS=}}T1~oZ`v`}Dy92E}oA>ChzDYkvLL1ig4
zOv(3rW9ZAuv6d2h=&2r42SE!EGt0eQm9lbP6hFjT6bV)4Lo}#U_Ld)1VZKt=V7EaI
z&czZnY~({eylR-7ZpRRi%MEo4%Xh-tsLra4~*rp;5Rm{U2XD1RLsK5;gC
zWL7kr|G+`i{s$s!mgbPcpAuMH_1VYp=zaeGwl{*?ai?5%P}+3b-`Zy?sup9)ZwXVG
zOz@KzjcjQZ^NtOJIrCV1Y&YI-)=`J_sbBng=4aVHV+Z0c;-;5ow19jZG9yf>x1kw8^?iYy@gO$t2eMBF4@XW0n-^+D;#ZSLN8
zIt|KZ(0b#QX`!>wl8@Pn)AluXV4?O?#=GrN+4$H?+)~rCkuXK|QL_b;LpZX~PERnJi(FzDEQ!R8FCACs@J@ik>kO`<%bClD87EWOF@%Hye2nR4Rf4fJb^ZeI`$_Z+1Rdp6IcTT{
z3Y7%KqifTnsxecskqQ91clEcWcAbef?q^=1syBAlv_2YWbhm|SN3zG%Xe+k@CHuXL
z2VSfhiLL9J0qlH_w&zCo(}Kh#gna9>VVMyK)Q+2BLrkx7H$tA5-`6#}!vP6-UW>8O
zd!kQTSZvY)#OZ=%lGH^vfFPn!_lpz1f!_V_ZZ{4u4(>1WGa@LG;eSr}fgX}RHZ95y
z!g=YR&6*Hj(w=BKhd;T~`aZQ=RCDEaNDSUFcetk9?(}zem+im0x!VEx3FPWLs(6V=
z)@?Fm{YHDCfTx#;h6Y>1vxSgfX{V~z$14)YF1zfHXzY5y
zQkg@l?tiwQ>N1C7p9~*E2%+rUv8wJ&)g!ia~EF&Iov>o!%3l5V8*O79E*F-lPkiv
zyhd2}^Zw25bC^e?DxkMVb1}7JYW&k>l|VBdsu|a|Eg>HU#A9={5^oG2T(o3ShBGHz
zf}$0|Iit~-lWB-%@qN>cLy!d3r`7+wD7jv`hs^VuzcUqC3E<_u|11!#E0$e$wmmY
zDINOg#v({L_;`%({-qPepU8-2-n$x)eES*5xP*AS`*X(L&uFGJ
z8vUuKtzVXDN4uTJi*>alF7zF~t*;N;$Bj_+4t^_=2chLsbvJ3gE9?EuWZV>owa7AL
zj^UL`XWiI8MVjY=UQbPROe-LwqDZ$`ZlDY36*@=RnqjY7eHQQfH{_S*z^bq`lQv$6
z>tCoLa+4t^6F6#otA{cdUld|DQ=#>+@f?oZ`KzYWDWqP2V{Xgz-cx%74-Cp1Oeas~
zcTd@3Tnw_z6UCpW&C8%khUtu&)7W{oJYP)X3X!e5BC=>MfLx?K;tM{$lIdu#hP@G;
z>yn_W8upeCo6Ieq|3RCSS6k7rxj$JAcSS&`;IDYwQ=uxg1&dfd!=X9^&Fd!;0ef12
z7ULW2fJU+|TxJ2HvkqH_Sz>>i#6czN
zqC5K}tG4^n0*L2Yy%GkwM-^y)xo8XD)ej{-rFkl1d0Rs&M-nBl*=jdoN#7GH4{Anb
zm9bKb2Bg0=TAoj{mip0Pp(n
z8<3w^Uco|#2Ql`O{6*=dIVSfyfqm+|Ek(I4hA_yHSG<+B7V|e?bXMu<%B>%aKx}C1
zM;Wk<;ukOT`wTcvPfG(qgx>s;h8LeOCCBcroRUC)lILczRk1e49(G~}#hUD26*{Zw
z8L~Nnt%1EyQ=hu?Td3`0r|N4xI%DgYBz2aq4i4zNbl&L*?U^u|(ED4HSc`uY5))&*
z7q2^wXA>2c6nn(6^6q0!xBYA12nXMFc}4h3cZB;hUCH*Q
zRz>7Xn)6@6;H=_rk+{mch(31lCN@Ajkrj!NFKbM($%d8FGe;FJ-I0CoTLaw3m=O#+
z3C|`{yRtN|vXD+9&92|m1U`WzI|w-p>vARqBBF4M9A8W(2`sM~9=TSx43v9|%dCFN
zTj8xfp|3gkJ&ZFqYS>J#9;)|`5{0EC8E*<;weBWuhJo)|!1j6IBLSynubrmc)YgW~
zCK7|P!fhQu;<;%3*yFSM?C5E!v!48Vh*g^7Nm*%gS?O+*G{Bkp_uRI_Un?QKY1WdY
zD5L1au)EYSvyv@enESjIi~q!!*nlkmhA{~{J6S*+XlR+$ES+pj-|ni;##W{#PH!#I
zw+cz8w`X!7HB&88SIal3$>I%>a&Ry;BeesWScA+>X_=KAAm5x#OdV+KzL~v=F|smo
zurRT5(lX1My1POgzB$s+BK_-4mJqPmTL~s9jTjFb3oAD(D+?DZ2P+GJg`S0#nuUe>
zO->&2?Y|RIa{$@dnSOh#=LFd}nj$ePsflYbNjTfs7=yszw>nQ6H495e(zokh_AroY
zm^wJVO-;(q^fnuSn~f8|$i~F^f9!mdfBQB$OBYkpH=_Wwq}FyICksbY5UI<5>2Wf#
zF|pGAQ!;7_{`L1yK(xX0
zKUjQotp}-u9e%k&X$^}uY2S|X-A{56xL8%-LWveea@Kv!X@>WDRpP`y2lq|dJ#70*
zP?#2cF&|+E!N2^3H^M;qd@&dihl~;hH=XVtfC~*oi$ZMke^14Mi6r+P143jTMW$_o
z7bT#*vklt=_a>1hv
z;TQAVpsv^oV|cx4G1Dy(HS
zcu{wr`7l7{MQJs}!`zX7yl)$;8iVn>-iJFQF(A$&$^Q05hS=tsT=a&Z
zQT97NyR}GF#Z>#SZ8=*KJ20#+8zy7F
z2!GJY`Qez2e&l@;>JW&|=9ih9B!T6@v%62b+3N@X2|OkgII!rZanGkgn~T>MU;kyQ
zOvt-XvrpaPgPikbf4h^V1UUUq2!x9{gZ=ibo=vI)&v-=fl7HjNf6R2^PhA_EIGhdH
zojs2CXepH}r$xLw%FbMU{wg6yWh-DjF!MLUEioHG0eV-&-%8sC`uj!_ybmqBZ_U8=l{uH*(56EO_DHUstMQB3)$)A(nK)OJZAZt}E
zu8!6IGo@ubRvd+`O}d!$(gH9vuCDs>gko%dBtwtmXrK%sqR+k>7}qC_q!R~}7MH~s
z?05K@943q{zp%KfWmb2)6j>legnJw`t;UKqpu`HJLj5SZN8u^8a$vYIp_QR1>=+Wrgh)hl3ba}(`Qk-vSJvv+x$YZ#HMeXayBM%nIu|B6i
zV7o1c6h}|Gf5RHaK`u^UK(=xyFhX!+^e!igx)Y$C9#u&B-^6YoQU%duj?*vrj>;E1c8!_mi{{TvuryrE?oTONL-Z;?C4&)czi$~P;8==mV(pDGyC~K2RAWlZR7@na-4-g~z
zjO*@PX4f@DskRvtD_bn3m66;?1_~FCZfJW^57aVV_pu`SOu}7HrEhDH#zQEAWH>du
z+hR2}i#?Mcd;SZYE8T|BC9Rc3tsP(0@%<~v`{msf?bk+-YsO&T?-oO=tsmZhLs_F+
zUZ17bjt@>PI#rk0>FTIeEk2<*-QSummwDdkk_l)zPo{|pUq55M;MrDA7%Wdl9tl#O
zuDn}%u=Nx6L<+dOd4F1hyHv}xP*gc@Z&_zjXAY`T98;&Qu`OEE5U8vFJobBy`l#~z
zgLbsfO$hi>xr3{ss1cf0`be?rGsb!QuHhN;dnPf{FY|54@#3YZH?QHu{qg0q%_S4^
zpc?KUK^0M^#s8m={!bsHrDAG^#4HW|X6i<&%R^IzvrTr6z=@@@Zve1FOWS%|zcUhsY6xVHIx0j@JdU3PM=
zpjh`
z)2v&8Md>&SsAb#%i^V%h4Kxhs^HWmHXk46TCL0=qw2XI(KOuEtcRhBW`u9k!K1l5b
z&-ku1yA6C*>bi_@cic(%__mmlkP#OfcR8%JFe?pcB)tIc+Q*olrq*dIgL&T1p^DfB
zhTzL?YZ}lqfbtvz1#4jex#9b1cF)uzx}fqf`AbowADt3vDP#NePV~%oA|vYJ_pM#v
zDufw7xa{sJ&s85A<^_bi0J8~;!{<>~RDaR$D(?#5nRjV`AEE*<+p^Xt#3zgodnrlm
zFw)-@b$^*a`t9$y!Dj)7+eK>8==4&JdH|D={mChSc;eF^Awj&VoBoKbCiq|mZ-?pR
zUL*Jje^!>yR4}E1(z3V$e|nR07YoXw4Vz
zi{6M>X|%wH$;_5IPP8QFE?cQ~@_M-wau!tc$h_(U*ncGeG>u+o_2a!Ppr0&qOt@r-
zv4KBJVSp4j=zl);1~T{HD%V&s81n3jPD&n-|1;DACR8T6(pEMSYtMld@*8vdib8Ga%DK=wxG@o{qo{1|5Jd-)mK*mg5x`51Jc
z*}{MUQFex?1?atH}R?1tY^
zeil%K4ke!T#IHX&byPG-@-DsQ052LR79m2Dbfsf`$e$(tM6NNHiggggYLHh*yn-jJ
z@@J*W{?%doR|j2~Y&}4#^G!NP0x1#;pO!<#qP+2Be7!HMr*?-#B%)FIKewr)6Uf2I
W?JXWg0 0 and bestAttrId in attributeIds:
+					toRemove = attributeIds.index(bestAttrId)
+					attributeIds.pop(toRemove)
+				child.next = self.id3Recv(
+					childSampleIds, attributeIds, child.next)
+		return root
+
+	def printTree(self):
+		if self.root:
+			roots = deque()
+			roots.append(self.root)
+			while len(roots) > 0:
+				root = roots.popleft()
+				print(root.value)
+				if root.childs:
+					for child in root.childs:
+						print('({})'.format(child.value))
+						roots.append(child.next)
+				elif root.next:
+					print(root.next)
+
+
+def test():
+	f = open('DataFiles/rideclass.csv')
+	attributes = f.readline().split(',')
+	attributes = attributes[1:len(attributes)-1]
+	print(attributes)
+	sample = f.readlines()
+	f.close()
+	for i in range(len(sample)):
+		sample[i] = re.sub('\d+,', '', sample[i])
+		sample[i] = sample[i].strip().split(',')
+	labels = []
+	for s in sample:
+		labels.append(s.pop())
+	# print(sample)
+	# print(labels)
+	decisionTree = DecisionTree(sample, attributes, labels)
+	print("System entropy {}".format(decisionTree.entropy))
+	decisionTree.id3()
+	decisionTree.printTree()
+
+
+if __name__ == '__main__':
+	test()
+\end{minted}
+
+% !split
+\subsection*{Cancer Data again now with Decision Trees and other Methods}
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import  train_test_split 
+from sklearn.datasets import load_breast_cancer
+from sklearn.svm import SVC
+from sklearn.linear_model import LogisticRegression
+from sklearn.tree import DecisionTreeClassifier
+
+# Load the data
+cancer = load_breast_cancer()
+
+X_train, X_test, y_train, y_test = train_test_split(cancer.data,cancer.target,random_state=0)
+print(X_train.shape)
+print(X_test.shape)
+# Logistic Regression
+logreg = LogisticRegression(solver='lbfgs')
+logreg.fit(X_train, y_train)
+print("Test set accuracy with Logistic Regression: {:.2f}".format(logreg.score(X_test,y_test)))
+# Support vector machine
+svm = SVC(gamma='auto', C=100)
+svm.fit(X_train, y_train)
+print("Test set accuracy with SVM: {:.2f}".format(svm.score(X_test,y_test)))
+# Decision Trees
+deep_tree_clf = DecisionTreeClassifier(max_depth=None)
+deep_tree_clf.fit(X_train, y_train)
+print("Test set accuracy with Decision Trees: {:.2f}".format(deep_tree_clf.score(X_test,y_test)))
+#now scale the data
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+# Logistic Regression
+logreg.fit(X_train_scaled, y_train)
+print("Test set accuracy Logistic Regression with scaled data: {:.2f}".format(logreg.score(X_test_scaled,y_test)))
+# Support Vector Machine
+svm.fit(X_train_scaled, y_train)
+print("Test set accuracy SVM with scaled data: {:.2f}".format(logreg.score(X_test_scaled,y_test)))
+# Decision Trees
+deep_tree_clf.fit(X_train_scaled, y_train)
+print("Test set accuracy with Decision Trees and scaled data: {:.2f}".format(deep_tree_clf.score(X_test_scaled,y_test)))
+
+\end{minted}
+
+
+% !split
+\subsection*{Another example, the moons again}
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+from __future__ import division, print_function, unicode_literals
+
+# Common imports
+import numpy as np
+import os
+
+# to make this notebook's output stable across runs
+np.random.seed(42)
+
+# To plot pretty figures
+import matplotlib
+import matplotlib.pyplot as plt
+from matplotlib.colors import ListedColormap
+plt.rcParams['axes.labelsize'] = 14
+plt.rcParams['xtick.labelsize'] = 12
+plt.rcParams['ytick.labelsize'] = 12
+
+
+from sklearn.svm import SVC
+from sklearn import datasets
+from sklearn.tree import DecisionTreeClassifier
+from sklearn.datasets import make_moons
+from sklearn.tree import export_graphviz
+
+Xm, ym = make_moons(n_samples=100, noise=0.25, random_state=53)
+
+deep_tree_clf1 = DecisionTreeClassifier(random_state=42)
+deep_tree_clf2 = DecisionTreeClassifier(min_samples_leaf=4, random_state=42)
+deep_tree_clf1.fit(Xm, ym)
+deep_tree_clf2.fit(Xm, ym)
+
+
+def plot_decision_boundary(clf, X, y, axes=[0, 7.5, 0, 3], iris=True, legend=False, plot_training=True):
+    x1s = np.linspace(axes[0], axes[1], 100)
+    x2s = np.linspace(axes[2], axes[3], 100)
+    x1, x2 = np.meshgrid(x1s, x2s)
+    X_new = np.c_[x1.ravel(), x2.ravel()]
+    y_pred = clf.predict(X_new).reshape(x1.shape)
+    custom_cmap = ListedColormap(['#fafab0','#9898ff','#a0faa0'])
+    plt.contourf(x1, x2, y_pred, alpha=0.3, cmap=custom_cmap)
+    if not iris:
+        custom_cmap2 = ListedColormap(['#7d7d58','#4c4c7f','#507d50'])
+        plt.contour(x1, x2, y_pred, cmap=custom_cmap2, alpha=0.8)
+    if plot_training:
+        plt.plot(X[:, 0][y==0], X[:, 1][y==0], "yo", label="Iris-Setosa")
+        plt.plot(X[:, 0][y==1], X[:, 1][y==1], "bs", label="Iris-Versicolor")
+        plt.plot(X[:, 0][y==2], X[:, 1][y==2], "g^", label="Iris-Virginica")
+        plt.axis(axes)
+    if iris:
+        plt.xlabel("Petal length", fontsize=14)
+        plt.ylabel("Petal width", fontsize=14)
+    else:
+        plt.xlabel(r"$x_1$", fontsize=18)
+        plt.ylabel(r"$x_2$", fontsize=18, rotation=0)
+    if legend:
+        plt.legend(loc="lower right", fontsize=14)
+plt.figure(figsize=(11, 4))
+plt.subplot(121)
+plot_decision_boundary(deep_tree_clf1, Xm, ym, axes=[-1.5, 2.5, -1, 1.5], iris=False)
+plt.title("No restrictions", fontsize=16)
+plt.subplot(122)
+plot_decision_boundary(deep_tree_clf2, Xm, ym, axes=[-1.5, 2.5, -1, 1.5], iris=False)
+plt.title("min_samples_leaf = {}".format(deep_tree_clf2.min_samples_leaf), fontsize=14)
+plt.show()
+
+\end{minted}
+
+% !split
+\subsection*{Playing around with regions}
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+np.random.seed(6)
+Xs = np.random.rand(100, 2) - 0.5
+ys = (Xs[:, 0] > 0).astype(np.float32) * 2
+
+angle = np.pi/4
+rotation_matrix = np.array([[np.cos(angle), -np.sin(angle)], [np.sin(angle), np.cos(angle)]])
+Xsr = Xs.dot(rotation_matrix)
+
+tree_clf_s = DecisionTreeClassifier(random_state=42)
+tree_clf_s.fit(Xs, ys)
+tree_clf_sr = DecisionTreeClassifier(random_state=42)
+tree_clf_sr.fit(Xsr, ys)
+
+plt.figure(figsize=(11, 4))
+plt.subplot(121)
+plot_decision_boundary(tree_clf_s, Xs, ys, axes=[-0.7, 0.7, -0.7, 0.7], iris=False)
+plt.subplot(122)
+plot_decision_boundary(tree_clf_sr, Xsr, ys, axes=[-0.7, 0.7, -0.7, 0.7], iris=False)
+
+plt.show()
+\end{minted}
+
+% !split
+\subsection*{Regression trees}
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+# Quadratic training set + noise
+np.random.seed(42)
+m = 200
+X = np.random.rand(m, 1)
+y = 4 * (X - 0.5) ** 2
+y = y + np.random.randn(m, 1) / 10
+\end{minted}
+
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+from sklearn.tree import DecisionTreeRegressor
+
+tree_reg = DecisionTreeRegressor(max_depth=2, random_state=42)
+tree_reg.fit(X, y)
+\end{minted}
+
+% !split
+\subsection*{Final regressor code}
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+from sklearn.tree import DecisionTreeRegressor
+
+tree_reg1 = DecisionTreeRegressor(random_state=42, max_depth=2)
+tree_reg2 = DecisionTreeRegressor(random_state=42, max_depth=3)
+tree_reg1.fit(X, y)
+tree_reg2.fit(X, y)
+
+def plot_regression_predictions(tree_reg, X, y, axes=[0, 1, -0.2, 1], ylabel="$y$"):
+    x1 = np.linspace(axes[0], axes[1], 500).reshape(-1, 1)
+    y_pred = tree_reg.predict(x1)
+    plt.axis(axes)
+    plt.xlabel("$x_1$", fontsize=18)
+    if ylabel:
+        plt.ylabel(ylabel, fontsize=18, rotation=0)
+    plt.plot(X, y, "b.")
+    plt.plot(x1, y_pred, "r.-", linewidth=2, label=r"$\hat{y}$")
+
+plt.figure(figsize=(11, 4))
+plt.subplot(121)
+plot_regression_predictions(tree_reg1, X, y)
+for split, style in ((0.1973, "k-"), (0.0917, "k--"), (0.7718, "k--")):
+    plt.plot([split, split], [-0.2, 1], style, linewidth=2)
+plt.text(0.21, 0.65, "Depth=0", fontsize=15)
+plt.text(0.01, 0.2, "Depth=1", fontsize=13)
+plt.text(0.65, 0.8, "Depth=1", fontsize=13)
+plt.legend(loc="upper center", fontsize=18)
+plt.title("max_depth=2", fontsize=14)
+
+plt.subplot(122)
+plot_regression_predictions(tree_reg2, X, y, ylabel=None)
+for split, style in ((0.1973, "k-"), (0.0917, "k--"), (0.7718, "k--")):
+    plt.plot([split, split], [-0.2, 1], style, linewidth=2)
+for split in (0.0458, 0.1298, 0.2873, 0.9040):
+    plt.plot([split, split], [-0.2, 1], "k:", linewidth=1)
+plt.text(0.3, 0.5, "Depth=2", fontsize=13)
+plt.title("max_depth=3", fontsize=14)
+
+plt.show()
+\end{minted}
+
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+tree_reg1 = DecisionTreeRegressor(random_state=42)
+tree_reg2 = DecisionTreeRegressor(random_state=42, min_samples_leaf=10)
+tree_reg1.fit(X, y)
+tree_reg2.fit(X, y)
+
+x1 = np.linspace(0, 1, 500).reshape(-1, 1)
+y_pred1 = tree_reg1.predict(x1)
+y_pred2 = tree_reg2.predict(x1)
+
+plt.figure(figsize=(11, 4))
+
+plt.subplot(121)
+plt.plot(X, y, "b.")
+plt.plot(x1, y_pred1, "r.-", linewidth=2, label=r"$\hat{y}$")
+plt.axis([0, 1, -0.2, 1.1])
+plt.xlabel("$x_1$", fontsize=18)
+plt.ylabel("$y$", fontsize=18, rotation=0)
+plt.legend(loc="upper center", fontsize=18)
+plt.title("No restrictions", fontsize=14)
+
+plt.subplot(122)
+plt.plot(X, y, "b.")
+plt.plot(x1, y_pred2, "r.-", linewidth=2, label=r"$\hat{y}$")
+plt.axis([0, 1, -0.2, 1.1])
+plt.xlabel("$x_1$", fontsize=18)
+plt.title("min_samples_leaf={}".format(tree_reg2.min_samples_leaf), fontsize=14)
+
+plt.show()
+\end{minted}
+
+
+
+% !split
+\subsection*{Pros and cons of trees, pros}
+
+\begin{itemize}
+\item White box, easy to interpret model. Some people believe that decision trees more closely mirror human decision-making than do the regression and classification approaches discussed earlier (think of support vector machines)
+
+\item Trees are very easy to explain to people. In fact, they are even easier to explain than linear regression!
+
+\item No feature normalization needed
+
+\item Tree models can handle both continuous and categorical data (Classification and Regression Trees)
+
+\item Can model nonlinear relationships
+
+\item Can model interactions between the different descriptive features
+
+\item Trees can be displayed graphically, and are easily interpreted even by a non-expert (especially if they are small)
+\end{itemize}
+
+\noindent
+% !split
+\subsection*{Disadvantages}
+
+\begin{itemize}
+\item Unfortunately, trees generally do not have the same level of predictive accuracy as some of the other regression and classification approaches
+
+\item If continuous features are used the tree may become quite large and hence less interpretable
+
+\item Decision trees are prone to overfit the training data and hence do not well generalize the data if no stopping criteria or improvements like pruning, boosting or bagging are implemented
+
+\item Small changes in the data may lead to a completely different tree. This issue can be addressed by using ensemble methods like bagging, boosting or random forests
+
+\item Unbalanced datasets where some target feature values occur much more frequently than others may lead to biased trees since the frequently occurring feature values are preferred over the less frequently occurring ones. 
+
+\item If the number of features is relatively large (high dimensional) and the number of instances is relatively low, the tree might overfit the data
+
+\item Features with many levels may be preferred over features with less levels since for them it is \emph{more easy} to split the dataset such that the sub datasets only contain pure target feature values. This issue can be addressed by preferring for instance the information gain ratio as splitting criteria over information gain
+\end{itemize}
+
+\noindent
+However, by aggregating many decision trees, using methods like bagging, random forests, and boosting, the predictive performance of trees can be substantially improved. 
+
+
+% !split
+\subsection*{Ensemble Methods: From a Single Tree to Many Trees and Extreme Boosting, Meet the Jungle of Methods}
+
+As stated above and seen in many of the examples discussed here about
+a single decision tree, we often end up overfitting our training
+data. This normally means that we have a high variance. Can we reduce
+the variance of a statistical learning method?
+
+This leads us to a set of different methods that can combine different
+machine learning algorithms or just use one of them to construct forests and jungles of trees, homogeneous ones or heterogenous ones. These methods are recognized by different names which we will try to explain here. These are
+\begin{enumerate}
+\item Voting classifiers
+
+\item Bagging and Pasting
+
+\item Random forests
+
+\item Boosting methods, from adaptive to Extreme Gradient Boosting (XGBoost)
+\end{enumerate}
+
+\noindent
+We discuss these methods here.
+
+
+% !split
+\subsection*{An Overview of Ensemble Methods}
+
+
+
+\vspace{6mm}
+
+% inline figure
+\centerline{\includegraphics[width=0.8\linewidth]{DataFiles/ensembleoverview.png}}
+
+\vspace{6mm}
+
+
+
+
+
+% !split
+\subsection*{Bagging}
+
+The \textbf{plain} decision trees suffer from high
+variance. This means that if we split the training data into two parts
+at random, and fit a decision tree to both halves, the results that we
+get could be quite different. In contrast, a procedure with low
+variance will yield similar results if applied repeatedly to distinct
+data sets; linear regression tends to have low variance, if the ratio
+of $n$ to $p$ is moderately large. 
+
+\textbf{Bootstrap aggregation}, or just \textbf{bagging}, is a
+general-purpose procedure for reducing the variance of a statistical
+learning method. 
+
+
+% !split
+\subsection*{More bagging}
+
+Bagging typically results in improved accuracy
+over prediction using a single tree. Unfortunately, however, it can be
+difficult to interpret the resulting model. Recall that one of the
+advantages of decision trees is the attractive and easily interpreted
+diagram that results.
+
+However, when we bag a large number of trees, it is no longer
+possible to represent the resulting statistical learning procedure
+using a single tree, and it is no longer clear which variables are
+most important to the procedure. Thus, bagging improves prediction
+accuracy at the expense of interpretability.  Although the collection
+of bagged trees is much more difficult to interpret than a single
+tree, one can obtain an overall summary of the importance of each
+predictor using the MSE (for bagging regression trees) or the Gini
+index (for bagging classification trees). In the case of bagging
+regression trees, we can record the total amount that the MSE is
+decreased due to splits over a given predictor, averaged over all $B$ possible
+trees. A large value indicates an important predictor. Similarly, in
+the context of bagging classification trees, we can add up the total
+amount that the Gini index  is decreased by splits over a given
+predictor, averaged over all $B$ trees.
+
+% !split
+\subsection*{Simple Voting Example, head or tail}
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+heads_proba = 0.51
+coin_tosses = (np.random.rand(10000, 10) < heads_proba).astype(np.int32)
+cumulative_heads_ratio = np.cumsum(coin_tosses, axis=0) / np.arange(1, 10001).reshape(-1, 1)
+plt.figure(figsize=(8,3.5))
+plt.plot(cumulative_heads_ratio)
+plt.plot([0, 10000], [0.51, 0.51], "k--", linewidth=2, label="51%")
+plt.plot([0, 10000], [0.5, 0.5], "k-", label="50%")
+plt.xlabel("Number of coin tosses")
+plt.ylabel("Heads ratio")
+plt.legend(loc="lower right")
+plt.axis([0, 10000, 0.42, 0.58])
+plt.show()
+
+\end{minted}
+
+% !split
+\subsection*{Using the Voting Classifier}
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+from sklearn.model_selection import train_test_split
+from sklearn.datasets import make_moons
+
+X, y = make_moons(n_samples=500, noise=0.30, random_state=42)
+X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=42)
+
+from sklearn.ensemble import RandomForestClassifier
+from sklearn.ensemble import VotingClassifier
+from sklearn.linear_model import LogisticRegression
+from sklearn.svm import SVC
+
+log_clf = LogisticRegression(solver="liblinear", random_state=42)
+rnd_clf = RandomForestClassifier(n_estimators=10, random_state=42)
+svm_clf = SVC(gamma="auto", random_state=42)
+
+voting_clf = VotingClassifier(
+    estimators=[('lr', log_clf), ('rf', rnd_clf), ('svc', svm_clf)],
+    voting='hard')
+
+voting_clf.fit(X_train, y_train)
+
+from sklearn.metrics import accuracy_score
+
+for clf in (log_clf, rnd_clf, svm_clf, voting_clf):
+    clf.fit(X_train, y_train)
+    y_pred = clf.predict(X_test)
+    print(clf.__class__.__name__, accuracy_score(y_test, y_pred))
+
+log_clf = LogisticRegression(solver="liblinear", random_state=42)
+rnd_clf = RandomForestClassifier(n_estimators=10, random_state=42)
+svm_clf = SVC(gamma="auto", probability=True, random_state=42)
+
+voting_clf = VotingClassifier(
+    estimators=[('lr', log_clf), ('rf', rnd_clf), ('svc', svm_clf)],
+    voting='soft')
+voting_clf.fit(X_train, y_train)
+
+from sklearn.metrics import accuracy_score
+
+for clf in (log_clf, rnd_clf, svm_clf, voting_clf):
+    clf.fit(X_train, y_train)
+    y_pred = clf.predict(X_test)
+    print(clf.__class__.__name__, accuracy_score(y_test, y_pred))
+
+\end{minted}
+
+% !split
+\subsection*{Please, not the moons again! Voting and Bagging}
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+from sklearn.model_selection import train_test_split
+from sklearn.datasets import make_moons
+
+X, y = make_moons(n_samples=500, noise=0.30, random_state=42)
+X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=42)
+from sklearn.ensemble import RandomForestClassifier
+from sklearn.ensemble import VotingClassifier
+from sklearn.linear_model import LogisticRegression
+from sklearn.svm import SVC
+
+log_clf = LogisticRegression(random_state=42)
+rnd_clf = RandomForestClassifier(random_state=42)
+svm_clf = SVC(random_state=42)
+
+voting_clf = VotingClassifier(
+    estimators=[('lr', log_clf), ('rf', rnd_clf), ('svc', svm_clf)],
+    voting='hard')
+voting_clf.fit(X_train, y_train)
+\end{minted}
+
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+from sklearn.metrics import accuracy_score
+
+for clf in (log_clf, rnd_clf, svm_clf, voting_clf):
+    clf.fit(X_train, y_train)
+    y_pred = clf.predict(X_test)
+    print(clf.__class__.__name__, accuracy_score(y_test, y_pred))
+\end{minted}
+
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+log_clf = LogisticRegression(random_state=42)
+rnd_clf = RandomForestClassifier(random_state=42)
+svm_clf = SVC(probability=True, random_state=42)
+
+voting_clf = VotingClassifier(
+    estimators=[('lr', log_clf), ('rf', rnd_clf), ('svc', svm_clf)],
+    voting='soft')
+voting_clf.fit(X_train, y_train)
+\end{minted}
+
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+from sklearn.metrics import accuracy_score
+
+for clf in (log_clf, rnd_clf, svm_clf, voting_clf):
+    clf.fit(X_train, y_train)
+    y_pred = clf.predict(X_test)
+    print(clf.__class__.__name__, accuracy_score(y_test, y_pred))
+\end{minted}
+
+% !split
+\subsection*{Now Bagging}
+
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+from sklearn.ensemble import BaggingClassifier
+from sklearn.tree import DecisionTreeClassifier
+
+bag_clf = BaggingClassifier(
+    DecisionTreeClassifier(random_state=42), n_estimators=500,
+    max_samples=100, bootstrap=True, n_jobs=-1, random_state=42)
+bag_clf.fit(X_train, y_train)
+y_pred = bag_clf.predict(X_test)
+\end{minted}
+
+
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+from sklearn.metrics import accuracy_score
+print(accuracy_score(y_test, y_pred))
+\end{minted}
+
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+tree_clf = DecisionTreeClassifier(random_state=42)
+tree_clf.fit(X_train, y_train)
+y_pred_tree = tree_clf.predict(X_test)
+print(accuracy_score(y_test, y_pred_tree))
+\end{minted}
+
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+from matplotlib.colors import ListedColormap
+
+def plot_decision_boundary(clf, X, y, axes=[-1.5, 2.5, -1, 1.5], alpha=0.5, contour=True):
+    x1s = np.linspace(axes[0], axes[1], 100)
+    x2s = np.linspace(axes[2], axes[3], 100)
+    x1, x2 = np.meshgrid(x1s, x2s)
+    X_new = np.c_[x1.ravel(), x2.ravel()]
+    y_pred = clf.predict(X_new).reshape(x1.shape)
+    custom_cmap = ListedColormap(['#fafab0','#9898ff','#a0faa0'])
+    plt.contourf(x1, x2, y_pred, alpha=0.3, cmap=custom_cmap)
+    if contour:
+        custom_cmap2 = ListedColormap(['#7d7d58','#4c4c7f','#507d50'])
+        plt.contour(x1, x2, y_pred, cmap=custom_cmap2, alpha=0.8)
+    plt.plot(X[:, 0][y==0], X[:, 1][y==0], "yo", alpha=alpha)
+    plt.plot(X[:, 0][y==1], X[:, 1][y==1], "bs", alpha=alpha)
+    plt.axis(axes)
+    plt.xlabel(r"$x_1$", fontsize=18)
+    plt.ylabel(r"$x_2$", fontsize=18, rotation=0)
+plt.figure(figsize=(11,4))
+plt.subplot(121)
+plot_decision_boundary(tree_clf, X, y)
+plt.title("Decision Tree", fontsize=14)
+plt.subplot(122)
+plot_decision_boundary(bag_clf, X, y)
+plt.title("Decision Trees with Bagging", fontsize=14)
+plt.show()
+\end{minted}
+
+
+% !split
+\subsection*{Making our own Bagging with Bootstrap}
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import train_test_split
+from sklearn.pipeline import make_pipeline
+from sklearn.utils import resample
+from sklearn.tree import DecisionTreeRegressor
+
+
+np.random.seed(2018)
+
+n = 40
+n_boostraps = 100
+maxdegree = 14
+
+# Make data set.
+x = np.linspace(-3, 3, n).reshape(-1, 1)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
+error = np.zeros(maxdegree)
+bias = np.zeros(maxdegree)
+variance = np.zeros(maxdegree)
+polydegree = np.zeros(maxdegree)
+X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
+
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+for degree in range(maxdegree):
+    model = DecisionTreeRegressor(max_depth=5) 
+    y_pred = np.empty((y_test.shape[0], n_boostraps))
+    for i in range(n_boostraps):
+        x_, y_ = resample(X_train_scaled, y_train)
+        model.fit(x_, y_)
+        y_pred[:, i] = model.predict(X_test_scaled).ravel()
+
+    polydegree[degree] = degree
+    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )
+    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )
+    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )
+    print('Polynomial degree:', degree)
+    print('Error:', error[degree])
+    print('Bias^2:', bias[degree])
+    print('Var:', variance[degree])
+    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
+
+plt.plot(polydegree, error, label='Error')
+plt.plot(polydegree, bias, label='bias')
+plt.plot(polydegree, variance, label='Variance')
+plt.legend()
+plt.show()
+
+\end{minted}
+
+
+% !split
+\subsection*{Changing the Level of the Decision Tree}
+
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+
+import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import train_test_split
+from sklearn.pipeline import make_pipeline
+from sklearn.utils import resample
+from sklearn.tree import DecisionTreeRegressor
+
+n = 100
+n_boostraps = 100
+maxdepth = 8
+
+# Make data set.
+x = np.linspace(-3, 3, n).reshape(-1, 1)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
+error = np.zeros(maxdepth)
+bias = np.zeros(maxdepth)
+variance = np.zeros(maxdepth)
+polydegree = np.zeros(maxdepth)
+X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
+
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+for degree in range(1,maxdepth):
+    model = DecisionTreeRegressor(max_depth=degree) 
+    y_pred = np.empty((y_test.shape[0], n_boostraps))
+    for i in range(n_boostraps):
+        x_, y_ = resample(X_train_scaled, y_train)
+        model.fit(x_, y_)
+        y_pred[:, i] = model.predict(X_test_scaled)#.ravel()
+
+    polydegree[degree] = degree
+    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )
+    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )
+    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )
+    print('Polynomial degree:', degree)
+    print('Error:', error[degree])
+    print('Bias^2:', bias[degree])
+    print('Var:', variance[degree])
+    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
+
+plt.xlim(1,maxdepth)
+plt.plot(polydegree, error, label='Error')
+plt.plot(polydegree, bias, label='bias')
+plt.plot(polydegree, variance, label='Variance')
+plt.legend()
+plt.show()
+
+
+
+
+\end{minted}
+
+
+
+
+% !split
+\subsection*{Random forests}
+
+Random forests provide an improvement over bagged trees by way of a
+small tweak that decorrelates the trees. 
+
+As in bagging, we build a
+number of decision trees on bootstrapped training samples. But when
+building these decision trees, each time a split in a tree is
+considered, a random sample of $m$ predictors is chosen as split
+candidates from the full set of $p$ predictors. The split is allowed to
+use only one of those $m$ predictors. 
+
+A fresh sample of $m$ predictors is
+taken at each split, and typically we choose 
+
+\[
+m\approx \sqrt{p}.
+\]
+
+In building a random forest, at
+each split in the tree, the algorithm is not even allowed to consider
+a majority of the available predictors. 
+
+The reason for this is rather clever. Suppose that there is one very
+strong predictor in the data set, along with a number of other
+moderately strong predictors. Then in the collection of bagged
+variable importance random forest trees, most or all of the trees will
+use this strong predictor in the top split. Consequently, all of the
+bagged trees will look quite similar to each other. Hence the
+predictions from the bagged trees will be highly correlated.
+Unfortunately, averaging many highly correlated quantities does not
+lead to as large of a reduction in variance as averaging many
+uncorrelated quanti- ties. In particular, this means that bagging will
+not lead to a substantial reduction in variance over a single tree in
+this setting.
+
+
+% !split
+\subsection*{Random Forest Algorithm}
+The algorithm described here can be applied to both classification and regression problems.
+
+We will grow of forest of say $M$ trees.
+\begin{enumerate}
+\item For $m=1:M$ we
+\begin{itemize}
+
+  \item Draw a bootstrap sample of from the training data organized in our $\bm{X}$ matrix.
+
+  \item We grow then a random forest tree $T_m$ based on the bootstrapped data by repeating the steps outlined till we reach the maximum node size is reached
+\begin{enumerate}
+
+   \item we select $m \le p$ varibales at random from the $p$ predictors/features
+
+   \item pick the best split point among the $m$ features using either the CART algorithm or the ID3 for classification and create a new node
+
+   \item split the node into daughter nodes
+
+\end{enumerate}
+
+\noindent
+\end{itemize}
+
+\noindent
+\item Output then the ensemble of trees $\{T_m\}_1^{M}$ and make predictions for either a regression type of problem or a classification type of problem. 
+\end{enumerate}
+
+\noindent
+% !split
+\subsection*{Bootstrap with Random Forests Instead of a Single Tree, own Bagging}
+
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+
+import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import train_test_split
+from sklearn.pipeline import make_pipeline
+from sklearn.utils import resample
+from sklearn.ensemble import RandomForestRegressor
+
+np.random.seed(2018)
+
+n = 100
+n_boostraps = 100
+maxdegree = 14
+
+# Make data set.
+x = np.linspace(-3, 3, n).reshape(-1, 1)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
+error = np.zeros(maxdegree)
+bias = np.zeros(maxdegree)
+variance = np.zeros(maxdegree)
+polydegree = np.zeros(maxdegree)
+X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
+
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+for degree in range(maxdegree):
+    model = RandomForestRegressor()
+    y_pred = np.empty((y_test.shape[0], n_boostraps))
+    for i in range(n_boostraps):
+        x_, y_ = resample(X_train_scaled, y_train)
+        model.fit(x_, y_.ravel())
+        y_pred[:, i] = model.predict(X_test_scaled).ravel()
+
+    polydegree[degree] = degree
+    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )
+    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )
+    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )
+    print('Polynomial degree:', degree)
+    print('Error:', error[degree])
+    print('Bias^2:', bias[degree])
+    print('Var:', variance[degree])
+    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
+
+plt.plot(polydegree, error, label='Error')
+plt.plot(polydegree, bias, label='bias')
+plt.plot(polydegree, variance, label='Variance')
+plt.legend()
+plt.show()
+
+
+
+\end{minted}
+
+
+
+
+% !split
+\subsection*{Random Forests Compared with other Methods on the Cancer Data}
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import  train_test_split 
+from sklearn.datasets import load_breast_cancer
+from sklearn.svm import SVC
+from sklearn.linear_model import LogisticRegression
+from sklearn.tree import DecisionTreeClassifier
+
+# Load the data
+cancer = load_breast_cancer()
+
+X_train, X_test, y_train, y_test = train_test_split(cancer.data,cancer.target,random_state=0)
+print(X_train.shape)
+print(X_test.shape)
+# Logistic Regression
+logreg = LogisticRegression(solver='lbfgs')
+logreg.fit(X_train, y_train)
+print("Test set accuracy with Logistic Regression: {:.2f}".format(logreg.score(X_test,y_test)))
+# Support vector machine
+svm = SVC(gamma='auto', C=100)
+svm.fit(X_train, y_train)
+print("Test set accuracy with SVM: {:.2f}".format(svm.score(X_test,y_test)))
+# Decision Trees
+deep_tree_clf = DecisionTreeClassifier(max_depth=None)
+deep_tree_clf.fit(X_train, y_train)
+print("Test set accuracy with Decision Trees: {:.2f}".format(deep_tree_clf.score(X_test,y_test)))
+#now scale the data
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+# Logistic Regression
+logreg.fit(X_train_scaled, y_train)
+print("Test set accuracy Logistic Regression with scaled data: {:.2f}".format(logreg.score(X_test_scaled,y_test)))
+# Support Vector Machine
+svm.fit(X_train_scaled, y_train)
+print("Test set accuracy SVM with scaled data: {:.2f}".format(logreg.score(X_test_scaled,y_test)))
+# Decision Trees
+deep_tree_clf.fit(X_train_scaled, y_train)
+print("Test set accuracy with Decision Trees and scaled data: {:.2f}".format(deep_tree_clf.score(X_test_scaled,y_test)))
+
+
+from sklearn.ensemble import RandomForestClassifier
+from sklearn.preprocessing import LabelEncoder
+from sklearn.model_selection import cross_validate
+# Data set not specificied
+#Instantiate the model with 500 trees and entropy as splitting criteria
+Random_Forest_model = RandomForestClassifier(n_estimators=500,criterion="entropy")
+Random_Forest_model.fit(X_train_scaled, y_train)
+#Cross validation
+accuracy = cross_validate(Random_Forest_model,X_test_scaled,y_test,cv=10)['test_score']
+print(accuracy)
+print("Test set accuracy with Random Forests and scaled data: {:.2f}".format(Random_Forest_model.score(X_test_scaled,y_test)))
+
+
+import scikitplot as skplt
+y_pred = Random_Forest_model.predict(X_test_scaled)
+skplt.metrics.plot_confusion_matrix(y_test, y_pred, normalize=True)
+plt.show()
+y_probas = Random_Forest_model.predict_proba(X_test_scaled)
+skplt.metrics.plot_roc(y_test, y_probas)
+plt.show()
+skplt.metrics.plot_cumulative_gain(y_test, y_probas)
+plt.show()
+
+\end{minted}
+
+
+% !split
+\subsection*{Compare  Bagging on Trees with Random Forests}
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+bag_clf = BaggingClassifier(
+    DecisionTreeClassifier(splitter="random", max_leaf_nodes=16, random_state=42),
+    n_estimators=500, max_samples=1.0, bootstrap=True, n_jobs=-1, random_state=42)
+\end{minted}
+
+
+
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+bag_clf.fit(X_train, y_train)
+y_pred = bag_clf.predict(X_test)
+from sklearn.ensemble import RandomForestClassifier
+rnd_clf = RandomForestClassifier(n_estimators=500, max_leaf_nodes=16, n_jobs=-1, random_state=42)
+rnd_clf.fit(X_train, y_train)
+y_pred_rf = rnd_clf.predict(X_test)
+np.sum(y_pred == y_pred_rf) / len(y_pred) 
+\end{minted}
+
+
+
+
+
+% !split
+\subsection*{Boosting, a Bird'e Eye}
+
+The basic idea is to combine weak classifiers in order to create a good
+classifier. With a weak classifier we often intend a classifier which
+produces results which are only slightly better than we would get by
+random guesses.
+
+This is done by applying in an iterative way a weak (or a standard
+classifier like decision trees) to modify the data. In each iteration
+we emphasize those observations which are misclassified by weighting
+them with a factor.
+
+
+% !split
+\subsection*{Adaptive boosting: AdaBoost, Basic Algorithm}
+
+The algorithm here is rather straightforward. Assume that our weak
+classifier is a decision tree and we consider a binary set of outputs
+with $y_i \in \{-1,1\}$ and $i=0,1,2,\dots,n-1$ as our set of
+observations. Our design matrix is given in terms of the
+feature/predictor vectors
+$\bm{X}=[\bm{x}_0\bm{x}_1\dots\bm{x}_{p-1}$. Finally, we define also a
+classifier determined by our data via a function $G(\bm{X})$. This function tells us how well we are able to classify our outputs/targets $\bm{y}$. 
+
+We can then define the misclassification error $\mathrm{err}$ as
+\[
+\mathrm{err}=\frac{1}{n}\sum_{i=0}^{n-1}I(y_i\ne G(\bm{X}_{i*}),
+\]
+where the function $I()$ is one if we misclassify and zero if we classify correctly. 
+
+% !split
+\subsection*{Basic Steps of AdaBoost}
+
+With the above definitions we are now ready to set up the algorithm for AdaBoost.
+The basic idea is to set up weights which will be used to scale the correctly classified and the misclassified cases.
+\begin{enumerate}
+\item We start by initializing all weights to $w_i = 1/n$, with $i=0,1,2,\dots n-1$. It is to see then that $\sum_{i=0}^{n-1}w_i = 1$.
+
+\item We rewrite the misclassification error as 
+\end{enumerate}
+
+\noindent
+\[
+\mathrm{err}=\frac{\sum_{i=0}^{n-1}w_iI(y_i\ne G(\bm{X}_{i*})}{\sum_{i=0}^{n-1}w_i},
+\]
+\begin{enumerate}
+\item Then we start looping over all attempts at classifying, namely we start an iterative process for $m=1:M$, where $M$ is the final number of classifications. Our given classifier could for example be a plain decision tree.
+\begin{enumerate}
+
+ \item Fit then a given classifier to the training using the weights $w_i$.
+
+ \item Compute then $\mathrm{err}$ and figure out which events are classified properly and which are classified wrongly.
+
+ \item Define a quantity $\alpha_{m} = \log{(1-\mathrm{err})/\mathrm{err}}
+
+ \item Set the new weights to $w_i = w_i\times \exp{(\alpha_m I(y_i\ne G(\bm{X}_{i*})}.
+
+\end{enumerate}
+
+\noindent
+\item Compute the new classifier $G(\bm{X})= \sum_{i=0}^{n-1}\alpha_m I(y_i\ne G(\bm{X}_{i*}).
+\end{enumerate}
+
+\noindent
+For the iterations with $m \le 2$ the weights are modified
+individually at each steps. The obersvations which were misclassified
+at iteration $m-1$ have a weight which is larger than those which were
+classified properly. As this proceeds, the observations which were
+difficult to classifiy correctly are given a larger influence. Each
+new classification step $m$ is then forced to concentrate on those
+observations that are missed in the previous iterations.
+
+
+
+% !split
+\subsection*{AdaBoost Examples}
+
+Using \textbf{Scikit-Learn} it is easy to appply the adaptive boosting algorithm, as done here.
+
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+from sklearn.ensemble import AdaBoostClassifier
+
+ada_clf = AdaBoostClassifier(
+    DecisionTreeClassifier(max_depth=1), n_estimators=200,
+    algorithm="SAMME.R", learning_rate=0.5, random_state=42)
+ada_clf.fit(X_train, y_train)
+
+plot_decision_boundary(ada_clf, X, y)
+
+m = len(X_train)
+
+plt.figure(figsize=(11, 4))
+for subplot, learning_rate in ((121, 1), (122, 0.5)):
+    sample_weights = np.ones(m)
+    plt.subplot(subplot)
+    for i in range(5):
+        svm_clf = SVC(kernel="rbf", C=0.05, gamma="auto", random_state=42)
+        svm_clf.fit(X_train, y_train, sample_weight=sample_weights)
+        y_pred = svm_clf.predict(X_train)
+        sample_weights[y_pred != y_train] *= (1 + learning_rate)
+        plot_decision_boundary(svm_clf, X, y, alpha=0.2)
+        plt.title("learning_rate = {}".format(learning_rate), fontsize=16)
+    if subplot == 121:
+        plt.text(-0.7, -0.65, "1", fontsize=14)
+        plt.text(-0.6, -0.10, "2", fontsize=14)
+        plt.text(-0.5,  0.10, "3", fontsize=14)
+        plt.text(-0.4,  0.55, "4", fontsize=14)
+        plt.text(-0.3,  0.90, "5", fontsize=14)
+
+save_fig("boosting_plot")
+plt.show()
+\end{minted}
+
+
+% !split
+\subsection*{Gradient boosting: Basics}
+
+Gradient boosting is again a similar technique to Adapative boosting,
+it combines so-called weak classifiers or regressors into a strong
+method via a series of iterations.
+
+In order to understand the method, let us illustrate its basics by
+bringing back the essential steps in linear regression, where our cost
+function was the least squares function.
+
+% !split
+\subsection*{Gradient Boosting, algorithm}
+
+Suppose we have a cost function $C(f)=\sum_{i=0}^{n-1}L(y_i, f(x_i))$ where $y_i$ is our target and $f(x_i)$ the function which is meant to model $y_i$. The above cost function could be our standard least squares  function
+\[
+C(\bm{y},\bm{f})=\frac{1}{n}\sum_{i=0}^{n-1}(y_i-f(x_i))^2.
+\]
+
+The way we proceed in an iterative fashion is to
+\begin{enumerate}
+\item Initialize our estimate by $f_0(x)=0$.
+
+\item For $m=1:M$, we
+\begin{enumerate}
+
+ \item compute the negative gradient vector $\bm{u}_m = -\partial C(\bm{y},\bm{f})/\partial \bm{f}(x)$ at $f(x) = f_{m-1}(x);
+
+ \item fit the so-called base-learner to the negative gradient $h_m(u_m,x)$;
+
+ \item update the estimate $f_m(x) = f_{m-1}(x)+\nu h_m(u_m,x)$;
+
+\end{enumerate}
+
+\noindent
+\item The final estimate is then $f_M(x) = \sum_{m=1}^M\nu h_m(u_m,x)$.
+\end{enumerate}
+
+\noindent
+% !split
+\subsection*{Gradient Boosting, Examples}
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+np.random.seed(42)
+X = np.random.rand(100, 1) - 0.5
+y = 3*X[:, 0]**2 + 0.05 * np.random.randn(100)
+
+from sklearn.tree import DecisionTreeRegressor
+
+tree_reg1 = DecisionTreeRegressor(max_depth=2, random_state=42)
+tree_reg1.fit(X, y)
+
+y2 = y - tree_reg1.predict(X)
+tree_reg2 = DecisionTreeRegressor(max_depth=2, random_state=42)
+tree_reg2.fit(X, y2)
+
+y3 = y2 - tree_reg2.predict(X)
+tree_reg3 = DecisionTreeRegressor(max_depth=2, random_state=42)
+tree_reg3.fit(X, y3)
+
+X_new = np.array([[0.8]])
+y_pred = sum(tree.predict(X_new) for tree in (tree_reg1, tree_reg2, tree_reg3))
+
+def plot_predictions(regressors, X, y, axes, label=None, style="r-", data_style="b.", data_label=None):
+    x1 = np.linspace(axes[0], axes[1], 500)
+    y_pred = sum(regressor.predict(x1.reshape(-1, 1)) for regressor in regressors)
+    plt.plot(X[:, 0], y, data_style, label=data_label)
+    plt.plot(x1, y_pred, style, linewidth=2, label=label)
+    if label or data_label:
+        plt.legend(loc="upper center", fontsize=16)
+    plt.axis(axes)
+
+plt.figure(figsize=(11,11))
+
+plt.subplot(321)
+plot_predictions([tree_reg1], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="$h_1(x_1)$", style="g-", data_label="Training set")
+plt.ylabel("$y$", fontsize=16, rotation=0)
+plt.title("Residuals and tree predictions", fontsize=16)
+
+plt.subplot(322)
+plot_predictions([tree_reg1], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="$h(x_1) = h_1(x_1)$", data_label="Training set")
+plt.ylabel("$y$", fontsize=16, rotation=0)
+plt.title("Ensemble predictions", fontsize=16)
+
+plt.subplot(323)
+plot_predictions([tree_reg2], X, y2, axes=[-0.5, 0.5, -0.5, 0.5], label="$h_2(x_1)$", style="g-", data_style="k+", data_label="Residuals")
+plt.ylabel("$y - h_1(x_1)$", fontsize=16)
+
+plt.subplot(324)
+plot_predictions([tree_reg1, tree_reg2], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="$h(x_1) = h_1(x_1) + h_2(x_1)$")
+plt.ylabel("$y$", fontsize=16, rotation=0)
+
+plt.subplot(325)
+plot_predictions([tree_reg3], X, y3, axes=[-0.5, 0.5, -0.5, 0.5], label="$h_3(x_1)$", style="g-", data_style="k+")
+plt.ylabel("$y - h_1(x_1) - h_2(x_1)$", fontsize=16)
+plt.xlabel("$x_1$", fontsize=16)
+
+plt.subplot(326)
+plot_predictions([tree_reg1, tree_reg2, tree_reg3], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="$h(x_1) = h_1(x_1) + h_2(x_1) + h_3(x_1)$")
+plt.xlabel("$x_1$", fontsize=16)
+plt.ylabel("$y$", fontsize=16, rotation=0)
+
+save_fig("gradient_boosting_plot")
+plt.show()
+
+from sklearn.ensemble import GradientBoostingRegressor
+
+gbrt = GradientBoostingRegressor(max_depth=2, n_estimators=3, learning_rate=1.0, random_state=42)
+gbrt.fit(X, y)
+
+gbrt_slow = GradientBoostingRegressor(max_depth=2, n_estimators=200, learning_rate=0.1, random_state=42)
+gbrt_slow.fit(X, y)
+
+plt.figure(figsize=(11,4))
+
+plt.subplot(121)
+plot_predictions([gbrt], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="Ensemble predictions")
+plt.title("learning_rate={}, n_estimators={}".format(gbrt.learning_rate, gbrt.n_estimators), fontsize=14)
+
+plt.subplot(122)
+plot_predictions([gbrt_slow], X, y, axes=[-0.5, 0.5, -0.1, 0.8])
+plt.title("learning_rate={}, n_estimators={}".format(gbrt_slow.learning_rate, gbrt_slow.n_estimators), fontsize=14)
+
+save_fig("gbrt_learning_rate_plot")
+plt.show()
+
+\end{minted}
+
+
+% !split
+\subsection*{Gradient Boots with Early Stopping}
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+
+from sklearn.model_selection import train_test_split
+from sklearn.metrics import mean_squared_error
+
+X_train, X_val, y_train, y_val = train_test_split(X, y, random_state=49)
+
+gbrt = GradientBoostingRegressor(max_depth=2, n_estimators=120, random_state=42)
+gbrt.fit(X_train, y_train)
+
+errors = [mean_squared_error(y_val, y_pred)
+          for y_pred in gbrt.staged_predict(X_val)]
+bst_n_estimators = np.argmin(errors) + 1
+
+gbrt_best = GradientBoostingRegressor(max_depth=2,n_estimators=bst_n_estimators, random_state=42)
+gbrt_best.fit(X_train, y_train)
+
+min_error = np.min(errors)
+plt.figure(figsize=(11, 4))
+
+plt.subplot(121)
+plt.plot(errors, "b.-")
+plt.plot([bst_n_estimators, bst_n_estimators], [0, min_error], "k--")
+plt.plot([0, 120], [min_error, min_error], "k--")
+plt.plot(bst_n_estimators, min_error, "ko")
+plt.text(bst_n_estimators, min_error*1.2, "Minimum", ha="center", fontsize=14)
+plt.axis([0, 120, 0, 0.01])
+plt.xlabel("Number of trees")
+plt.title("Validation error", fontsize=14)
+
+plt.subplot(122)
+plot_predictions([gbrt_best], X, y, axes=[-0.5, 0.5, -0.1, 0.8])
+plt.title("Best model (%d trees)" % bst_n_estimators, fontsize=14)
+
+save_fig("early_stopping_gbrt_plot")
+plt.show()
+
+
+gbrt = GradientBoostingRegressor(max_depth=2, warm_start=True, random_state=42)
+
+min_val_error = float("inf")
+error_going_up = 0
+for n_estimators in range(1, 120):
+    gbrt.n_estimators = n_estimators
+    gbrt.fit(X_train, y_train)
+    y_pred = gbrt.predict(X_val)
+    val_error = mean_squared_error(y_val, y_pred)
+    if val_error < min_val_error:
+        min_val_error = val_error
+        error_going_up = 0
+    else:
+        error_going_up += 1
+        if error_going_up == 5:
+            break  # early stopping
+
+
+print(gbrt.n_estimators)
+print("Minimum validation MSE:", min_val_error)
+\end{minted}
+
+% !split
+\subsection*{XGBoost: Extreme Gradient Boosting}
+
+
+\href{{https://github.com/dmlc/xgboost}}{XGBoost} or Extreme Gradient
+Boosting, is an optimized distributed gradient boosting library
+designed to be highly efficient, flexible and portable. It implements
+machine learning algorithms under the Gradient Boosting
+framework. XGBoost provides a parallel tree boosting that solve many
+data science problems in a fast and accurate way. See the \href{{https://arxiv.org/abs/1603.02754}}{article by Chen and Guestrin}.
+
+The authors design and build a highly scalable end-to-end tree
+boosting system. It has  a theoretically justified weighted quantile
+sketch for efficient proposal calculation. It introduces a novel sparsity-aware algorithm for parallel tree learning and an effective cache-aware block structure for out-of-core tree learning.
+
+It is now the algorithm which wins essentially all ML competitions!!!
+
+% !split
+\subsection*{Regression Case}
+
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import train_test_split
+import xgboost as xgb
+from sklearn.preprocessing import StandardScaler
+import scikitplot as skplt
+from sklearn.metrics import mean_squared_error
+
+n = 100
+maxdegree = 6
+
+# Make data set.
+x = np.linspace(-3, 3, n).reshape(-1, 1)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
+
+error = np.zeros(maxdegree)
+bias = np.zeros(maxdegree)
+variance = np.zeros(maxdegree)
+polydegree = np.zeros(maxdegree)
+X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+for degree in range(maxdegree):
+    model =  xgb.XGBRegressor(objective ='reg:squarederror', colsample_bytree = 0.3, learning_rate = 0.1,
+                max_depth = degree, alpha = 10, n_estimators = 10)
+    model.fit(X_train_scaled,y_train)
+    y_pred = model.predict(X_test_scaled)
+    polydegree[degree] = degree
+    error[degree] = np.mean( np.mean((y_test - y_pred)**2) )
+    bias[degree] = np.mean( (y_test - np.mean(y_pred))**2 )
+    variance[degree] = np.mean( np.var(y_pred) )
+    print('Max depth:', degree)
+    print('Error:', error[degree])
+    print('Bias^2:', bias[degree])
+    print('Var:', variance[degree])
+    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
+
+plt.xlim(1,maxdegree-1)
+plt.plot(polydegree, error, label='Error')
+plt.plot(polydegree, bias, label='bias')
+plt.plot(polydegree, variance, label='Variance')
+plt.legend()
+plt.show()
+
+
+\end{minted}
+
+
+
+% !split
+\subsection*{Xgboost on the Cancer Data}
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import  train_test_split 
+from sklearn.datasets import load_breast_cancer
+from sklearn.preprocessing import LabelEncoder
+from sklearn.model_selection import cross_validate
+import scikitplot as skplt
+import xgboost as xgb
+# Load the data
+cancer = load_breast_cancer()
+
+X_train, X_test, y_train, y_test = train_test_split(cancer.data,cancer.target,random_state=0)
+print(X_train.shape)
+print(X_test.shape)
+#now scale the data
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+xg_clf = xgb.XGBClassifier()
+xg_clf.fit(X_train_scaled,y_train)
+xgb.plot_tree(xg_clf,num_trees=0)
+plt.rcParams['figure.figsize'] = [50, 10]
+plt.show()
+xgb.plot_importance(xg_clf)
+plt.rcParams['figure.figsize'] = [5, 5]
+plt.show()
+\end{minted}
+
+% ------------------- end of main content ---------------
+
+\end{document}
+
diff --git a/doc/src/DecisionTrees/DecisionTrees-reveal.html b/doc/src/DecisionTrees/DecisionTrees-reveal.html
new file mode 100644
index 000000000..0d1365a1e
--- /dev/null
+++ b/doc/src/DecisionTrees/DecisionTrees-reveal.html
@@ -0,0 +1,2657 @@
+
+
+
+
+
+
+
+Data Analysis and Machine Learning: From Decision Trees to Forests and all that
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
    
+
+
+
+
    
+
+
+
+
+
+
+
+
+
+
+    
+
+
+
+
    
+
+
+
+
+
    Data Analysis and Machine Learning: From Decision Trees to Forests and all that
      
+
+
    
+
+
+
    
+Morten Hjorth-Jensen [1, 2]
+
    
+
+
     
    
+
+
+
    [1] Department of Physics, University of Oslo
    
+
    [2] Department of Physics and Astronomy and National Superconducting Cyclotron Laboratory, Michigan State University
    
+
    
+
     
    
+
    Nov 7, 2019
     
+
    
+
    
+
+
    
+ © 1999-2019, Morten Hjorth-Jensen. Released under CC Attribution-NonCommercial 4.0 license
+
    
+
    
+
+
+
    
+
    Decision trees, overarching aims  
    
+
+
    
+Decision trees are supervised learning algorithms used for both,
+classification and regression tasks.
+
+
    
+The main idea of decision trees
+is to find those descriptive features which contain the most
+information regarding the target feature and then split the dataset
+along the values of these features such that the target feature values
+for the resulting underlying datasets are as pure as possible.
+
+
    
+The descriptive features which reproduce best the target/output features are normally  said
+to be the most informative ones. The process of finding the most
+informative feature is done until we accomplish a stopping criteria
+where we then finally end up in so called leaf nodes.
+
+
    
+A decision tree is typically divided into a root node, the interior nodes,
+and the final leaf nodes or just leaves. These entities are then connected by so-called branches.
+
+
    
+The leaf nodes
+contain the predictions we will make for new query instances presented
+to our trained model. This is possible since the model has 
+learned the underlying structure of the training data and hence can,
+given some assumptions, make predictions about the target feature value
+(class) of unseen query instances.
+
    
+
+
+
    
+
    A typical Decision Tree with its pertinent Jargon, Classification Problem 
    
+
+
    
+



    
+
+
    
+This tree was produced using the Wisconsin cancer data (discussed here as well, see code examples below) using Scikit-Learn's decision tree classifier. Here we have used the so-called gini index (see below) to split the various branches.
+
    
+
+
+
    
+
    General Features 
    
+
+
    
+The overarching approach to decision trees is a top-down approach.
+
+
      
+
    •  A leaf provides the classification of a given instance.
      • 
+
    •  A node specifies a test of some attribute of the instance.
      • 
+
    •  A branch corresponds to a possible values of an attribute.
      • 
+
    •  An instance is classified by starting at the root node of the tree, testing the attribute specified by this node, then moving down the tree branch corresponding to the value of the attribute in the given example.
      • 
+
    
+
    
+
+This process is then repeated for the subtree rooted at the new
+node.
+
    
+
+
+
    
+
    How do we set it up? 
    
+
+
    
+In simplified terms, the process of training a decision tree and
+predicting the target features of query instances is as follows:
+
+
      
+
    1.  Present a dataset containing of a number of training instances characterized by a number of descriptive features and a target feature
      2. 
+
    2.  Train the decision tree model by continuously splitting the target feature along the values of the descriptive features using a measure of information gain during the training process
      3. 
+
    3.  Grow the tree until we accomplish a stopping criteria create leaf nodes which represent the predictions we want to make for new query instances
      4. 
+
    4.  Show query instances to the tree and run down the tree until we arrive at leaf nodes
      5. 
+
    
+
    
+
+Then we are essentially done!
+
    
+
+
+
    
+
    Decision trees and Regression  
    
+
    
+
+
+
    import numpy as np
+import matplotlib.pyplot as plt
+from sklearn.preprocessing import PolynomialFeatures
+from sklearn.linear_model import LinearRegression
+
+steps=250
+
+distance=0
+x=0
+distance_list=[]
+steps_list=[]
+while x<steps:
+    distance+=np.random.randint(-1,2)
+    distance_list.append(distance)
+    x+=1
+    steps_list.append(x)
+plt.plot(steps_list,distance_list, color='green', label="Random Walk Data")
+
+steps_list=np.asarray(steps_list)
+distance_list=np.asarray(distance_list)
+
+X=steps_list[:,np.newaxis]
+
+#Polynomial fits
+
+#Degree 2
+poly_features=PolynomialFeatures(degree=2, include_bias=False)
+X_poly=poly_features.fit_transform(X)
+
+lin_reg=LinearRegression()
+poly_fit=lin_reg.fit(X_poly,distance_list)
+b=lin_reg.coef_
+c=lin_reg.intercept_
+print ("2nd degree coefficients:")
+print ("zero power: ",c)
+print ("first power: ", b[0])
+print ("second power: ",b[1])
+
+z = np.arange(0, steps, .01)
+z_mod=b[1]*z**2+b[0]*z+c
+
+fit_mod=b[1]*X**2+b[0]*X+c
+plt.plot(z, z_mod, color='r', label="2nd Degree Fit")
+plt.title("Polynomial Regression")
+
+plt.xlabel("Steps")
+plt.ylabel("Distance")
+
+#Degree 10
+poly_features10=PolynomialFeatures(degree=10, include_bias=False)
+X_poly10=poly_features10.fit_transform(X)
+
+poly_fit10=lin_reg.fit(X_poly10,distance_list)
+
+y_plot=poly_fit10.predict(X_poly10)
+plt.plot(X, y_plot, color='black', label="10th Degree Fit")
+
+plt.legend()
+plt.show()
+
+
+#Decision Tree Regression
+from sklearn.tree import DecisionTreeRegressor
+regr_1=DecisionTreeRegressor(max_depth=2)
+regr_2=DecisionTreeRegressor(max_depth=5)
+regr_3=DecisionTreeRegressor(max_depth=7)
+regr_1.fit(X, distance_list)
+regr_2.fit(X, distance_list)
+regr_3.fit(X, distance_list)
+
+X_test = np.arange(0.0, steps, 0.01)[:, np.newaxis]
+y_1 = regr_1.predict(X_test)
+y_2 = regr_2.predict(X_test)
+y_3=regr_3.predict(X_test)
+
+# Plot the results
+plt.figure()
+plt.scatter(X, distance_list, s=2.5, c="black", label="data")
+plt.plot(X_test, y_1, color="red",
+         label="max_depth=2", linewidth=2)
+plt.plot(X_test, y_2, color="green", label="max_depth=5", linewidth=2)
+plt.plot(X_test, y_3, color="m", label="max_depth=7", linewidth=2)
+
+plt.xlabel("Data")
+plt.ylabel("Darget")
+plt.title("Decision Tree Regression")
+plt.legend()
+plt.show()
+
    
+
    
+
+
+
    
+
    Building a tree, regression 
    
+
+
    
+There are mainly two steps
+
+
      
+
    1.  We split the predictor space (the set of possible values \( x_1,x_2,\dots, x_p \)) into \( J \) distinct and non-non-overlapping regions, \( R_1,R_2,\dots,R_J \).
      2. 
+
+
    2.  For every observation that falls into the region \( R_j \) , we make the same prediction, which is simply the mean of the response values for the training observations in \( R_j \).
      3. 
+
    
+
    
+
+How do we construct the regions \( R_1,\dots,R_J \)?  In theory, the
+regions could have any shape. However, we choose to divide the
+predictor space into high-dimensional rectangles, or boxes, for
+simplicity and for ease of interpretation of the resulting predictive
+model. The goal is to find boxes \( R_1,\dots,R_J \) that minimize the
+MSE, given by
+
+
     
    
+$$
+\sum_{j=1}^J\sum_{i\in R_j}(y_i-\overline{y}_{R_j})^2,
+$$
+
     
    
+
+
    
+where \( \overline{y}_{R_j} \)  is the mean response for the training observations 
+within box \( j \).
+
    
+
+
+
    
+
    A top-down approach, recursive binary splitting 
    
+
+
    
+Unfortunately, it is computationally infeasible to consider every
+possible partition of the feature space into \( J \) boxes.  The common
+strategy is to take a top-down approach
+
+
    
+The approach is top-down because it begins at the top of the tree (all
+observations belong to a single region) and then successively splits
+the predictor space; each split is indicated via two new branches
+further down on the tree. It is greedy because at each step of the
+tree-building process, the best split is made at that particular step,
+rather than looking ahead and picking a split that will lead to a
+better tree in some future step.
+
    
+
+
+
    
+
    Making a tree 
    
+
+
    
+In order to implement the recursive binary splitting we start by selecting
+the predictor \( x_j \) and a cutpoint \( s \) that splits the predictor space into two regions \( R_1 \) and \( R_2 \)
+
     
    
+$$
+\left\{X\vert x_j < s\right\},
+$$
+
     
    
+
+and
+
     
    
+$$
+\left\{X\vert x_j \geq s\right\},
+$$
+
     
    
+
+so that we obtain the lowest MSE, that is
+
     
    
+$$
+\sum_{i:x_i\in R_j}(y_i-\overline{y}_{R_1})^2+\sum_{i:x_i\in R_2}(y_i-\overline{y}_{R_2})^2,
+$$
+
     
    
+
+
    
+which we want to minimize by considering all predictors
+\( x_1,x_2,\dots,x_p \).  We consider also all possible values of \( s \) for
+each predictor. These values could be determined by randomly assigned
+numbers or by starting at the midpoint and then proceed till we find
+an optimal value.
+
+
    
+For any \( j \) and \( s \), we define the pair of half-planes where
+\( \overline{y}_{R_1} \) is the mean response for the training
+observations in \( R_1(j,s) \), and \( \overline{y}_{R_2} \) is the mean
+response for the training observations in \( R_2(j,s) \).
+
+
    
+Finding the values of \( j \) and \( s \) that minimize the above equation can be
+done quite quickly, especially when the number of features \( p \) is not
+too large.
+
+
    
+Next, we repeat the process, looking
+for the best predictor and best cutpoint in order to split the data
+further so as to minimize the MSE within each of the resulting
+regions. However, this time, instead of splitting the entire predictor
+space, we split one of the two previously identified regions. We now
+have three regions. Again, we look to split one of these three regions
+further, so as to minimize the MSE. The process continues until a
+stopping criterion is reached; for instance, we may continue until no
+region contains more than five observations.
+
    
+
+
+
    
+
    Pruning the tree 
    
+
+
    
+The above procedure is rather straightforward, but leads often to
+overfitting and unnecessarily large and complicated trees. The basic
+idea is to grow a large tree \( T_0 \) and then prune it back in order to
+obtain a subtree. A smaller tree with fewer splits (fewer regions) can
+lead to smaller variance and better interpretation at the cost of a
+little more bias.
+
+
    
+The so-called Cost complexity pruning algorithm gives us a
+way to do just this. Rather than considering every possible subtree,
+we consider a sequence of trees indexed by a nonnegative tuning
+parameter \( \alpha \).
+
    
+
+
+
    
+
    Cost complexity pruning 
    
+For each value of \( \alpha \)  there corresponds a subtree \( T \in T_0 \) such that
+
     
    
+$$
+\sum_{m=1}^{\overline{T}}\sum_{i:x_i\in R_m}(y_i-\overline{y}_{R_m})^2+\alpha\overline{T},
+$$
+
     
    
+
+is as small as possible. Here \( \overline{T} \) is 
+the number of terminal nodes of the tree \( T \) , \( R_m \) is the
+rectangle (i.e. the subset of predictor space)  corresponding to the \( m \)-th terminal node.
+
+
    
+The tuning parameter \( \alpha \) controls a trade-off between the subtree’s
+com- plexity and its fit to the training data. When \( \alpha = 0 \), then the
+subtree \( T \) will simply equal \( T_0 \), 
+because then the above equation just measures the
+training error. 
+However, as \( \alpha \) increases, there is a price to pay for
+having a tree with many terminal nodes. The above equation will
+tend to be minimized for a smaller subtree.
+
+
    
+It turns out that as we increase \( \alpha \) from zero
+branches get pruned from the tree in a nested and predictable fashion,
+so obtaining the whole sequence of subtrees as a function of \( \alpha \) is
+easy. We can select a value of \( \alpha \) using a validation set or using
+cross-validation. We then return to the full data set and obtain the
+subtree corresponding to \( \alpha \).
+
    
+
+
+
    
+
    Schematic Regression Procedure 
    
+
+
    
+
    
+Building a Regression Tree.
+
      
+
    1.  Use recursive binary splitting to grow a large tree on the training data, stopping only when each terminal node has fewer than some minimum number of observations.
      2. 
+
    2.  Apply cost complexity pruning to the large tree in order to obtain a sequence of best subtrees, as a function of \( \alpha \).
      3. 
+
    3.  Use for example \( K \)-fold cross-validation to choose \( \alpha \). Divide the training observations into \( K \) folds. For each \( k=1,2,\dots,K \) we:
      4.  
+
+
        
+
+
      •  repeat steps 1 and 2 on all but the \( k \)-th fold of the training data.
        • 
+
+
      •  Then we valuate the mean squared prediction error on the data in the left-out \( k \)-th fold, as a function of \( \alpha \).
        • 
+
+
      •  Finally  we average the results for each value of \( \alpha \), and pick \( \alpha \) to minimize the average error.
        • 
+
      
+
    4.  Return the subtree from Step 2 that corresponds to the chosen value of \( \alpha \).
      5.  
+
    
+
    
+
    
+
+
+
    
+
    A Classification Tree 
    
+
+
    
+A classification tree is very similar to a regression tree, except
+that it is used to predict a qualitative response rather than a
+quantitative one. Recall that for a regression tree, the predicted
+response for an observation is given by the mean response of the
+training observations that belong to the same terminal node. In
+contrast, for a classification tree, we predict that each observation
+belongs to the most commonly occurring class of training observations
+in the region to which it belongs. In interpreting the results of a
+classification tree, we are often interested not only in the class
+prediction corresponding to a particular terminal node region, but
+also in the class proportions among the training observations that
+fall into that region.
+
    
+
+
+
    
+
    Growing a classification tree 
    
+
+
    
+The task of growing a
+classification tree is quite similar to the task of growing a
+regression tree. Just as in the regression setting, we use recursive
+binary splitting to grow a classification tree. However, in the
+classification setting, the MSE cannot be used as a criterion for making
+the binary splits.  A natural alternative to MSE is the classification
+error rate. Since we plan to assign an observation in a given region
+to the most commonly occurring error rate class of training
+observations in that region, the classification error rate is simply
+the fraction of the training observations in that region that do not
+belong to the most common class.
+
+
    
+When building a classification tree, either the Gini index or the
+entropy are typically used to evaluate the quality of a particular
+split, since these two approaches are more sensitive to node purity
+than is the classification error rate.
+
    
+
+
+
    
+
    Classification tree, how to split nodes 
    
+
+
    
+If our targets are the outcome of a classification process that takes
+for example \( k=1,2,\dots,K \) values, the only thing we need to think of
+is to set up the splitting criteria for each node.
+
+
    
+We define a PDF \( p_{mk} \) that represents the number of observations of
+a class \( k \) in a region \( R_m \) with \( N_m \) observations. We represent
+this likelihood function in terms of the proportion \( I(y_i=k) \) of
+observations of this class in the region \( R_m \) as
+
+
     
    
+$$
+p_{mk} = \frac{1}{N_m}\sum_{x_i\in R_m}I(y_i=k).
+$$
+
     
    
+
+
    
+We let \( p_{mk} \) represent the majority class of observations in region
+\( m \). The three most common ways of splitting a node are given by
+
+
      
+
    •  Misclassification error
      •  
+
    
+
     
    
+$$
+p_{mk} = \frac{1}{N_m}\sum_{x_i\in R_m}I(y_i\ne k) = 1-p_{mk}.
+$$
+
     
    
+
+
+
      
+
    •  Gini index \( g \)
      • 
+
    
+
     
    
+$$
+g = \sum_{k=1}^K p_{mk}(1-p_{mk}).
+$$
+
     
    
+
+
+
      
+
    •  Information entropy or just entropy \( s \)
      • 
+
    
+
     
    
+$$
+s = -\sum_{k=1}^K p_{mk}\log{p_{mk}}.
+$$
+
     
    
+
    
+
+
+
    
+
    Visualizing the Tree, Classification 
    
+
    
+
+
+
    import os
+from sklearn.datasets import load_breast_cancer
+from sklearn.tree import DecisionTreeClassifier
+from sklearn.model_selection import train_test_split
+from sklearn.metrics import confusion_matrix
+from sklearn.tree import export_graphviz
+
+from IPython.display import Image 
+from pydot import graph_from_dot_data
+import pandas as pd
+import numpy as np
+
+
+cancer = load_breast_cancer()
+X = pd.DataFrame(cancer.data, columns=cancer.feature_names)
+print(X)
+y = pd.Categorical.from_codes(cancer.target, cancer.target_names)
+y = pd.get_dummies(y)
+print(y)
+X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=1)
+tree_clf = DecisionTreeClassifier(max_depth=5)
+tree_clf.fit(X_train, y_train)
+
+export_graphviz(
+    tree_clf,
+    out_file="DataFiles/cancer.dot",
+    feature_names=cancer.feature_names,
+    class_names=cancer.target_names,
+    rounded=True,
+    filled=True
+)
+cmd = 'dot -Tpng DataFiles/cancer.dot -o DataFiles/cancer.png'
+os.system(cmd)
+
    
+
    
+
+
+
    
+
    Visualizing the Tree, The Moons  
    
+
    
+
+
+
    # Common imports
+import numpy as np
+from sklearn.model_selection import  train_test_split 
+from sklearn.tree import DecisionTreeClassifier
+from sklearn.datasets import make_moons
+from sklearn.tree import export_graphviz
+from pydot import graph_from_dot_data
+import pandas as pd
+import os
+
+np.random.seed(42)
+X, y = make_moons(n_samples=100, noise=0.25, random_state=53)
+X_train, X_test, y_train, y_test = train_test_split(X,y,random_state=0)
+tree_clf = DecisionTreeClassifier(max_depth=5)
+tree_clf.fit(X_train, y_train)
+
+export_graphviz(
+    tree_clf,
+    out_file="DataFiles/moons.dot",
+    rounded=True,
+    filled=True
+)
+cmd = 'dot -Tpng DataFiles/moons.dot -o DataFiles/moons.png'
+os.system(cmd)
+
    
+
    
+
+
+
    
+
    Algorithms for Setting up Decision Trees 
    
+
+
    
+Two algorithms stand out in the set up of decision trees:
+
+
      
+
    1.  The CART (Classification And Regression Tree) algorithm for both classification and regression
      2. 
+
    2.  The ID3 algorithm based on the computation of the information gain for classification
      3. 
+
    
+
    
+
+We discuss both algorithms with applications here. The popular library Scikit-Learn uses the CART algorithm. For classification problems you can use either the gini index or the entropy to split a tree in two branches.
+
    
+
+
+
    
+
    The CART algorithm for Classification 
    
+
    
+
+
+
    
+
    The CART algorithm for Regression 
    
+
    
+
+
+
    
+
    Computing the Gini index 
    
+
+
    
+The example we will look at is a classical one in many Machine
+Learning applications. Based on various meteorological features, we
+have several so-called attributes which decide whether we at the end
+will do some outdoor activity like skiing, going for a bike ride etc
+etc.  The table here contains the feautures outlook, temperature,
+humidity and wind.  The target or output is whether we ride
+(True=1) or whether we do something else that day (False=0). The
+attributes for each feature are then sunny, overcast and rain for the
+outlook, hot, cold and mild for temperature, high and normal for
+humidity and weak and strong for wind.
+
+
    
+The table here summarizes the various attributes and
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
    Day Outlook  Temperature Humidity  Wind  Ride 
       1          Sunny           Hot                High            Weak          0        
       2          Sunny           Hot                High            Strong        1        
       3          Overcast        Hot                High            Weak          1        
       4          Rain            Mild               High            Weak          1        
       5          Rain            Cool               Normal          Weak          1        
       6          Rain            Cool               Normal          Strong        0        
       7          Overcast        Cool               Normal          Strong        1        
       8          Sunny           Mild               High            Weak          0        
       9          Sunny           Cool               Normal          Weak          1        
       10         Rain            Mild               Normal          Weak          1        
       11         Sunny           Mild               Normal          Strong        1        
       12         Overcast        Mild               High            Strong        1        
       13         Overcast        Hot                Normal          Weak          1        
       14         Rain            Mild               High            Strong        0        
    
+
    
+
+
+
    
+
    Simple Python Code to read in Data and perform Classification 
    
+
+
    
+
+
+
    # Common imports
+import numpy as np
+import pandas as pd
+import matplotlib.pyplot as plt
+from sklearn.tree import DecisionTreeClassifier
+from sklearn.model_selection import train_test_split
+from sklearn.tree import export_graphviz
+from sklearn.preprocessing import StandardScaler, OneHotEncoder
+from sklearn.compose import ColumnTransformer
+from IPython.display import Image 
+from pydot import graph_from_dot_data
+import os
+
+# Where to save the figures and data files
+PROJECT_ROOT_DIR = "Results"
+FIGURE_ID = "Results/FigureFiles"
+DATA_ID = "DataFiles/"
+
+if not os.path.exists(PROJECT_ROOT_DIR):
+    os.mkdir(PROJECT_ROOT_DIR)
+
+if not os.path.exists(FIGURE_ID):
+    os.makedirs(FIGURE_ID)
+
+if not os.path.exists(DATA_ID):
+    os.makedirs(DATA_ID)
+
+def image_path(fig_id):
+    return os.path.join(FIGURE_ID, fig_id)
+
+def data_path(dat_id):
+    return os.path.join(DATA_ID, dat_id)
+
+def save_fig(fig_id):
+    plt.savefig(image_path(fig_id) + ".png", format='png')
+
+infile = open(data_path("rideclass.csv"),'r')
+
+# Read the experimental data with Pandas
+from IPython.display import display
+ridedata = pd.read_csv(infile,names = ('Outlook','Temperature','Humidity','Wind','Ride'))
+ridedata = pd.DataFrame(ridedata)
+
+# Features and targets
+X = ridedata.loc[:, ridedata.columns != 'Ride'].values
+y = ridedata.loc[:, ridedata.columns == 'Ride'].values
+
+# Create the encoder.
+encoder = OneHotEncoder(handle_unknown="ignore")
+# Assume for simplicity all features are categorical.
+encoder.fit(X)    
+# Apply the encoder.
+X = encoder.transform(X)
+print(X)
+# Then do a Classification tree
+tree_clf = DecisionTreeClassifier(max_depth=2)
+tree_clf.fit(X, y)
+print("Train set accuracy with Decision Tree: {:.2f}".format(tree_clf.score(X,y)))
+#transfer to a decision tree graph
+export_graphviz(
+    tree_clf,
+    out_file="DataFiles/ride.dot",
+    rounded=True,
+    filled=True
+)
+cmd = 'dot -Tpng DataFiles/cancer.dot -o DataFiles/cancer.png'
+os.system(cmd)
+
    
+
    
+
+
+
    
+
    Computing the Gini Factor  
    
+
+
    
+The above functions (gini, entropy and misclassification error) are
+important components of the so-called CART algorithm. We will discuss
+this algorithm below after we have discussed the information gain
+algorithm ID3.
+
+
    
+In the example here we have converted all our attributes into numerical values \( 0,1,2 \) etc.
+
+
    
+
+
+
    # Split a dataset based on an attribute and an attribute value
+def test_split(index, value, dataset):
+	left, right = list(), list()
+	for row in dataset:
+		if row[index] < value:
+			left.append(row)
+		else:
+			right.append(row)
+	return left, right
+ 
+# Calculate the Gini index for a split dataset
+def gini_index(groups, classes):
+	# count all samples at split point
+	n_instances = float(sum([len(group) for group in groups]))
+	# sum weighted Gini index for each group
+	gini = 0.0
+	for group in groups:
+		size = float(len(group))
+		# avoid divide by zero
+		if size == 0:
+			continue
+		score = 0.0
+		# score the group based on the score for each class
+		for class_val in classes:
+			p = [row[-1] for row in group].count(class_val) / size
+			score += p * p
+		# weight the group score by its relative size
+		gini += (1.0 - score) * (size / n_instances)
+	return gini
+
+# Select the best split point for a dataset
+def get_split(dataset):
+	class_values = list(set(row[-1] for row in dataset))
+	b_index, b_value, b_score, b_groups = 999, 999, 999, None
+	for index in range(len(dataset[0])-1):
+		for row in dataset:
+			groups = test_split(index, row[index], dataset)
+			gini = gini_index(groups, class_values)
+			print('X%d < %.3f Gini=%.3f' % ((index+1), row[index], gini))
+			if gini < b_score:
+				b_index, b_value, b_score, b_groups = index, row[index], gini, groups
+	return {'index':b_index, 'value':b_value, 'groups':b_groups}
+ 
+dataset = [[0,0,0,0,0],
+            [0,0,0,1,1],
+            [1,0,0,0,1],
+            [2,1,0,0,1],
+            [2,2,1,0,1],
+            [2,2,1,1,0],
+            [1,2,1,1,1],
+            [0,1,0,0,0],
+            [0,2,1,0,1],
+            [2,1,1,0,1],
+            [0,1,1,1,1],
+            [1,1,0,1,1],
+            [1,0,1,0,1],
+            [2,1,0,1,0]]
+
+split = get_split(dataset)
+print('Split: [X%d < %.3f]' % ((split['index']+1), split['value']))
+
    
+
    
+
+
+
    
+
    Entropy and the ID3 algorithm 
    
+
+
    
+ID3, learns decision trees by constructing
+them topdown, beginning with the question which attribute should be tested at the root of the tree?
+
+
      
+
    1.  Each instance attribute is evaluated using a statistical test to determine how well it alone classifies the training examples.
      2. 
+
    2.  The best attribute is selected and used as the test at the root node of the tree.
      3. 
+
    3.  A descendant of the root node is then created for each possible value of this attribute.
      4. 
+
    4.  Training examples are sorted to the appropriate descendant node.
      5. 
+
    5.  The entire process is then repeated using the training examples associated with each descendant node to select the best attribute to test at that point in the tree.
      6. 
+
    6.  This forms a greedy search for an acceptable decision tree, in which the algorithm never backtracks to reconsider earlier choices.
      7.  
+
    
+
    
+
+The ID3 algorithm selects, which attribute to test at each node in the
+tree.
+
+
    
+We would like to select the attribute that is most useful for classifying
+examples.
+
+
    
+What is a good quantitative measure of the worth of an attribute?
+
+
    
+Information gain measures how well a given attribute separates the
+training examples according to their target classification.
+
+
    
+The ID3 algorithm uses this information gain measure to select among the candidate
+attributes at each step while growing the tree.
+
    
+
+
+
    
+
    Implementing the ID3 Algorithm 
    
+
+
    
+
+
+
    import re
+import math
+from collections import deque
+
+# x is examples in training set
+# y is set of targets
+# label is target attributes
+# Node is a class which has properties values, childs, and next
+# root is top node in the decision tree
+
+class Node(object):
+	def __init__(self):
+		self.value = None
+		self.next = None
+		self.childs = None
+
+# Simple class of Decision Tree
+# Aimed for who want to learn Decision Tree, so it is not optimized
+class DecisionTree(object):
+	def __init__(self, sample, attributes, labels):
+		self.sample = sample
+		self.attributes = attributes
+		self.labels = labels
+		self.labelCodes = None
+		self.labelCodesCount = None
+		self.initLabelCodes()
+		# print(self.labelCodes)
+		self.root = None
+		self.entropy = self.getEntropy([x for x in range(len(self.labels))])
+
+	def initLabelCodes(self):
+		self.labelCodes = []
+		self.labelCodesCount = []
+		for l in self.labels:
+			if l not in self.labelCodes:
+				self.labelCodes.append(l)
+				self.labelCodesCount.append(0)
+			self.labelCodesCount[self.labelCodes.index(l)] += 1
+
+	def getLabelCodeId(self, sampleId):
+		return self.labelCodes.index(self.labels[sampleId])
+
+	def getAttributeValues(self, sampleIds, attributeId):
+		vals = []
+		for sid in sampleIds:
+			val = self.sample[sid][attributeId]
+			if val not in vals:
+				vals.append(val)
+		# print(vals)
+		return vals
+
+	def getEntropy(self, sampleIds):
+		entropy = 0
+		labelCount = [0] * len(self.labelCodes)
+		for sid in sampleIds:
+			labelCount[self.getLabelCodeId(sid)] += 1
+		# print("-ge", labelCount)
+		for lv in labelCount:
+			# print(lv)
+			if lv != 0:
+				entropy += -lv/len(sampleIds) * math.log(lv/len(sampleIds), 2)
+			else:
+				entropy += 0
+		return entropy
+
+	def getDominantLabel(self, sampleIds):
+		labelCodesCount = [0] * len(self.labelCodes)
+		for sid in sampleIds:
+			labelCodesCount[self.labelCodes.index(self.labels[sid])] += 1
+		return self.labelCodes[labelCodesCount.index(max(labelCodesCount))]
+
+	def getInformationGain(self, sampleIds, attributeId):
+		gain = self.getEntropy(sampleIds)
+		attributeVals = []
+		attributeValsCount = []
+		attributeValsIds = []
+		for sid in sampleIds:
+			val = self.sample[sid][attributeId]
+			if val not in attributeVals:
+				attributeVals.append(val)
+				attributeValsCount.append(0)
+				attributeValsIds.append([])
+			vid = attributeVals.index(val)
+			attributeValsCount[vid] += 1
+			attributeValsIds[vid].append(sid)
+		# print("-gig", self.attributes[attributeId])
+		for vc, vids in zip(attributeValsCount, attributeValsIds):
+			# print("-gig", vids)
+			gain -= vc/len(sampleIds) * self.getEntropy(vids)
+		return gain
+
+	def getAttributeMaxInformationGain(self, sampleIds, attributeIds):
+		attributesEntropy = [0] * len(attributeIds)
+		for i, attId in zip(range(len(attributeIds)), attributeIds):
+			attributesEntropy[i] = self.getInformationGain(sampleIds, attId)
+		maxId = attributeIds[attributesEntropy.index(max(attributesEntropy))]
+		return self.attributes[maxId], maxId
+
+	def isSingleLabeled(self, sampleIds):
+		label = self.labels[sampleIds[0]]
+		for sid in sampleIds:
+			if self.labels[sid] != label:
+				return False
+		return True
+
+	def getLabel(self, sampleId):
+		return self.labels[sampleId]
+
+	def id3(self):
+		sampleIds = [x for x in range(len(self.sample))]
+		attributeIds = [x for x in range(len(self.attributes))]
+		self.root = self.id3Recv(sampleIds, attributeIds, self.root)
+
+	def id3Recv(self, sampleIds, attributeIds, root):
+		root = Node() # Initialize current root
+		if self.isSingleLabeled(sampleIds):
+			root.value = self.labels[sampleIds[0]]
+			return root
+		# print(attributeIds)
+		if len(attributeIds) == 0:
+			root.value = self.getDominantLabel(sampleIds)
+			return root
+		bestAttrName, bestAttrId = self.getAttributeMaxInformationGain(
+			sampleIds, attributeIds)
+		# print(bestAttrName)
+		root.value = bestAttrName
+		root.childs = []  # Create list of children
+		for value in self.getAttributeValues(sampleIds, bestAttrId):
+			# print(value)
+			child = Node()
+			child.value = value
+			root.childs.append(child)  # Append new child node to current
+									   # root
+			childSampleIds = []
+			for sid in sampleIds:
+				if self.sample[sid][bestAttrId] == value:
+					childSampleIds.append(sid)
+			if len(childSampleIds) == 0:
+				child.next = self.getDominantLabel(sampleIds)
+			else:
+				# print(bestAttrName, bestAttrId)
+				# print(attributeIds)
+				if len(attributeIds) > 0 and bestAttrId in attributeIds:
+					toRemove = attributeIds.index(bestAttrId)
+					attributeIds.pop(toRemove)
+				child.next = self.id3Recv(
+					childSampleIds, attributeIds, child.next)
+		return root
+
+	def printTree(self):
+		if self.root:
+			roots = deque()
+			roots.append(self.root)
+			while len(roots) > 0:
+				root = roots.popleft()
+				print(root.value)
+				if root.childs:
+					for child in root.childs:
+						print('({})'.format(child.value))
+						roots.append(child.next)
+				elif root.next:
+					print(root.next)
+
+
+def test():
+	f = open('DataFiles/rideclass.csv')
+	attributes = f.readline().split(',')
+	attributes = attributes[1:len(attributes)-1]
+	print(attributes)
+	sample = f.readlines()
+	f.close()
+	for i in range(len(sample)):
+		sample[i] = re.sub('\d+,', '', sample[i])
+		sample[i] = sample[i].strip().split(',')
+	labels = []
+	for s in sample:
+		labels.append(s.pop())
+	# print(sample)
+	# print(labels)
+	decisionTree = DecisionTree(sample, attributes, labels)
+	print("System entropy {}".format(decisionTree.entropy))
+	decisionTree.id3()
+	decisionTree.printTree()
+
+
+if __name__ == '__main__':
+	test()
+
    
+
    
+
+
+
    
+
    Cancer Data again now with Decision Trees and other Methods 
    
+
    
+
+
+
    import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import  train_test_split 
+from sklearn.datasets import load_breast_cancer
+from sklearn.svm import SVC
+from sklearn.linear_model import LogisticRegression
+from sklearn.tree import DecisionTreeClassifier
+
+# Load the data
+cancer = load_breast_cancer()
+
+X_train, X_test, y_train, y_test = train_test_split(cancer.data,cancer.target,random_state=0)
+print(X_train.shape)
+print(X_test.shape)
+# Logistic Regression
+logreg = LogisticRegression(solver='lbfgs')
+logreg.fit(X_train, y_train)
+print("Test set accuracy with Logistic Regression: {:.2f}".format(logreg.score(X_test,y_test)))
+# Support vector machine
+svm = SVC(gamma='auto', C=100)
+svm.fit(X_train, y_train)
+print("Test set accuracy with SVM: {:.2f}".format(svm.score(X_test,y_test)))
+# Decision Trees
+deep_tree_clf = DecisionTreeClassifier(max_depth=None)
+deep_tree_clf.fit(X_train, y_train)
+print("Test set accuracy with Decision Trees: {:.2f}".format(deep_tree_clf.score(X_test,y_test)))
+#now scale the data
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+# Logistic Regression
+logreg.fit(X_train_scaled, y_train)
+print("Test set accuracy Logistic Regression with scaled data: {:.2f}".format(logreg.score(X_test_scaled,y_test)))
+# Support Vector Machine
+svm.fit(X_train_scaled, y_train)
+print("Test set accuracy SVM with scaled data: {:.2f}".format(logreg.score(X_test_scaled,y_test)))
+# Decision Trees
+deep_tree_clf.fit(X_train_scaled, y_train)
+print("Test set accuracy with Decision Trees and scaled data: {:.2f}".format(deep_tree_clf.score(X_test_scaled,y_test)))
+
    
+
    
+
+
+
    
+
    Another example, the moons again 
    
+
    
+
+
+
    from __future__ import division, print_function, unicode_literals
+
+# Common imports
+import numpy as np
+import os
+
+# to make this notebook's output stable across runs
+np.random.seed(42)
+
+# To plot pretty figures
+import matplotlib
+import matplotlib.pyplot as plt
+from matplotlib.colors import ListedColormap
+plt.rcParams['axes.labelsize'] = 14
+plt.rcParams['xtick.labelsize'] = 12
+plt.rcParams['ytick.labelsize'] = 12
+
+
+from sklearn.svm import SVC
+from sklearn import datasets
+from sklearn.tree import DecisionTreeClassifier
+from sklearn.datasets import make_moons
+from sklearn.tree import export_graphviz
+
+Xm, ym = make_moons(n_samples=100, noise=0.25, random_state=53)
+
+deep_tree_clf1 = DecisionTreeClassifier(random_state=42)
+deep_tree_clf2 = DecisionTreeClassifier(min_samples_leaf=4, random_state=42)
+deep_tree_clf1.fit(Xm, ym)
+deep_tree_clf2.fit(Xm, ym)
+
+
+def plot_decision_boundary(clf, X, y, axes=[0, 7.5, 0, 3], iris=True, legend=False, plot_training=True):
+    x1s = np.linspace(axes[0], axes[1], 100)
+    x2s = np.linspace(axes[2], axes[3], 100)
+    x1, x2 = np.meshgrid(x1s, x2s)
+    X_new = np.c_[x1.ravel(), x2.ravel()]
+    y_pred = clf.predict(X_new).reshape(x1.shape)
+    custom_cmap = ListedColormap(['#fafab0','#9898ff','#a0faa0'])
+    plt.contourf(x1, x2, y_pred, alpha=0.3, cmap=custom_cmap)
+    if not iris:
+        custom_cmap2 = ListedColormap(['#7d7d58','#4c4c7f','#507d50'])
+        plt.contour(x1, x2, y_pred, cmap=custom_cmap2, alpha=0.8)
+    if plot_training:
+        plt.plot(X[:, 0][y==0], X[:, 1][y==0], "yo", label="Iris-Setosa")
+        plt.plot(X[:, 0][y==1], X[:, 1][y==1], "bs", label="Iris-Versicolor")
+        plt.plot(X[:, 0][y==2], X[:, 1][y==2], "g^", label="Iris-Virginica")
+        plt.axis(axes)
+    if iris:
+        plt.xlabel("Petal length", fontsize=14)
+        plt.ylabel("Petal width", fontsize=14)
+    else:
+        plt.xlabel(r"$x_1$", fontsize=18)
+        plt.ylabel(r"$x_2$", fontsize=18, rotation=0)
+    if legend:
+        plt.legend(loc="lower right", fontsize=14)
+plt.figure(figsize=(11, 4))
+plt.subplot(121)
+plot_decision_boundary(deep_tree_clf1, Xm, ym, axes=[-1.5, 2.5, -1, 1.5], iris=False)
+plt.title("No restrictions", fontsize=16)
+plt.subplot(122)
+plot_decision_boundary(deep_tree_clf2, Xm, ym, axes=[-1.5, 2.5, -1, 1.5], iris=False)
+plt.title("min_samples_leaf = {}".format(deep_tree_clf2.min_samples_leaf), fontsize=14)
+plt.show()
+
    
+
    
+
+
+
    
+
    Playing around with regions 
    
+
    
+
+
+
    np.random.seed(6)
+Xs = np.random.rand(100, 2) - 0.5
+ys = (Xs[:, 0] > 0).astype(np.float32) * 2
+
+angle = np.pi/4
+rotation_matrix = np.array([[np.cos(angle), -np.sin(angle)], [np.sin(angle), np.cos(angle)]])
+Xsr = Xs.dot(rotation_matrix)
+
+tree_clf_s = DecisionTreeClassifier(random_state=42)
+tree_clf_s.fit(Xs, ys)
+tree_clf_sr = DecisionTreeClassifier(random_state=42)
+tree_clf_sr.fit(Xsr, ys)
+
+plt.figure(figsize=(11, 4))
+plt.subplot(121)
+plot_decision_boundary(tree_clf_s, Xs, ys, axes=[-0.7, 0.7, -0.7, 0.7], iris=False)
+plt.subplot(122)
+plot_decision_boundary(tree_clf_sr, Xsr, ys, axes=[-0.7, 0.7, -0.7, 0.7], iris=False)
+
+plt.show()
+
    
+
    
+
+
+
    
+
    Regression trees 
    
+
    
+
+
+
    # Quadratic training set + noise
+np.random.seed(42)
+m = 200
+X = np.random.rand(m, 1)
+y = 4 * (X - 0.5) ** 2
+y = y + np.random.randn(m, 1) / 10
+
    
+
    
+
+
+
    from sklearn.tree import DecisionTreeRegressor
+
+tree_reg = DecisionTreeRegressor(max_depth=2, random_state=42)
+tree_reg.fit(X, y)
+
    
+
    
+
+
+
    
+
    Final regressor code 
    
+
    
+
+
+
    from sklearn.tree import DecisionTreeRegressor
+
+tree_reg1 = DecisionTreeRegressor(random_state=42, max_depth=2)
+tree_reg2 = DecisionTreeRegressor(random_state=42, max_depth=3)
+tree_reg1.fit(X, y)
+tree_reg2.fit(X, y)
+
+def plot_regression_predictions(tree_reg, X, y, axes=[0, 1, -0.2, 1], ylabel="$y$"):
+    x1 = np.linspace(axes[0], axes[1], 500).reshape(-1, 1)
+    y_pred = tree_reg.predict(x1)
+    plt.axis(axes)
+    plt.xlabel("$x_1$", fontsize=18)
+    if ylabel:
+        plt.ylabel(ylabel, fontsize=18, rotation=0)
+    plt.plot(X, y, "b.")
+    plt.plot(x1, y_pred, "r.-", linewidth=2, label=r"$\hat{y}$")
+
+plt.figure(figsize=(11, 4))
+plt.subplot(121)
+plot_regression_predictions(tree_reg1, X, y)
+for split, style in ((0.1973, "k-"), (0.0917, "k--"), (0.7718, "k--")):
+    plt.plot([split, split], [-0.2, 1], style, linewidth=2)
+plt.text(0.21, 0.65, "Depth=0", fontsize=15)
+plt.text(0.01, 0.2, "Depth=1", fontsize=13)
+plt.text(0.65, 0.8, "Depth=1", fontsize=13)
+plt.legend(loc="upper center", fontsize=18)
+plt.title("max_depth=2", fontsize=14)
+
+plt.subplot(122)
+plot_regression_predictions(tree_reg2, X, y, ylabel=None)
+for split, style in ((0.1973, "k-"), (0.0917, "k--"), (0.7718, "k--")):
+    plt.plot([split, split], [-0.2, 1], style, linewidth=2)
+for split in (0.0458, 0.1298, 0.2873, 0.9040):
+    plt.plot([split, split], [-0.2, 1], "k:", linewidth=1)
+plt.text(0.3, 0.5, "Depth=2", fontsize=13)
+plt.title("max_depth=3", fontsize=14)
+
+plt.show()
+
    
+
    
+
+
+
    tree_reg1 = DecisionTreeRegressor(random_state=42)
+tree_reg2 = DecisionTreeRegressor(random_state=42, min_samples_leaf=10)
+tree_reg1.fit(X, y)
+tree_reg2.fit(X, y)
+
+x1 = np.linspace(0, 1, 500).reshape(-1, 1)
+y_pred1 = tree_reg1.predict(x1)
+y_pred2 = tree_reg2.predict(x1)
+
+plt.figure(figsize=(11, 4))
+
+plt.subplot(121)
+plt.plot(X, y, "b.")
+plt.plot(x1, y_pred1, "r.-", linewidth=2, label=r"$\hat{y}$")
+plt.axis([0, 1, -0.2, 1.1])
+plt.xlabel("$x_1$", fontsize=18)
+plt.ylabel("$y$", fontsize=18, rotation=0)
+plt.legend(loc="upper center", fontsize=18)
+plt.title("No restrictions", fontsize=14)
+
+plt.subplot(122)
+plt.plot(X, y, "b.")
+plt.plot(x1, y_pred2, "r.-", linewidth=2, label=r"$\hat{y}$")
+plt.axis([0, 1, -0.2, 1.1])
+plt.xlabel("$x_1$", fontsize=18)
+plt.title("min_samples_leaf={}".format(tree_reg2.min_samples_leaf), fontsize=14)
+
+plt.show()
+
    
+
    
+
+
+
    
+
    Pros and cons of trees, pros 
    
+
+
      
+
    •  White box, easy to interpret model. Some people believe that decision trees more closely mirror human decision-making than do the regression and classification approaches discussed earlier (think of support vector machines)
      • 
+
    •  Trees are very easy to explain to people. In fact, they are even easier to explain than linear regression!
      • 
+
    •  No feature normalization needed
      • 
+
    •  Tree models can handle both continuous and categorical data (Classification and Regression Trees)
      • 
+
    •  Can model nonlinear relationships
      • 
+
    •  Can model interactions between the different descriptive features
      • 
+
    •  Trees can be displayed graphically, and are easily interpreted even by a non-expert (especially if they are small)
      • 
+
    
+
    
+
+
+
    
+
    Disadvantages 
    
+
+
      
+
    •  Unfortunately, trees generally do not have the same level of predictive accuracy as some of the other regression and classification approaches
      • 
+
    •  If continuous features are used the tree may become quite large and hence less interpretable
      • 
+
    •  Decision trees are prone to overfit the training data and hence do not well generalize the data if no stopping criteria or improvements like pruning, boosting or bagging are implemented
      • 
+
    •  Small changes in the data may lead to a completely different tree. This issue can be addressed by using ensemble methods like bagging, boosting or random forests
      • 
+
    •  Unbalanced datasets where some target feature values occur much more frequently than others may lead to biased trees since the frequently occurring feature values are preferred over the less frequently occurring ones.
      •  
+
    •  If the number of features is relatively large (high dimensional) and the number of instances is relatively low, the tree might overfit the data
      • 
+
    •  Features with many levels may be preferred over features with less levels since for them it is more easy to split the dataset such that the sub datasets only contain pure target feature values. This issue can be addressed by preferring for instance the information gain ratio as splitting criteria over information gain
      • 
+
    
+
    
+
+However, by aggregating many decision trees, using methods like bagging, random forests, and boosting, the predictive performance of trees can be substantially improved.
+
    
+
+
+
    
+
    Ensemble Methods: From a Single Tree to Many Trees and Extreme Boosting, Meet the Jungle of Methods 
    
+
+
    
+As stated above and seen in many of the examples discussed here about
+a single decision tree, we often end up overfitting our training
+data. This normally means that we have a high variance. Can we reduce
+the variance of a statistical learning method?
+
+
    
+This leads us to a set of different methods that can combine different
+machine learning algorithms or just use one of them to construct forests and jungles of trees, homogeneous ones or heterogenous ones. These methods are recognized by different names which we will try to explain here. These are
+
+
      
+
    1.  Voting classifiers
      2. 
+
    2.  Bagging and Pasting
      3. 
+
    3.  Random forests
      4. 
+
    4.  Boosting methods, from adaptive to Extreme Gradient Boosting (XGBoost)
      5. 
+
    
+
    
+
+We discuss these methods here.
+
    
+
+
+
    
+
    An Overview of Ensemble Methods  
    
+
+
    
+



    
+
    
+
+
+
    
+
    Bagging 
    
+
+
    
+The plain decision trees suffer from high
+variance. This means that if we split the training data into two parts
+at random, and fit a decision tree to both halves, the results that we
+get could be quite different. In contrast, a procedure with low
+variance will yield similar results if applied repeatedly to distinct
+data sets; linear regression tends to have low variance, if the ratio
+of \( n \) to \( p \) is moderately large.
+
+
    
+Bootstrap aggregation, or just bagging, is a
+general-purpose procedure for reducing the variance of a statistical
+learning method.
+
    
+
+
+
    
+
    More bagging 
    
+
+
    
+Bagging typically results in improved accuracy
+over prediction using a single tree. Unfortunately, however, it can be
+difficult to interpret the resulting model. Recall that one of the
+advantages of decision trees is the attractive and easily interpreted
+diagram that results.
+
+
    
+However, when we bag a large number of trees, it is no longer
+possible to represent the resulting statistical learning procedure
+using a single tree, and it is no longer clear which variables are
+most important to the procedure. Thus, bagging improves prediction
+accuracy at the expense of interpretability.  Although the collection
+of bagged trees is much more difficult to interpret than a single
+tree, one can obtain an overall summary of the importance of each
+predictor using the MSE (for bagging regression trees) or the Gini
+index (for bagging classification trees). In the case of bagging
+regression trees, we can record the total amount that the MSE is
+decreased due to splits over a given predictor, averaged over all \( B \) possible
+trees. A large value indicates an important predictor. Similarly, in
+the context of bagging classification trees, we can add up the total
+amount that the Gini index  is decreased by splits over a given
+predictor, averaged over all \( B \) trees.
+
    
+
+
+
    
+
    Simple Voting Example, head or tail 
    
+
    
+
+
+
    heads_proba = 0.51
+coin_tosses = (np.random.rand(10000, 10) < heads_proba).astype(np.int32)
+cumulative_heads_ratio = np.cumsum(coin_tosses, axis=0) / np.arange(1, 10001).reshape(-1, 1)
+plt.figure(figsize=(8,3.5))
+plt.plot(cumulative_heads_ratio)
+plt.plot([0, 10000], [0.51, 0.51], "k--", linewidth=2, label="51%")
+plt.plot([0, 10000], [0.5, 0.5], "k-", label="50%")
+plt.xlabel("Number of coin tosses")
+plt.ylabel("Heads ratio")
+plt.legend(loc="lower right")
+plt.axis([0, 10000, 0.42, 0.58])
+plt.show()
+
    
+
    
+
+
+
    
+
    Using the Voting Classifier 
    
+
    
+
+
+
    from sklearn.model_selection import train_test_split
+from sklearn.datasets import make_moons
+
+X, y = make_moons(n_samples=500, noise=0.30, random_state=42)
+X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=42)
+
+from sklearn.ensemble import RandomForestClassifier
+from sklearn.ensemble import VotingClassifier
+from sklearn.linear_model import LogisticRegression
+from sklearn.svm import SVC
+
+log_clf = LogisticRegression(solver="liblinear", random_state=42)
+rnd_clf = RandomForestClassifier(n_estimators=10, random_state=42)
+svm_clf = SVC(gamma="auto", random_state=42)
+
+voting_clf = VotingClassifier(
+    estimators=[('lr', log_clf), ('rf', rnd_clf), ('svc', svm_clf)],
+    voting='hard')
+
+voting_clf.fit(X_train, y_train)
+
+from sklearn.metrics import accuracy_score
+
+for clf in (log_clf, rnd_clf, svm_clf, voting_clf):
+    clf.fit(X_train, y_train)
+    y_pred = clf.predict(X_test)
+    print(clf.__class__.__name__, accuracy_score(y_test, y_pred))
+
+log_clf = LogisticRegression(solver="liblinear", random_state=42)
+rnd_clf = RandomForestClassifier(n_estimators=10, random_state=42)
+svm_clf = SVC(gamma="auto", probability=True, random_state=42)
+
+voting_clf = VotingClassifier(
+    estimators=[('lr', log_clf), ('rf', rnd_clf), ('svc', svm_clf)],
+    voting='soft')
+voting_clf.fit(X_train, y_train)
+
+from sklearn.metrics import accuracy_score
+
+for clf in (log_clf, rnd_clf, svm_clf, voting_clf):
+    clf.fit(X_train, y_train)
+    y_pred = clf.predict(X_test)
+    print(clf.__class__.__name__, accuracy_score(y_test, y_pred))
+
    
+
    
+
+
+
    
+
    Please, not the moons again! Voting and Bagging 
    
+
    
+
+
+
    from sklearn.model_selection import train_test_split
+from sklearn.datasets import make_moons
+
+X, y = make_moons(n_samples=500, noise=0.30, random_state=42)
+X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=42)
+from sklearn.ensemble import RandomForestClassifier
+from sklearn.ensemble import VotingClassifier
+from sklearn.linear_model import LogisticRegression
+from sklearn.svm import SVC
+
+log_clf = LogisticRegression(random_state=42)
+rnd_clf = RandomForestClassifier(random_state=42)
+svm_clf = SVC(random_state=42)
+
+voting_clf = VotingClassifier(
+    estimators=[('lr', log_clf), ('rf', rnd_clf), ('svc', svm_clf)],
+    voting='hard')
+voting_clf.fit(X_train, y_train)
+
    
+
    
+
+
+
    from sklearn.metrics import accuracy_score
+
+for clf in (log_clf, rnd_clf, svm_clf, voting_clf):
+    clf.fit(X_train, y_train)
+    y_pred = clf.predict(X_test)
+    print(clf.__class__.__name__, accuracy_score(y_test, y_pred))
+
    
+
    
+
+
+
    log_clf = LogisticRegression(random_state=42)
+rnd_clf = RandomForestClassifier(random_state=42)
+svm_clf = SVC(probability=True, random_state=42)
+
+voting_clf = VotingClassifier(
+    estimators=[('lr', log_clf), ('rf', rnd_clf), ('svc', svm_clf)],
+    voting='soft')
+voting_clf.fit(X_train, y_train)
+
    
+
    
+
+
+
    from sklearn.metrics import accuracy_score
+
+for clf in (log_clf, rnd_clf, svm_clf, voting_clf):
+    clf.fit(X_train, y_train)
+    y_pred = clf.predict(X_test)
+    print(clf.__class__.__name__, accuracy_score(y_test, y_pred))
+
    
+
    
+
+
+
    
+
    Now Bagging  
    
+
+
    
+
+
+
    from sklearn.ensemble import BaggingClassifier
+from sklearn.tree import DecisionTreeClassifier
+
+bag_clf = BaggingClassifier(
+    DecisionTreeClassifier(random_state=42), n_estimators=500,
+    max_samples=100, bootstrap=True, n_jobs=-1, random_state=42)
+bag_clf.fit(X_train, y_train)
+y_pred = bag_clf.predict(X_test)
+
    
+
    
+
+
+
    from sklearn.metrics import accuracy_score
+print(accuracy_score(y_test, y_pred))
+
    
+
    
+
+
+
    tree_clf = DecisionTreeClassifier(random_state=42)
+tree_clf.fit(X_train, y_train)
+y_pred_tree = tree_clf.predict(X_test)
+print(accuracy_score(y_test, y_pred_tree))
+
    
+
    
+
+
+
    from matplotlib.colors import ListedColormap
+
+def plot_decision_boundary(clf, X, y, axes=[-1.5, 2.5, -1, 1.5], alpha=0.5, contour=True):
+    x1s = np.linspace(axes[0], axes[1], 100)
+    x2s = np.linspace(axes[2], axes[3], 100)
+    x1, x2 = np.meshgrid(x1s, x2s)
+    X_new = np.c_[x1.ravel(), x2.ravel()]
+    y_pred = clf.predict(X_new).reshape(x1.shape)
+    custom_cmap = ListedColormap(['#fafab0','#9898ff','#a0faa0'])
+    plt.contourf(x1, x2, y_pred, alpha=0.3, cmap=custom_cmap)
+    if contour:
+        custom_cmap2 = ListedColormap(['#7d7d58','#4c4c7f','#507d50'])
+        plt.contour(x1, x2, y_pred, cmap=custom_cmap2, alpha=0.8)
+    plt.plot(X[:, 0][y==0], X[:, 1][y==0], "yo", alpha=alpha)
+    plt.plot(X[:, 0][y==1], X[:, 1][y==1], "bs", alpha=alpha)
+    plt.axis(axes)
+    plt.xlabel(r"$x_1$", fontsize=18)
+    plt.ylabel(r"$x_2$", fontsize=18, rotation=0)
+plt.figure(figsize=(11,4))
+plt.subplot(121)
+plot_decision_boundary(tree_clf, X, y)
+plt.title("Decision Tree", fontsize=14)
+plt.subplot(122)
+plot_decision_boundary(bag_clf, X, y)
+plt.title("Decision Trees with Bagging", fontsize=14)
+plt.show()
+
    
+
    
+
+
+
    
+
    Making our own Bagging with Bootstrap 
    
+
    
+
+
+
    import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import train_test_split
+from sklearn.pipeline import make_pipeline
+from sklearn.utils import resample
+from sklearn.tree import DecisionTreeRegressor
+
+
+np.random.seed(2018)
+
+n = 40
+n_boostraps = 100
+maxdegree = 14
+
+# Make data set.
+x = np.linspace(-3, 3, n).reshape(-1, 1)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
+error = np.zeros(maxdegree)
+bias = np.zeros(maxdegree)
+variance = np.zeros(maxdegree)
+polydegree = np.zeros(maxdegree)
+X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
+
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+for degree in range(maxdegree):
+    model = DecisionTreeRegressor(max_depth=5) 
+    y_pred = np.empty((y_test.shape[0], n_boostraps))
+    for i in range(n_boostraps):
+        x_, y_ = resample(X_train_scaled, y_train)
+        model.fit(x_, y_)
+        y_pred[:, i] = model.predict(X_test_scaled).ravel()
+
+    polydegree[degree] = degree
+    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )
+    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )
+    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )
+    print('Polynomial degree:', degree)
+    print('Error:', error[degree])
+    print('Bias^2:', bias[degree])
+    print('Var:', variance[degree])
+    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
+
+plt.plot(polydegree, error, label='Error')
+plt.plot(polydegree, bias, label='bias')
+plt.plot(polydegree, variance, label='Variance')
+plt.legend()
+plt.show()
+
    
+
    
+
+
+
    
+
    Changing the Level of the Decision Tree 
    
+
+
    
+
+
+
    import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import train_test_split
+from sklearn.pipeline import make_pipeline
+from sklearn.utils import resample
+from sklearn.tree import DecisionTreeRegressor
+
+n = 100
+n_boostraps = 100
+maxdepth = 8
+
+# Make data set.
+x = np.linspace(-3, 3, n).reshape(-1, 1)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
+error = np.zeros(maxdepth)
+bias = np.zeros(maxdepth)
+variance = np.zeros(maxdepth)
+polydegree = np.zeros(maxdepth)
+X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
+
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+for degree in range(1,maxdepth):
+    model = DecisionTreeRegressor(max_depth=degree) 
+    y_pred = np.empty((y_test.shape[0], n_boostraps))
+    for i in range(n_boostraps):
+        x_, y_ = resample(X_train_scaled, y_train)
+        model.fit(x_, y_)
+        y_pred[:, i] = model.predict(X_test_scaled)#.ravel()
+
+    polydegree[degree] = degree
+    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )
+    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )
+    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )
+    print('Polynomial degree:', degree)
+    print('Error:', error[degree])
+    print('Bias^2:', bias[degree])
+    print('Var:', variance[degree])
+    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
+
+plt.xlim(1,maxdepth)
+plt.plot(polydegree, error, label='Error')
+plt.plot(polydegree, bias, label='bias')
+plt.plot(polydegree, variance, label='Variance')
+plt.legend()
+plt.show()
+
    
+
    
+
+
+
    
+
    Random forests 
    
+
+
    
+Random forests provide an improvement over bagged trees by way of a
+small tweak that decorrelates the trees.
+
+
    
+As in bagging, we build a
+number of decision trees on bootstrapped training samples. But when
+building these decision trees, each time a split in a tree is
+considered, a random sample of \( m \) predictors is chosen as split
+candidates from the full set of \( p \) predictors. The split is allowed to
+use only one of those \( m \) predictors.
+
+
    
+A fresh sample of \( m \) predictors is
+taken at each split, and typically we choose
+
+
     
    
+$$
+m\approx \sqrt{p}.
+$$
+
     
    
+
+
    
+In building a random forest, at
+each split in the tree, the algorithm is not even allowed to consider
+a majority of the available predictors.
+
+
    
+The reason for this is rather clever. Suppose that there is one very
+strong predictor in the data set, along with a number of other
+moderately strong predictors. Then in the collection of bagged
+variable importance random forest trees, most or all of the trees will
+use this strong predictor in the top split. Consequently, all of the
+bagged trees will look quite similar to each other. Hence the
+predictions from the bagged trees will be highly correlated.
+Unfortunately, averaging many highly correlated quantities does not
+lead to as large of a reduction in variance as averaging many
+uncorrelated quanti- ties. In particular, this means that bagging will
+not lead to a substantial reduction in variance over a single tree in
+this setting.
+
    
+
+
+
    
+
    Random Forest Algorithm 
    
+The algorithm described here can be applied to both classification and regression problems.
+
+
    
+We will grow of forest of say \( M \) trees.
+
+
      
+
    1.  For \( m=1:M \) we
      2. 
+
+
        
+
+
      •  Draw a bootstrap sample of from the training data organized in our \( \boldsymbol{X} \) matrix.
        • 
+
+
      •  We grow then a random forest tree \( T_m \) based on the bootstrapped data by repeating the steps outlined till we reach the maximum node size is reached
        • 
+
+
          
+
+
        1.  we select \( m \le p \) varibales at random from the \( p \) predictors/features
          2. 
+
+
        2.  pick the best split point among the \( m \) features using either the CART algorithm or the ID3 for classification and create a new node
          3. 
+
+
        3.  split the node into daughter nodes
          4. 
+
        
+
        
+
      
+
    2.  Output then the ensemble of trees \( \{T_m\}_1^{M} \) and make predictions for either a regression type of problem or a classification type of problem.
      3.  
+
    
+
    
+
+
+
    
+
    Bootstrap with Random Forests Instead of a Single Tree, own Bagging 
    
+
+
    
+
+
+
    import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import train_test_split
+from sklearn.pipeline import make_pipeline
+from sklearn.utils import resample
+from sklearn.ensemble import RandomForestRegressor
+
+np.random.seed(2018)
+
+n = 100
+n_boostraps = 100
+maxdegree = 14
+
+# Make data set.
+x = np.linspace(-3, 3, n).reshape(-1, 1)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
+error = np.zeros(maxdegree)
+bias = np.zeros(maxdegree)
+variance = np.zeros(maxdegree)
+polydegree = np.zeros(maxdegree)
+X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
+
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+for degree in range(maxdegree):
+    model = RandomForestRegressor()
+    y_pred = np.empty((y_test.shape[0], n_boostraps))
+    for i in range(n_boostraps):
+        x_, y_ = resample(X_train_scaled, y_train)
+        model.fit(x_, y_.ravel())
+        y_pred[:, i] = model.predict(X_test_scaled).ravel()
+
+    polydegree[degree] = degree
+    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )
+    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )
+    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )
+    print('Polynomial degree:', degree)
+    print('Error:', error[degree])
+    print('Bias^2:', bias[degree])
+    print('Var:', variance[degree])
+    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
+
+plt.plot(polydegree, error, label='Error')
+plt.plot(polydegree, bias, label='bias')
+plt.plot(polydegree, variance, label='Variance')
+plt.legend()
+plt.show()
+
    
+
    
+
+
+
    
+
    Random Forests Compared with other Methods on the Cancer Data 
    
+
    
+
+
+
    import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import  train_test_split 
+from sklearn.datasets import load_breast_cancer
+from sklearn.svm import SVC
+from sklearn.linear_model import LogisticRegression
+from sklearn.tree import DecisionTreeClassifier
+
+# Load the data
+cancer = load_breast_cancer()
+
+X_train, X_test, y_train, y_test = train_test_split(cancer.data,cancer.target,random_state=0)
+print(X_train.shape)
+print(X_test.shape)
+# Logistic Regression
+logreg = LogisticRegression(solver='lbfgs')
+logreg.fit(X_train, y_train)
+print("Test set accuracy with Logistic Regression: {:.2f}".format(logreg.score(X_test,y_test)))
+# Support vector machine
+svm = SVC(gamma='auto', C=100)
+svm.fit(X_train, y_train)
+print("Test set accuracy with SVM: {:.2f}".format(svm.score(X_test,y_test)))
+# Decision Trees
+deep_tree_clf = DecisionTreeClassifier(max_depth=None)
+deep_tree_clf.fit(X_train, y_train)
+print("Test set accuracy with Decision Trees: {:.2f}".format(deep_tree_clf.score(X_test,y_test)))
+#now scale the data
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+# Logistic Regression
+logreg.fit(X_train_scaled, y_train)
+print("Test set accuracy Logistic Regression with scaled data: {:.2f}".format(logreg.score(X_test_scaled,y_test)))
+# Support Vector Machine
+svm.fit(X_train_scaled, y_train)
+print("Test set accuracy SVM with scaled data: {:.2f}".format(logreg.score(X_test_scaled,y_test)))
+# Decision Trees
+deep_tree_clf.fit(X_train_scaled, y_train)
+print("Test set accuracy with Decision Trees and scaled data: {:.2f}".format(deep_tree_clf.score(X_test_scaled,y_test)))
+
+
+from sklearn.ensemble import RandomForestClassifier
+from sklearn.preprocessing import LabelEncoder
+from sklearn.model_selection import cross_validate
+# Data set not specificied
+#Instantiate the model with 500 trees and entropy as splitting criteria
+Random_Forest_model = RandomForestClassifier(n_estimators=500,criterion="entropy")
+Random_Forest_model.fit(X_train_scaled, y_train)
+#Cross validation
+accuracy = cross_validate(Random_Forest_model,X_test_scaled,y_test,cv=10)['test_score']
+print(accuracy)
+print("Test set accuracy with Random Forests and scaled data: {:.2f}".format(Random_Forest_model.score(X_test_scaled,y_test)))
+
+
+import scikitplot as skplt
+y_pred = Random_Forest_model.predict(X_test_scaled)
+skplt.metrics.plot_confusion_matrix(y_test, y_pred, normalize=True)
+plt.show()
+y_probas = Random_Forest_model.predict_proba(X_test_scaled)
+skplt.metrics.plot_roc(y_test, y_probas)
+plt.show()
+skplt.metrics.plot_cumulative_gain(y_test, y_probas)
+plt.show()
+
    
+
    
+
+
+
    
+
    Compare  Bagging on Trees with Random Forests  
    
+
    
+
+
+
    bag_clf = BaggingClassifier(
+    DecisionTreeClassifier(splitter="random", max_leaf_nodes=16, random_state=42),
+    n_estimators=500, max_samples=1.0, bootstrap=True, n_jobs=-1, random_state=42)
+
    
+
    
+
+
+
    bag_clf.fit(X_train, y_train)
+y_pred = bag_clf.predict(X_test)
+from sklearn.ensemble import RandomForestClassifier
+rnd_clf = RandomForestClassifier(n_estimators=500, max_leaf_nodes=16, n_jobs=-1, random_state=42)
+rnd_clf.fit(X_train, y_train)
+y_pred_rf = rnd_clf.predict(X_test)
+np.sum(y_pred == y_pred_rf) / len(y_pred) 
+
    
+
    
+
+
+
    
+
    Boosting, a Bird'e Eye 
    
+
+
    
+The basic idea is to combine weak classifiers in order to create a good
+classifier. With a weak classifier we often intend a classifier which
+produces results which are only slightly better than we would get by
+random guesses.
+
+
    
+This is done by applying in an iterative way a weak (or a standard
+classifier like decision trees) to modify the data. In each iteration
+we emphasize those observations which are misclassified by weighting
+them with a factor.
+
    
+
+
+
    
+
    Adaptive boosting: AdaBoost, Basic Algorithm 
    
+
+
    
+The algorithm here is rather straightforward. Assume that our weak
+classifier is a decision tree and we consider a binary set of outputs
+with \( y_i \in \{-1,1\} \) and \( i=0,1,2,\dots,n-1 \) as our set of
+observations. Our design matrix is given in terms of the
+feature/predictor vectors
+\( \boldsymbol{X}=[\boldsymbol{x}_0\boldsymbol{x}_1\dots\boldsymbol{x}_{p-1} \). Finally, we define also a
+classifier determined by our data via a function \( G(\boldsymbol{X}) \). This function tells us how well we are able to classify our outputs/targets \( \boldsymbol{y} \).
+
+
    
+We can then define the misclassification error \( \mathrm{err} \) as
+
     
    
+$$
+\mathrm{err}=\frac{1}{n}\sum_{i=0}^{n-1}I(y_i\ne G(\boldsymbol{X}_{i*}),
+$$
+
     
    
+
+where the function \( I() \) is one if we misclassify and zero if we classify correctly.
+
    
+
+
+
    
+
    Basic Steps of AdaBoost 
    
+
+
    
+With the above definitions we are now ready to set up the algorithm for AdaBoost.
+The basic idea is to set up weights which will be used to scale the correctly classified and the misclassified cases.
+
+
      
+
    1.  We start by initializing all weights to \( w_i = 1/n \), with \( i=0,1,2,\dots n-1 \). It is to see then that \( \sum_{i=0}^{n-1}w_i = 1 \).
      2. 
+
    2.  We rewrite the misclassification error as
      3.  
+
    
+
     
    
+$$
+\mathrm{err}=\frac{\sum_{i=0}^{n-1}w_iI(y_i\ne G(\boldsymbol{X}_{i*})}{\sum_{i=0}^{n-1}w_i},
+$$
+
     
    
+
+
+
      
+
    1.  Then we start looping over all attempts at classifying, namely we start an iterative process for \( m=1:M \), where \( M \) is the final number of classifications. Our given classifier could for example be a plain decision tree.
+
+
        
+ 
      1.  Fit then a given classifier to the training using the weights \( w_i \).
        2. 
+ 
      2.  Compute then \( \mathrm{err} \) and figure out which events are classified properly and which are classified wrongly.
        3. 
+ 
      3.  Define a quantity $\alpha_{m} = \log{(1-\mathrm{err})/\mathrm{err}}
        4. 
+ 
      4.  Set the new weights to $w_i = w_i\times \exp{(\alpha_m I(y_i\ne G(\boldsymbol{X}_{i*})}.
        5. 
+
      
+
    2.  Compute the new classifier $G(\boldsymbol{X})= \sum_{i=0}^{n-1}\alpha_m I(y_i\ne G(\boldsymbol{X}_{i*}).
      3. 
+
    
+
    
+
+For the iterations with \( m \le 2 \) the weights are modified
+individually at each steps. The obersvations which were misclassified
+at iteration \( m-1 \) have a weight which is larger than those which were
+classified properly. As this proceeds, the observations which were
+difficult to classifiy correctly are given a larger influence. Each
+new classification step \( m \) is then forced to concentrate on those
+observations that are missed in the previous iterations.
+
    
+
+
+
    
+
    AdaBoost Examples 
    
+
+
    
+Using Scikit-Learn it is easy to appply the adaptive boosting algorithm, as done here.
+
+
    
+
+
+
    from sklearn.ensemble import AdaBoostClassifier
+
+ada_clf = AdaBoostClassifier(
+    DecisionTreeClassifier(max_depth=1), n_estimators=200,
+    algorithm="SAMME.R", learning_rate=0.5, random_state=42)
+ada_clf.fit(X_train, y_train)
+
+plot_decision_boundary(ada_clf, X, y)
+
+m = len(X_train)
+
+plt.figure(figsize=(11, 4))
+for subplot, learning_rate in ((121, 1), (122, 0.5)):
+    sample_weights = np.ones(m)
+    plt.subplot(subplot)
+    for i in range(5):
+        svm_clf = SVC(kernel="rbf", C=0.05, gamma="auto", random_state=42)
+        svm_clf.fit(X_train, y_train, sample_weight=sample_weights)
+        y_pred = svm_clf.predict(X_train)
+        sample_weights[y_pred != y_train] *= (1 + learning_rate)
+        plot_decision_boundary(svm_clf, X, y, alpha=0.2)
+        plt.title("learning_rate = {}".format(learning_rate), fontsize=16)
+    if subplot == 121:
+        plt.text(-0.7, -0.65, "1", fontsize=14)
+        plt.text(-0.6, -0.10, "2", fontsize=14)
+        plt.text(-0.5,  0.10, "3", fontsize=14)
+        plt.text(-0.4,  0.55, "4", fontsize=14)
+        plt.text(-0.3,  0.90, "5", fontsize=14)
+
+save_fig("boosting_plot")
+plt.show()
+
    
+
    
+
+
+
    
+
    Gradient boosting: Basics 
    
+
+
    
+Gradient boosting is again a similar technique to Adapative boosting,
+it combines so-called weak classifiers or regressors into a strong
+method via a series of iterations.
+
+
    
+In order to understand the method, let us illustrate its basics by
+bringing back the essential steps in linear regression, where our cost
+function was the least squares function.
+
    
+
+
+
    
+
    Gradient Boosting, algorithm 
    
+
+
    
+Suppose we have a cost function \( C(f)=\sum_{i=0}^{n-1}L(y_i, f(x_i)) \) where \( y_i \) is our target and \( f(x_i) \) the function which is meant to model \( y_i \). The above cost function could be our standard least squares  function
+
     
    
+$$
+C(\boldsymbol{y},\boldsymbol{f})=\frac{1}{n}\sum_{i=0}^{n-1}(y_i-f(x_i))^2.
+$$
+
     
    
+
+
    
+The way we proceed in an iterative fashion is to
+
+
      
+
    1.  Initialize our estimate by \( f_0(x)=0 \).
      2. 
+
    2.  For \( m=1:M \), we
+
+
        
+ 
      1.  compute the negative gradient vector \( \boldsymbol{u}_m = -\partial C(\boldsymbol{y},\boldsymbol{f})/\partial \boldsymbol{f}(x) \) at $f(x) = f_{m-1}(x);
        2. 
+ 
      2.  fit the so-called base-learner to the negative gradient \( h_m(u_m,x) \);
        3. 
+ 
      3.  update the estimate \( f_m(x) = f_{m-1}(x)+\nu h_m(u_m,x) \);
        4. 
+
      
+
    3.  The final estimate is then \( f_M(x) = \sum_{m=1}^M\nu h_m(u_m,x) \).
      4. 
+
    
+
    
+
+
+
    
+
    Gradient Boosting, Examples 
    
+
    
+
+
+
    np.random.seed(42)
+X = np.random.rand(100, 1) - 0.5
+y = 3*X[:, 0]**2 + 0.05 * np.random.randn(100)
+
+from sklearn.tree import DecisionTreeRegressor
+
+tree_reg1 = DecisionTreeRegressor(max_depth=2, random_state=42)
+tree_reg1.fit(X, y)
+
+y2 = y - tree_reg1.predict(X)
+tree_reg2 = DecisionTreeRegressor(max_depth=2, random_state=42)
+tree_reg2.fit(X, y2)
+
+y3 = y2 - tree_reg2.predict(X)
+tree_reg3 = DecisionTreeRegressor(max_depth=2, random_state=42)
+tree_reg3.fit(X, y3)
+
+X_new = np.array([[0.8]])
+y_pred = sum(tree.predict(X_new) for tree in (tree_reg1, tree_reg2, tree_reg3))
+
+def plot_predictions(regressors, X, y, axes, label=None, style="r-", data_style="b.", data_label=None):
+    x1 = np.linspace(axes[0], axes[1], 500)
+    y_pred = sum(regressor.predict(x1.reshape(-1, 1)) for regressor in regressors)
+    plt.plot(X[:, 0], y, data_style, label=data_label)
+    plt.plot(x1, y_pred, style, linewidth=2, label=label)
+    if label or data_label:
+        plt.legend(loc="upper center", fontsize=16)
+    plt.axis(axes)
+
+plt.figure(figsize=(11,11))
+
+plt.subplot(321)
+plot_predictions([tree_reg1], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="$h_1(x_1)$", style="g-", data_label="Training set")
+plt.ylabel("$y$", fontsize=16, rotation=0)
+plt.title("Residuals and tree predictions", fontsize=16)
+
+plt.subplot(322)
+plot_predictions([tree_reg1], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="$h(x_1) = h_1(x_1)$", data_label="Training set")
+plt.ylabel("$y$", fontsize=16, rotation=0)
+plt.title("Ensemble predictions", fontsize=16)
+
+plt.subplot(323)
+plot_predictions([tree_reg2], X, y2, axes=[-0.5, 0.5, -0.5, 0.5], label="$h_2(x_1)$", style="g-", data_style="k+", data_label="Residuals")
+plt.ylabel("$y - h_1(x_1)$", fontsize=16)
+
+plt.subplot(324)
+plot_predictions([tree_reg1, tree_reg2], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="$h(x_1) = h_1(x_1) + h_2(x_1)$")
+plt.ylabel("$y$", fontsize=16, rotation=0)
+
+plt.subplot(325)
+plot_predictions([tree_reg3], X, y3, axes=[-0.5, 0.5, -0.5, 0.5], label="$h_3(x_1)$", style="g-", data_style="k+")
+plt.ylabel("$y - h_1(x_1) - h_2(x_1)$", fontsize=16)
+plt.xlabel("$x_1$", fontsize=16)
+
+plt.subplot(326)
+plot_predictions([tree_reg1, tree_reg2, tree_reg3], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="$h(x_1) = h_1(x_1) + h_2(x_1) + h_3(x_1)$")
+plt.xlabel("$x_1$", fontsize=16)
+plt.ylabel("$y$", fontsize=16, rotation=0)
+
+save_fig("gradient_boosting_plot")
+plt.show()
+
+from sklearn.ensemble import GradientBoostingRegressor
+
+gbrt = GradientBoostingRegressor(max_depth=2, n_estimators=3, learning_rate=1.0, random_state=42)
+gbrt.fit(X, y)
+
+gbrt_slow = GradientBoostingRegressor(max_depth=2, n_estimators=200, learning_rate=0.1, random_state=42)
+gbrt_slow.fit(X, y)
+
+plt.figure(figsize=(11,4))
+
+plt.subplot(121)
+plot_predictions([gbrt], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="Ensemble predictions")
+plt.title("learning_rate={}, n_estimators={}".format(gbrt.learning_rate, gbrt.n_estimators), fontsize=14)
+
+plt.subplot(122)
+plot_predictions([gbrt_slow], X, y, axes=[-0.5, 0.5, -0.1, 0.8])
+plt.title("learning_rate={}, n_estimators={}".format(gbrt_slow.learning_rate, gbrt_slow.n_estimators), fontsize=14)
+
+save_fig("gbrt_learning_rate_plot")
+plt.show()
+
    
+
    
+
+
+
    
+
    Gradient Boots with Early Stopping 
    
+
    
+
+
+
    from sklearn.model_selection import train_test_split
+from sklearn.metrics import mean_squared_error
+
+X_train, X_val, y_train, y_val = train_test_split(X, y, random_state=49)
+
+gbrt = GradientBoostingRegressor(max_depth=2, n_estimators=120, random_state=42)
+gbrt.fit(X_train, y_train)
+
+errors = [mean_squared_error(y_val, y_pred)
+          for y_pred in gbrt.staged_predict(X_val)]
+bst_n_estimators = np.argmin(errors) + 1
+
+gbrt_best = GradientBoostingRegressor(max_depth=2,n_estimators=bst_n_estimators, random_state=42)
+gbrt_best.fit(X_train, y_train)
+
+min_error = np.min(errors)
+plt.figure(figsize=(11, 4))
+
+plt.subplot(121)
+plt.plot(errors, "b.-")
+plt.plot([bst_n_estimators, bst_n_estimators], [0, min_error], "k--")
+plt.plot([0, 120], [min_error, min_error], "k--")
+plt.plot(bst_n_estimators, min_error, "ko")
+plt.text(bst_n_estimators, min_error*1.2, "Minimum", ha="center", fontsize=14)
+plt.axis([0, 120, 0, 0.01])
+plt.xlabel("Number of trees")
+plt.title("Validation error", fontsize=14)
+
+plt.subplot(122)
+plot_predictions([gbrt_best], X, y, axes=[-0.5, 0.5, -0.1, 0.8])
+plt.title("Best model (%d trees)" % bst_n_estimators, fontsize=14)
+
+save_fig("early_stopping_gbrt_plot")
+plt.show()
+
+
+gbrt = GradientBoostingRegressor(max_depth=2, warm_start=True, random_state=42)
+
+min_val_error = float("inf")
+error_going_up = 0
+for n_estimators in range(1, 120):
+    gbrt.n_estimators = n_estimators
+    gbrt.fit(X_train, y_train)
+    y_pred = gbrt.predict(X_val)
+    val_error = mean_squared_error(y_val, y_pred)
+    if val_error < min_val_error:
+        min_val_error = val_error
+        error_going_up = 0
+    else:
+        error_going_up += 1
+        if error_going_up == 5:
+            break  # early stopping
+
+
+print(gbrt.n_estimators)
+print("Minimum validation MSE:", min_val_error)
+
    
+
    
+
+
+
    
+
    XGBoost: Extreme Gradient Boosting 
    
+
+
    
+XGBoost or Extreme Gradient
+Boosting, is an optimized distributed gradient boosting library
+designed to be highly efficient, flexible and portable. It implements
+machine learning algorithms under the Gradient Boosting
+framework. XGBoost provides a parallel tree boosting that solve many
+data science problems in a fast and accurate way. See the article by Chen and Guestrin.
+
+
    
+The authors design and build a highly scalable end-to-end tree
+boosting system. It has  a theoretically justified weighted quantile
+sketch for efficient proposal calculation. It introduces a novel sparsity-aware algorithm for parallel tree learning and an effective cache-aware block structure for out-of-core tree learning.
+
+
    
+It is now the algorithm which wins essentially all ML competitions!!!
+
    
+
+
+
    
+
    Regression Case 
    
+
+
    
+
+
+
    import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import train_test_split
+import xgboost as xgb
+from sklearn.preprocessing import StandardScaler
+import scikitplot as skplt
+from sklearn.metrics import mean_squared_error
+
+n = 100
+maxdegree = 6
+
+# Make data set.
+x = np.linspace(-3, 3, n).reshape(-1, 1)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
+
+error = np.zeros(maxdegree)
+bias = np.zeros(maxdegree)
+variance = np.zeros(maxdegree)
+polydegree = np.zeros(maxdegree)
+X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+for degree in range(maxdegree):
+    model =  xgb.XGBRegressor(objective ='reg:squarederror', colsample_bytree = 0.3, learning_rate = 0.1,
+                max_depth = degree, alpha = 10, n_estimators = 10)
+    model.fit(X_train_scaled,y_train)
+    y_pred = model.predict(X_test_scaled)
+    polydegree[degree] = degree
+    error[degree] = np.mean( np.mean((y_test - y_pred)**2) )
+    bias[degree] = np.mean( (y_test - np.mean(y_pred))**2 )
+    variance[degree] = np.mean( np.var(y_pred) )
+    print('Max depth:', degree)
+    print('Error:', error[degree])
+    print('Bias^2:', bias[degree])
+    print('Var:', variance[degree])
+    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
+
+plt.xlim(1,maxdegree-1)
+plt.plot(polydegree, error, label='Error')
+plt.plot(polydegree, bias, label='bias')
+plt.plot(polydegree, variance, label='Variance')
+plt.legend()
+plt.show()
+
    
+
    
+
+
+
    
+
    Xgboost on the Cancer Data 
    
+
    
+
+
+
    import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import  train_test_split 
+from sklearn.datasets import load_breast_cancer
+from sklearn.preprocessing import LabelEncoder
+from sklearn.model_selection import cross_validate
+import scikitplot as skplt
+import xgboost as xgb
+# Load the data
+cancer = load_breast_cancer()
+
+X_train, X_test, y_train, y_test = train_test_split(cancer.data,cancer.target,random_state=0)
+print(X_train.shape)
+print(X_test.shape)
+#now scale the data
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+xg_clf = xgb.XGBClassifier()
+xg_clf.fit(X_train_scaled,y_train)
+xgb.plot_tree(xg_clf,num_trees=0)
+plt.rcParams['figure.figsize'] = [50, 10]
+plt.show()
+xgb.plot_importance(xg_clf)
+plt.rcParams['figure.figsize'] = [5, 5]
+plt.show()
+
    
+
    
+
+
+
+
     
+
     
+
+
+
+
+
+
+
+
+
+
+
+
diff --git a/doc/src/DecisionTrees/DecisionTrees-solarized.html b/doc/src/DecisionTrees/DecisionTrees-solarized.html
new file mode 100644
index 000000000..1d6e730c8
--- /dev/null
+++ b/doc/src/DecisionTrees/DecisionTrees-solarized.html
@@ -0,0 +1,2502 @@
+
+
+
+
+
+
+
+
+Data Analysis and Machine Learning: From Decision Trees to Forests and all that
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+    
+
+
+
+
+
    Data Analysis and Machine Learning: From Decision Trees to Forests and all that
      
+
+
    
+
+
+
    
+Morten Hjorth-Jensen [1, 2]
+
    
+
+
    
+
+
+
    [1] Department of Physics, University of Oslo
    
+
    [2] Department of Physics and Astronomy and National Superconducting Cyclotron Laboratory, Michigan State University
    
+
    
+
    
+
    Nov 7, 2019
     
+
    
+
    
+









    
+
+
    Decision trees, overarching aims  
    
+
+
    
+Decision trees are supervised learning algorithms used for both,
+classification and regression tasks.
+
+
    
+The main idea of decision trees
+is to find those descriptive features which contain the most
+information regarding the target feature and then split the dataset
+along the values of these features such that the target feature values
+for the resulting underlying datasets are as pure as possible.
+
+
    
+The descriptive features which reproduce best the target/output features are normally  said
+to be the most informative ones. The process of finding the most
+informative feature is done until we accomplish a stopping criteria
+where we then finally end up in so called leaf nodes.
+
+
    
+A decision tree is typically divided into a root node, the interior nodes,
+and the final leaf nodes or just leaves. These entities are then connected by so-called branches.
+
+
    
+The leaf nodes
+contain the predictions we will make for new query instances presented
+to our trained model. This is possible since the model has 
+learned the underlying structure of the training data and hence can,
+given some assumptions, make predictions about the target feature value
+(class) of unseen query instances.
+
+
    
+









    
+
+
    A typical Decision Tree with its pertinent Jargon, Classification Problem 
    
+
+
    
+



    
+
+
    
+This tree was produced using the Wisconsin cancer data (discussed here as well, see code examples below) using Scikit-Learn's decision tree classifier. Here we have used the so-called gini index (see below) to split the various branches.
+
+
    
+









    
+
+
    General Features 
    
+
+
    
+The overarching approach to decision trees is a top-down approach.
+
+
      
+
    •  A leaf provides the classification of a given instance.
      • 
+
    •  A node specifies a test of some attribute of the instance.
      • 
+
    •  A branch corresponds to a possible values of an attribute.
      • 
+
    •  An instance is classified by starting at the root node of the tree, testing the attribute specified by this node, then moving down the tree branch corresponding to the value of the attribute in the given example.
      • 
+
    
+
+This process is then repeated for the subtree rooted at the new
+node.
+
+
    
+









    
+
+
    How do we set it up? 
    
+
+
    
+In simplified terms, the process of training a decision tree and
+predicting the target features of query instances is as follows:
+
+
      
+
    1.  Present a dataset containing of a number of training instances characterized by a number of descriptive features and a target feature
      2. 
+
    2.  Train the decision tree model by continuously splitting the target feature along the values of the descriptive features using a measure of information gain during the training process
      3. 
+
    3.  Grow the tree until we accomplish a stopping criteria create leaf nodes which represent the predictions we want to make for new query instances
      4. 
+
    4.  Show query instances to the tree and run down the tree until we arrive at leaf nodes
      5. 
+
    
+
+Then we are essentially done!
+
+
    
+









    
+
+
    Decision trees and Regression  
    
+
    
+
+
+
    import numpy as np
+import matplotlib.pyplot as plt
+from sklearn.preprocessing import PolynomialFeatures
+from sklearn.linear_model import LinearRegression
+
+steps=250
+
+distance=0
+x=0
+distance_list=[]
+steps_list=[]
+while x<steps:
+    distance+=np.random.randint(-1,2)
+    distance_list.append(distance)
+    x+=1
+    steps_list.append(x)
+plt.plot(steps_list,distance_list, color='green', label="Random Walk Data")
+
+steps_list=np.asarray(steps_list)
+distance_list=np.asarray(distance_list)
+
+X=steps_list[:,np.newaxis]
+
+#Polynomial fits
+
+#Degree 2
+poly_features=PolynomialFeatures(degree=2, include_bias=False)
+X_poly=poly_features.fit_transform(X)
+
+lin_reg=LinearRegression()
+poly_fit=lin_reg.fit(X_poly,distance_list)
+b=lin_reg.coef_
+c=lin_reg.intercept_
+print ("2nd degree coefficients:")
+print ("zero power: ",c)
+print ("first power: ", b[0])
+print ("second power: ",b[1])
+
+z = np.arange(0, steps, .01)
+z_mod=b[1]*z**2+b[0]*z+c
+
+fit_mod=b[1]*X**2+b[0]*X+c
+plt.plot(z, z_mod, color='r', label="2nd Degree Fit")
+plt.title("Polynomial Regression")
+
+plt.xlabel("Steps")
+plt.ylabel("Distance")
+
+#Degree 10
+poly_features10=PolynomialFeatures(degree=10, include_bias=False)
+X_poly10=poly_features10.fit_transform(X)
+
+poly_fit10=lin_reg.fit(X_poly10,distance_list)
+
+y_plot=poly_fit10.predict(X_poly10)
+plt.plot(X, y_plot, color='black', label="10th Degree Fit")
+
+plt.legend()
+plt.show()
+
+
+#Decision Tree Regression
+from sklearn.tree import DecisionTreeRegressor
+regr_1=DecisionTreeRegressor(max_depth=2)
+regr_2=DecisionTreeRegressor(max_depth=5)
+regr_3=DecisionTreeRegressor(max_depth=7)
+regr_1.fit(X, distance_list)
+regr_2.fit(X, distance_list)
+regr_3.fit(X, distance_list)
+
+X_test = np.arange(0.0, steps, 0.01)[:, np.newaxis]
+y_1 = regr_1.predict(X_test)
+y_2 = regr_2.predict(X_test)
+y_3=regr_3.predict(X_test)
+
+# Plot the results
+plt.figure()
+plt.scatter(X, distance_list, s=2.5, c="black", label="data")
+plt.plot(X_test, y_1, color="red",
+         label="max_depth=2", linewidth=2)
+plt.plot(X_test, y_2, color="green", label="max_depth=5", linewidth=2)
+plt.plot(X_test, y_3, color="m", label="max_depth=7", linewidth=2)
+
+plt.xlabel("Data")
+plt.ylabel("Darget")
+plt.title("Decision Tree Regression")
+plt.legend()
+plt.show()
+
    
+
    
+









    
+
+
    Building a tree, regression 
    
+
+
    
+There are mainly two steps
+
+
      
+
    1.  We split the predictor space (the set of possible values \( x_1,x_2,\dots, x_p \)) into \( J \) distinct and non-non-overlapping regions, \( R_1,R_2,\dots,R_J \).
      2.   
+
    2.  For every observation that falls into the region \( R_j \) , we make the same prediction, which is simply the mean of the response values for the training observations in \( R_j \).
      3. 
+
    
+
+How do we construct the regions \( R_1,\dots,R_J \)?  In theory, the
+regions could have any shape. However, we choose to divide the
+predictor space into high-dimensional rectangles, or boxes, for
+simplicity and for ease of interpretation of the resulting predictive
+model. The goal is to find boxes \( R_1,\dots,R_J \) that minimize the
+MSE, given by
+
+$$
+\sum_{j=1}^J\sum_{i\in R_j}(y_i-\overline{y}_{R_j})^2,
+$$
+
+
    
+where \( \overline{y}_{R_j} \)  is the mean response for the training observations 
+within box \( j \).
+
+
    
+









    
+
+
    A top-down approach, recursive binary splitting 
    
+
+
    
+Unfortunately, it is computationally infeasible to consider every
+possible partition of the feature space into \( J \) boxes.  The common
+strategy is to take a top-down approach
+
+
    
+The approach is top-down because it begins at the top of the tree (all
+observations belong to a single region) and then successively splits
+the predictor space; each split is indicated via two new branches
+further down on the tree. It is greedy because at each step of the
+tree-building process, the best split is made at that particular step,
+rather than looking ahead and picking a split that will lead to a
+better tree in some future step.
+
+
    
+









    
+
+
    Making a tree 
    
+
+
    
+In order to implement the recursive binary splitting we start by selecting
+the predictor \( x_j \) and a cutpoint \( s \) that splits the predictor space into two regions \( R_1 \) and \( R_2 \)
+$$
+\left\{X\vert x_j < s\right\},
+$$
+
+and
+$$
+\left\{X\vert x_j \geq s\right\},
+$$
+
+so that we obtain the lowest MSE, that is
+$$
+\sum_{i:x_i\in R_j}(y_i-\overline{y}_{R_1})^2+\sum_{i:x_i\in R_2}(y_i-\overline{y}_{R_2})^2,
+$$
+
+
    
+which we want to minimize by considering all predictors
+\( x_1,x_2,\dots,x_p \).  We consider also all possible values of \( s \) for
+each predictor. These values could be determined by randomly assigned
+numbers or by starting at the midpoint and then proceed till we find
+an optimal value.
+
+
    
+For any \( j \) and \( s \), we define the pair of half-planes where
+\( \overline{y}_{R_1} \) is the mean response for the training
+observations in \( R_1(j,s) \), and \( \overline{y}_{R_2} \) is the mean
+response for the training observations in \( R_2(j,s) \).
+
+
    
+Finding the values of \( j \) and \( s \) that minimize the above equation can be
+done quite quickly, especially when the number of features \( p \) is not
+too large.
+
+
    
+Next, we repeat the process, looking
+for the best predictor and best cutpoint in order to split the data
+further so as to minimize the MSE within each of the resulting
+regions. However, this time, instead of splitting the entire predictor
+space, we split one of the two previously identified regions. We now
+have three regions. Again, we look to split one of these three regions
+further, so as to minimize the MSE. The process continues until a
+stopping criterion is reached; for instance, we may continue until no
+region contains more than five observations.
+
+
    
+
+
+
    Pruning the tree 
    
+
+
    
+The above procedure is rather straightforward, but leads often to
+overfitting and unnecessarily large and complicated trees. The basic
+idea is to grow a large tree \( T_0 \) and then prune it back in order to
+obtain a subtree. A smaller tree with fewer splits (fewer regions) can
+lead to smaller variance and better interpretation at the cost of a
+little more bias.
+
+
    
+The so-called Cost complexity pruning algorithm gives us a
+way to do just this. Rather than considering every possible subtree,
+we consider a sequence of trees indexed by a nonnegative tuning
+parameter \( \alpha \).
+
+
    
+









    
+
+
    Cost complexity pruning 
    
+For each value of \( \alpha \)  there corresponds a subtree \( T \in T_0 \) such that
+$$
+\sum_{m=1}^{\overline{T}}\sum_{i:x_i\in R_m}(y_i-\overline{y}_{R_m})^2+\alpha\overline{T},
+$$
+
+is as small as possible. Here \( \overline{T} \) is 
+the number of terminal nodes of the tree \( T \) , \( R_m \) is the
+rectangle (i.e. the subset of predictor space)  corresponding to the \( m \)-th terminal node.
+
+
    
+The tuning parameter \( \alpha \) controls a trade-off between the subtree’s
+com- plexity and its fit to the training data. When \( \alpha = 0 \), then the
+subtree \( T \) will simply equal \( T_0 \), 
+because then the above equation just measures the
+training error. 
+However, as \( \alpha \) increases, there is a price to pay for
+having a tree with many terminal nodes. The above equation will
+tend to be minimized for a smaller subtree.
+
+
    
+It turns out that as we increase \( \alpha \) from zero
+branches get pruned from the tree in a nested and predictable fashion,
+so obtaining the whole sequence of subtrees as a function of \( \alpha \) is
+easy. We can select a value of \( \alpha \) using a validation set or using
+cross-validation. We then return to the full data set and obtain the
+subtree corresponding to \( \alpha \).
+
+
    
+









    
+
+
    Schematic Regression Procedure 
    
+
+
    
+
    
+Building a Regression Tree.
+
    
+
+
      
+
    1.  Use recursive binary splitting to grow a large tree on the training data, stopping only when each terminal node has fewer than some minimum number of observations.
      2. 
+
    2.  Apply cost complexity pruning to the large tree in order to obtain a sequence of best subtrees, as a function of \( \alpha \).
      3. 
+
    3.  Use for example \( K \)-fold cross-validation to choose \( \alpha \). Divide the training observations into \( K \) folds. For each \( k=1,2,\dots,K \) we:
      4.  
+
+
        
+  
      •  repeat steps 1 and 2 on all but the \( k \)-th fold of the training data.
        •  
+  
      •  Then we valuate the mean squared prediction error on the data in the left-out \( k \)-th fold, as a function of \( \alpha \).
        • 
+  
      •  Finally  we average the results for each value of \( \alpha \), and pick \( \alpha \) to minimize the average error.
        • 
+
      
+
+
    4.  Return the subtree from Step 2 that corresponds to the chosen value of \( \alpha \).
      5.  
+
    
+
    
+
+
+
    
+









    
+
+
    A Classification Tree 
    
+
+
    
+A classification tree is very similar to a regression tree, except
+that it is used to predict a qualitative response rather than a
+quantitative one. Recall that for a regression tree, the predicted
+response for an observation is given by the mean response of the
+training observations that belong to the same terminal node. In
+contrast, for a classification tree, we predict that each observation
+belongs to the most commonly occurring class of training observations
+in the region to which it belongs. In interpreting the results of a
+classification tree, we are often interested not only in the class
+prediction corresponding to a particular terminal node region, but
+also in the class proportions among the training observations that
+fall into that region.
+
+
    
+









    
+
+
    Growing a classification tree 
    
+
+
    
+The task of growing a
+classification tree is quite similar to the task of growing a
+regression tree. Just as in the regression setting, we use recursive
+binary splitting to grow a classification tree. However, in the
+classification setting, the MSE cannot be used as a criterion for making
+the binary splits.  A natural alternative to MSE is the classification
+error rate. Since we plan to assign an observation in a given region
+to the most commonly occurring error rate class of training
+observations in that region, the classification error rate is simply
+the fraction of the training observations in that region that do not
+belong to the most common class.
+
+
    
+When building a classification tree, either the Gini index or the
+entropy are typically used to evaluate the quality of a particular
+split, since these two approaches are more sensitive to node purity
+than is the classification error rate.
+
+
    
+









    
+
+
    Classification tree, how to split nodes 
    
+
+
    
+If our targets are the outcome of a classification process that takes
+for example \( k=1,2,\dots,K \) values, the only thing we need to think of
+is to set up the splitting criteria for each node.
+
+
    
+We define a PDF \( p_{mk} \) that represents the number of observations of
+a class \( k \) in a region \( R_m \) with \( N_m \) observations. We represent
+this likelihood function in terms of the proportion \( I(y_i=k) \) of
+observations of this class in the region \( R_m \) as
+
+$$
+p_{mk} = \frac{1}{N_m}\sum_{x_i\in R_m}I(y_i=k).
+$$
+
+
    
+We let \( p_{mk} \) represent the majority class of observations in region
+\( m \). The three most common ways of splitting a node are given by
+
+
      
+
    •  Misclassification error
      •  
+
    
+
+$$
+p_{mk} = \frac{1}{N_m}\sum_{x_i\in R_m}I(y_i\ne k) = 1-p_{mk}.
+$$
+
+
+
      
+
    •  Gini index \( g \)
      • 
+
    
+
+$$
+g = \sum_{k=1}^K p_{mk}(1-p_{mk}).
+$$
+
+
+
      
+
    •  Information entropy or just entropy \( s \)
      • 
+
    
+
+$$
+s = -\sum_{k=1}^K p_{mk}\log{p_{mk}}.
+$$
+
+
    
+









    
+
+
    Visualizing the Tree, Classification 
    
+
    
+
+
+
    import os
+from sklearn.datasets import load_breast_cancer
+from sklearn.tree import DecisionTreeClassifier
+from sklearn.model_selection import train_test_split
+from sklearn.metrics import confusion_matrix
+from sklearn.tree import export_graphviz
+
+from IPython.display import Image 
+from pydot import graph_from_dot_data
+import pandas as pd
+import numpy as np
+
+
+cancer = load_breast_cancer()
+X = pd.DataFrame(cancer.data, columns=cancer.feature_names)
+print(X)
+y = pd.Categorical.from_codes(cancer.target, cancer.target_names)
+y = pd.get_dummies(y)
+print(y)
+X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=1)
+tree_clf = DecisionTreeClassifier(max_depth=5)
+tree_clf.fit(X_train, y_train)
+
+export_graphviz(
+    tree_clf,
+    out_file="DataFiles/cancer.dot",
+    feature_names=cancer.feature_names,
+    class_names=cancer.target_names,
+    rounded=True,
+    filled=True
+)
+cmd = 'dot -Tpng DataFiles/cancer.dot -o DataFiles/cancer.png'
+os.system(cmd)
+
    
+
    
+









    
+
+
    Visualizing the Tree, The Moons  
    
+
    
+
+
+
    # Common imports
+import numpy as np
+from sklearn.model_selection import  train_test_split 
+from sklearn.tree import DecisionTreeClassifier
+from sklearn.datasets import make_moons
+from sklearn.tree import export_graphviz
+from pydot import graph_from_dot_data
+import pandas as pd
+import os
+
+np.random.seed(42)
+X, y = make_moons(n_samples=100, noise=0.25, random_state=53)
+X_train, X_test, y_train, y_test = train_test_split(X,y,random_state=0)
+tree_clf = DecisionTreeClassifier(max_depth=5)
+tree_clf.fit(X_train, y_train)
+
+export_graphviz(
+    tree_clf,
+    out_file="DataFiles/moons.dot",
+    rounded=True,
+    filled=True
+)
+cmd = 'dot -Tpng DataFiles/moons.dot -o DataFiles/moons.png'
+os.system(cmd)
+
    
+
    
+









    
+
+
    Algorithms for Setting up Decision Trees 
    
+
+
    
+Two algorithms stand out in the set up of decision trees:
+
+
      
+
    1.  The CART (Classification And Regression Tree) algorithm for both classification and regression
      2. 
+
    2.  The ID3 algorithm based on the computation of the information gain for classification
      3. 
+
    
+
+We discuss both algorithms with applications here. The popular library Scikit-Learn uses the CART algorithm. For classification problems you can use either the gini index or the entropy to split a tree in two branches.
+
+
    
+









    
+
+
    The CART algorithm for Classification 
    
+
+
    
+









    
+
+
    The CART algorithm for Regression 
    
+
+
    
+









    
+
+
    Computing the Gini index 
    
+
+
    
+The example we will look at is a classical one in many Machine
+Learning applications. Based on various meteorological features, we
+have several so-called attributes which decide whether we at the end
+will do some outdoor activity like skiing, going for a bike ride etc
+etc.  The table here contains the feautures outlook, temperature,
+humidity and wind.  The target or output is whether we ride
+(True=1) or whether we do something else that day (False=0). The
+attributes for each feature are then sunny, overcast and rain for the
+outlook, hot, cold and mild for temperature, high and normal for
+humidity and weak and strong for wind.
+
+
    
+The table here summarizes the various attributes and
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
    Day Outlook  Temperature Humidity  Wind  Ride 
       1          Sunny           Hot                High            Weak          0        
       2          Sunny           Hot                High            Strong        1        
       3          Overcast        Hot                High            Weak          1        
       4          Rain            Mild               High            Weak          1        
       5          Rain            Cool               Normal          Weak          1        
       6          Rain            Cool               Normal          Strong        0        
       7          Overcast        Cool               Normal          Strong        1        
       8          Sunny           Mild               High            Weak          0        
       9          Sunny           Cool               Normal          Weak          1        
       10         Rain            Mild               Normal          Weak          1        
       11         Sunny           Mild               Normal          Strong        1        
       12         Overcast        Mild               High            Strong        1        
       13         Overcast        Hot                Normal          Weak          1        
       14         Rain            Mild               High            Strong        0        
    
+
    
+









    
+
+
    Simple Python Code to read in Data and perform Classification 
    
+
+
    
+
+
+
    # Common imports
+import numpy as np
+import pandas as pd
+import matplotlib.pyplot as plt
+from sklearn.tree import DecisionTreeClassifier
+from sklearn.model_selection import train_test_split
+from sklearn.tree import export_graphviz
+from sklearn.preprocessing import StandardScaler, OneHotEncoder
+from sklearn.compose import ColumnTransformer
+from IPython.display import Image 
+from pydot import graph_from_dot_data
+import os
+
+# Where to save the figures and data files
+PROJECT_ROOT_DIR = "Results"
+FIGURE_ID = "Results/FigureFiles"
+DATA_ID = "DataFiles/"
+
+if not os.path.exists(PROJECT_ROOT_DIR):
+    os.mkdir(PROJECT_ROOT_DIR)
+
+if not os.path.exists(FIGURE_ID):
+    os.makedirs(FIGURE_ID)
+
+if not os.path.exists(DATA_ID):
+    os.makedirs(DATA_ID)
+
+def image_path(fig_id):
+    return os.path.join(FIGURE_ID, fig_id)
+
+def data_path(dat_id):
+    return os.path.join(DATA_ID, dat_id)
+
+def save_fig(fig_id):
+    plt.savefig(image_path(fig_id) + ".png", format='png')
+
+infile = open(data_path("rideclass.csv"),'r')
+
+# Read the experimental data with Pandas
+from IPython.display import display
+ridedata = pd.read_csv(infile,names = ('Outlook','Temperature','Humidity','Wind','Ride'))
+ridedata = pd.DataFrame(ridedata)
+
+# Features and targets
+X = ridedata.loc[:, ridedata.columns != 'Ride'].values
+y = ridedata.loc[:, ridedata.columns == 'Ride'].values
+
+# Create the encoder.
+encoder = OneHotEncoder(handle_unknown="ignore")
+# Assume for simplicity all features are categorical.
+encoder.fit(X)    
+# Apply the encoder.
+X = encoder.transform(X)
+print(X)
+# Then do a Classification tree
+tree_clf = DecisionTreeClassifier(max_depth=2)
+tree_clf.fit(X, y)
+print("Train set accuracy with Decision Tree: {:.2f}".format(tree_clf.score(X,y)))
+#transfer to a decision tree graph
+export_graphviz(
+    tree_clf,
+    out_file="DataFiles/ride.dot",
+    rounded=True,
+    filled=True
+)
+cmd = 'dot -Tpng DataFiles/cancer.dot -o DataFiles/cancer.png'
+os.system(cmd)
+
    
+
    
+









    
+
+
    Computing the Gini Factor  
    
+
+
    
+The above functions (gini, entropy and misclassification error) are
+important components of the so-called CART algorithm. We will discuss
+this algorithm below after we have discussed the information gain
+algorithm ID3.
+
+
    
+In the example here we have converted all our attributes into numerical values \( 0,1,2 \) etc.
+
+
    
+
+
+
    # Split a dataset based on an attribute and an attribute value
+def test_split(index, value, dataset):
+	left, right = list(), list()
+	for row in dataset:
+		if row[index] < value:
+			left.append(row)
+		else:
+			right.append(row)
+	return left, right
+ 
+# Calculate the Gini index for a split dataset
+def gini_index(groups, classes):
+	# count all samples at split point
+	n_instances = float(sum([len(group) for group in groups]))
+	# sum weighted Gini index for each group
+	gini = 0.0
+	for group in groups:
+		size = float(len(group))
+		# avoid divide by zero
+		if size == 0:
+			continue
+		score = 0.0
+		# score the group based on the score for each class
+		for class_val in classes:
+			p = [row[-1] for row in group].count(class_val) / size
+			score += p * p
+		# weight the group score by its relative size
+		gini += (1.0 - score) * (size / n_instances)
+	return gini
+
+# Select the best split point for a dataset
+def get_split(dataset):
+	class_values = list(set(row[-1] for row in dataset))
+	b_index, b_value, b_score, b_groups = 999, 999, 999, None
+	for index in range(len(dataset[0])-1):
+		for row in dataset:
+			groups = test_split(index, row[index], dataset)
+			gini = gini_index(groups, class_values)
+			print('X%d < %.3f Gini=%.3f' % ((index+1), row[index], gini))
+			if gini < b_score:
+				b_index, b_value, b_score, b_groups = index, row[index], gini, groups
+	return {'index':b_index, 'value':b_value, 'groups':b_groups}
+ 
+dataset = [[0,0,0,0,0],
+            [0,0,0,1,1],
+            [1,0,0,0,1],
+            [2,1,0,0,1],
+            [2,2,1,0,1],
+            [2,2,1,1,0],
+            [1,2,1,1,1],
+            [0,1,0,0,0],
+            [0,2,1,0,1],
+            [2,1,1,0,1],
+            [0,1,1,1,1],
+            [1,1,0,1,1],
+            [1,0,1,0,1],
+            [2,1,0,1,0]]
+
+split = get_split(dataset)
+print('Split: [X%d < %.3f]' % ((split['index']+1), split['value']))
+
    
+
    
+









    
+
+
    Entropy and the ID3 algorithm 
    
+
+
    
+ID3, learns decision trees by constructing
+them topdown, beginning with the question which attribute should be tested at the root of the tree?
+
+
      
+
    1.  Each instance attribute is evaluated using a statistical test to determine how well it alone classifies the training examples.
      2. 
+
    2.  The best attribute is selected and used as the test at the root node of the tree.
      3. 
+
    3.  A descendant of the root node is then created for each possible value of this attribute.
      4. 
+
    4.  Training examples are sorted to the appropriate descendant node.
      5. 
+
    5.  The entire process is then repeated using the training examples associated with each descendant node to select the best attribute to test at that point in the tree.
      6. 
+
    6.  This forms a greedy search for an acceptable decision tree, in which the algorithm never backtracks to reconsider earlier choices.
      7.  
+
    
+
+The ID3 algorithm selects, which attribute to test at each node in the
+tree.
+
+
    
+We would like to select the attribute that is most useful for classifying
+examples.
+
+
    
+What is a good quantitative measure of the worth of an attribute?
+
+
    
+Information gain measures how well a given attribute separates the
+training examples according to their target classification.
+
+
    
+The ID3 algorithm uses this information gain measure to select among the candidate
+attributes at each step while growing the tree.
+
+
    
+









    
+
+
    Implementing the ID3 Algorithm 
    
+
+
    
+
+
+
    import re
+import math
+from collections import deque
+
+# x is examples in training set
+# y is set of targets
+# label is target attributes
+# Node is a class which has properties values, childs, and next
+# root is top node in the decision tree
+
+class Node(object):
+	def __init__(self):
+		self.value = None
+		self.next = None
+		self.childs = None
+
+# Simple class of Decision Tree
+# Aimed for who want to learn Decision Tree, so it is not optimized
+class DecisionTree(object):
+	def __init__(self, sample, attributes, labels):
+		self.sample = sample
+		self.attributes = attributes
+		self.labels = labels
+		self.labelCodes = None
+		self.labelCodesCount = None
+		self.initLabelCodes()
+		# print(self.labelCodes)
+		self.root = None
+		self.entropy = self.getEntropy([x for x in range(len(self.labels))])
+
+	def initLabelCodes(self):
+		self.labelCodes = []
+		self.labelCodesCount = []
+		for l in self.labels:
+			if l not in self.labelCodes:
+				self.labelCodes.append(l)
+				self.labelCodesCount.append(0)
+			self.labelCodesCount[self.labelCodes.index(l)] += 1
+
+	def getLabelCodeId(self, sampleId):
+		return self.labelCodes.index(self.labels[sampleId])
+
+	def getAttributeValues(self, sampleIds, attributeId):
+		vals = []
+		for sid in sampleIds:
+			val = self.sample[sid][attributeId]
+			if val not in vals:
+				vals.append(val)
+		# print(vals)
+		return vals
+
+	def getEntropy(self, sampleIds):
+		entropy = 0
+		labelCount = [0] * len(self.labelCodes)
+		for sid in sampleIds:
+			labelCount[self.getLabelCodeId(sid)] += 1
+		# print("-ge", labelCount)
+		for lv in labelCount:
+			# print(lv)
+			if lv != 0:
+				entropy += -lv/len(sampleIds) * math.log(lv/len(sampleIds), 2)
+			else:
+				entropy += 0
+		return entropy
+
+	def getDominantLabel(self, sampleIds):
+		labelCodesCount = [0] * len(self.labelCodes)
+		for sid in sampleIds:
+			labelCodesCount[self.labelCodes.index(self.labels[sid])] += 1
+		return self.labelCodes[labelCodesCount.index(max(labelCodesCount))]
+
+	def getInformationGain(self, sampleIds, attributeId):
+		gain = self.getEntropy(sampleIds)
+		attributeVals = []
+		attributeValsCount = []
+		attributeValsIds = []
+		for sid in sampleIds:
+			val = self.sample[sid][attributeId]
+			if val not in attributeVals:
+				attributeVals.append(val)
+				attributeValsCount.append(0)
+				attributeValsIds.append([])
+			vid = attributeVals.index(val)
+			attributeValsCount[vid] += 1
+			attributeValsIds[vid].append(sid)
+		# print("-gig", self.attributes[attributeId])
+		for vc, vids in zip(attributeValsCount, attributeValsIds):
+			# print("-gig", vids)
+			gain -= vc/len(sampleIds) * self.getEntropy(vids)
+		return gain
+
+	def getAttributeMaxInformationGain(self, sampleIds, attributeIds):
+		attributesEntropy = [0] * len(attributeIds)
+		for i, attId in zip(range(len(attributeIds)), attributeIds):
+			attributesEntropy[i] = self.getInformationGain(sampleIds, attId)
+		maxId = attributeIds[attributesEntropy.index(max(attributesEntropy))]
+		return self.attributes[maxId], maxId
+
+	def isSingleLabeled(self, sampleIds):
+		label = self.labels[sampleIds[0]]
+		for sid in sampleIds:
+			if self.labels[sid] != label:
+				return False
+		return True
+
+	def getLabel(self, sampleId):
+		return self.labels[sampleId]
+
+	def id3(self):
+		sampleIds = [x for x in range(len(self.sample))]
+		attributeIds = [x for x in range(len(self.attributes))]
+		self.root = self.id3Recv(sampleIds, attributeIds, self.root)
+
+	def id3Recv(self, sampleIds, attributeIds, root):
+		root = Node() # Initialize current root
+		if self.isSingleLabeled(sampleIds):
+			root.value = self.labels[sampleIds[0]]
+			return root
+		# print(attributeIds)
+		if len(attributeIds) == 0:
+			root.value = self.getDominantLabel(sampleIds)
+			return root
+		bestAttrName, bestAttrId = self.getAttributeMaxInformationGain(
+			sampleIds, attributeIds)
+		# print(bestAttrName)
+		root.value = bestAttrName
+		root.childs = []  # Create list of children
+		for value in self.getAttributeValues(sampleIds, bestAttrId):
+			# print(value)
+			child = Node()
+			child.value = value
+			root.childs.append(child)  # Append new child node to current
+									   # root
+			childSampleIds = []
+			for sid in sampleIds:
+				if self.sample[sid][bestAttrId] == value:
+					childSampleIds.append(sid)
+			if len(childSampleIds) == 0:
+				child.next = self.getDominantLabel(sampleIds)
+			else:
+				# print(bestAttrName, bestAttrId)
+				# print(attributeIds)
+				if len(attributeIds) > 0 and bestAttrId in attributeIds:
+					toRemove = attributeIds.index(bestAttrId)
+					attributeIds.pop(toRemove)
+				child.next = self.id3Recv(
+					childSampleIds, attributeIds, child.next)
+		return root
+
+	def printTree(self):
+		if self.root:
+			roots = deque()
+			roots.append(self.root)
+			while len(roots) > 0:
+				root = roots.popleft()
+				print(root.value)
+				if root.childs:
+					for child in root.childs:
+						print('({})'.format(child.value))
+						roots.append(child.next)
+				elif root.next:
+					print(root.next)
+
+
+def test():
+	f = open('DataFiles/rideclass.csv')
+	attributes = f.readline().split(',')
+	attributes = attributes[1:len(attributes)-1]
+	print(attributes)
+	sample = f.readlines()
+	f.close()
+	for i in range(len(sample)):
+		sample[i] = re.sub('\d+,', '', sample[i])
+		sample[i] = sample[i].strip().split(',')
+	labels = []
+	for s in sample:
+		labels.append(s.pop())
+	# print(sample)
+	# print(labels)
+	decisionTree = DecisionTree(sample, attributes, labels)
+	print("System entropy {}".format(decisionTree.entropy))
+	decisionTree.id3()
+	decisionTree.printTree()
+
+
+if __name__ == '__main__':
+	test()
+
    
+
    
+









    
+
+
    Cancer Data again now with Decision Trees and other Methods 
    
+
    
+
+
+
    import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import  train_test_split 
+from sklearn.datasets import load_breast_cancer
+from sklearn.svm import SVC
+from sklearn.linear_model import LogisticRegression
+from sklearn.tree import DecisionTreeClassifier
+
+# Load the data
+cancer = load_breast_cancer()
+
+X_train, X_test, y_train, y_test = train_test_split(cancer.data,cancer.target,random_state=0)
+print(X_train.shape)
+print(X_test.shape)
+# Logistic Regression
+logreg = LogisticRegression(solver='lbfgs')
+logreg.fit(X_train, y_train)
+print("Test set accuracy with Logistic Regression: {:.2f}".format(logreg.score(X_test,y_test)))
+# Support vector machine
+svm = SVC(gamma='auto', C=100)
+svm.fit(X_train, y_train)
+print("Test set accuracy with SVM: {:.2f}".format(svm.score(X_test,y_test)))
+# Decision Trees
+deep_tree_clf = DecisionTreeClassifier(max_depth=None)
+deep_tree_clf.fit(X_train, y_train)
+print("Test set accuracy with Decision Trees: {:.2f}".format(deep_tree_clf.score(X_test,y_test)))
+#now scale the data
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+# Logistic Regression
+logreg.fit(X_train_scaled, y_train)
+print("Test set accuracy Logistic Regression with scaled data: {:.2f}".format(logreg.score(X_test_scaled,y_test)))
+# Support Vector Machine
+svm.fit(X_train_scaled, y_train)
+print("Test set accuracy SVM with scaled data: {:.2f}".format(logreg.score(X_test_scaled,y_test)))
+# Decision Trees
+deep_tree_clf.fit(X_train_scaled, y_train)
+print("Test set accuracy with Decision Trees and scaled data: {:.2f}".format(deep_tree_clf.score(X_test_scaled,y_test)))
+
    
+
    
+









    
+
+
    Another example, the moons again 
    
+
    
+
+
+
    from __future__ import division, print_function, unicode_literals
+
+# Common imports
+import numpy as np
+import os
+
+# to make this notebook's output stable across runs
+np.random.seed(42)
+
+# To plot pretty figures
+import matplotlib
+import matplotlib.pyplot as plt
+from matplotlib.colors import ListedColormap
+plt.rcParams['axes.labelsize'] = 14
+plt.rcParams['xtick.labelsize'] = 12
+plt.rcParams['ytick.labelsize'] = 12
+
+
+from sklearn.svm import SVC
+from sklearn import datasets
+from sklearn.tree import DecisionTreeClassifier
+from sklearn.datasets import make_moons
+from sklearn.tree import export_graphviz
+
+Xm, ym = make_moons(n_samples=100, noise=0.25, random_state=53)
+
+deep_tree_clf1 = DecisionTreeClassifier(random_state=42)
+deep_tree_clf2 = DecisionTreeClassifier(min_samples_leaf=4, random_state=42)
+deep_tree_clf1.fit(Xm, ym)
+deep_tree_clf2.fit(Xm, ym)
+
+
+def plot_decision_boundary(clf, X, y, axes=[0, 7.5, 0, 3], iris=True, legend=False, plot_training=True):
+    x1s = np.linspace(axes[0], axes[1], 100)
+    x2s = np.linspace(axes[2], axes[3], 100)
+    x1, x2 = np.meshgrid(x1s, x2s)
+    X_new = np.c_[x1.ravel(), x2.ravel()]
+    y_pred = clf.predict(X_new).reshape(x1.shape)
+    custom_cmap = ListedColormap(['#fafab0','#9898ff','#a0faa0'])
+    plt.contourf(x1, x2, y_pred, alpha=0.3, cmap=custom_cmap)
+    if not iris:
+        custom_cmap2 = ListedColormap(['#7d7d58','#4c4c7f','#507d50'])
+        plt.contour(x1, x2, y_pred, cmap=custom_cmap2, alpha=0.8)
+    if plot_training:
+        plt.plot(X[:, 0][y==0], X[:, 1][y==0], "yo", label="Iris-Setosa")
+        plt.plot(X[:, 0][y==1], X[:, 1][y==1], "bs", label="Iris-Versicolor")
+        plt.plot(X[:, 0][y==2], X[:, 1][y==2], "g^", label="Iris-Virginica")
+        plt.axis(axes)
+    if iris:
+        plt.xlabel("Petal length", fontsize=14)
+        plt.ylabel("Petal width", fontsize=14)
+    else:
+        plt.xlabel(r"$x_1$", fontsize=18)
+        plt.ylabel(r"$x_2$", fontsize=18, rotation=0)
+    if legend:
+        plt.legend(loc="lower right", fontsize=14)
+plt.figure(figsize=(11, 4))
+plt.subplot(121)
+plot_decision_boundary(deep_tree_clf1, Xm, ym, axes=[-1.5, 2.5, -1, 1.5], iris=False)
+plt.title("No restrictions", fontsize=16)
+plt.subplot(122)
+plot_decision_boundary(deep_tree_clf2, Xm, ym, axes=[-1.5, 2.5, -1, 1.5], iris=False)
+plt.title("min_samples_leaf = {}".format(deep_tree_clf2.min_samples_leaf), fontsize=14)
+plt.show()
+
    
+
    
+









    
+
+
    Playing around with regions 
    
+
    
+
+
+
    np.random.seed(6)
+Xs = np.random.rand(100, 2) - 0.5
+ys = (Xs[:, 0] > 0).astype(np.float32) * 2
+
+angle = np.pi/4
+rotation_matrix = np.array([[np.cos(angle), -np.sin(angle)], [np.sin(angle), np.cos(angle)]])
+Xsr = Xs.dot(rotation_matrix)
+
+tree_clf_s = DecisionTreeClassifier(random_state=42)
+tree_clf_s.fit(Xs, ys)
+tree_clf_sr = DecisionTreeClassifier(random_state=42)
+tree_clf_sr.fit(Xsr, ys)
+
+plt.figure(figsize=(11, 4))
+plt.subplot(121)
+plot_decision_boundary(tree_clf_s, Xs, ys, axes=[-0.7, 0.7, -0.7, 0.7], iris=False)
+plt.subplot(122)
+plot_decision_boundary(tree_clf_sr, Xsr, ys, axes=[-0.7, 0.7, -0.7, 0.7], iris=False)
+
+plt.show()
+
    
+
    
+









    
+
+
    Regression trees 
    
+
    
+
+
+
    # Quadratic training set + noise
+np.random.seed(42)
+m = 200
+X = np.random.rand(m, 1)
+y = 4 * (X - 0.5) ** 2
+y = y + np.random.randn(m, 1) / 10
+
    
+
    
+
+
+
    from sklearn.tree import DecisionTreeRegressor
+
+tree_reg = DecisionTreeRegressor(max_depth=2, random_state=42)
+tree_reg.fit(X, y)
+
    
+
    
+









    
+
+
    Final regressor code 
    
+
    
+
+
+
    from sklearn.tree import DecisionTreeRegressor
+
+tree_reg1 = DecisionTreeRegressor(random_state=42, max_depth=2)
+tree_reg2 = DecisionTreeRegressor(random_state=42, max_depth=3)
+tree_reg1.fit(X, y)
+tree_reg2.fit(X, y)
+
+def plot_regression_predictions(tree_reg, X, y, axes=[0, 1, -0.2, 1], ylabel="$y$"):
+    x1 = np.linspace(axes[0], axes[1], 500).reshape(-1, 1)
+    y_pred = tree_reg.predict(x1)
+    plt.axis(axes)
+    plt.xlabel("$x_1$", fontsize=18)
+    if ylabel:
+        plt.ylabel(ylabel, fontsize=18, rotation=0)
+    plt.plot(X, y, "b.")
+    plt.plot(x1, y_pred, "r.-", linewidth=2, label=r"$\hat{y}$")
+
+plt.figure(figsize=(11, 4))
+plt.subplot(121)
+plot_regression_predictions(tree_reg1, X, y)
+for split, style in ((0.1973, "k-"), (0.0917, "k--"), (0.7718, "k--")):
+    plt.plot([split, split], [-0.2, 1], style, linewidth=2)
+plt.text(0.21, 0.65, "Depth=0", fontsize=15)
+plt.text(0.01, 0.2, "Depth=1", fontsize=13)
+plt.text(0.65, 0.8, "Depth=1", fontsize=13)
+plt.legend(loc="upper center", fontsize=18)
+plt.title("max_depth=2", fontsize=14)
+
+plt.subplot(122)
+plot_regression_predictions(tree_reg2, X, y, ylabel=None)
+for split, style in ((0.1973, "k-"), (0.0917, "k--"), (0.7718, "k--")):
+    plt.plot([split, split], [-0.2, 1], style, linewidth=2)
+for split in (0.0458, 0.1298, 0.2873, 0.9040):
+    plt.plot([split, split], [-0.2, 1], "k:", linewidth=1)
+plt.text(0.3, 0.5, "Depth=2", fontsize=13)
+plt.title("max_depth=3", fontsize=14)
+
+plt.show()
+
    
+
    
+
+
+
    tree_reg1 = DecisionTreeRegressor(random_state=42)
+tree_reg2 = DecisionTreeRegressor(random_state=42, min_samples_leaf=10)
+tree_reg1.fit(X, y)
+tree_reg2.fit(X, y)
+
+x1 = np.linspace(0, 1, 500).reshape(-1, 1)
+y_pred1 = tree_reg1.predict(x1)
+y_pred2 = tree_reg2.predict(x1)
+
+plt.figure(figsize=(11, 4))
+
+plt.subplot(121)
+plt.plot(X, y, "b.")
+plt.plot(x1, y_pred1, "r.-", linewidth=2, label=r"$\hat{y}$")
+plt.axis([0, 1, -0.2, 1.1])
+plt.xlabel("$x_1$", fontsize=18)
+plt.ylabel("$y$", fontsize=18, rotation=0)
+plt.legend(loc="upper center", fontsize=18)
+plt.title("No restrictions", fontsize=14)
+
+plt.subplot(122)
+plt.plot(X, y, "b.")
+plt.plot(x1, y_pred2, "r.-", linewidth=2, label=r"$\hat{y}$")
+plt.axis([0, 1, -0.2, 1.1])
+plt.xlabel("$x_1$", fontsize=18)
+plt.title("min_samples_leaf={}".format(tree_reg2.min_samples_leaf), fontsize=14)
+
+plt.show()
+
    
+
    
+









    
+
+
    Pros and cons of trees, pros 
    
+
+
      
+
    •  White box, easy to interpret model. Some people believe that decision trees more closely mirror human decision-making than do the regression and classification approaches discussed earlier (think of support vector machines)
      • 
+
    •  Trees are very easy to explain to people. In fact, they are even easier to explain than linear regression!
      • 
+
    •  No feature normalization needed
      • 
+
    •  Tree models can handle both continuous and categorical data (Classification and Regression Trees)
      • 
+
    •  Can model nonlinear relationships
      • 
+
    •  Can model interactions between the different descriptive features
      • 
+
    •  Trees can be displayed graphically, and are easily interpreted even by a non-expert (especially if they are small)
      • 
+
    
+
+









    
+
+
    Disadvantages 
    
+
+
      
+
    •  Unfortunately, trees generally do not have the same level of predictive accuracy as some of the other regression and classification approaches
      • 
+
    •  If continuous features are used the tree may become quite large and hence less interpretable
      • 
+
    •  Decision trees are prone to overfit the training data and hence do not well generalize the data if no stopping criteria or improvements like pruning, boosting or bagging are implemented
      • 
+
    •  Small changes in the data may lead to a completely different tree. This issue can be addressed by using ensemble methods like bagging, boosting or random forests
      • 
+
    •  Unbalanced datasets where some target feature values occur much more frequently than others may lead to biased trees since the frequently occurring feature values are preferred over the less frequently occurring ones.
      •  
+
    •  If the number of features is relatively large (high dimensional) and the number of instances is relatively low, the tree might overfit the data
      • 
+
    •  Features with many levels may be preferred over features with less levels since for them it is more easy to split the dataset such that the sub datasets only contain pure target feature values. This issue can be addressed by preferring for instance the information gain ratio as splitting criteria over information gain
      • 
+
    
+
+However, by aggregating many decision trees, using methods like bagging, random forests, and boosting, the predictive performance of trees can be substantially improved.
+
+
    
+









    
+
+
    Ensemble Methods: From a Single Tree to Many Trees and Extreme Boosting, Meet the Jungle of Methods 
    
+
+
    
+As stated above and seen in many of the examples discussed here about
+a single decision tree, we often end up overfitting our training
+data. This normally means that we have a high variance. Can we reduce
+the variance of a statistical learning method?
+
+
    
+This leads us to a set of different methods that can combine different
+machine learning algorithms or just use one of them to construct forests and jungles of trees, homogeneous ones or heterogenous ones. These methods are recognized by different names which we will try to explain here. These are
+
+
      
+
    1.  Voting classifiers
      2. 
+
    2.  Bagging and Pasting
      3. 
+
    3.  Random forests
      4. 
+
    4.  Boosting methods, from adaptive to Extreme Gradient Boosting (XGBoost)
      5. 
+
    
+
+We discuss these methods here.
+
+
    
+









    
+
+
    An Overview of Ensemble Methods  
    
+
+
    
+



    
+
+
    
+









    
+
+
    Bagging 
    
+
+
    
+The plain decision trees suffer from high
+variance. This means that if we split the training data into two parts
+at random, and fit a decision tree to both halves, the results that we
+get could be quite different. In contrast, a procedure with low
+variance will yield similar results if applied repeatedly to distinct
+data sets; linear regression tends to have low variance, if the ratio
+of \( n \) to \( p \) is moderately large.
+
+
    
+Bootstrap aggregation, or just bagging, is a
+general-purpose procedure for reducing the variance of a statistical
+learning method.
+
+
    
+









    
+
+
    More bagging 
    
+
+
    
+Bagging typically results in improved accuracy
+over prediction using a single tree. Unfortunately, however, it can be
+difficult to interpret the resulting model. Recall that one of the
+advantages of decision trees is the attractive and easily interpreted
+diagram that results.
+
+
    
+However, when we bag a large number of trees, it is no longer
+possible to represent the resulting statistical learning procedure
+using a single tree, and it is no longer clear which variables are
+most important to the procedure. Thus, bagging improves prediction
+accuracy at the expense of interpretability.  Although the collection
+of bagged trees is much more difficult to interpret than a single
+tree, one can obtain an overall summary of the importance of each
+predictor using the MSE (for bagging regression trees) or the Gini
+index (for bagging classification trees). In the case of bagging
+regression trees, we can record the total amount that the MSE is
+decreased due to splits over a given predictor, averaged over all \( B \) possible
+trees. A large value indicates an important predictor. Similarly, in
+the context of bagging classification trees, we can add up the total
+amount that the Gini index  is decreased by splits over a given
+predictor, averaged over all \( B \) trees.
+
+
    
+









    
+
+
    Simple Voting Example, head or tail 
    
+
    
+
+
+
    heads_proba = 0.51
+coin_tosses = (np.random.rand(10000, 10) < heads_proba).astype(np.int32)
+cumulative_heads_ratio = np.cumsum(coin_tosses, axis=0) / np.arange(1, 10001).reshape(-1, 1)
+plt.figure(figsize=(8,3.5))
+plt.plot(cumulative_heads_ratio)
+plt.plot([0, 10000], [0.51, 0.51], "k--", linewidth=2, label="51%")
+plt.plot([0, 10000], [0.5, 0.5], "k-", label="50%")
+plt.xlabel("Number of coin tosses")
+plt.ylabel("Heads ratio")
+plt.legend(loc="lower right")
+plt.axis([0, 10000, 0.42, 0.58])
+plt.show()
+
    
+
    
+









    
+
+
    Using the Voting Classifier 
    
+
    
+
+
+
    from sklearn.model_selection import train_test_split
+from sklearn.datasets import make_moons
+
+X, y = make_moons(n_samples=500, noise=0.30, random_state=42)
+X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=42)
+
+from sklearn.ensemble import RandomForestClassifier
+from sklearn.ensemble import VotingClassifier
+from sklearn.linear_model import LogisticRegression
+from sklearn.svm import SVC
+
+log_clf = LogisticRegression(solver="liblinear", random_state=42)
+rnd_clf = RandomForestClassifier(n_estimators=10, random_state=42)
+svm_clf = SVC(gamma="auto", random_state=42)
+
+voting_clf = VotingClassifier(
+    estimators=[('lr', log_clf), ('rf', rnd_clf), ('svc', svm_clf)],
+    voting='hard')
+
+voting_clf.fit(X_train, y_train)
+
+from sklearn.metrics import accuracy_score
+
+for clf in (log_clf, rnd_clf, svm_clf, voting_clf):
+    clf.fit(X_train, y_train)
+    y_pred = clf.predict(X_test)
+    print(clf.__class__.__name__, accuracy_score(y_test, y_pred))
+
+log_clf = LogisticRegression(solver="liblinear", random_state=42)
+rnd_clf = RandomForestClassifier(n_estimators=10, random_state=42)
+svm_clf = SVC(gamma="auto", probability=True, random_state=42)
+
+voting_clf = VotingClassifier(
+    estimators=[('lr', log_clf), ('rf', rnd_clf), ('svc', svm_clf)],
+    voting='soft')
+voting_clf.fit(X_train, y_train)
+
+from sklearn.metrics import accuracy_score
+
+for clf in (log_clf, rnd_clf, svm_clf, voting_clf):
+    clf.fit(X_train, y_train)
+    y_pred = clf.predict(X_test)
+    print(clf.__class__.__name__, accuracy_score(y_test, y_pred))
+
    
+
    
+









    
+
+
    Please, not the moons again! Voting and Bagging 
    
+
    
+
+
+
    from sklearn.model_selection import train_test_split
+from sklearn.datasets import make_moons
+
+X, y = make_moons(n_samples=500, noise=0.30, random_state=42)
+X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=42)
+from sklearn.ensemble import RandomForestClassifier
+from sklearn.ensemble import VotingClassifier
+from sklearn.linear_model import LogisticRegression
+from sklearn.svm import SVC
+
+log_clf = LogisticRegression(random_state=42)
+rnd_clf = RandomForestClassifier(random_state=42)
+svm_clf = SVC(random_state=42)
+
+voting_clf = VotingClassifier(
+    estimators=[('lr', log_clf), ('rf', rnd_clf), ('svc', svm_clf)],
+    voting='hard')
+voting_clf.fit(X_train, y_train)
+
    
+
    
+
+
+
    from sklearn.metrics import accuracy_score
+
+for clf in (log_clf, rnd_clf, svm_clf, voting_clf):
+    clf.fit(X_train, y_train)
+    y_pred = clf.predict(X_test)
+    print(clf.__class__.__name__, accuracy_score(y_test, y_pred))
+
    
+
    
+
+
+
    log_clf = LogisticRegression(random_state=42)
+rnd_clf = RandomForestClassifier(random_state=42)
+svm_clf = SVC(probability=True, random_state=42)
+
+voting_clf = VotingClassifier(
+    estimators=[('lr', log_clf), ('rf', rnd_clf), ('svc', svm_clf)],
+    voting='soft')
+voting_clf.fit(X_train, y_train)
+
    
+
    
+
+
+
    from sklearn.metrics import accuracy_score
+
+for clf in (log_clf, rnd_clf, svm_clf, voting_clf):
+    clf.fit(X_train, y_train)
+    y_pred = clf.predict(X_test)
+    print(clf.__class__.__name__, accuracy_score(y_test, y_pred))
+
    
+
    
+









    
+
+
    Now Bagging  
    
+
+
    
+
+
+
    from sklearn.ensemble import BaggingClassifier
+from sklearn.tree import DecisionTreeClassifier
+
+bag_clf = BaggingClassifier(
+    DecisionTreeClassifier(random_state=42), n_estimators=500,
+    max_samples=100, bootstrap=True, n_jobs=-1, random_state=42)
+bag_clf.fit(X_train, y_train)
+y_pred = bag_clf.predict(X_test)
+
    
+
    
+
+
+
    from sklearn.metrics import accuracy_score
+print(accuracy_score(y_test, y_pred))
+
    
+
    
+
+
+
    tree_clf = DecisionTreeClassifier(random_state=42)
+tree_clf.fit(X_train, y_train)
+y_pred_tree = tree_clf.predict(X_test)
+print(accuracy_score(y_test, y_pred_tree))
+
    
+
    
+
+
+
    from matplotlib.colors import ListedColormap
+
+def plot_decision_boundary(clf, X, y, axes=[-1.5, 2.5, -1, 1.5], alpha=0.5, contour=True):
+    x1s = np.linspace(axes[0], axes[1], 100)
+    x2s = np.linspace(axes[2], axes[3], 100)
+    x1, x2 = np.meshgrid(x1s, x2s)
+    X_new = np.c_[x1.ravel(), x2.ravel()]
+    y_pred = clf.predict(X_new).reshape(x1.shape)
+    custom_cmap = ListedColormap(['#fafab0','#9898ff','#a0faa0'])
+    plt.contourf(x1, x2, y_pred, alpha=0.3, cmap=custom_cmap)
+    if contour:
+        custom_cmap2 = ListedColormap(['#7d7d58','#4c4c7f','#507d50'])
+        plt.contour(x1, x2, y_pred, cmap=custom_cmap2, alpha=0.8)
+    plt.plot(X[:, 0][y==0], X[:, 1][y==0], "yo", alpha=alpha)
+    plt.plot(X[:, 0][y==1], X[:, 1][y==1], "bs", alpha=alpha)
+    plt.axis(axes)
+    plt.xlabel(r"$x_1$", fontsize=18)
+    plt.ylabel(r"$x_2$", fontsize=18, rotation=0)
+plt.figure(figsize=(11,4))
+plt.subplot(121)
+plot_decision_boundary(tree_clf, X, y)
+plt.title("Decision Tree", fontsize=14)
+plt.subplot(122)
+plot_decision_boundary(bag_clf, X, y)
+plt.title("Decision Trees with Bagging", fontsize=14)
+plt.show()
+
    
+
    
+









    
+
+
    Making our own Bagging with Bootstrap 
    
+
    
+
+
+
    import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import train_test_split
+from sklearn.pipeline import make_pipeline
+from sklearn.utils import resample
+from sklearn.tree import DecisionTreeRegressor
+
+
+np.random.seed(2018)
+
+n = 40
+n_boostraps = 100
+maxdegree = 14
+
+# Make data set.
+x = np.linspace(-3, 3, n).reshape(-1, 1)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
+error = np.zeros(maxdegree)
+bias = np.zeros(maxdegree)
+variance = np.zeros(maxdegree)
+polydegree = np.zeros(maxdegree)
+X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
+
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+for degree in range(maxdegree):
+    model = DecisionTreeRegressor(max_depth=5) 
+    y_pred = np.empty((y_test.shape[0], n_boostraps))
+    for i in range(n_boostraps):
+        x_, y_ = resample(X_train_scaled, y_train)
+        model.fit(x_, y_)
+        y_pred[:, i] = model.predict(X_test_scaled).ravel()
+
+    polydegree[degree] = degree
+    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )
+    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )
+    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )
+    print('Polynomial degree:', degree)
+    print('Error:', error[degree])
+    print('Bias^2:', bias[degree])
+    print('Var:', variance[degree])
+    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
+
+plt.plot(polydegree, error, label='Error')
+plt.plot(polydegree, bias, label='bias')
+plt.plot(polydegree, variance, label='Variance')
+plt.legend()
+plt.show()
+
    
+
    
+









    
+
+
    Changing the Level of the Decision Tree 
    
+
+
    
+
+
+
    import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import train_test_split
+from sklearn.pipeline import make_pipeline
+from sklearn.utils import resample
+from sklearn.tree import DecisionTreeRegressor
+
+n = 100
+n_boostraps = 100
+maxdepth = 8
+
+# Make data set.
+x = np.linspace(-3, 3, n).reshape(-1, 1)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
+error = np.zeros(maxdepth)
+bias = np.zeros(maxdepth)
+variance = np.zeros(maxdepth)
+polydegree = np.zeros(maxdepth)
+X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
+
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+for degree in range(1,maxdepth):
+    model = DecisionTreeRegressor(max_depth=degree) 
+    y_pred = np.empty((y_test.shape[0], n_boostraps))
+    for i in range(n_boostraps):
+        x_, y_ = resample(X_train_scaled, y_train)
+        model.fit(x_, y_)
+        y_pred[:, i] = model.predict(X_test_scaled)#.ravel()
+
+    polydegree[degree] = degree
+    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )
+    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )
+    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )
+    print('Polynomial degree:', degree)
+    print('Error:', error[degree])
+    print('Bias^2:', bias[degree])
+    print('Var:', variance[degree])
+    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
+
+plt.xlim(1,maxdepth)
+plt.plot(polydegree, error, label='Error')
+plt.plot(polydegree, bias, label='bias')
+plt.plot(polydegree, variance, label='Variance')
+plt.legend()
+plt.show()
+
    
+
    
+









    
+
+
    Random forests 
    
+
+
    
+Random forests provide an improvement over bagged trees by way of a
+small tweak that decorrelates the trees.
+
+
    
+As in bagging, we build a
+number of decision trees on bootstrapped training samples. But when
+building these decision trees, each time a split in a tree is
+considered, a random sample of \( m \) predictors is chosen as split
+candidates from the full set of \( p \) predictors. The split is allowed to
+use only one of those \( m \) predictors.
+
+
    
+A fresh sample of \( m \) predictors is
+taken at each split, and typically we choose 
+
+$$
+m\approx \sqrt{p}.
+$$
+
+
    
+In building a random forest, at
+each split in the tree, the algorithm is not even allowed to consider
+a majority of the available predictors.
+
+
    
+The reason for this is rather clever. Suppose that there is one very
+strong predictor in the data set, along with a number of other
+moderately strong predictors. Then in the collection of bagged
+variable importance random forest trees, most or all of the trees will
+use this strong predictor in the top split. Consequently, all of the
+bagged trees will look quite similar to each other. Hence the
+predictions from the bagged trees will be highly correlated.
+Unfortunately, averaging many highly correlated quantities does not
+lead to as large of a reduction in variance as averaging many
+uncorrelated quanti- ties. In particular, this means that bagging will
+not lead to a substantial reduction in variance over a single tree in
+this setting.
+
+
    
+









    
+
+
    Random Forest Algorithm 
    
+The algorithm described here can be applied to both classification and regression problems.
+
+
    
+We will grow of forest of say \( M \) trees.
+
+
      
+
    1.  For \( m=1:M \) we
      2. 
+
+
        
+  
      •  Draw a bootstrap sample of from the training data organized in our \( \boldsymbol{X} \) matrix.
        • 
+  
      •  We grow then a random forest tree \( T_m \) based on the bootstrapped data by repeating the steps outlined till we reach the maximum node size is reached
        • 
+
+
          
+   
        1.  we select \( m \le p \) varibales at random from the \( p \) predictors/features
          2. 
+   
        2.  pick the best split point among the \( m \) features using either the CART algorithm or the ID3 for classification and create a new node
          3. 
+   
        3.  split the node into daughter nodes
          4. 
+
        
+
+
      
+
+
    2.  Output then the ensemble of trees \( \{T_m\}_1^{M} \) and make predictions for either a regression type of problem or a classification type of problem.
      3.  
+
    
+
+









    
+
+
    Bootstrap with Random Forests Instead of a Single Tree, own Bagging 
    
+
+
    
+
+
+
    import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import train_test_split
+from sklearn.pipeline import make_pipeline
+from sklearn.utils import resample
+from sklearn.ensemble import RandomForestRegressor
+
+np.random.seed(2018)
+
+n = 100
+n_boostraps = 100
+maxdegree = 14
+
+# Make data set.
+x = np.linspace(-3, 3, n).reshape(-1, 1)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
+error = np.zeros(maxdegree)
+bias = np.zeros(maxdegree)
+variance = np.zeros(maxdegree)
+polydegree = np.zeros(maxdegree)
+X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
+
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+for degree in range(maxdegree):
+    model = RandomForestRegressor()
+    y_pred = np.empty((y_test.shape[0], n_boostraps))
+    for i in range(n_boostraps):
+        x_, y_ = resample(X_train_scaled, y_train)
+        model.fit(x_, y_.ravel())
+        y_pred[:, i] = model.predict(X_test_scaled).ravel()
+
+    polydegree[degree] = degree
+    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )
+    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )
+    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )
+    print('Polynomial degree:', degree)
+    print('Error:', error[degree])
+    print('Bias^2:', bias[degree])
+    print('Var:', variance[degree])
+    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
+
+plt.plot(polydegree, error, label='Error')
+plt.plot(polydegree, bias, label='bias')
+plt.plot(polydegree, variance, label='Variance')
+plt.legend()
+plt.show()
+
    
+
    
+









    
+
+
    Random Forests Compared with other Methods on the Cancer Data 
    
+
    
+
+
+
    import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import  train_test_split 
+from sklearn.datasets import load_breast_cancer
+from sklearn.svm import SVC
+from sklearn.linear_model import LogisticRegression
+from sklearn.tree import DecisionTreeClassifier
+
+# Load the data
+cancer = load_breast_cancer()
+
+X_train, X_test, y_train, y_test = train_test_split(cancer.data,cancer.target,random_state=0)
+print(X_train.shape)
+print(X_test.shape)
+# Logistic Regression
+logreg = LogisticRegression(solver='lbfgs')
+logreg.fit(X_train, y_train)
+print("Test set accuracy with Logistic Regression: {:.2f}".format(logreg.score(X_test,y_test)))
+# Support vector machine
+svm = SVC(gamma='auto', C=100)
+svm.fit(X_train, y_train)
+print("Test set accuracy with SVM: {:.2f}".format(svm.score(X_test,y_test)))
+# Decision Trees
+deep_tree_clf = DecisionTreeClassifier(max_depth=None)
+deep_tree_clf.fit(X_train, y_train)
+print("Test set accuracy with Decision Trees: {:.2f}".format(deep_tree_clf.score(X_test,y_test)))
+#now scale the data
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+# Logistic Regression
+logreg.fit(X_train_scaled, y_train)
+print("Test set accuracy Logistic Regression with scaled data: {:.2f}".format(logreg.score(X_test_scaled,y_test)))
+# Support Vector Machine
+svm.fit(X_train_scaled, y_train)
+print("Test set accuracy SVM with scaled data: {:.2f}".format(logreg.score(X_test_scaled,y_test)))
+# Decision Trees
+deep_tree_clf.fit(X_train_scaled, y_train)
+print("Test set accuracy with Decision Trees and scaled data: {:.2f}".format(deep_tree_clf.score(X_test_scaled,y_test)))
+
+
+from sklearn.ensemble import RandomForestClassifier
+from sklearn.preprocessing import LabelEncoder
+from sklearn.model_selection import cross_validate
+# Data set not specificied
+#Instantiate the model with 500 trees and entropy as splitting criteria
+Random_Forest_model = RandomForestClassifier(n_estimators=500,criterion="entropy")
+Random_Forest_model.fit(X_train_scaled, y_train)
+#Cross validation
+accuracy = cross_validate(Random_Forest_model,X_test_scaled,y_test,cv=10)['test_score']
+print(accuracy)
+print("Test set accuracy with Random Forests and scaled data: {:.2f}".format(Random_Forest_model.score(X_test_scaled,y_test)))
+
+
+import scikitplot as skplt
+y_pred = Random_Forest_model.predict(X_test_scaled)
+skplt.metrics.plot_confusion_matrix(y_test, y_pred, normalize=True)
+plt.show()
+y_probas = Random_Forest_model.predict_proba(X_test_scaled)
+skplt.metrics.plot_roc(y_test, y_probas)
+plt.show()
+skplt.metrics.plot_cumulative_gain(y_test, y_probas)
+plt.show()
+
    
+
    
+









    
+
+
    Compare  Bagging on Trees with Random Forests  
    
+
    
+
+
+
    bag_clf = BaggingClassifier(
+    DecisionTreeClassifier(splitter="random", max_leaf_nodes=16, random_state=42),
+    n_estimators=500, max_samples=1.0, bootstrap=True, n_jobs=-1, random_state=42)
+
    
+
    
+
+
+
    bag_clf.fit(X_train, y_train)
+y_pred = bag_clf.predict(X_test)
+from sklearn.ensemble import RandomForestClassifier
+rnd_clf = RandomForestClassifier(n_estimators=500, max_leaf_nodes=16, n_jobs=-1, random_state=42)
+rnd_clf.fit(X_train, y_train)
+y_pred_rf = rnd_clf.predict(X_test)
+np.sum(y_pred == y_pred_rf) / len(y_pred) 
+
    
+
    
+









    
+
+
    Boosting, a Bird'e Eye 
    
+
+
    
+The basic idea is to combine weak classifiers in order to create a good
+classifier. With a weak classifier we often intend a classifier which
+produces results which are only slightly better than we would get by
+random guesses.
+
+
    
+This is done by applying in an iterative way a weak (or a standard
+classifier like decision trees) to modify the data. In each iteration
+we emphasize those observations which are misclassified by weighting
+them with a factor.
+
+
    
+









    
+
+
    Adaptive boosting: AdaBoost, Basic Algorithm 
    
+
+
    
+The algorithm here is rather straightforward. Assume that our weak
+classifier is a decision tree and we consider a binary set of outputs
+with \( y_i \in \{-1,1\} \) and \( i=0,1,2,\dots,n-1 \) as our set of
+observations. Our design matrix is given in terms of the
+feature/predictor vectors
+\( \boldsymbol{X}=[\boldsymbol{x}_0\boldsymbol{x}_1\dots\boldsymbol{x}_{p-1} \). Finally, we define also a
+classifier determined by our data via a function \( G(\boldsymbol{X}) \). This function tells us how well we are able to classify our outputs/targets \( \boldsymbol{y} \).
+
+
    
+We can then define the misclassification error \( \mathrm{err} \) as
+$$
+\mathrm{err}=\frac{1}{n}\sum_{i=0}^{n-1}I(y_i\ne G(\boldsymbol{X}_{i*}),
+$$
+
+where the function \( I() \) is one if we misclassify and zero if we classify correctly.
+
+
    
+









    
+
+
    Basic Steps of AdaBoost 
    
+
+
    
+With the above definitions we are now ready to set up the algorithm for AdaBoost.
+The basic idea is to set up weights which will be used to scale the correctly classified and the misclassified cases.
+
+
      
+
    1.  We start by initializing all weights to \( w_i = 1/n \), with \( i=0,1,2,\dots n-1 \). It is to see then that \( \sum_{i=0}^{n-1}w_i = 1 \).
      2. 
+
    2.  We rewrite the misclassification error as
      3.  
+
    
+
+$$
+\mathrm{err}=\frac{\sum_{i=0}^{n-1}w_iI(y_i\ne G(\boldsymbol{X}_{i*})}{\sum_{i=0}^{n-1}w_i},
+$$
+
+
+
      
+
    1.  Then we start looping over all attempts at classifying, namely we start an iterative process for \( m=1:M \), where \( M \) is the final number of classifications. Our given classifier could for example be a plain decision tree.
+
+
        
+ 
      1.  Fit then a given classifier to the training using the weights \( w_i \).
        2. 
+ 
      2.  Compute then \( \mathrm{err} \) and figure out which events are classified properly and which are classified wrongly.
        3. 
+ 
      3.  Define a quantity $\alpha_{m} = \log{(1-\mathrm{err})/\mathrm{err}}
        4. 
+ 
      4.  Set the new weights to $w_i = w_i\times \exp{(\alpha_m I(y_i\ne G(\boldsymbol{X}_{i*})}.
        5. 
+
      
+
+
    2.  Compute the new classifier $G(\boldsymbol{X})= \sum_{i=0}^{n-1}\alpha_m I(y_i\ne G(\boldsymbol{X}_{i*}).
      3. 
+
    
+
+For the iterations with \( m \le 2 \) the weights are modified
+individually at each steps. The obersvations which were misclassified
+at iteration \( m-1 \) have a weight which is larger than those which were
+classified properly. As this proceeds, the observations which were
+difficult to classifiy correctly are given a larger influence. Each
+new classification step \( m \) is then forced to concentrate on those
+observations that are missed in the previous iterations.
+
+
    
+









    
+
+
    AdaBoost Examples 
    
+
+
    
+Using Scikit-Learn it is easy to appply the adaptive boosting algorithm, as done here.
+
+
    
+
+
+
    from sklearn.ensemble import AdaBoostClassifier
+
+ada_clf = AdaBoostClassifier(
+    DecisionTreeClassifier(max_depth=1), n_estimators=200,
+    algorithm="SAMME.R", learning_rate=0.5, random_state=42)
+ada_clf.fit(X_train, y_train)
+
+plot_decision_boundary(ada_clf, X, y)
+
+m = len(X_train)
+
+plt.figure(figsize=(11, 4))
+for subplot, learning_rate in ((121, 1), (122, 0.5)):
+    sample_weights = np.ones(m)
+    plt.subplot(subplot)
+    for i in range(5):
+        svm_clf = SVC(kernel="rbf", C=0.05, gamma="auto", random_state=42)
+        svm_clf.fit(X_train, y_train, sample_weight=sample_weights)
+        y_pred = svm_clf.predict(X_train)
+        sample_weights[y_pred != y_train] *= (1 + learning_rate)
+        plot_decision_boundary(svm_clf, X, y, alpha=0.2)
+        plt.title("learning_rate = {}".format(learning_rate), fontsize=16)
+    if subplot == 121:
+        plt.text(-0.7, -0.65, "1", fontsize=14)
+        plt.text(-0.6, -0.10, "2", fontsize=14)
+        plt.text(-0.5,  0.10, "3", fontsize=14)
+        plt.text(-0.4,  0.55, "4", fontsize=14)
+        plt.text(-0.3,  0.90, "5", fontsize=14)
+
+save_fig("boosting_plot")
+plt.show()
+
    
+
    
+









    
+
+
    Gradient boosting: Basics 
    
+
+
    
+Gradient boosting is again a similar technique to Adapative boosting,
+it combines so-called weak classifiers or regressors into a strong
+method via a series of iterations.
+
+
    
+In order to understand the method, let us illustrate its basics by
+bringing back the essential steps in linear regression, where our cost
+function was the least squares function.
+
+
    
+









    
+
+
    Gradient Boosting, algorithm 
    
+
+
    
+Suppose we have a cost function \( C(f)=\sum_{i=0}^{n-1}L(y_i, f(x_i)) \) where \( y_i \) is our target and \( f(x_i) \) the function which is meant to model \( y_i \). The above cost function could be our standard least squares  function
+$$
+C(\boldsymbol{y},\boldsymbol{f})=\frac{1}{n}\sum_{i=0}^{n-1}(y_i-f(x_i))^2.
+$$
+
+
    
+The way we proceed in an iterative fashion is to
+
+
      
+
    1.  Initialize our estimate by \( f_0(x)=0 \).
      2. 
+
    2.  For \( m=1:M \), we
+
+
        
+ 
      1.  compute the negative gradient vector \( \boldsymbol{u}_m = -\partial C(\boldsymbol{y},\boldsymbol{f})/\partial \boldsymbol{f}(x) \) at $f(x) = f_{m-1}(x);
        2. 
+ 
      2.  fit the so-called base-learner to the negative gradient \( h_m(u_m,x) \);
        3. 
+ 
      3.  update the estimate \( f_m(x) = f_{m-1}(x)+\nu h_m(u_m,x) \);
        4. 
+
      
+
+
    3.  The final estimate is then \( f_M(x) = \sum_{m=1}^M\nu h_m(u_m,x) \).
      4. 
+
    
+
+









    
+
+
    Gradient Boosting, Examples 
    
+
    
+
+
+
    np.random.seed(42)
+X = np.random.rand(100, 1) - 0.5
+y = 3*X[:, 0]**2 + 0.05 * np.random.randn(100)
+
+from sklearn.tree import DecisionTreeRegressor
+
+tree_reg1 = DecisionTreeRegressor(max_depth=2, random_state=42)
+tree_reg1.fit(X, y)
+
+y2 = y - tree_reg1.predict(X)
+tree_reg2 = DecisionTreeRegressor(max_depth=2, random_state=42)
+tree_reg2.fit(X, y2)
+
+y3 = y2 - tree_reg2.predict(X)
+tree_reg3 = DecisionTreeRegressor(max_depth=2, random_state=42)
+tree_reg3.fit(X, y3)
+
+X_new = np.array([[0.8]])
+y_pred = sum(tree.predict(X_new) for tree in (tree_reg1, tree_reg2, tree_reg3))
+
+def plot_predictions(regressors, X, y, axes, label=None, style="r-", data_style="b.", data_label=None):
+    x1 = np.linspace(axes[0], axes[1], 500)
+    y_pred = sum(regressor.predict(x1.reshape(-1, 1)) for regressor in regressors)
+    plt.plot(X[:, 0], y, data_style, label=data_label)
+    plt.plot(x1, y_pred, style, linewidth=2, label=label)
+    if label or data_label:
+        plt.legend(loc="upper center", fontsize=16)
+    plt.axis(axes)
+
+plt.figure(figsize=(11,11))
+
+plt.subplot(321)
+plot_predictions([tree_reg1], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="$h_1(x_1)$", style="g-", data_label="Training set")
+plt.ylabel("$y$", fontsize=16, rotation=0)
+plt.title("Residuals and tree predictions", fontsize=16)
+
+plt.subplot(322)
+plot_predictions([tree_reg1], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="$h(x_1) = h_1(x_1)$", data_label="Training set")
+plt.ylabel("$y$", fontsize=16, rotation=0)
+plt.title("Ensemble predictions", fontsize=16)
+
+plt.subplot(323)
+plot_predictions([tree_reg2], X, y2, axes=[-0.5, 0.5, -0.5, 0.5], label="$h_2(x_1)$", style="g-", data_style="k+", data_label="Residuals")
+plt.ylabel("$y - h_1(x_1)$", fontsize=16)
+
+plt.subplot(324)
+plot_predictions([tree_reg1, tree_reg2], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="$h(x_1) = h_1(x_1) + h_2(x_1)$")
+plt.ylabel("$y$", fontsize=16, rotation=0)
+
+plt.subplot(325)
+plot_predictions([tree_reg3], X, y3, axes=[-0.5, 0.5, -0.5, 0.5], label="$h_3(x_1)$", style="g-", data_style="k+")
+plt.ylabel("$y - h_1(x_1) - h_2(x_1)$", fontsize=16)
+plt.xlabel("$x_1$", fontsize=16)
+
+plt.subplot(326)
+plot_predictions([tree_reg1, tree_reg2, tree_reg3], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="$h(x_1) = h_1(x_1) + h_2(x_1) + h_3(x_1)$")
+plt.xlabel("$x_1$", fontsize=16)
+plt.ylabel("$y$", fontsize=16, rotation=0)
+
+save_fig("gradient_boosting_plot")
+plt.show()
+
+from sklearn.ensemble import GradientBoostingRegressor
+
+gbrt = GradientBoostingRegressor(max_depth=2, n_estimators=3, learning_rate=1.0, random_state=42)
+gbrt.fit(X, y)
+
+gbrt_slow = GradientBoostingRegressor(max_depth=2, n_estimators=200, learning_rate=0.1, random_state=42)
+gbrt_slow.fit(X, y)
+
+plt.figure(figsize=(11,4))
+
+plt.subplot(121)
+plot_predictions([gbrt], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="Ensemble predictions")
+plt.title("learning_rate={}, n_estimators={}".format(gbrt.learning_rate, gbrt.n_estimators), fontsize=14)
+
+plt.subplot(122)
+plot_predictions([gbrt_slow], X, y, axes=[-0.5, 0.5, -0.1, 0.8])
+plt.title("learning_rate={}, n_estimators={}".format(gbrt_slow.learning_rate, gbrt_slow.n_estimators), fontsize=14)
+
+save_fig("gbrt_learning_rate_plot")
+plt.show()
+
    
+
    
+









    
+
+
    Gradient Boots with Early Stopping 
    
+
    
+
+
+
    from sklearn.model_selection import train_test_split
+from sklearn.metrics import mean_squared_error
+
+X_train, X_val, y_train, y_val = train_test_split(X, y, random_state=49)
+
+gbrt = GradientBoostingRegressor(max_depth=2, n_estimators=120, random_state=42)
+gbrt.fit(X_train, y_train)
+
+errors = [mean_squared_error(y_val, y_pred)
+          for y_pred in gbrt.staged_predict(X_val)]
+bst_n_estimators = np.argmin(errors) + 1
+
+gbrt_best = GradientBoostingRegressor(max_depth=2,n_estimators=bst_n_estimators, random_state=42)
+gbrt_best.fit(X_train, y_train)
+
+min_error = np.min(errors)
+plt.figure(figsize=(11, 4))
+
+plt.subplot(121)
+plt.plot(errors, "b.-")
+plt.plot([bst_n_estimators, bst_n_estimators], [0, min_error], "k--")
+plt.plot([0, 120], [min_error, min_error], "k--")
+plt.plot(bst_n_estimators, min_error, "ko")
+plt.text(bst_n_estimators, min_error*1.2, "Minimum", ha="center", fontsize=14)
+plt.axis([0, 120, 0, 0.01])
+plt.xlabel("Number of trees")
+plt.title("Validation error", fontsize=14)
+
+plt.subplot(122)
+plot_predictions([gbrt_best], X, y, axes=[-0.5, 0.5, -0.1, 0.8])
+plt.title("Best model (%d trees)" % bst_n_estimators, fontsize=14)
+
+save_fig("early_stopping_gbrt_plot")
+plt.show()
+
+
+gbrt = GradientBoostingRegressor(max_depth=2, warm_start=True, random_state=42)
+
+min_val_error = float("inf")
+error_going_up = 0
+for n_estimators in range(1, 120):
+    gbrt.n_estimators = n_estimators
+    gbrt.fit(X_train, y_train)
+    y_pred = gbrt.predict(X_val)
+    val_error = mean_squared_error(y_val, y_pred)
+    if val_error < min_val_error:
+        min_val_error = val_error
+        error_going_up = 0
+    else:
+        error_going_up += 1
+        if error_going_up == 5:
+            break  # early stopping
+
+
+print(gbrt.n_estimators)
+print("Minimum validation MSE:", min_val_error)
+
    
+
    
+









    
+
+
    XGBoost: Extreme Gradient Boosting 
    
+
+
    
+XGBoost or Extreme Gradient
+Boosting, is an optimized distributed gradient boosting library
+designed to be highly efficient, flexible and portable. It implements
+machine learning algorithms under the Gradient Boosting
+framework. XGBoost provides a parallel tree boosting that solve many
+data science problems in a fast and accurate way. See the article by Chen and Guestrin.
+
+
    
+The authors design and build a highly scalable end-to-end tree
+boosting system. It has  a theoretically justified weighted quantile
+sketch for efficient proposal calculation. It introduces a novel sparsity-aware algorithm for parallel tree learning and an effective cache-aware block structure for out-of-core tree learning.
+
+
    
+It is now the algorithm which wins essentially all ML competitions!!!
+
+
    
+









    
+
+
    Regression Case 
    
+
+
    
+
+
+
    import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import train_test_split
+import xgboost as xgb
+from sklearn.preprocessing import StandardScaler
+import scikitplot as skplt
+from sklearn.metrics import mean_squared_error
+
+n = 100
+maxdegree = 6
+
+# Make data set.
+x = np.linspace(-3, 3, n).reshape(-1, 1)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
+
+error = np.zeros(maxdegree)
+bias = np.zeros(maxdegree)
+variance = np.zeros(maxdegree)
+polydegree = np.zeros(maxdegree)
+X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+for degree in range(maxdegree):
+    model =  xgb.XGBRegressor(objective ='reg:squarederror', colsample_bytree = 0.3, learning_rate = 0.1,
+                max_depth = degree, alpha = 10, n_estimators = 10)
+    model.fit(X_train_scaled,y_train)
+    y_pred = model.predict(X_test_scaled)
+    polydegree[degree] = degree
+    error[degree] = np.mean( np.mean((y_test - y_pred)**2) )
+    bias[degree] = np.mean( (y_test - np.mean(y_pred))**2 )
+    variance[degree] = np.mean( np.var(y_pred) )
+    print('Max depth:', degree)
+    print('Error:', error[degree])
+    print('Bias^2:', bias[degree])
+    print('Var:', variance[degree])
+    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
+
+plt.xlim(1,maxdegree-1)
+plt.plot(polydegree, error, label='Error')
+plt.plot(polydegree, bias, label='bias')
+plt.plot(polydegree, variance, label='Variance')
+plt.legend()
+plt.show()
+
    
+
    
+









    
+
+
    Xgboost on the Cancer Data 
    
+
    
+
+
+
    import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import  train_test_split 
+from sklearn.datasets import load_breast_cancer
+from sklearn.preprocessing import LabelEncoder
+from sklearn.model_selection import cross_validate
+import scikitplot as skplt
+import xgboost as xgb
+# Load the data
+cancer = load_breast_cancer()
+
+X_train, X_test, y_train, y_test = train_test_split(cancer.data,cancer.target,random_state=0)
+print(X_train.shape)
+print(X_test.shape)
+#now scale the data
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+xg_clf = xgb.XGBClassifier()
+xg_clf.fit(X_train_scaled,y_train)
+xgb.plot_tree(xg_clf,num_trees=0)
+plt.rcParams['figure.figsize'] = [50, 10]
+plt.show()
+xgb.plot_importance(xg_clf)
+plt.rcParams['figure.figsize'] = [5, 5]
+plt.show()
+
    
+
    
+
+
+
+
+
    
+ © 1999-2019, Morten Hjorth-Jensen. Released under CC Attribution-NonCommercial 4.0 license
+
    
+
+
+
+
+    
+
diff --git a/doc/src/DecisionTrees/DecisionTrees.do.txt b/doc/src/DecisionTrees/DecisionTrees.do.txt
index 59a5a1bf7..cbdaccecb 100644
--- a/doc/src/DecisionTrees/DecisionTrees.do.txt
+++ b/doc/src/DecisionTrees/DecisionTrees.do.txt
@@ -1567,6 +1567,70 @@ o For $m=1:M$ we
 o Output then the ensemble of trees $\{T_m\}_1^{M}$ and make predictions for either a regression type of problem or a classification type of problem. 
 
 
+!split
+===== Bootstrap with Random Forests Instead of a Single Tree, own Bagging =====
+
+!bc pycod
+
+import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import train_test_split
+from sklearn.pipeline import make_pipeline
+from sklearn.utils import resample
+from sklearn.ensemble import RandomForestRegressor
+
+np.random.seed(2018)
+
+n = 100
+n_boostraps = 100
+maxdegree = 14
+
+# Make data set.
+x = np.linspace(-3, 3, n).reshape(-1, 1)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
+error = np.zeros(maxdegree)
+bias = np.zeros(maxdegree)
+variance = np.zeros(maxdegree)
+polydegree = np.zeros(maxdegree)
+X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
+
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+for degree in range(maxdegree):
+    model = RandomForestRegressor()
+    y_pred = np.empty((y_test.shape[0], n_boostraps))
+    for i in range(n_boostraps):
+        x_, y_ = resample(X_train_scaled, y_train)
+        model.fit(x_, y_.ravel())
+        y_pred[:, i] = model.predict(X_test_scaled).ravel()
+
+    polydegree[degree] = degree
+    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )
+    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )
+    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )
+    print('Polynomial degree:', degree)
+    print('Error:', error[degree])
+    print('Bias^2:', bias[degree])
+    print('Var:', variance[degree])
+    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
+
+plt.plot(polydegree, error, label='Error')
+plt.plot(polydegree, bias, label='bias')
+plt.plot(polydegree, variance, label='Variance')
+plt.legend()
+plt.show()
+
+
+
+!ec
+
+
+
+
 !split
 ===== Random Forests Compared with other Methods on the Cancer Data =====
 !bc pycod 
@@ -1661,67 +1725,6 @@ np.sum(y_pred == y_pred_rf) / len(y_pred)
 
 
 
-!split
-===== Bootstrap with Random Forests Instead of a Single Tree, own Bagging =====
-
-!bc pycod
-
-import matplotlib.pyplot as plt
-import numpy as np
-from sklearn.model_selection import train_test_split
-from sklearn.pipeline import make_pipeline
-from sklearn.utils import resample
-from sklearn.ensemble import RandomForestRegressor
-
-np.random.seed(2018)
-
-n = 100
-n_boostraps = 100
-maxdegree = 14
-
-# Make data set.
-x = np.linspace(-3, 3, n).reshape(-1, 1)
-y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
-error = np.zeros(maxdegree)
-bias = np.zeros(maxdegree)
-variance = np.zeros(maxdegree)
-polydegree = np.zeros(maxdegree)
-X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
-
-from sklearn.preprocessing import StandardScaler
-scaler = StandardScaler()
-scaler.fit(X_train)
-X_train_scaled = scaler.transform(X_train)
-X_test_scaled = scaler.transform(X_test)
-
-for degree in range(maxdegree):
-    model = RandomForestRegressor()
-    y_pred = np.empty((y_test.shape[0], n_boostraps))
-    for i in range(n_boostraps):
-        x_, y_ = resample(X_train_scaled, y_train)
-        model.fit(x_, y_.ravel())
-        y_pred[:, i] = model.predict(X_test_scaled).ravel()
-
-    polydegree[degree] = degree
-    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )
-    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )
-    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )
-    print('Polynomial degree:', degree)
-    print('Error:', error[degree])
-    print('Bias^2:', bias[degree])
-    print('Var:', variance[degree])
-    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
-
-plt.plot(polydegree, error, label='Error')
-plt.plot(polydegree, bias, label='bias')
-plt.plot(polydegree, variance, label='Variance')
-plt.legend()
-plt.show()
-
-
-
-!ec
-
 
 
 !split
diff --git a/doc/src/DecisionTrees/DecisionTrees.html b/doc/src/DecisionTrees/DecisionTrees.html
new file mode 100644
index 000000000..e66d392b2
--- /dev/null
+++ b/doc/src/DecisionTrees/DecisionTrees.html
@@ -0,0 +1,2507 @@
+
+
+
+
+
+
+
+
+Data Analysis and Machine Learning: From Decision Trees to Forests and all that
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+    
+
+
+
+
+
    Data Analysis and Machine Learning: From Decision Trees to Forests and all that
      
+
+
    
+
+
+
    
+Morten Hjorth-Jensen [1, 2]
+
    
+
+
    
+
+
+
    [1] Department of Physics, University of Oslo
    
+
    [2] Department of Physics and Astronomy and National Superconducting Cyclotron Laboratory, Michigan State University
    
+
    
+
    
+
    Nov 7, 2019
     
+
    
+
    
+









    
+
+
    Decision trees, overarching aims  
    
+
+
    
+Decision trees are supervised learning algorithms used for both,
+classification and regression tasks.
+
+
    
+The main idea of decision trees
+is to find those descriptive features which contain the most
+information regarding the target feature and then split the dataset
+along the values of these features such that the target feature values
+for the resulting underlying datasets are as pure as possible.
+
+
    
+The descriptive features which reproduce best the target/output features are normally  said
+to be the most informative ones. The process of finding the most
+informative feature is done until we accomplish a stopping criteria
+where we then finally end up in so called leaf nodes.
+
+
    
+A decision tree is typically divided into a root node, the interior nodes,
+and the final leaf nodes or just leaves. These entities are then connected by so-called branches.
+
+
    
+The leaf nodes
+contain the predictions we will make for new query instances presented
+to our trained model. This is possible since the model has 
+learned the underlying structure of the training data and hence can,
+given some assumptions, make predictions about the target feature value
+(class) of unseen query instances.
+
+
    
+









    
+
+
    A typical Decision Tree with its pertinent Jargon, Classification Problem 
    
+
+
    
+



    
+
+
    
+This tree was produced using the Wisconsin cancer data (discussed here as well, see code examples below) using Scikit-Learn's decision tree classifier. Here we have used the so-called gini index (see below) to split the various branches.
+
+
    
+









    
+
+
    General Features 
    
+
+
    
+The overarching approach to decision trees is a top-down approach.
+
+
      
+
    •  A leaf provides the classification of a given instance.
      • 
+
    •  A node specifies a test of some attribute of the instance.
      • 
+
    •  A branch corresponds to a possible values of an attribute.
      • 
+
    •  An instance is classified by starting at the root node of the tree, testing the attribute specified by this node, then moving down the tree branch corresponding to the value of the attribute in the given example.
      • 
+
    
+
+This process is then repeated for the subtree rooted at the new
+node.
+
+
    
+









    
+
+
    How do we set it up? 
    
+
+
    
+In simplified terms, the process of training a decision tree and
+predicting the target features of query instances is as follows:
+
+
      
+
    1.  Present a dataset containing of a number of training instances characterized by a number of descriptive features and a target feature
      2. 
+
    2.  Train the decision tree model by continuously splitting the target feature along the values of the descriptive features using a measure of information gain during the training process
      3. 
+
    3.  Grow the tree until we accomplish a stopping criteria create leaf nodes which represent the predictions we want to make for new query instances
      4. 
+
    4.  Show query instances to the tree and run down the tree until we arrive at leaf nodes
      5. 
+
    
+
+Then we are essentially done!
+
+
    
+









    
+
+
    Decision trees and Regression  
    
+
    
+
+
+
    import numpy as np
+import matplotlib.pyplot as plt
+from sklearn.preprocessing import PolynomialFeatures
+from sklearn.linear_model import LinearRegression
+
+steps=250
+
+distance=0
+x=0
+distance_list=[]
+steps_list=[]
+while x<steps:
+    distance+=np.random.randint(-1,2)
+    distance_list.append(distance)
+    x+=1
+    steps_list.append(x)
+plt.plot(steps_list,distance_list, color='green', label="Random Walk Data")
+
+steps_list=np.asarray(steps_list)
+distance_list=np.asarray(distance_list)
+
+X=steps_list[:,np.newaxis]
+
+#Polynomial fits
+
+#Degree 2
+poly_features=PolynomialFeatures(degree=2, include_bias=False)
+X_poly=poly_features.fit_transform(X)
+
+lin_reg=LinearRegression()
+poly_fit=lin_reg.fit(X_poly,distance_list)
+b=lin_reg.coef_
+c=lin_reg.intercept_
+print ("2nd degree coefficients:")
+print ("zero power: ",c)
+print ("first power: ", b[0])
+print ("second power: ",b[1])
+
+z = np.arange(0, steps, .01)
+z_mod=b[1]*z**2+b[0]*z+c
+
+fit_mod=b[1]*X**2+b[0]*X+c
+plt.plot(z, z_mod, color='r', label="2nd Degree Fit")
+plt.title("Polynomial Regression")
+
+plt.xlabel("Steps")
+plt.ylabel("Distance")
+
+#Degree 10
+poly_features10=PolynomialFeatures(degree=10, include_bias=False)
+X_poly10=poly_features10.fit_transform(X)
+
+poly_fit10=lin_reg.fit(X_poly10,distance_list)
+
+y_plot=poly_fit10.predict(X_poly10)
+plt.plot(X, y_plot, color='black', label="10th Degree Fit")
+
+plt.legend()
+plt.show()
+
+
+#Decision Tree Regression
+from sklearn.tree import DecisionTreeRegressor
+regr_1=DecisionTreeRegressor(max_depth=2)
+regr_2=DecisionTreeRegressor(max_depth=5)
+regr_3=DecisionTreeRegressor(max_depth=7)
+regr_1.fit(X, distance_list)
+regr_2.fit(X, distance_list)
+regr_3.fit(X, distance_list)
+
+X_test = np.arange(0.0, steps, 0.01)[:, np.newaxis]
+y_1 = regr_1.predict(X_test)
+y_2 = regr_2.predict(X_test)
+y_3=regr_3.predict(X_test)
+
+# Plot the results
+plt.figure()
+plt.scatter(X, distance_list, s=2.5, c="black", label="data")
+plt.plot(X_test, y_1, color="red",
+         label="max_depth=2", linewidth=2)
+plt.plot(X_test, y_2, color="green", label="max_depth=5", linewidth=2)
+plt.plot(X_test, y_3, color="m", label="max_depth=7", linewidth=2)
+
+plt.xlabel("Data")
+plt.ylabel("Darget")
+plt.title("Decision Tree Regression")
+plt.legend()
+plt.show()
+
    
+
    
+









    
+
+
    Building a tree, regression 
    
+
+
    
+There are mainly two steps
+
+
      
+
    1.  We split the predictor space (the set of possible values \( x_1,x_2,\dots, x_p \)) into \( J \) distinct and non-non-overlapping regions, \( R_1,R_2,\dots,R_J \).
      2.   
+
    2.  For every observation that falls into the region \( R_j \) , we make the same prediction, which is simply the mean of the response values for the training observations in \( R_j \).
      3. 
+
    
+
+How do we construct the regions \( R_1,\dots,R_J \)?  In theory, the
+regions could have any shape. However, we choose to divide the
+predictor space into high-dimensional rectangles, or boxes, for
+simplicity and for ease of interpretation of the resulting predictive
+model. The goal is to find boxes \( R_1,\dots,R_J \) that minimize the
+MSE, given by
+
+$$
+\sum_{j=1}^J\sum_{i\in R_j}(y_i-\overline{y}_{R_j})^2,
+$$
+
+
    
+where \( \overline{y}_{R_j} \)  is the mean response for the training observations 
+within box \( j \).
+
+
    
+









    
+
+
    A top-down approach, recursive binary splitting 
    
+
+
    
+Unfortunately, it is computationally infeasible to consider every
+possible partition of the feature space into \( J \) boxes.  The common
+strategy is to take a top-down approach
+
+
    
+The approach is top-down because it begins at the top of the tree (all
+observations belong to a single region) and then successively splits
+the predictor space; each split is indicated via two new branches
+further down on the tree. It is greedy because at each step of the
+tree-building process, the best split is made at that particular step,
+rather than looking ahead and picking a split that will lead to a
+better tree in some future step.
+
+
    
+









    
+
+
    Making a tree 
    
+
+
    
+In order to implement the recursive binary splitting we start by selecting
+the predictor \( x_j \) and a cutpoint \( s \) that splits the predictor space into two regions \( R_1 \) and \( R_2 \)
+$$
+\left\{X\vert x_j < s\right\},
+$$
+
+and
+$$
+\left\{X\vert x_j \geq s\right\},
+$$
+
+so that we obtain the lowest MSE, that is
+$$
+\sum_{i:x_i\in R_j}(y_i-\overline{y}_{R_1})^2+\sum_{i:x_i\in R_2}(y_i-\overline{y}_{R_2})^2,
+$$
+
+
    
+which we want to minimize by considering all predictors
+\( x_1,x_2,\dots,x_p \).  We consider also all possible values of \( s \) for
+each predictor. These values could be determined by randomly assigned
+numbers or by starting at the midpoint and then proceed till we find
+an optimal value.
+
+
    
+For any \( j \) and \( s \), we define the pair of half-planes where
+\( \overline{y}_{R_1} \) is the mean response for the training
+observations in \( R_1(j,s) \), and \( \overline{y}_{R_2} \) is the mean
+response for the training observations in \( R_2(j,s) \).
+
+
    
+Finding the values of \( j \) and \( s \) that minimize the above equation can be
+done quite quickly, especially when the number of features \( p \) is not
+too large.
+
+
    
+Next, we repeat the process, looking
+for the best predictor and best cutpoint in order to split the data
+further so as to minimize the MSE within each of the resulting
+regions. However, this time, instead of splitting the entire predictor
+space, we split one of the two previously identified regions. We now
+have three regions. Again, we look to split one of these three regions
+further, so as to minimize the MSE. The process continues until a
+stopping criterion is reached; for instance, we may continue until no
+region contains more than five observations.
+
+
    
+
+
+
    Pruning the tree 
    
+
+
    
+The above procedure is rather straightforward, but leads often to
+overfitting and unnecessarily large and complicated trees. The basic
+idea is to grow a large tree \( T_0 \) and then prune it back in order to
+obtain a subtree. A smaller tree with fewer splits (fewer regions) can
+lead to smaller variance and better interpretation at the cost of a
+little more bias.
+
+
    
+The so-called Cost complexity pruning algorithm gives us a
+way to do just this. Rather than considering every possible subtree,
+we consider a sequence of trees indexed by a nonnegative tuning
+parameter \( \alpha \).
+
+
    
+









    
+
+
    Cost complexity pruning 
    
+For each value of \( \alpha \)  there corresponds a subtree \( T \in T_0 \) such that
+$$
+\sum_{m=1}^{\overline{T}}\sum_{i:x_i\in R_m}(y_i-\overline{y}_{R_m})^2+\alpha\overline{T},
+$$
+
+is as small as possible. Here \( \overline{T} \) is 
+the number of terminal nodes of the tree \( T \) , \( R_m \) is the
+rectangle (i.e. the subset of predictor space)  corresponding to the \( m \)-th terminal node.
+
+
    
+The tuning parameter \( \alpha \) controls a trade-off between the subtree’s
+com- plexity and its fit to the training data. When \( \alpha = 0 \), then the
+subtree \( T \) will simply equal \( T_0 \), 
+because then the above equation just measures the
+training error. 
+However, as \( \alpha \) increases, there is a price to pay for
+having a tree with many terminal nodes. The above equation will
+tend to be minimized for a smaller subtree.
+
+
    
+It turns out that as we increase \( \alpha \) from zero
+branches get pruned from the tree in a nested and predictable fashion,
+so obtaining the whole sequence of subtrees as a function of \( \alpha \) is
+easy. We can select a value of \( \alpha \) using a validation set or using
+cross-validation. We then return to the full data set and obtain the
+subtree corresponding to \( \alpha \).
+
+
    
+









    
+
+
    Schematic Regression Procedure 
    
+
+
    
+
    
+Building a Regression Tree.
+
    
+
+
      
+
    1.  Use recursive binary splitting to grow a large tree on the training data, stopping only when each terminal node has fewer than some minimum number of observations.
      2. 
+
    2.  Apply cost complexity pruning to the large tree in order to obtain a sequence of best subtrees, as a function of \( \alpha \).
      3. 
+
    3.  Use for example \( K \)-fold cross-validation to choose \( \alpha \). Divide the training observations into \( K \) folds. For each \( k=1,2,\dots,K \) we:
      4.  
+
+
        
+  
      •  repeat steps 1 and 2 on all but the \( k \)-th fold of the training data.
        •  
+  
      •  Then we valuate the mean squared prediction error on the data in the left-out \( k \)-th fold, as a function of \( \alpha \).
        • 
+  
      •  Finally  we average the results for each value of \( \alpha \), and pick \( \alpha \) to minimize the average error.
        • 
+
      
+
+
    4.  Return the subtree from Step 2 that corresponds to the chosen value of \( \alpha \).
      5.  
+
    
+
    
+
+
+
    
+









    
+
+
    A Classification Tree 
    
+
+
    
+A classification tree is very similar to a regression tree, except
+that it is used to predict a qualitative response rather than a
+quantitative one. Recall that for a regression tree, the predicted
+response for an observation is given by the mean response of the
+training observations that belong to the same terminal node. In
+contrast, for a classification tree, we predict that each observation
+belongs to the most commonly occurring class of training observations
+in the region to which it belongs. In interpreting the results of a
+classification tree, we are often interested not only in the class
+prediction corresponding to a particular terminal node region, but
+also in the class proportions among the training observations that
+fall into that region.
+
+
    
+









    
+
+
    Growing a classification tree 
    
+
+
    
+The task of growing a
+classification tree is quite similar to the task of growing a
+regression tree. Just as in the regression setting, we use recursive
+binary splitting to grow a classification tree. However, in the
+classification setting, the MSE cannot be used as a criterion for making
+the binary splits.  A natural alternative to MSE is the classification
+error rate. Since we plan to assign an observation in a given region
+to the most commonly occurring error rate class of training
+observations in that region, the classification error rate is simply
+the fraction of the training observations in that region that do not
+belong to the most common class.
+
+
    
+When building a classification tree, either the Gini index or the
+entropy are typically used to evaluate the quality of a particular
+split, since these two approaches are more sensitive to node purity
+than is the classification error rate.
+
+
    
+









    
+
+
    Classification tree, how to split nodes 
    
+
+
    
+If our targets are the outcome of a classification process that takes
+for example \( k=1,2,\dots,K \) values, the only thing we need to think of
+is to set up the splitting criteria for each node.
+
+
    
+We define a PDF \( p_{mk} \) that represents the number of observations of
+a class \( k \) in a region \( R_m \) with \( N_m \) observations. We represent
+this likelihood function in terms of the proportion \( I(y_i=k) \) of
+observations of this class in the region \( R_m \) as
+
+$$
+p_{mk} = \frac{1}{N_m}\sum_{x_i\in R_m}I(y_i=k).
+$$
+
+
    
+We let \( p_{mk} \) represent the majority class of observations in region
+\( m \). The three most common ways of splitting a node are given by
+
+
      
+
    •  Misclassification error
      •  
+
    
+
+$$
+p_{mk} = \frac{1}{N_m}\sum_{x_i\in R_m}I(y_i\ne k) = 1-p_{mk}.
+$$
+
+
+
      
+
    •  Gini index \( g \)
      • 
+
    
+
+$$
+g = \sum_{k=1}^K p_{mk}(1-p_{mk}).
+$$
+
+
+
      
+
    •  Information entropy or just entropy \( s \)
      • 
+
    
+
+$$
+s = -\sum_{k=1}^K p_{mk}\log{p_{mk}}.
+$$
+
+
    
+









    
+
+
    Visualizing the Tree, Classification 
    
+
    
+
+
+
    import os
+from sklearn.datasets import load_breast_cancer
+from sklearn.tree import DecisionTreeClassifier
+from sklearn.model_selection import train_test_split
+from sklearn.metrics import confusion_matrix
+from sklearn.tree import export_graphviz
+
+from IPython.display import Image 
+from pydot import graph_from_dot_data
+import pandas as pd
+import numpy as np
+
+
+cancer = load_breast_cancer()
+X = pd.DataFrame(cancer.data, columns=cancer.feature_names)
+print(X)
+y = pd.Categorical.from_codes(cancer.target, cancer.target_names)
+y = pd.get_dummies(y)
+print(y)
+X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=1)
+tree_clf = DecisionTreeClassifier(max_depth=5)
+tree_clf.fit(X_train, y_train)
+
+export_graphviz(
+    tree_clf,
+    out_file="DataFiles/cancer.dot",
+    feature_names=cancer.feature_names,
+    class_names=cancer.target_names,
+    rounded=True,
+    filled=True
+)
+cmd = 'dot -Tpng DataFiles/cancer.dot -o DataFiles/cancer.png'
+os.system(cmd)
+
    
+
    
+









    
+
+
    Visualizing the Tree, The Moons  
    
+
    
+
+
+
    # Common imports
+import numpy as np
+from sklearn.model_selection import  train_test_split 
+from sklearn.tree import DecisionTreeClassifier
+from sklearn.datasets import make_moons
+from sklearn.tree import export_graphviz
+from pydot import graph_from_dot_data
+import pandas as pd
+import os
+
+np.random.seed(42)
+X, y = make_moons(n_samples=100, noise=0.25, random_state=53)
+X_train, X_test, y_train, y_test = train_test_split(X,y,random_state=0)
+tree_clf = DecisionTreeClassifier(max_depth=5)
+tree_clf.fit(X_train, y_train)
+
+export_graphviz(
+    tree_clf,
+    out_file="DataFiles/moons.dot",
+    rounded=True,
+    filled=True
+)
+cmd = 'dot -Tpng DataFiles/moons.dot -o DataFiles/moons.png'
+os.system(cmd)
+
    
+
    
+









    
+
+
    Algorithms for Setting up Decision Trees 
    
+
+
    
+Two algorithms stand out in the set up of decision trees:
+
+
      
+
    1.  The CART (Classification And Regression Tree) algorithm for both classification and regression
      2. 
+
    2.  The ID3 algorithm based on the computation of the information gain for classification
      3. 
+
    
+
+We discuss both algorithms with applications here. The popular library Scikit-Learn uses the CART algorithm. For classification problems you can use either the gini index or the entropy to split a tree in two branches.
+
+
    
+









    
+
+
    The CART algorithm for Classification 
    
+
+
    
+









    
+
+
    The CART algorithm for Regression 
    
+
+
    
+









    
+
+
    Computing the Gini index 
    
+
+
    
+The example we will look at is a classical one in many Machine
+Learning applications. Based on various meteorological features, we
+have several so-called attributes which decide whether we at the end
+will do some outdoor activity like skiing, going for a bike ride etc
+etc.  The table here contains the feautures outlook, temperature,
+humidity and wind.  The target or output is whether we ride
+(True=1) or whether we do something else that day (False=0). The
+attributes for each feature are then sunny, overcast and rain for the
+outlook, hot, cold and mild for temperature, high and normal for
+humidity and weak and strong for wind.
+
+
    
+The table here summarizes the various attributes and
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
    Day Outlook  Temperature Humidity  Wind  Ride 
       1          Sunny           Hot                High            Weak          0        
       2          Sunny           Hot                High            Strong        1        
       3          Overcast        Hot                High            Weak          1        
       4          Rain            Mild               High            Weak          1        
       5          Rain            Cool               Normal          Weak          1        
       6          Rain            Cool               Normal          Strong        0        
       7          Overcast        Cool               Normal          Strong        1        
       8          Sunny           Mild               High            Weak          0        
       9          Sunny           Cool               Normal          Weak          1        
       10         Rain            Mild               Normal          Weak          1        
       11         Sunny           Mild               Normal          Strong        1        
       12         Overcast        Mild               High            Strong        1        
       13         Overcast        Hot                Normal          Weak          1        
       14         Rain            Mild               High            Strong        0        
    
+
    
+









    
+
+
    Simple Python Code to read in Data and perform Classification 
    
+
+
    
+
+
+
    # Common imports
+import numpy as np
+import pandas as pd
+import matplotlib.pyplot as plt
+from sklearn.tree import DecisionTreeClassifier
+from sklearn.model_selection import train_test_split
+from sklearn.tree import export_graphviz
+from sklearn.preprocessing import StandardScaler, OneHotEncoder
+from sklearn.compose import ColumnTransformer
+from IPython.display import Image 
+from pydot import graph_from_dot_data
+import os
+
+# Where to save the figures and data files
+PROJECT_ROOT_DIR = "Results"
+FIGURE_ID = "Results/FigureFiles"
+DATA_ID = "DataFiles/"
+
+if not os.path.exists(PROJECT_ROOT_DIR):
+    os.mkdir(PROJECT_ROOT_DIR)
+
+if not os.path.exists(FIGURE_ID):
+    os.makedirs(FIGURE_ID)
+
+if not os.path.exists(DATA_ID):
+    os.makedirs(DATA_ID)
+
+def image_path(fig_id):
+    return os.path.join(FIGURE_ID, fig_id)
+
+def data_path(dat_id):
+    return os.path.join(DATA_ID, dat_id)
+
+def save_fig(fig_id):
+    plt.savefig(image_path(fig_id) + ".png", format='png')
+
+infile = open(data_path("rideclass.csv"),'r')
+
+# Read the experimental data with Pandas
+from IPython.display import display
+ridedata = pd.read_csv(infile,names = ('Outlook','Temperature','Humidity','Wind','Ride'))
+ridedata = pd.DataFrame(ridedata)
+
+# Features and targets
+X = ridedata.loc[:, ridedata.columns != 'Ride'].values
+y = ridedata.loc[:, ridedata.columns == 'Ride'].values
+
+# Create the encoder.
+encoder = OneHotEncoder(handle_unknown="ignore")
+# Assume for simplicity all features are categorical.
+encoder.fit(X)    
+# Apply the encoder.
+X = encoder.transform(X)
+print(X)
+# Then do a Classification tree
+tree_clf = DecisionTreeClassifier(max_depth=2)
+tree_clf.fit(X, y)
+print("Train set accuracy with Decision Tree: {:.2f}".format(tree_clf.score(X,y)))
+#transfer to a decision tree graph
+export_graphviz(
+    tree_clf,
+    out_file="DataFiles/ride.dot",
+    rounded=True,
+    filled=True
+)
+cmd = 'dot -Tpng DataFiles/cancer.dot -o DataFiles/cancer.png'
+os.system(cmd)
+
    
+
    
+









    
+
+
    Computing the Gini Factor  
    
+
+
    
+The above functions (gini, entropy and misclassification error) are
+important components of the so-called CART algorithm. We will discuss
+this algorithm below after we have discussed the information gain
+algorithm ID3.
+
+
    
+In the example here we have converted all our attributes into numerical values \( 0,1,2 \) etc.
+
+
    
+
+
+
    # Split a dataset based on an attribute and an attribute value
+def test_split(index, value, dataset):
+	left, right = list(), list()
+	for row in dataset:
+		if row[index] < value:
+			left.append(row)
+		else:
+			right.append(row)
+	return left, right
+ 
+# Calculate the Gini index for a split dataset
+def gini_index(groups, classes):
+	# count all samples at split point
+	n_instances = float(sum([len(group) for group in groups]))
+	# sum weighted Gini index for each group
+	gini = 0.0
+	for group in groups:
+		size = float(len(group))
+		# avoid divide by zero
+		if size == 0:
+			continue
+		score = 0.0
+		# score the group based on the score for each class
+		for class_val in classes:
+			p = [row[-1] for row in group].count(class_val) / size
+			score += p * p
+		# weight the group score by its relative size
+		gini += (1.0 - score) * (size / n_instances)
+	return gini
+
+# Select the best split point for a dataset
+def get_split(dataset):
+	class_values = list(set(row[-1] for row in dataset))
+	b_index, b_value, b_score, b_groups = 999, 999, 999, None
+	for index in range(len(dataset[0])-1):
+		for row in dataset:
+			groups = test_split(index, row[index], dataset)
+			gini = gini_index(groups, class_values)
+			print('X%d < %.3f Gini=%.3f' % ((index+1), row[index], gini))
+			if gini < b_score:
+				b_index, b_value, b_score, b_groups = index, row[index], gini, groups
+	return {'index':b_index, 'value':b_value, 'groups':b_groups}
+ 
+dataset = [[0,0,0,0,0],
+            [0,0,0,1,1],
+            [1,0,0,0,1],
+            [2,1,0,0,1],
+            [2,2,1,0,1],
+            [2,2,1,1,0],
+            [1,2,1,1,1],
+            [0,1,0,0,0],
+            [0,2,1,0,1],
+            [2,1,1,0,1],
+            [0,1,1,1,1],
+            [1,1,0,1,1],
+            [1,0,1,0,1],
+            [2,1,0,1,0]]
+
+split = get_split(dataset)
+print('Split: [X%d < %.3f]' % ((split['index']+1), split['value']))
+
    
+
    
+









    
+
+
    Entropy and the ID3 algorithm 
    
+
+
    
+ID3, learns decision trees by constructing
+them topdown, beginning with the question which attribute should be tested at the root of the tree?
+
+
      
+
    1.  Each instance attribute is evaluated using a statistical test to determine how well it alone classifies the training examples.
      2. 
+
    2.  The best attribute is selected and used as the test at the root node of the tree.
      3. 
+
    3.  A descendant of the root node is then created for each possible value of this attribute.
      4. 
+
    4.  Training examples are sorted to the appropriate descendant node.
      5. 
+
    5.  The entire process is then repeated using the training examples associated with each descendant node to select the best attribute to test at that point in the tree.
      6. 
+
    6.  This forms a greedy search for an acceptable decision tree, in which the algorithm never backtracks to reconsider earlier choices.
      7.  
+
    
+
+The ID3 algorithm selects, which attribute to test at each node in the
+tree.
+
+
    
+We would like to select the attribute that is most useful for classifying
+examples.
+
+
    
+What is a good quantitative measure of the worth of an attribute?
+
+
    
+Information gain measures how well a given attribute separates the
+training examples according to their target classification.
+
+
    
+The ID3 algorithm uses this information gain measure to select among the candidate
+attributes at each step while growing the tree.
+
+
    
+









    
+
+
    Implementing the ID3 Algorithm 
    
+
+
    
+
+
+
    import re
+import math
+from collections import deque
+
+# x is examples in training set
+# y is set of targets
+# label is target attributes
+# Node is a class which has properties values, childs, and next
+# root is top node in the decision tree
+
+class Node(object):
+	def __init__(self):
+		self.value = None
+		self.next = None
+		self.childs = None
+
+# Simple class of Decision Tree
+# Aimed for who want to learn Decision Tree, so it is not optimized
+class DecisionTree(object):
+	def __init__(self, sample, attributes, labels):
+		self.sample = sample
+		self.attributes = attributes
+		self.labels = labels
+		self.labelCodes = None
+		self.labelCodesCount = None
+		self.initLabelCodes()
+		# print(self.labelCodes)
+		self.root = None
+		self.entropy = self.getEntropy([x for x in range(len(self.labels))])
+
+	def initLabelCodes(self):
+		self.labelCodes = []
+		self.labelCodesCount = []
+		for l in self.labels:
+			if l not in self.labelCodes:
+				self.labelCodes.append(l)
+				self.labelCodesCount.append(0)
+			self.labelCodesCount[self.labelCodes.index(l)] += 1
+
+	def getLabelCodeId(self, sampleId):
+		return self.labelCodes.index(self.labels[sampleId])
+
+	def getAttributeValues(self, sampleIds, attributeId):
+		vals = []
+		for sid in sampleIds:
+			val = self.sample[sid][attributeId]
+			if val not in vals:
+				vals.append(val)
+		# print(vals)
+		return vals
+
+	def getEntropy(self, sampleIds):
+		entropy = 0
+		labelCount = [0] * len(self.labelCodes)
+		for sid in sampleIds:
+			labelCount[self.getLabelCodeId(sid)] += 1
+		# print("-ge", labelCount)
+		for lv in labelCount:
+			# print(lv)
+			if lv != 0:
+				entropy += -lv/len(sampleIds) * math.log(lv/len(sampleIds), 2)
+			else:
+				entropy += 0
+		return entropy
+
+	def getDominantLabel(self, sampleIds):
+		labelCodesCount = [0] * len(self.labelCodes)
+		for sid in sampleIds:
+			labelCodesCount[self.labelCodes.index(self.labels[sid])] += 1
+		return self.labelCodes[labelCodesCount.index(max(labelCodesCount))]
+
+	def getInformationGain(self, sampleIds, attributeId):
+		gain = self.getEntropy(sampleIds)
+		attributeVals = []
+		attributeValsCount = []
+		attributeValsIds = []
+		for sid in sampleIds:
+			val = self.sample[sid][attributeId]
+			if val not in attributeVals:
+				attributeVals.append(val)
+				attributeValsCount.append(0)
+				attributeValsIds.append([])
+			vid = attributeVals.index(val)
+			attributeValsCount[vid] += 1
+			attributeValsIds[vid].append(sid)
+		# print("-gig", self.attributes[attributeId])
+		for vc, vids in zip(attributeValsCount, attributeValsIds):
+			# print("-gig", vids)
+			gain -= vc/len(sampleIds) * self.getEntropy(vids)
+		return gain
+
+	def getAttributeMaxInformationGain(self, sampleIds, attributeIds):
+		attributesEntropy = [0] * len(attributeIds)
+		for i, attId in zip(range(len(attributeIds)), attributeIds):
+			attributesEntropy[i] = self.getInformationGain(sampleIds, attId)
+		maxId = attributeIds[attributesEntropy.index(max(attributesEntropy))]
+		return self.attributes[maxId], maxId
+
+	def isSingleLabeled(self, sampleIds):
+		label = self.labels[sampleIds[0]]
+		for sid in sampleIds:
+			if self.labels[sid] != label:
+				return False
+		return True
+
+	def getLabel(self, sampleId):
+		return self.labels[sampleId]
+
+	def id3(self):
+		sampleIds = [x for x in range(len(self.sample))]
+		attributeIds = [x for x in range(len(self.attributes))]
+		self.root = self.id3Recv(sampleIds, attributeIds, self.root)
+
+	def id3Recv(self, sampleIds, attributeIds, root):
+		root = Node() # Initialize current root
+		if self.isSingleLabeled(sampleIds):
+			root.value = self.labels[sampleIds[0]]
+			return root
+		# print(attributeIds)
+		if len(attributeIds) == 0:
+			root.value = self.getDominantLabel(sampleIds)
+			return root
+		bestAttrName, bestAttrId = self.getAttributeMaxInformationGain(
+			sampleIds, attributeIds)
+		# print(bestAttrName)
+		root.value = bestAttrName
+		root.childs = []  # Create list of children
+		for value in self.getAttributeValues(sampleIds, bestAttrId):
+			# print(value)
+			child = Node()
+			child.value = value
+			root.childs.append(child)  # Append new child node to current
+									   # root
+			childSampleIds = []
+			for sid in sampleIds:
+				if self.sample[sid][bestAttrId] == value:
+					childSampleIds.append(sid)
+			if len(childSampleIds) == 0:
+				child.next = self.getDominantLabel(sampleIds)
+			else:
+				# print(bestAttrName, bestAttrId)
+				# print(attributeIds)
+				if len(attributeIds) > 0 and bestAttrId in attributeIds:
+					toRemove = attributeIds.index(bestAttrId)
+					attributeIds.pop(toRemove)
+				child.next = self.id3Recv(
+					childSampleIds, attributeIds, child.next)
+		return root
+
+	def printTree(self):
+		if self.root:
+			roots = deque()
+			roots.append(self.root)
+			while len(roots) > 0:
+				root = roots.popleft()
+				print(root.value)
+				if root.childs:
+					for child in root.childs:
+						print('({})'.format(child.value))
+						roots.append(child.next)
+				elif root.next:
+					print(root.next)
+
+
+def test():
+	f = open('DataFiles/rideclass.csv')
+	attributes = f.readline().split(',')
+	attributes = attributes[1:len(attributes)-1]
+	print(attributes)
+	sample = f.readlines()
+	f.close()
+	for i in range(len(sample)):
+		sample[i] = re.sub('\d+,', '', sample[i])
+		sample[i] = sample[i].strip().split(',')
+	labels = []
+	for s in sample:
+		labels.append(s.pop())
+	# print(sample)
+	# print(labels)
+	decisionTree = DecisionTree(sample, attributes, labels)
+	print("System entropy {}".format(decisionTree.entropy))
+	decisionTree.id3()
+	decisionTree.printTree()
+
+
+if __name__ == '__main__':
+	test()
+
    
+
    
+









    
+
+
    Cancer Data again now with Decision Trees and other Methods 
    
+
    
+
+
+
    import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import  train_test_split 
+from sklearn.datasets import load_breast_cancer
+from sklearn.svm import SVC
+from sklearn.linear_model import LogisticRegression
+from sklearn.tree import DecisionTreeClassifier
+
+# Load the data
+cancer = load_breast_cancer()
+
+X_train, X_test, y_train, y_test = train_test_split(cancer.data,cancer.target,random_state=0)
+print(X_train.shape)
+print(X_test.shape)
+# Logistic Regression
+logreg = LogisticRegression(solver='lbfgs')
+logreg.fit(X_train, y_train)
+print("Test set accuracy with Logistic Regression: {:.2f}".format(logreg.score(X_test,y_test)))
+# Support vector machine
+svm = SVC(gamma='auto', C=100)
+svm.fit(X_train, y_train)
+print("Test set accuracy with SVM: {:.2f}".format(svm.score(X_test,y_test)))
+# Decision Trees
+deep_tree_clf = DecisionTreeClassifier(max_depth=None)
+deep_tree_clf.fit(X_train, y_train)
+print("Test set accuracy with Decision Trees: {:.2f}".format(deep_tree_clf.score(X_test,y_test)))
+#now scale the data
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+# Logistic Regression
+logreg.fit(X_train_scaled, y_train)
+print("Test set accuracy Logistic Regression with scaled data: {:.2f}".format(logreg.score(X_test_scaled,y_test)))
+# Support Vector Machine
+svm.fit(X_train_scaled, y_train)
+print("Test set accuracy SVM with scaled data: {:.2f}".format(logreg.score(X_test_scaled,y_test)))
+# Decision Trees
+deep_tree_clf.fit(X_train_scaled, y_train)
+print("Test set accuracy with Decision Trees and scaled data: {:.2f}".format(deep_tree_clf.score(X_test_scaled,y_test)))
+
    
+
    
+









    
+
+
    Another example, the moons again 
    
+
    
+
+
+
    from __future__ import division, print_function, unicode_literals
+
+# Common imports
+import numpy as np
+import os
+
+# to make this notebook's output stable across runs
+np.random.seed(42)
+
+# To plot pretty figures
+import matplotlib
+import matplotlib.pyplot as plt
+from matplotlib.colors import ListedColormap
+plt.rcParams['axes.labelsize'] = 14
+plt.rcParams['xtick.labelsize'] = 12
+plt.rcParams['ytick.labelsize'] = 12
+
+
+from sklearn.svm import SVC
+from sklearn import datasets
+from sklearn.tree import DecisionTreeClassifier
+from sklearn.datasets import make_moons
+from sklearn.tree import export_graphviz
+
+Xm, ym = make_moons(n_samples=100, noise=0.25, random_state=53)
+
+deep_tree_clf1 = DecisionTreeClassifier(random_state=42)
+deep_tree_clf2 = DecisionTreeClassifier(min_samples_leaf=4, random_state=42)
+deep_tree_clf1.fit(Xm, ym)
+deep_tree_clf2.fit(Xm, ym)
+
+
+def plot_decision_boundary(clf, X, y, axes=[0, 7.5, 0, 3], iris=True, legend=False, plot_training=True):
+    x1s = np.linspace(axes[0], axes[1], 100)
+    x2s = np.linspace(axes[2], axes[3], 100)
+    x1, x2 = np.meshgrid(x1s, x2s)
+    X_new = np.c_[x1.ravel(), x2.ravel()]
+    y_pred = clf.predict(X_new).reshape(x1.shape)
+    custom_cmap = ListedColormap(['#fafab0','#9898ff','#a0faa0'])
+    plt.contourf(x1, x2, y_pred, alpha=0.3, cmap=custom_cmap)
+    if not iris:
+        custom_cmap2 = ListedColormap(['#7d7d58','#4c4c7f','#507d50'])
+        plt.contour(x1, x2, y_pred, cmap=custom_cmap2, alpha=0.8)
+    if plot_training:
+        plt.plot(X[:, 0][y==0], X[:, 1][y==0], "yo", label="Iris-Setosa")
+        plt.plot(X[:, 0][y==1], X[:, 1][y==1], "bs", label="Iris-Versicolor")
+        plt.plot(X[:, 0][y==2], X[:, 1][y==2], "g^", label="Iris-Virginica")
+        plt.axis(axes)
+    if iris:
+        plt.xlabel("Petal length", fontsize=14)
+        plt.ylabel("Petal width", fontsize=14)
+    else:
+        plt.xlabel(r"$x_1$", fontsize=18)
+        plt.ylabel(r"$x_2$", fontsize=18, rotation=0)
+    if legend:
+        plt.legend(loc="lower right", fontsize=14)
+plt.figure(figsize=(11, 4))
+plt.subplot(121)
+plot_decision_boundary(deep_tree_clf1, Xm, ym, axes=[-1.5, 2.5, -1, 1.5], iris=False)
+plt.title("No restrictions", fontsize=16)
+plt.subplot(122)
+plot_decision_boundary(deep_tree_clf2, Xm, ym, axes=[-1.5, 2.5, -1, 1.5], iris=False)
+plt.title("min_samples_leaf = {}".format(deep_tree_clf2.min_samples_leaf), fontsize=14)
+plt.show()
+
    
+
    
+









    
+
+
    Playing around with regions 
    
+
    
+
+
+
    np.random.seed(6)
+Xs = np.random.rand(100, 2) - 0.5
+ys = (Xs[:, 0] > 0).astype(np.float32) * 2
+
+angle = np.pi/4
+rotation_matrix = np.array([[np.cos(angle), -np.sin(angle)], [np.sin(angle), np.cos(angle)]])
+Xsr = Xs.dot(rotation_matrix)
+
+tree_clf_s = DecisionTreeClassifier(random_state=42)
+tree_clf_s.fit(Xs, ys)
+tree_clf_sr = DecisionTreeClassifier(random_state=42)
+tree_clf_sr.fit(Xsr, ys)
+
+plt.figure(figsize=(11, 4))
+plt.subplot(121)
+plot_decision_boundary(tree_clf_s, Xs, ys, axes=[-0.7, 0.7, -0.7, 0.7], iris=False)
+plt.subplot(122)
+plot_decision_boundary(tree_clf_sr, Xsr, ys, axes=[-0.7, 0.7, -0.7, 0.7], iris=False)
+
+plt.show()
+
    
+
    
+









    
+
+
    Regression trees 
    
+
    
+
+
+
    # Quadratic training set + noise
+np.random.seed(42)
+m = 200
+X = np.random.rand(m, 1)
+y = 4 * (X - 0.5) ** 2
+y = y + np.random.randn(m, 1) / 10
+
    
+
    
+
+
+
    from sklearn.tree import DecisionTreeRegressor
+
+tree_reg = DecisionTreeRegressor(max_depth=2, random_state=42)
+tree_reg.fit(X, y)
+
    
+
    
+









    
+
+
    Final regressor code 
    
+
    
+
+
+
    from sklearn.tree import DecisionTreeRegressor
+
+tree_reg1 = DecisionTreeRegressor(random_state=42, max_depth=2)
+tree_reg2 = DecisionTreeRegressor(random_state=42, max_depth=3)
+tree_reg1.fit(X, y)
+tree_reg2.fit(X, y)
+
+def plot_regression_predictions(tree_reg, X, y, axes=[0, 1, -0.2, 1], ylabel="$y$"):
+    x1 = np.linspace(axes[0], axes[1], 500).reshape(-1, 1)
+    y_pred = tree_reg.predict(x1)
+    plt.axis(axes)
+    plt.xlabel("$x_1$", fontsize=18)
+    if ylabel:
+        plt.ylabel(ylabel, fontsize=18, rotation=0)
+    plt.plot(X, y, "b.")
+    plt.plot(x1, y_pred, "r.-", linewidth=2, label=r"$\hat{y}$")
+
+plt.figure(figsize=(11, 4))
+plt.subplot(121)
+plot_regression_predictions(tree_reg1, X, y)
+for split, style in ((0.1973, "k-"), (0.0917, "k--"), (0.7718, "k--")):
+    plt.plot([split, split], [-0.2, 1], style, linewidth=2)
+plt.text(0.21, 0.65, "Depth=0", fontsize=15)
+plt.text(0.01, 0.2, "Depth=1", fontsize=13)
+plt.text(0.65, 0.8, "Depth=1", fontsize=13)
+plt.legend(loc="upper center", fontsize=18)
+plt.title("max_depth=2", fontsize=14)
+
+plt.subplot(122)
+plot_regression_predictions(tree_reg2, X, y, ylabel=None)
+for split, style in ((0.1973, "k-"), (0.0917, "k--"), (0.7718, "k--")):
+    plt.plot([split, split], [-0.2, 1], style, linewidth=2)
+for split in (0.0458, 0.1298, 0.2873, 0.9040):
+    plt.plot([split, split], [-0.2, 1], "k:", linewidth=1)
+plt.text(0.3, 0.5, "Depth=2", fontsize=13)
+plt.title("max_depth=3", fontsize=14)
+
+plt.show()
+
    
+
    
+
+
+
    tree_reg1 = DecisionTreeRegressor(random_state=42)
+tree_reg2 = DecisionTreeRegressor(random_state=42, min_samples_leaf=10)
+tree_reg1.fit(X, y)
+tree_reg2.fit(X, y)
+
+x1 = np.linspace(0, 1, 500).reshape(-1, 1)
+y_pred1 = tree_reg1.predict(x1)
+y_pred2 = tree_reg2.predict(x1)
+
+plt.figure(figsize=(11, 4))
+
+plt.subplot(121)
+plt.plot(X, y, "b.")
+plt.plot(x1, y_pred1, "r.-", linewidth=2, label=r"$\hat{y}$")
+plt.axis([0, 1, -0.2, 1.1])
+plt.xlabel("$x_1$", fontsize=18)
+plt.ylabel("$y$", fontsize=18, rotation=0)
+plt.legend(loc="upper center", fontsize=18)
+plt.title("No restrictions", fontsize=14)
+
+plt.subplot(122)
+plt.plot(X, y, "b.")
+plt.plot(x1, y_pred2, "r.-", linewidth=2, label=r"$\hat{y}$")
+plt.axis([0, 1, -0.2, 1.1])
+plt.xlabel("$x_1$", fontsize=18)
+plt.title("min_samples_leaf={}".format(tree_reg2.min_samples_leaf), fontsize=14)
+
+plt.show()
+
    
+
    
+









    
+
+
    Pros and cons of trees, pros 
    
+
+
      
+
    •  White box, easy to interpret model. Some people believe that decision trees more closely mirror human decision-making than do the regression and classification approaches discussed earlier (think of support vector machines)
      • 
+
    •  Trees are very easy to explain to people. In fact, they are even easier to explain than linear regression!
      • 
+
    •  No feature normalization needed
      • 
+
    •  Tree models can handle both continuous and categorical data (Classification and Regression Trees)
      • 
+
    •  Can model nonlinear relationships
      • 
+
    •  Can model interactions between the different descriptive features
      • 
+
    •  Trees can be displayed graphically, and are easily interpreted even by a non-expert (especially if they are small)
      • 
+
    
+
+









    
+
+
    Disadvantages 
    
+
+
      
+
    •  Unfortunately, trees generally do not have the same level of predictive accuracy as some of the other regression and classification approaches
      • 
+
    •  If continuous features are used the tree may become quite large and hence less interpretable
      • 
+
    •  Decision trees are prone to overfit the training data and hence do not well generalize the data if no stopping criteria or improvements like pruning, boosting or bagging are implemented
      • 
+
    •  Small changes in the data may lead to a completely different tree. This issue can be addressed by using ensemble methods like bagging, boosting or random forests
      • 
+
    •  Unbalanced datasets where some target feature values occur much more frequently than others may lead to biased trees since the frequently occurring feature values are preferred over the less frequently occurring ones.
      •  
+
    •  If the number of features is relatively large (high dimensional) and the number of instances is relatively low, the tree might overfit the data
      • 
+
    •  Features with many levels may be preferred over features with less levels since for them it is more easy to split the dataset such that the sub datasets only contain pure target feature values. This issue can be addressed by preferring for instance the information gain ratio as splitting criteria over information gain
      • 
+
    
+
+However, by aggregating many decision trees, using methods like bagging, random forests, and boosting, the predictive performance of trees can be substantially improved.
+
+
    
+









    
+
+
    Ensemble Methods: From a Single Tree to Many Trees and Extreme Boosting, Meet the Jungle of Methods 
    
+
+
    
+As stated above and seen in many of the examples discussed here about
+a single decision tree, we often end up overfitting our training
+data. This normally means that we have a high variance. Can we reduce
+the variance of a statistical learning method?
+
+
    
+This leads us to a set of different methods that can combine different
+machine learning algorithms or just use one of them to construct forests and jungles of trees, homogeneous ones or heterogenous ones. These methods are recognized by different names which we will try to explain here. These are
+
+
      
+
    1.  Voting classifiers
      2. 
+
    2.  Bagging and Pasting
      3. 
+
    3.  Random forests
      4. 
+
    4.  Boosting methods, from adaptive to Extreme Gradient Boosting (XGBoost)
      5. 
+
    
+
+We discuss these methods here.
+
+
    
+









    
+
+
    An Overview of Ensemble Methods  
    
+
+
    
+



    
+
+
    
+









    
+
+
    Bagging 
    
+
+
    
+The plain decision trees suffer from high
+variance. This means that if we split the training data into two parts
+at random, and fit a decision tree to both halves, the results that we
+get could be quite different. In contrast, a procedure with low
+variance will yield similar results if applied repeatedly to distinct
+data sets; linear regression tends to have low variance, if the ratio
+of \( n \) to \( p \) is moderately large.
+
+
    
+Bootstrap aggregation, or just bagging, is a
+general-purpose procedure for reducing the variance of a statistical
+learning method.
+
+
    
+









    
+
+
    More bagging 
    
+
+
    
+Bagging typically results in improved accuracy
+over prediction using a single tree. Unfortunately, however, it can be
+difficult to interpret the resulting model. Recall that one of the
+advantages of decision trees is the attractive and easily interpreted
+diagram that results.
+
+
    
+However, when we bag a large number of trees, it is no longer
+possible to represent the resulting statistical learning procedure
+using a single tree, and it is no longer clear which variables are
+most important to the procedure. Thus, bagging improves prediction
+accuracy at the expense of interpretability.  Although the collection
+of bagged trees is much more difficult to interpret than a single
+tree, one can obtain an overall summary of the importance of each
+predictor using the MSE (for bagging regression trees) or the Gini
+index (for bagging classification trees). In the case of bagging
+regression trees, we can record the total amount that the MSE is
+decreased due to splits over a given predictor, averaged over all \( B \) possible
+trees. A large value indicates an important predictor. Similarly, in
+the context of bagging classification trees, we can add up the total
+amount that the Gini index  is decreased by splits over a given
+predictor, averaged over all \( B \) trees.
+
+
    
+









    
+
+
    Simple Voting Example, head or tail 
    
+
    
+
+
+
    heads_proba = 0.51
+coin_tosses = (np.random.rand(10000, 10) < heads_proba).astype(np.int32)
+cumulative_heads_ratio = np.cumsum(coin_tosses, axis=0) / np.arange(1, 10001).reshape(-1, 1)
+plt.figure(figsize=(8,3.5))
+plt.plot(cumulative_heads_ratio)
+plt.plot([0, 10000], [0.51, 0.51], "k--", linewidth=2, label="51%")
+plt.plot([0, 10000], [0.5, 0.5], "k-", label="50%")
+plt.xlabel("Number of coin tosses")
+plt.ylabel("Heads ratio")
+plt.legend(loc="lower right")
+plt.axis([0, 10000, 0.42, 0.58])
+plt.show()
+
    
+
    
+









    
+
+
    Using the Voting Classifier 
    
+
    
+
+
+
    from sklearn.model_selection import train_test_split
+from sklearn.datasets import make_moons
+
+X, y = make_moons(n_samples=500, noise=0.30, random_state=42)
+X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=42)
+
+from sklearn.ensemble import RandomForestClassifier
+from sklearn.ensemble import VotingClassifier
+from sklearn.linear_model import LogisticRegression
+from sklearn.svm import SVC
+
+log_clf = LogisticRegression(solver="liblinear", random_state=42)
+rnd_clf = RandomForestClassifier(n_estimators=10, random_state=42)
+svm_clf = SVC(gamma="auto", random_state=42)
+
+voting_clf = VotingClassifier(
+    estimators=[('lr', log_clf), ('rf', rnd_clf), ('svc', svm_clf)],
+    voting='hard')
+
+voting_clf.fit(X_train, y_train)
+
+from sklearn.metrics import accuracy_score
+
+for clf in (log_clf, rnd_clf, svm_clf, voting_clf):
+    clf.fit(X_train, y_train)
+    y_pred = clf.predict(X_test)
+    print(clf.__class__.__name__, accuracy_score(y_test, y_pred))
+
+log_clf = LogisticRegression(solver="liblinear", random_state=42)
+rnd_clf = RandomForestClassifier(n_estimators=10, random_state=42)
+svm_clf = SVC(gamma="auto", probability=True, random_state=42)
+
+voting_clf = VotingClassifier(
+    estimators=[('lr', log_clf), ('rf', rnd_clf), ('svc', svm_clf)],
+    voting='soft')
+voting_clf.fit(X_train, y_train)
+
+from sklearn.metrics import accuracy_score
+
+for clf in (log_clf, rnd_clf, svm_clf, voting_clf):
+    clf.fit(X_train, y_train)
+    y_pred = clf.predict(X_test)
+    print(clf.__class__.__name__, accuracy_score(y_test, y_pred))
+
    
+
    
+









    
+
+
    Please, not the moons again! Voting and Bagging 
    
+
    
+
+
+
    from sklearn.model_selection import train_test_split
+from sklearn.datasets import make_moons
+
+X, y = make_moons(n_samples=500, noise=0.30, random_state=42)
+X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=42)
+from sklearn.ensemble import RandomForestClassifier
+from sklearn.ensemble import VotingClassifier
+from sklearn.linear_model import LogisticRegression
+from sklearn.svm import SVC
+
+log_clf = LogisticRegression(random_state=42)
+rnd_clf = RandomForestClassifier(random_state=42)
+svm_clf = SVC(random_state=42)
+
+voting_clf = VotingClassifier(
+    estimators=[('lr', log_clf), ('rf', rnd_clf), ('svc', svm_clf)],
+    voting='hard')
+voting_clf.fit(X_train, y_train)
+
    
+
    
+
+
+
    from sklearn.metrics import accuracy_score
+
+for clf in (log_clf, rnd_clf, svm_clf, voting_clf):
+    clf.fit(X_train, y_train)
+    y_pred = clf.predict(X_test)
+    print(clf.__class__.__name__, accuracy_score(y_test, y_pred))
+
    
+
    
+
+
+
    log_clf = LogisticRegression(random_state=42)
+rnd_clf = RandomForestClassifier(random_state=42)
+svm_clf = SVC(probability=True, random_state=42)
+
+voting_clf = VotingClassifier(
+    estimators=[('lr', log_clf), ('rf', rnd_clf), ('svc', svm_clf)],
+    voting='soft')
+voting_clf.fit(X_train, y_train)
+
    
+
    
+
+
+
    from sklearn.metrics import accuracy_score
+
+for clf in (log_clf, rnd_clf, svm_clf, voting_clf):
+    clf.fit(X_train, y_train)
+    y_pred = clf.predict(X_test)
+    print(clf.__class__.__name__, accuracy_score(y_test, y_pred))
+
    
+
    
+









    
+
+
    Now Bagging  
    
+
+
    
+
+
+
    from sklearn.ensemble import BaggingClassifier
+from sklearn.tree import DecisionTreeClassifier
+
+bag_clf = BaggingClassifier(
+    DecisionTreeClassifier(random_state=42), n_estimators=500,
+    max_samples=100, bootstrap=True, n_jobs=-1, random_state=42)
+bag_clf.fit(X_train, y_train)
+y_pred = bag_clf.predict(X_test)
+
    
+
    
+
+
+
    from sklearn.metrics import accuracy_score
+print(accuracy_score(y_test, y_pred))
+
    
+
    
+
+
+
    tree_clf = DecisionTreeClassifier(random_state=42)
+tree_clf.fit(X_train, y_train)
+y_pred_tree = tree_clf.predict(X_test)
+print(accuracy_score(y_test, y_pred_tree))
+
    
+
    
+
+
+
    from matplotlib.colors import ListedColormap
+
+def plot_decision_boundary(clf, X, y, axes=[-1.5, 2.5, -1, 1.5], alpha=0.5, contour=True):
+    x1s = np.linspace(axes[0], axes[1], 100)
+    x2s = np.linspace(axes[2], axes[3], 100)
+    x1, x2 = np.meshgrid(x1s, x2s)
+    X_new = np.c_[x1.ravel(), x2.ravel()]
+    y_pred = clf.predict(X_new).reshape(x1.shape)
+    custom_cmap = ListedColormap(['#fafab0','#9898ff','#a0faa0'])
+    plt.contourf(x1, x2, y_pred, alpha=0.3, cmap=custom_cmap)
+    if contour:
+        custom_cmap2 = ListedColormap(['#7d7d58','#4c4c7f','#507d50'])
+        plt.contour(x1, x2, y_pred, cmap=custom_cmap2, alpha=0.8)
+    plt.plot(X[:, 0][y==0], X[:, 1][y==0], "yo", alpha=alpha)
+    plt.plot(X[:, 0][y==1], X[:, 1][y==1], "bs", alpha=alpha)
+    plt.axis(axes)
+    plt.xlabel(r"$x_1$", fontsize=18)
+    plt.ylabel(r"$x_2$", fontsize=18, rotation=0)
+plt.figure(figsize=(11,4))
+plt.subplot(121)
+plot_decision_boundary(tree_clf, X, y)
+plt.title("Decision Tree", fontsize=14)
+plt.subplot(122)
+plot_decision_boundary(bag_clf, X, y)
+plt.title("Decision Trees with Bagging", fontsize=14)
+plt.show()
+
    
+
    
+









    
+
+
    Making our own Bagging with Bootstrap 
    
+
    
+
+
+
    import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import train_test_split
+from sklearn.pipeline import make_pipeline
+from sklearn.utils import resample
+from sklearn.tree import DecisionTreeRegressor
+
+
+np.random.seed(2018)
+
+n = 40
+n_boostraps = 100
+maxdegree = 14
+
+# Make data set.
+x = np.linspace(-3, 3, n).reshape(-1, 1)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
+error = np.zeros(maxdegree)
+bias = np.zeros(maxdegree)
+variance = np.zeros(maxdegree)
+polydegree = np.zeros(maxdegree)
+X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
+
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+for degree in range(maxdegree):
+    model = DecisionTreeRegressor(max_depth=5) 
+    y_pred = np.empty((y_test.shape[0], n_boostraps))
+    for i in range(n_boostraps):
+        x_, y_ = resample(X_train_scaled, y_train)
+        model.fit(x_, y_)
+        y_pred[:, i] = model.predict(X_test_scaled).ravel()
+
+    polydegree[degree] = degree
+    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )
+    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )
+    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )
+    print('Polynomial degree:', degree)
+    print('Error:', error[degree])
+    print('Bias^2:', bias[degree])
+    print('Var:', variance[degree])
+    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
+
+plt.plot(polydegree, error, label='Error')
+plt.plot(polydegree, bias, label='bias')
+plt.plot(polydegree, variance, label='Variance')
+plt.legend()
+plt.show()
+
    
+
    
+









    
+
+
    Changing the Level of the Decision Tree 
    
+
+
    
+
+
+
    import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import train_test_split
+from sklearn.pipeline import make_pipeline
+from sklearn.utils import resample
+from sklearn.tree import DecisionTreeRegressor
+
+n = 100
+n_boostraps = 100
+maxdepth = 8
+
+# Make data set.
+x = np.linspace(-3, 3, n).reshape(-1, 1)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
+error = np.zeros(maxdepth)
+bias = np.zeros(maxdepth)
+variance = np.zeros(maxdepth)
+polydegree = np.zeros(maxdepth)
+X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
+
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+for degree in range(1,maxdepth):
+    model = DecisionTreeRegressor(max_depth=degree) 
+    y_pred = np.empty((y_test.shape[0], n_boostraps))
+    for i in range(n_boostraps):
+        x_, y_ = resample(X_train_scaled, y_train)
+        model.fit(x_, y_)
+        y_pred[:, i] = model.predict(X_test_scaled)#.ravel()
+
+    polydegree[degree] = degree
+    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )
+    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )
+    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )
+    print('Polynomial degree:', degree)
+    print('Error:', error[degree])
+    print('Bias^2:', bias[degree])
+    print('Var:', variance[degree])
+    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
+
+plt.xlim(1,maxdepth)
+plt.plot(polydegree, error, label='Error')
+plt.plot(polydegree, bias, label='bias')
+plt.plot(polydegree, variance, label='Variance')
+plt.legend()
+plt.show()
+
    
+
    
+









    
+
+
    Random forests 
    
+
+
    
+Random forests provide an improvement over bagged trees by way of a
+small tweak that decorrelates the trees.
+
+
    
+As in bagging, we build a
+number of decision trees on bootstrapped training samples. But when
+building these decision trees, each time a split in a tree is
+considered, a random sample of \( m \) predictors is chosen as split
+candidates from the full set of \( p \) predictors. The split is allowed to
+use only one of those \( m \) predictors.
+
+
    
+A fresh sample of \( m \) predictors is
+taken at each split, and typically we choose 
+
+$$
+m\approx \sqrt{p}.
+$$
+
+
    
+In building a random forest, at
+each split in the tree, the algorithm is not even allowed to consider
+a majority of the available predictors.
+
+
    
+The reason for this is rather clever. Suppose that there is one very
+strong predictor in the data set, along with a number of other
+moderately strong predictors. Then in the collection of bagged
+variable importance random forest trees, most or all of the trees will
+use this strong predictor in the top split. Consequently, all of the
+bagged trees will look quite similar to each other. Hence the
+predictions from the bagged trees will be highly correlated.
+Unfortunately, averaging many highly correlated quantities does not
+lead to as large of a reduction in variance as averaging many
+uncorrelated quanti- ties. In particular, this means that bagging will
+not lead to a substantial reduction in variance over a single tree in
+this setting.
+
+
    
+









    
+
+
    Random Forest Algorithm 
    
+The algorithm described here can be applied to both classification and regression problems.
+
+
    
+We will grow of forest of say \( M \) trees.
+
+
      
+
    1.  For \( m=1:M \) we
      2. 
+
+
        
+  
      •  Draw a bootstrap sample of from the training data organized in our \( \boldsymbol{X} \) matrix.
        • 
+  
      •  We grow then a random forest tree \( T_m \) based on the bootstrapped data by repeating the steps outlined till we reach the maximum node size is reached
        • 
+
+
          
+   
        1.  we select \( m \le p \) varibales at random from the \( p \) predictors/features
          2. 
+   
        2.  pick the best split point among the \( m \) features using either the CART algorithm or the ID3 for classification and create a new node
          3. 
+   
        3.  split the node into daughter nodes
          4. 
+
        
+
+
      
+
+
    2.  Output then the ensemble of trees \( \{T_m\}_1^{M} \) and make predictions for either a regression type of problem or a classification type of problem.
      3.  
+
    
+
+









    
+
+
    Bootstrap with Random Forests Instead of a Single Tree, own Bagging 
    
+
+
    
+
+
+
    import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import train_test_split
+from sklearn.pipeline import make_pipeline
+from sklearn.utils import resample
+from sklearn.ensemble import RandomForestRegressor
+
+np.random.seed(2018)
+
+n = 100
+n_boostraps = 100
+maxdegree = 14
+
+# Make data set.
+x = np.linspace(-3, 3, n).reshape(-1, 1)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
+error = np.zeros(maxdegree)
+bias = np.zeros(maxdegree)
+variance = np.zeros(maxdegree)
+polydegree = np.zeros(maxdegree)
+X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
+
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+for degree in range(maxdegree):
+    model = RandomForestRegressor()
+    y_pred = np.empty((y_test.shape[0], n_boostraps))
+    for i in range(n_boostraps):
+        x_, y_ = resample(X_train_scaled, y_train)
+        model.fit(x_, y_.ravel())
+        y_pred[:, i] = model.predict(X_test_scaled).ravel()
+
+    polydegree[degree] = degree
+    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )
+    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )
+    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )
+    print('Polynomial degree:', degree)
+    print('Error:', error[degree])
+    print('Bias^2:', bias[degree])
+    print('Var:', variance[degree])
+    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
+
+plt.plot(polydegree, error, label='Error')
+plt.plot(polydegree, bias, label='bias')
+plt.plot(polydegree, variance, label='Variance')
+plt.legend()
+plt.show()
+
    
+
    
+









    
+
+
    Random Forests Compared with other Methods on the Cancer Data 
    
+
    
+
+
+
    import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import  train_test_split 
+from sklearn.datasets import load_breast_cancer
+from sklearn.svm import SVC
+from sklearn.linear_model import LogisticRegression
+from sklearn.tree import DecisionTreeClassifier
+
+# Load the data
+cancer = load_breast_cancer()
+
+X_train, X_test, y_train, y_test = train_test_split(cancer.data,cancer.target,random_state=0)
+print(X_train.shape)
+print(X_test.shape)
+# Logistic Regression
+logreg = LogisticRegression(solver='lbfgs')
+logreg.fit(X_train, y_train)
+print("Test set accuracy with Logistic Regression: {:.2f}".format(logreg.score(X_test,y_test)))
+# Support vector machine
+svm = SVC(gamma='auto', C=100)
+svm.fit(X_train, y_train)
+print("Test set accuracy with SVM: {:.2f}".format(svm.score(X_test,y_test)))
+# Decision Trees
+deep_tree_clf = DecisionTreeClassifier(max_depth=None)
+deep_tree_clf.fit(X_train, y_train)
+print("Test set accuracy with Decision Trees: {:.2f}".format(deep_tree_clf.score(X_test,y_test)))
+#now scale the data
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+# Logistic Regression
+logreg.fit(X_train_scaled, y_train)
+print("Test set accuracy Logistic Regression with scaled data: {:.2f}".format(logreg.score(X_test_scaled,y_test)))
+# Support Vector Machine
+svm.fit(X_train_scaled, y_train)
+print("Test set accuracy SVM with scaled data: {:.2f}".format(logreg.score(X_test_scaled,y_test)))
+# Decision Trees
+deep_tree_clf.fit(X_train_scaled, y_train)
+print("Test set accuracy with Decision Trees and scaled data: {:.2f}".format(deep_tree_clf.score(X_test_scaled,y_test)))
+
+
+from sklearn.ensemble import RandomForestClassifier
+from sklearn.preprocessing import LabelEncoder
+from sklearn.model_selection import cross_validate
+# Data set not specificied
+#Instantiate the model with 500 trees and entropy as splitting criteria
+Random_Forest_model = RandomForestClassifier(n_estimators=500,criterion="entropy")
+Random_Forest_model.fit(X_train_scaled, y_train)
+#Cross validation
+accuracy = cross_validate(Random_Forest_model,X_test_scaled,y_test,cv=10)['test_score']
+print(accuracy)
+print("Test set accuracy with Random Forests and scaled data: {:.2f}".format(Random_Forest_model.score(X_test_scaled,y_test)))
+
+
+import scikitplot as skplt
+y_pred = Random_Forest_model.predict(X_test_scaled)
+skplt.metrics.plot_confusion_matrix(y_test, y_pred, normalize=True)
+plt.show()
+y_probas = Random_Forest_model.predict_proba(X_test_scaled)
+skplt.metrics.plot_roc(y_test, y_probas)
+plt.show()
+skplt.metrics.plot_cumulative_gain(y_test, y_probas)
+plt.show()
+
    
+
    
+









    
+
+
    Compare  Bagging on Trees with Random Forests  
    
+
    
+
+
+
    bag_clf = BaggingClassifier(
+    DecisionTreeClassifier(splitter="random", max_leaf_nodes=16, random_state=42),
+    n_estimators=500, max_samples=1.0, bootstrap=True, n_jobs=-1, random_state=42)
+
    
+
    
+
+
+
    bag_clf.fit(X_train, y_train)
+y_pred = bag_clf.predict(X_test)
+from sklearn.ensemble import RandomForestClassifier
+rnd_clf = RandomForestClassifier(n_estimators=500, max_leaf_nodes=16, n_jobs=-1, random_state=42)
+rnd_clf.fit(X_train, y_train)
+y_pred_rf = rnd_clf.predict(X_test)
+np.sum(y_pred == y_pred_rf) / len(y_pred) 
+
    
+
    
+









    
+
+
    Boosting, a Bird'e Eye 
    
+
+
    
+The basic idea is to combine weak classifiers in order to create a good
+classifier. With a weak classifier we often intend a classifier which
+produces results which are only slightly better than we would get by
+random guesses.
+
+
    
+This is done by applying in an iterative way a weak (or a standard
+classifier like decision trees) to modify the data. In each iteration
+we emphasize those observations which are misclassified by weighting
+them with a factor.
+
+
    
+









    
+
+
    Adaptive boosting: AdaBoost, Basic Algorithm 
    
+
+
    
+The algorithm here is rather straightforward. Assume that our weak
+classifier is a decision tree and we consider a binary set of outputs
+with \( y_i \in \{-1,1\} \) and \( i=0,1,2,\dots,n-1 \) as our set of
+observations. Our design matrix is given in terms of the
+feature/predictor vectors
+\( \boldsymbol{X}=[\boldsymbol{x}_0\boldsymbol{x}_1\dots\boldsymbol{x}_{p-1} \). Finally, we define also a
+classifier determined by our data via a function \( G(\boldsymbol{X}) \). This function tells us how well we are able to classify our outputs/targets \( \boldsymbol{y} \).
+
+
    
+We can then define the misclassification error \( \mathrm{err} \) as
+$$
+\mathrm{err}=\frac{1}{n}\sum_{i=0}^{n-1}I(y_i\ne G(\boldsymbol{X}_{i*}),
+$$
+
+where the function \( I() \) is one if we misclassify and zero if we classify correctly.
+
+
    
+









    
+
+
    Basic Steps of AdaBoost 
    
+
+
    
+With the above definitions we are now ready to set up the algorithm for AdaBoost.
+The basic idea is to set up weights which will be used to scale the correctly classified and the misclassified cases.
+
+
      
+
    1.  We start by initializing all weights to \( w_i = 1/n \), with \( i=0,1,2,\dots n-1 \). It is to see then that \( \sum_{i=0}^{n-1}w_i = 1 \).
      2. 
+
    2.  We rewrite the misclassification error as
      3.  
+
    
+
+$$
+\mathrm{err}=\frac{\sum_{i=0}^{n-1}w_iI(y_i\ne G(\boldsymbol{X}_{i*})}{\sum_{i=0}^{n-1}w_i},
+$$
+
+
+
      
+
    1.  Then we start looping over all attempts at classifying, namely we start an iterative process for \( m=1:M \), where \( M \) is the final number of classifications. Our given classifier could for example be a plain decision tree.
+
+
        
+ 
      1.  Fit then a given classifier to the training using the weights \( w_i \).
        2. 
+ 
      2.  Compute then \( \mathrm{err} \) and figure out which events are classified properly and which are classified wrongly.
        3. 
+ 
      3.  Define a quantity $\alpha_{m} = \log{(1-\mathrm{err})/\mathrm{err}}
        4. 
+ 
      4.  Set the new weights to $w_i = w_i\times \exp{(\alpha_m I(y_i\ne G(\boldsymbol{X}_{i*})}.
        5. 
+
      
+
+
    2.  Compute the new classifier $G(\boldsymbol{X})= \sum_{i=0}^{n-1}\alpha_m I(y_i\ne G(\boldsymbol{X}_{i*}).
      3. 
+
    
+
+For the iterations with \( m \le 2 \) the weights are modified
+individually at each steps. The obersvations which were misclassified
+at iteration \( m-1 \) have a weight which is larger than those which were
+classified properly. As this proceeds, the observations which were
+difficult to classifiy correctly are given a larger influence. Each
+new classification step \( m \) is then forced to concentrate on those
+observations that are missed in the previous iterations.
+
+
    
+









    
+
+
    AdaBoost Examples 
    
+
+
    
+Using Scikit-Learn it is easy to appply the adaptive boosting algorithm, as done here.
+
+
    
+
+
+
    from sklearn.ensemble import AdaBoostClassifier
+
+ada_clf = AdaBoostClassifier(
+    DecisionTreeClassifier(max_depth=1), n_estimators=200,
+    algorithm="SAMME.R", learning_rate=0.5, random_state=42)
+ada_clf.fit(X_train, y_train)
+
+plot_decision_boundary(ada_clf, X, y)
+
+m = len(X_train)
+
+plt.figure(figsize=(11, 4))
+for subplot, learning_rate in ((121, 1), (122, 0.5)):
+    sample_weights = np.ones(m)
+    plt.subplot(subplot)
+    for i in range(5):
+        svm_clf = SVC(kernel="rbf", C=0.05, gamma="auto", random_state=42)
+        svm_clf.fit(X_train, y_train, sample_weight=sample_weights)
+        y_pred = svm_clf.predict(X_train)
+        sample_weights[y_pred != y_train] *= (1 + learning_rate)
+        plot_decision_boundary(svm_clf, X, y, alpha=0.2)
+        plt.title("learning_rate = {}".format(learning_rate), fontsize=16)
+    if subplot == 121:
+        plt.text(-0.7, -0.65, "1", fontsize=14)
+        plt.text(-0.6, -0.10, "2", fontsize=14)
+        plt.text(-0.5,  0.10, "3", fontsize=14)
+        plt.text(-0.4,  0.55, "4", fontsize=14)
+        plt.text(-0.3,  0.90, "5", fontsize=14)
+
+save_fig("boosting_plot")
+plt.show()
+
    
+
    
+









    
+
+
    Gradient boosting: Basics 
    
+
+
    
+Gradient boosting is again a similar technique to Adapative boosting,
+it combines so-called weak classifiers or regressors into a strong
+method via a series of iterations.
+
+
    
+In order to understand the method, let us illustrate its basics by
+bringing back the essential steps in linear regression, where our cost
+function was the least squares function.
+
+
    
+









    
+
+
    Gradient Boosting, algorithm 
    
+
+
    
+Suppose we have a cost function \( C(f)=\sum_{i=0}^{n-1}L(y_i, f(x_i)) \) where \( y_i \) is our target and \( f(x_i) \) the function which is meant to model \( y_i \). The above cost function could be our standard least squares  function
+$$
+C(\boldsymbol{y},\boldsymbol{f})=\frac{1}{n}\sum_{i=0}^{n-1}(y_i-f(x_i))^2.
+$$
+
+
    
+The way we proceed in an iterative fashion is to
+
+
      
+
    1.  Initialize our estimate by \( f_0(x)=0 \).
      2. 
+
    2.  For \( m=1:M \), we
+
+
        
+ 
      1.  compute the negative gradient vector \( \boldsymbol{u}_m = -\partial C(\boldsymbol{y},\boldsymbol{f})/\partial \boldsymbol{f}(x) \) at $f(x) = f_{m-1}(x);
        2. 
+ 
      2.  fit the so-called base-learner to the negative gradient \( h_m(u_m,x) \);
        3. 
+ 
      3.  update the estimate \( f_m(x) = f_{m-1}(x)+\nu h_m(u_m,x) \);
        4. 
+
      
+
+
    3.  The final estimate is then \( f_M(x) = \sum_{m=1}^M\nu h_m(u_m,x) \).
      4. 
+
    
+
+









    
+
+
    Gradient Boosting, Examples 
    
+
    
+
+
+
    np.random.seed(42)
+X = np.random.rand(100, 1) - 0.5
+y = 3*X[:, 0]**2 + 0.05 * np.random.randn(100)
+
+from sklearn.tree import DecisionTreeRegressor
+
+tree_reg1 = DecisionTreeRegressor(max_depth=2, random_state=42)
+tree_reg1.fit(X, y)
+
+y2 = y - tree_reg1.predict(X)
+tree_reg2 = DecisionTreeRegressor(max_depth=2, random_state=42)
+tree_reg2.fit(X, y2)
+
+y3 = y2 - tree_reg2.predict(X)
+tree_reg3 = DecisionTreeRegressor(max_depth=2, random_state=42)
+tree_reg3.fit(X, y3)
+
+X_new = np.array([[0.8]])
+y_pred = sum(tree.predict(X_new) for tree in (tree_reg1, tree_reg2, tree_reg3))
+
+def plot_predictions(regressors, X, y, axes, label=None, style="r-", data_style="b.", data_label=None):
+    x1 = np.linspace(axes[0], axes[1], 500)
+    y_pred = sum(regressor.predict(x1.reshape(-1, 1)) for regressor in regressors)
+    plt.plot(X[:, 0], y, data_style, label=data_label)
+    plt.plot(x1, y_pred, style, linewidth=2, label=label)
+    if label or data_label:
+        plt.legend(loc="upper center", fontsize=16)
+    plt.axis(axes)
+
+plt.figure(figsize=(11,11))
+
+plt.subplot(321)
+plot_predictions([tree_reg1], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="$h_1(x_1)$", style="g-", data_label="Training set")
+plt.ylabel("$y$", fontsize=16, rotation=0)
+plt.title("Residuals and tree predictions", fontsize=16)
+
+plt.subplot(322)
+plot_predictions([tree_reg1], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="$h(x_1) = h_1(x_1)$", data_label="Training set")
+plt.ylabel("$y$", fontsize=16, rotation=0)
+plt.title("Ensemble predictions", fontsize=16)
+
+plt.subplot(323)
+plot_predictions([tree_reg2], X, y2, axes=[-0.5, 0.5, -0.5, 0.5], label="$h_2(x_1)$", style="g-", data_style="k+", data_label="Residuals")
+plt.ylabel("$y - h_1(x_1)$", fontsize=16)
+
+plt.subplot(324)
+plot_predictions([tree_reg1, tree_reg2], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="$h(x_1) = h_1(x_1) + h_2(x_1)$")
+plt.ylabel("$y$", fontsize=16, rotation=0)
+
+plt.subplot(325)
+plot_predictions([tree_reg3], X, y3, axes=[-0.5, 0.5, -0.5, 0.5], label="$h_3(x_1)$", style="g-", data_style="k+")
+plt.ylabel("$y - h_1(x_1) - h_2(x_1)$", fontsize=16)
+plt.xlabel("$x_1$", fontsize=16)
+
+plt.subplot(326)
+plot_predictions([tree_reg1, tree_reg2, tree_reg3], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="$h(x_1) = h_1(x_1) + h_2(x_1) + h_3(x_1)$")
+plt.xlabel("$x_1$", fontsize=16)
+plt.ylabel("$y$", fontsize=16, rotation=0)
+
+save_fig("gradient_boosting_plot")
+plt.show()
+
+from sklearn.ensemble import GradientBoostingRegressor
+
+gbrt = GradientBoostingRegressor(max_depth=2, n_estimators=3, learning_rate=1.0, random_state=42)
+gbrt.fit(X, y)
+
+gbrt_slow = GradientBoostingRegressor(max_depth=2, n_estimators=200, learning_rate=0.1, random_state=42)
+gbrt_slow.fit(X, y)
+
+plt.figure(figsize=(11,4))
+
+plt.subplot(121)
+plot_predictions([gbrt], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="Ensemble predictions")
+plt.title("learning_rate={}, n_estimators={}".format(gbrt.learning_rate, gbrt.n_estimators), fontsize=14)
+
+plt.subplot(122)
+plot_predictions([gbrt_slow], X, y, axes=[-0.5, 0.5, -0.1, 0.8])
+plt.title("learning_rate={}, n_estimators={}".format(gbrt_slow.learning_rate, gbrt_slow.n_estimators), fontsize=14)
+
+save_fig("gbrt_learning_rate_plot")
+plt.show()
+
    
+
    
+









    
+
+
    Gradient Boots with Early Stopping 
    
+
    
+
+
+
    from sklearn.model_selection import train_test_split
+from sklearn.metrics import mean_squared_error
+
+X_train, X_val, y_train, y_val = train_test_split(X, y, random_state=49)
+
+gbrt = GradientBoostingRegressor(max_depth=2, n_estimators=120, random_state=42)
+gbrt.fit(X_train, y_train)
+
+errors = [mean_squared_error(y_val, y_pred)
+          for y_pred in gbrt.staged_predict(X_val)]
+bst_n_estimators = np.argmin(errors) + 1
+
+gbrt_best = GradientBoostingRegressor(max_depth=2,n_estimators=bst_n_estimators, random_state=42)
+gbrt_best.fit(X_train, y_train)
+
+min_error = np.min(errors)
+plt.figure(figsize=(11, 4))
+
+plt.subplot(121)
+plt.plot(errors, "b.-")
+plt.plot([bst_n_estimators, bst_n_estimators], [0, min_error], "k--")
+plt.plot([0, 120], [min_error, min_error], "k--")
+plt.plot(bst_n_estimators, min_error, "ko")
+plt.text(bst_n_estimators, min_error*1.2, "Minimum", ha="center", fontsize=14)
+plt.axis([0, 120, 0, 0.01])
+plt.xlabel("Number of trees")
+plt.title("Validation error", fontsize=14)
+
+plt.subplot(122)
+plot_predictions([gbrt_best], X, y, axes=[-0.5, 0.5, -0.1, 0.8])
+plt.title("Best model (%d trees)" % bst_n_estimators, fontsize=14)
+
+save_fig("early_stopping_gbrt_plot")
+plt.show()
+
+
+gbrt = GradientBoostingRegressor(max_depth=2, warm_start=True, random_state=42)
+
+min_val_error = float("inf")
+error_going_up = 0
+for n_estimators in range(1, 120):
+    gbrt.n_estimators = n_estimators
+    gbrt.fit(X_train, y_train)
+    y_pred = gbrt.predict(X_val)
+    val_error = mean_squared_error(y_val, y_pred)
+    if val_error < min_val_error:
+        min_val_error = val_error
+        error_going_up = 0
+    else:
+        error_going_up += 1
+        if error_going_up == 5:
+            break  # early stopping
+
+
+print(gbrt.n_estimators)
+print("Minimum validation MSE:", min_val_error)
+
    
+
    
+









    
+
+
    XGBoost: Extreme Gradient Boosting 
    
+
+
    
+XGBoost or Extreme Gradient
+Boosting, is an optimized distributed gradient boosting library
+designed to be highly efficient, flexible and portable. It implements
+machine learning algorithms under the Gradient Boosting
+framework. XGBoost provides a parallel tree boosting that solve many
+data science problems in a fast and accurate way. See the article by Chen and Guestrin.
+
+
    
+The authors design and build a highly scalable end-to-end tree
+boosting system. It has  a theoretically justified weighted quantile
+sketch for efficient proposal calculation. It introduces a novel sparsity-aware algorithm for parallel tree learning and an effective cache-aware block structure for out-of-core tree learning.
+
+
    
+It is now the algorithm which wins essentially all ML competitions!!!
+
+
    
+









    
+
+
    Regression Case 
    
+
+
    
+
+
+
    import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import train_test_split
+import xgboost as xgb
+from sklearn.preprocessing import StandardScaler
+import scikitplot as skplt
+from sklearn.metrics import mean_squared_error
+
+n = 100
+maxdegree = 6
+
+# Make data set.
+x = np.linspace(-3, 3, n).reshape(-1, 1)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
+
+error = np.zeros(maxdegree)
+bias = np.zeros(maxdegree)
+variance = np.zeros(maxdegree)
+polydegree = np.zeros(maxdegree)
+X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+for degree in range(maxdegree):
+    model =  xgb.XGBRegressor(objective ='reg:squarederror', colsample_bytree = 0.3, learning_rate = 0.1,
+                max_depth = degree, alpha = 10, n_estimators = 10)
+    model.fit(X_train_scaled,y_train)
+    y_pred = model.predict(X_test_scaled)
+    polydegree[degree] = degree
+    error[degree] = np.mean( np.mean((y_test - y_pred)**2) )
+    bias[degree] = np.mean( (y_test - np.mean(y_pred))**2 )
+    variance[degree] = np.mean( np.var(y_pred) )
+    print('Max depth:', degree)
+    print('Error:', error[degree])
+    print('Bias^2:', bias[degree])
+    print('Var:', variance[degree])
+    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
+
+plt.xlim(1,maxdegree-1)
+plt.plot(polydegree, error, label='Error')
+plt.plot(polydegree, bias, label='bias')
+plt.plot(polydegree, variance, label='Variance')
+plt.legend()
+plt.show()
+
    
+
    
+









    
+
+
    Xgboost on the Cancer Data 
    
+
    
+
+
+
    import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import  train_test_split 
+from sklearn.datasets import load_breast_cancer
+from sklearn.preprocessing import LabelEncoder
+from sklearn.model_selection import cross_validate
+import scikitplot as skplt
+import xgboost as xgb
+# Load the data
+cancer = load_breast_cancer()
+
+X_train, X_test, y_train, y_test = train_test_split(cancer.data,cancer.target,random_state=0)
+print(X_train.shape)
+print(X_test.shape)
+#now scale the data
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+xg_clf = xgb.XGBClassifier()
+xg_clf.fit(X_train_scaled,y_train)
+xgb.plot_tree(xg_clf,num_trees=0)
+plt.rcParams['figure.figsize'] = [50, 10]
+plt.show()
+xgb.plot_importance(xg_clf)
+plt.rcParams['figure.figsize'] = [5, 5]
+plt.show()
+
    
+
    
+
+
+
+
+
    
+ © 1999-2019, Morten Hjorth-Jensen. Released under CC Attribution-NonCommercial 4.0 license
+
    
+
+
+
+
+    
+
diff --git a/doc/src/DecisionTrees/DecisionTrees.ipynb b/doc/src/DecisionTrees/DecisionTrees.ipynb
new file mode 100644
index 000000000..1161b80b6
--- /dev/null
+++ b/doc/src/DecisionTrees/DecisionTrees.ipynb
@@ -0,0 +1,2602 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "\n",
+    "# Data Analysis and Machine Learning: From Decision Trees to Forests and all that\n",
+    "\n",
+    "  \n",
+    "**Morten Hjorth-Jensen**, Department of Physics, University of Oslo and Department of Physics and Astronomy and National Superconducting Cyclotron Laboratory, Michigan State University\n",
+    "\n",
+    "Date: **Nov 7, 2019**\n",
+    "\n",
+    "Copyright 1999-2019, Morten Hjorth-Jensen. Released under CC Attribution-NonCommercial 4.0 license\n",
+    "\n",
+    "\n",
+    "\n",
+    "\n",
+    "## Decision trees, overarching aims\n",
+    "\n",
+    "\n",
+    "Decision trees are supervised learning algorithms used for both,\n",
+    "classification and regression tasks.\n",
+    "\n",
+    "\n",
+    "The main idea of decision trees\n",
+    "is to find those descriptive features which contain the most\n",
+    "**information** regarding the target feature and then split the dataset\n",
+    "along the values of these features such that the target feature values\n",
+    "for the resulting underlying datasets are as pure as possible.\n",
+    "\n",
+    "The descriptive features which reproduce best the target/output features are normally  said\n",
+    "to be the most informative ones. The process of finding the **most\n",
+    "informative** feature is done until we accomplish a stopping criteria\n",
+    "where we then finally end up in so called **leaf nodes**. \n",
+    "\n",
+    "A decision tree is typically divided into a **root node**, the **interior nodes**,\n",
+    "and the final **leaf nodes** or just **leaves**. These entities are then connected by so-called **branches**.\n",
+    "\n",
+    "The leaf nodes\n",
+    "contain the predictions we will make for new query instances presented\n",
+    "to our trained model. This is possible since the model has \n",
+    "learned the underlying structure of the training data and hence can,\n",
+    "given some assumptions, make predictions about the target feature value\n",
+    "(class) of unseen query instances.\n",
+    "\n",
+    "## A typical Decision Tree with its pertinent Jargon, Classification Problem\n",
+    "\n",
+    "\n",
+    "\n",
+    "\n",
+    "
    \n",
+    "\n",
+    "\n",
+    "\n",
+    "\n",
+    "\n",
+    "This tree was produced using the Wisconsin cancer data (discussed here as well, see code examples below) using **Scikit-Learn**'s decision tree classifier. Here we have used the so-called **gini** index (see below) to split the various branches.\n",
+    "\n",
+    "\n",
+    "\n",
+    "## General Features\n",
+    "\n",
+    "The overarching approach to decision trees is a top-down approach.\n",
+    "\n",
+    "* A leaf provides the classification of a given instance.\n",
+    "\n",
+    "* A node specifies a test of some attribute of the instance.\n",
+    "\n",
+    "* A branch corresponds to a possible values of an attribute.\n",
+    "\n",
+    "* An instance is classified by starting at the root node of the tree, testing the attribute specified by this node, then moving down the tree branch corresponding to the value of the attribute in the given example.\n",
+    "\n",
+    "This process is then repeated for the subtree rooted at the new\n",
+    "node.\n",
+    "\n",
+    "\n",
+    "## How do we set it up?\n",
+    "\n",
+    "\n",
+    "In simplified terms, the process of training a decision tree and\n",
+    "predicting the target features of query instances is as follows:\n",
+    "\n",
+    "1. Present a dataset containing of a number of training instances characterized by a number of descriptive features and a target feature\n",
+    "\n",
+    "2. Train the decision tree model by continuously splitting the target feature along the values of the descriptive features using a measure of information gain during the training process\n",
+    "\n",
+    "3. Grow the tree until we accomplish a stopping criteria create leaf nodes which represent the *predictions* we want to make for new query instances\n",
+    "\n",
+    "4. Show query instances to the tree and run down the tree until we arrive at leaf nodes\n",
+    "\n",
+    "Then we are essentially done!\n",
+    "\n",
+    "\n",
+    "\n",
+    "\n",
+    "\n",
+    "## Decision trees and Regression"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "%matplotlib inline\n",
+    "\n",
+    "import numpy as np\n",
+    "import matplotlib.pyplot as plt\n",
+    "from sklearn.preprocessing import PolynomialFeatures\n",
+    "from sklearn.linear_model import LinearRegression\n",
+    "\n",
+    "steps=250\n",
+    "\n",
+    "distance=0\n",
+    "x=0\n",
+    "distance_list=[]\n",
+    "steps_list=[]\n",
+    "while x\n",
+    "## Pruning the tree\n",
+    "\n",
+    "The above procedure is rather straightforward, but leads often to\n",
+    "overfitting and unnecessarily large and complicated trees. The basic\n",
+    "idea is to grow a large tree $T_0$ and then prune it back in order to\n",
+    "obtain a subtree. A smaller tree with fewer splits (fewer regions) can\n",
+    "lead to smaller variance and better interpretation at the cost of a\n",
+    "little more bias.\n",
+    "\n",
+    "The so-called Cost complexity pruning algorithm gives us a\n",
+    "way to do just this. Rather than considering every possible subtree,\n",
+    "we consider a sequence of trees indexed by a nonnegative tuning\n",
+    "parameter $\\alpha$.\n",
+    "\n",
+    "## Cost complexity pruning\n",
+    "For each value of $\\alpha$  there corresponds a subtree $T \\in T_0$ such that"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "$$\n",
+    "\\sum_{m=1}^{\\overline{T}}\\sum_{i:x_i\\in R_m}(y_i-\\overline{y}_{R_m})^2+\\alpha\\overline{T},\n",
+    "$$"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "is as small as possible. Here $\\overline{T}$ is \n",
+    "the number of terminal nodes of the tree $T$ , $R_m$ is the\n",
+    "rectangle (i.e. the subset of predictor space)  corresponding to the $m$-th terminal node.\n",
+    "\n",
+    "The tuning parameter $\\alpha$ controls a trade-off between the subtree’s\n",
+    "com- plexity and its fit to the training data. When $\\alpha = 0$, then the\n",
+    "subtree $T$ will simply equal $T_0$, \n",
+    "because then the above equation just measures the\n",
+    "training error. \n",
+    "However, as $\\alpha$ increases, there is a price to pay for\n",
+    "having a tree with many terminal nodes. The above equation will\n",
+    "tend to be minimized for a smaller subtree. \n",
+    "\n",
+    "\n",
+    "It turns out that as we increase $\\alpha$ from zero\n",
+    "branches get pruned from the tree in a nested and predictable fashion,\n",
+    "so obtaining the whole sequence of subtrees as a function of $\\alpha$ is\n",
+    "easy. We can select a value of $\\alpha$ using a validation set or using\n",
+    "cross-validation. We then return to the full data set and obtain the\n",
+    "subtree corresponding to $\\alpha$. \n",
+    "\n",
+    "\n",
+    "## Schematic Regression Procedure\n",
+    "\n",
+    "**Building a Regression Tree.**\n",
+    "\n",
+    "1. Use recursive binary splitting to grow a large tree on the training data, stopping only when each terminal node has fewer than some minimum number of observations.\n",
+    "\n",
+    "2. Apply cost complexity pruning to the large tree in order to obtain a sequence of best subtrees, as a function of $\\alpha$.\n",
+    "\n",
+    "3. Use for example $K$-fold cross-validation to choose $\\alpha$. Divide the training observations into $K$ folds. For each $k=1,2,\\dots,K$ we: \n",
+    "\n",
+    "  * repeat steps 1 and 2 on all but the $k$-th fold of the training data. \n",
+    "\n",
+    "  * Then we valuate the mean squared prediction error on the data in the left-out $k$-th fold, as a function of $\\alpha$.\n",
+    "\n",
+    "  * Finally  we average the results for each value of $\\alpha$, and pick $\\alpha$ to minimize the average error.\n",
+    "\n",
+    "\n",
+    "4. Return the subtree from Step 2 that corresponds to the chosen value of $\\alpha$.\n",
+    "\n",
+    "\n",
+    "\n",
+    "\n",
+    "## A Classification Tree\n",
+    "\n",
+    "A classification tree is very similar to a regression tree, except\n",
+    "that it is used to predict a qualitative response rather than a\n",
+    "quantitative one. Recall that for a regression tree, the predicted\n",
+    "response for an observation is given by the mean response of the\n",
+    "training observations that belong to the same terminal node. In\n",
+    "contrast, for a classification tree, we predict that each observation\n",
+    "belongs to the most commonly occurring class of training observations\n",
+    "in the region to which it belongs. In interpreting the results of a\n",
+    "classification tree, we are often interested not only in the class\n",
+    "prediction corresponding to a particular terminal node region, but\n",
+    "also in the class proportions among the training observations that\n",
+    "fall into that region.  \n",
+    "\n",
+    "## Growing a classification tree\n",
+    "\n",
+    "The task of growing a\n",
+    "classification tree is quite similar to the task of growing a\n",
+    "regression tree. Just as in the regression setting, we use recursive\n",
+    "binary splitting to grow a classification tree. However, in the\n",
+    "classification setting, the MSE cannot be used as a criterion for making\n",
+    "the binary splits.  A natural alternative to MSE is the **classification\n",
+    "error rate**. Since we plan to assign an observation in a given region\n",
+    "to the most commonly occurring error rate class of training\n",
+    "observations in that region, the classification error rate is simply\n",
+    "the fraction of the training observations in that region that do not\n",
+    "belong to the most common class. \n",
+    "\n",
+    "When building a classification tree, either the Gini index or the\n",
+    "entropy are typically used to evaluate the quality of a particular\n",
+    "split, since these two approaches are more sensitive to node purity\n",
+    "than is the classification error rate. \n",
+    "\n",
+    "\n",
+    "## Classification tree, how to split nodes\n",
+    "\n",
+    "If our targets are the outcome of a classification process that takes\n",
+    "for example $k=1,2,\\dots,K$ values, the only thing we need to think of\n",
+    "is to set up the splitting criteria for each node.\n",
+    "\n",
+    "We define a PDF $p_{mk}$ that represents the number of observations of\n",
+    "a class $k$ in a region $R_m$ with $N_m$ observations. We represent\n",
+    "this likelihood function in terms of the proportion $I(y_i=k)$ of\n",
+    "observations of this class in the region $R_m$ as"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "$$\n",
+    "p_{mk} = \\frac{1}{N_m}\\sum_{x_i\\in R_m}I(y_i=k).\n",
+    "$$"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "We let $p_{mk}$ represent the majority class of observations in region\n",
+    "$m$. The three most common ways of splitting a node are given by\n",
+    "\n",
+    "* Misclassification error"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "$$\n",
+    "p_{mk} = \\frac{1}{N_m}\\sum_{x_i\\in R_m}I(y_i\\ne k) = 1-p_{mk}.\n",
+    "$$"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "* Gini index $g$"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "$$\n",
+    "g = \\sum_{k=1}^K p_{mk}(1-p_{mk}).\n",
+    "$$"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "* Information entropy or just entropy $s$"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "$$\n",
+    "s = -\\sum_{k=1}^K p_{mk}\\log{p_{mk}}.\n",
+    "$$"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Visualizing the Tree, Classification"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "import os\n",
+    "from sklearn.datasets import load_breast_cancer\n",
+    "from sklearn.tree import DecisionTreeClassifier\n",
+    "from sklearn.model_selection import train_test_split\n",
+    "from sklearn.metrics import confusion_matrix\n",
+    "from sklearn.tree import export_graphviz\n",
+    "\n",
+    "from IPython.display import Image \n",
+    "from pydot import graph_from_dot_data\n",
+    "import pandas as pd\n",
+    "import numpy as np\n",
+    "\n",
+    "\n",
+    "cancer = load_breast_cancer()\n",
+    "X = pd.DataFrame(cancer.data, columns=cancer.feature_names)\n",
+    "print(X)\n",
+    "y = pd.Categorical.from_codes(cancer.target, cancer.target_names)\n",
+    "y = pd.get_dummies(y)\n",
+    "print(y)\n",
+    "X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=1)\n",
+    "tree_clf = DecisionTreeClassifier(max_depth=5)\n",
+    "tree_clf.fit(X_train, y_train)\n",
+    "\n",
+    "export_graphviz(\n",
+    "    tree_clf,\n",
+    "    out_file=\"DataFiles/cancer.dot\",\n",
+    "    feature_names=cancer.feature_names,\n",
+    "    class_names=cancer.target_names,\n",
+    "    rounded=True,\n",
+    "    filled=True\n",
+    ")\n",
+    "cmd = 'dot -Tpng DataFiles/cancer.dot -o DataFiles/cancer.png'\n",
+    "os.system(cmd)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Visualizing the Tree, The Moons"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "# Common imports\n",
+    "import numpy as np\n",
+    "from sklearn.model_selection import  train_test_split \n",
+    "from sklearn.tree import DecisionTreeClassifier\n",
+    "from sklearn.datasets import make_moons\n",
+    "from sklearn.tree import export_graphviz\n",
+    "from pydot import graph_from_dot_data\n",
+    "import pandas as pd\n",
+    "import os\n",
+    "\n",
+    "np.random.seed(42)\n",
+    "X, y = make_moons(n_samples=100, noise=0.25, random_state=53)\n",
+    "X_train, X_test, y_train, y_test = train_test_split(X,y,random_state=0)\n",
+    "tree_clf = DecisionTreeClassifier(max_depth=5)\n",
+    "tree_clf.fit(X_train, y_train)\n",
+    "\n",
+    "export_graphviz(\n",
+    "    tree_clf,\n",
+    "    out_file=\"DataFiles/moons.dot\",\n",
+    "    rounded=True,\n",
+    "    filled=True\n",
+    ")\n",
+    "cmd = 'dot -Tpng DataFiles/moons.dot -o DataFiles/moons.png'\n",
+    "os.system(cmd)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Algorithms for Setting up Decision Trees\n",
+    "\n",
+    "Two algorithms stand out in the set up of decision trees:\n",
+    "1. The CART (Classification And Regression Tree) algorithm for both classification and regression\n",
+    "\n",
+    "2. The ID3 algorithm based on the computation of the information gain for classification\n",
+    "\n",
+    "We discuss both algorithms with applications here. The popular library **Scikit-Learn** uses the CART algorithm. For classification problems you can use either the **gini** index or the **entropy** to split a tree in two branches.\n",
+    "\n",
+    "## The CART algorithm for Classification\n",
+    "\n",
+    "\n",
+    "## The CART algorithm for Regression\n",
+    "\n",
+    "\n",
+    "\n",
+    "## Computing the Gini index\n",
+    "\n",
+    "The example we will look at is a classical one in many Machine\n",
+    "Learning applications. Based on various meteorological features, we\n",
+    "have several so-called attributes which decide whether we at the end\n",
+    "will do some outdoor activity like skiing, going for a bike ride etc\n",
+    "etc.  The table here contains the feautures **outlook**, **temperature**,\n",
+    "**humidity** and **wind**.  The target or output is whether we ride\n",
+    "(True=1) or whether we do something else that day (False=0). The\n",
+    "attributes for each feature are then sunny, overcast and rain for the\n",
+    "outlook, hot, cold and mild for temperature, high and normal for\n",
+    "humidity and weak and strong for wind.\n",
+    "\n",
+    "The table here summarizes the various attributes and\n",
+    "\n",
+    "\n",
+    "\n",
+    "\n",
+    "\n",
+    "\n",
+    "\n",
+    "\n",
+    "\n",
+    "\n",
+    "\n",
+    "\n",
+    "\n",
+    "\n",
+    "\n",
+    "\n",
+    "\n",
+    "\n",
+    "\n",
+    "\n",
+    "
    Day Outlook  Temperature Humidity  Wind  Ride 
       1          Sunny           Hot                High            Weak          0        
       2          Sunny           Hot                High            Strong        1        
       3          Overcast        Hot                High            Weak          1        
       4          Rain            Mild               High            Weak          1        
       5          Rain            Cool               Normal          Weak          1        
       6          Rain            Cool               Normal          Strong        0        
       7          Overcast        Cool               Normal          Strong        1        
       8          Sunny           Mild               High            Weak          0        
       9          Sunny           Cool               Normal          Weak          1        
       10         Rain            Mild               Normal          Weak          1        
       11         Sunny           Mild               Normal          Strong        1        
       12         Overcast        Mild               High            Strong        1        
       13         Overcast        Hot                Normal          Weak          1        
       14         Rain            Mild               High            Strong        0        
    \n",
+    "\n",
+    "## Simple Python Code to read in Data and perform Classification"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "# Common imports\n",
+    "import numpy as np\n",
+    "import pandas as pd\n",
+    "import matplotlib.pyplot as plt\n",
+    "from sklearn.tree import DecisionTreeClassifier\n",
+    "from sklearn.model_selection import train_test_split\n",
+    "from sklearn.tree import export_graphviz\n",
+    "from sklearn.preprocessing import StandardScaler, OneHotEncoder\n",
+    "from sklearn.compose import ColumnTransformer\n",
+    "from IPython.display import Image \n",
+    "from pydot import graph_from_dot_data\n",
+    "import os\n",
+    "\n",
+    "# Where to save the figures and data files\n",
+    "PROJECT_ROOT_DIR = \"Results\"\n",
+    "FIGURE_ID = \"Results/FigureFiles\"\n",
+    "DATA_ID = \"DataFiles/\"\n",
+    "\n",
+    "if not os.path.exists(PROJECT_ROOT_DIR):\n",
+    "    os.mkdir(PROJECT_ROOT_DIR)\n",
+    "\n",
+    "if not os.path.exists(FIGURE_ID):\n",
+    "    os.makedirs(FIGURE_ID)\n",
+    "\n",
+    "if not os.path.exists(DATA_ID):\n",
+    "    os.makedirs(DATA_ID)\n",
+    "\n",
+    "def image_path(fig_id):\n",
+    "    return os.path.join(FIGURE_ID, fig_id)\n",
+    "\n",
+    "def data_path(dat_id):\n",
+    "    return os.path.join(DATA_ID, dat_id)\n",
+    "\n",
+    "def save_fig(fig_id):\n",
+    "    plt.savefig(image_path(fig_id) + \".png\", format='png')\n",
+    "\n",
+    "infile = open(data_path(\"rideclass.csv\"),'r')\n",
+    "\n",
+    "# Read the experimental data with Pandas\n",
+    "from IPython.display import display\n",
+    "ridedata = pd.read_csv(infile,names = ('Outlook','Temperature','Humidity','Wind','Ride'))\n",
+    "ridedata = pd.DataFrame(ridedata)\n",
+    "\n",
+    "# Features and targets\n",
+    "X = ridedata.loc[:, ridedata.columns != 'Ride'].values\n",
+    "y = ridedata.loc[:, ridedata.columns == 'Ride'].values\n",
+    "\n",
+    "# Create the encoder.\n",
+    "encoder = OneHotEncoder(handle_unknown=\"ignore\")\n",
+    "# Assume for simplicity all features are categorical.\n",
+    "encoder.fit(X)    \n",
+    "# Apply the encoder.\n",
+    "X = encoder.transform(X)\n",
+    "print(X)\n",
+    "# Then do a Classification tree\n",
+    "tree_clf = DecisionTreeClassifier(max_depth=2)\n",
+    "tree_clf.fit(X, y)\n",
+    "print(\"Train set accuracy with Decision Tree: {:.2f}\".format(tree_clf.score(X,y)))\n",
+    "#transfer to a decision tree graph\n",
+    "export_graphviz(\n",
+    "    tree_clf,\n",
+    "    out_file=\"DataFiles/ride.dot\",\n",
+    "    rounded=True,\n",
+    "    filled=True\n",
+    ")\n",
+    "cmd = 'dot -Tpng DataFiles/cancer.dot -o DataFiles/cancer.png'\n",
+    "os.system(cmd)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Computing the Gini Factor\n",
+    "\n",
+    "The above functions (gini, entropy and misclassification error) are\n",
+    "important components of the so-called CART algorithm. We will discuss\n",
+    "this algorithm below after we have discussed the information gain\n",
+    "algorithm ID3.\n",
+    "\n",
+    "In the example here we have converted all our attributes into numerical values $0,1,2$ etc."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "# Split a dataset based on an attribute and an attribute value\n",
+    "def test_split(index, value, dataset):\n",
+    "\tleft, right = list(), list()\n",
+    "\tfor row in dataset:\n",
+    "\t\tif row[index] < value:\n",
+    "\t\t\tleft.append(row)\n",
+    "\t\telse:\n",
+    "\t\t\tright.append(row)\n",
+    "\treturn left, right\n",
+    " \n",
+    "# Calculate the Gini index for a split dataset\n",
+    "def gini_index(groups, classes):\n",
+    "\t# count all samples at split point\n",
+    "\tn_instances = float(sum([len(group) for group in groups]))\n",
+    "\t# sum weighted Gini index for each group\n",
+    "\tgini = 0.0\n",
+    "\tfor group in groups:\n",
+    "\t\tsize = float(len(group))\n",
+    "\t\t# avoid divide by zero\n",
+    "\t\tif size == 0:\n",
+    "\t\t\tcontinue\n",
+    "\t\tscore = 0.0\n",
+    "\t\t# score the group based on the score for each class\n",
+    "\t\tfor class_val in classes:\n",
+    "\t\t\tp = [row[-1] for row in group].count(class_val) / size\n",
+    "\t\t\tscore += p * p\n",
+    "\t\t# weight the group score by its relative size\n",
+    "\t\tgini += (1.0 - score) * (size / n_instances)\n",
+    "\treturn gini\n",
+    "\n",
+    "# Select the best split point for a dataset\n",
+    "def get_split(dataset):\n",
+    "\tclass_values = list(set(row[-1] for row in dataset))\n",
+    "\tb_index, b_value, b_score, b_groups = 999, 999, 999, None\n",
+    "\tfor index in range(len(dataset[0])-1):\n",
+    "\t\tfor row in dataset:\n",
+    "\t\t\tgroups = test_split(index, row[index], dataset)\n",
+    "\t\t\tgini = gini_index(groups, class_values)\n",
+    "\t\t\tprint('X%d < %.3f Gini=%.3f' % ((index+1), row[index], gini))\n",
+    "\t\t\tif gini < b_score:\n",
+    "\t\t\t\tb_index, b_value, b_score, b_groups = index, row[index], gini, groups\n",
+    "\treturn {'index':b_index, 'value':b_value, 'groups':b_groups}\n",
+    " \n",
+    "dataset = [[0,0,0,0,0],\n",
+    "            [0,0,0,1,1],\n",
+    "            [1,0,0,0,1],\n",
+    "            [2,1,0,0,1],\n",
+    "            [2,2,1,0,1],\n",
+    "            [2,2,1,1,0],\n",
+    "            [1,2,1,1,1],\n",
+    "            [0,1,0,0,0],\n",
+    "            [0,2,1,0,1],\n",
+    "            [2,1,1,0,1],\n",
+    "            [0,1,1,1,1],\n",
+    "            [1,1,0,1,1],\n",
+    "            [1,0,1,0,1],\n",
+    "            [2,1,0,1,0]]\n",
+    "\n",
+    "split = get_split(dataset)\n",
+    "print('Split: [X%d < %.3f]' % ((split['index']+1), split['value']))"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Entropy and the ID3 algorithm\n",
+    "\n",
+    "ID3, learns decision trees by constructing\n",
+    "them topdown, beginning with the question **which attribute should be tested at the root of the tree**?\n",
+    "\n",
+    "1. Each instance attribute is evaluated using a statistical test to determine how well it alone classifies the training examples.\n",
+    "\n",
+    "2. The best attribute is selected and used as the test at the root node of the tree.\n",
+    "\n",
+    "3. A descendant of the root node is then created for each possible value of this attribute.\n",
+    "\n",
+    "4. Training examples are sorted to the appropriate descendant node.\n",
+    "\n",
+    "5. The entire process is then repeated using the training examples associated with each descendant node to select the best attribute to test at that point in the tree.\n",
+    "\n",
+    "6. This forms a greedy search for an acceptable decision tree, in which the algorithm never backtracks to reconsider earlier choices. \n",
+    "\n",
+    "The ID3 algorithm selects, which attribute to test at each node in the\n",
+    "tree.\n",
+    "\n",
+    "We would like to select the attribute that is most useful for classifying\n",
+    "examples.\n",
+    "\n",
+    "What is a good quantitative measure of the worth of an attribute?\n",
+    "\n",
+    "Information gain measures how well a given attribute separates the\n",
+    "training examples according to their target classification.\n",
+    "\n",
+    "The ID3 algorithm uses this information gain measure to select among the candidate\n",
+    "attributes at each step while growing the tree.\n",
+    "\n",
+    "## Implementing the ID3 Algorithm"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "import re\n",
+    "import math\n",
+    "from collections import deque\n",
+    "\n",
+    "# x is examples in training set\n",
+    "# y is set of targets\n",
+    "# label is target attributes\n",
+    "# Node is a class which has properties values, childs, and next\n",
+    "# root is top node in the decision tree\n",
+    "\n",
+    "class Node(object):\n",
+    "\tdef __init__(self):\n",
+    "\t\tself.value = None\n",
+    "\t\tself.next = None\n",
+    "\t\tself.childs = None\n",
+    "\n",
+    "# Simple class of Decision Tree\n",
+    "# Aimed for who want to learn Decision Tree, so it is not optimized\n",
+    "class DecisionTree(object):\n",
+    "\tdef __init__(self, sample, attributes, labels):\n",
+    "\t\tself.sample = sample\n",
+    "\t\tself.attributes = attributes\n",
+    "\t\tself.labels = labels\n",
+    "\t\tself.labelCodes = None\n",
+    "\t\tself.labelCodesCount = None\n",
+    "\t\tself.initLabelCodes()\n",
+    "\t\t# print(self.labelCodes)\n",
+    "\t\tself.root = None\n",
+    "\t\tself.entropy = self.getEntropy([x for x in range(len(self.labels))])\n",
+    "\n",
+    "\tdef initLabelCodes(self):\n",
+    "\t\tself.labelCodes = []\n",
+    "\t\tself.labelCodesCount = []\n",
+    "\t\tfor l in self.labels:\n",
+    "\t\t\tif l not in self.labelCodes:\n",
+    "\t\t\t\tself.labelCodes.append(l)\n",
+    "\t\t\t\tself.labelCodesCount.append(0)\n",
+    "\t\t\tself.labelCodesCount[self.labelCodes.index(l)] += 1\n",
+    "\n",
+    "\tdef getLabelCodeId(self, sampleId):\n",
+    "\t\treturn self.labelCodes.index(self.labels[sampleId])\n",
+    "\n",
+    "\tdef getAttributeValues(self, sampleIds, attributeId):\n",
+    "\t\tvals = []\n",
+    "\t\tfor sid in sampleIds:\n",
+    "\t\t\tval = self.sample[sid][attributeId]\n",
+    "\t\t\tif val not in vals:\n",
+    "\t\t\t\tvals.append(val)\n",
+    "\t\t# print(vals)\n",
+    "\t\treturn vals\n",
+    "\n",
+    "\tdef getEntropy(self, sampleIds):\n",
+    "\t\tentropy = 0\n",
+    "\t\tlabelCount = [0] * len(self.labelCodes)\n",
+    "\t\tfor sid in sampleIds:\n",
+    "\t\t\tlabelCount[self.getLabelCodeId(sid)] += 1\n",
+    "\t\t# print(\"-ge\", labelCount)\n",
+    "\t\tfor lv in labelCount:\n",
+    "\t\t\t# print(lv)\n",
+    "\t\t\tif lv != 0:\n",
+    "\t\t\t\tentropy += -lv/len(sampleIds) * math.log(lv/len(sampleIds), 2)\n",
+    "\t\t\telse:\n",
+    "\t\t\t\tentropy += 0\n",
+    "\t\treturn entropy\n",
+    "\n",
+    "\tdef getDominantLabel(self, sampleIds):\n",
+    "\t\tlabelCodesCount = [0] * len(self.labelCodes)\n",
+    "\t\tfor sid in sampleIds:\n",
+    "\t\t\tlabelCodesCount[self.labelCodes.index(self.labels[sid])] += 1\n",
+    "\t\treturn self.labelCodes[labelCodesCount.index(max(labelCodesCount))]\n",
+    "\n",
+    "\tdef getInformationGain(self, sampleIds, attributeId):\n",
+    "\t\tgain = self.getEntropy(sampleIds)\n",
+    "\t\tattributeVals = []\n",
+    "\t\tattributeValsCount = []\n",
+    "\t\tattributeValsIds = []\n",
+    "\t\tfor sid in sampleIds:\n",
+    "\t\t\tval = self.sample[sid][attributeId]\n",
+    "\t\t\tif val not in attributeVals:\n",
+    "\t\t\t\tattributeVals.append(val)\n",
+    "\t\t\t\tattributeValsCount.append(0)\n",
+    "\t\t\t\tattributeValsIds.append([])\n",
+    "\t\t\tvid = attributeVals.index(val)\n",
+    "\t\t\tattributeValsCount[vid] += 1\n",
+    "\t\t\tattributeValsIds[vid].append(sid)\n",
+    "\t\t# print(\"-gig\", self.attributes[attributeId])\n",
+    "\t\tfor vc, vids in zip(attributeValsCount, attributeValsIds):\n",
+    "\t\t\t# print(\"-gig\", vids)\n",
+    "\t\t\tgain -= vc/len(sampleIds) * self.getEntropy(vids)\n",
+    "\t\treturn gain\n",
+    "\n",
+    "\tdef getAttributeMaxInformationGain(self, sampleIds, attributeIds):\n",
+    "\t\tattributesEntropy = [0] * len(attributeIds)\n",
+    "\t\tfor i, attId in zip(range(len(attributeIds)), attributeIds):\n",
+    "\t\t\tattributesEntropy[i] = self.getInformationGain(sampleIds, attId)\n",
+    "\t\tmaxId = attributeIds[attributesEntropy.index(max(attributesEntropy))]\n",
+    "\t\treturn self.attributes[maxId], maxId\n",
+    "\n",
+    "\tdef isSingleLabeled(self, sampleIds):\n",
+    "\t\tlabel = self.labels[sampleIds[0]]\n",
+    "\t\tfor sid in sampleIds:\n",
+    "\t\t\tif self.labels[sid] != label:\n",
+    "\t\t\t\treturn False\n",
+    "\t\treturn True\n",
+    "\n",
+    "\tdef getLabel(self, sampleId):\n",
+    "\t\treturn self.labels[sampleId]\n",
+    "\n",
+    "\tdef id3(self):\n",
+    "\t\tsampleIds = [x for x in range(len(self.sample))]\n",
+    "\t\tattributeIds = [x for x in range(len(self.attributes))]\n",
+    "\t\tself.root = self.id3Recv(sampleIds, attributeIds, self.root)\n",
+    "\n",
+    "\tdef id3Recv(self, sampleIds, attributeIds, root):\n",
+    "\t\troot = Node() # Initialize current root\n",
+    "\t\tif self.isSingleLabeled(sampleIds):\n",
+    "\t\t\troot.value = self.labels[sampleIds[0]]\n",
+    "\t\t\treturn root\n",
+    "\t\t# print(attributeIds)\n",
+    "\t\tif len(attributeIds) == 0:\n",
+    "\t\t\troot.value = self.getDominantLabel(sampleIds)\n",
+    "\t\t\treturn root\n",
+    "\t\tbestAttrName, bestAttrId = self.getAttributeMaxInformationGain(\n",
+    "\t\t\tsampleIds, attributeIds)\n",
+    "\t\t# print(bestAttrName)\n",
+    "\t\troot.value = bestAttrName\n",
+    "\t\troot.childs = []  # Create list of children\n",
+    "\t\tfor value in self.getAttributeValues(sampleIds, bestAttrId):\n",
+    "\t\t\t# print(value)\n",
+    "\t\t\tchild = Node()\n",
+    "\t\t\tchild.value = value\n",
+    "\t\t\troot.childs.append(child)  # Append new child node to current\n",
+    "\t\t\t\t\t\t\t\t\t   # root\n",
+    "\t\t\tchildSampleIds = []\n",
+    "\t\t\tfor sid in sampleIds:\n",
+    "\t\t\t\tif self.sample[sid][bestAttrId] == value:\n",
+    "\t\t\t\t\tchildSampleIds.append(sid)\n",
+    "\t\t\tif len(childSampleIds) == 0:\n",
+    "\t\t\t\tchild.next = self.getDominantLabel(sampleIds)\n",
+    "\t\t\telse:\n",
+    "\t\t\t\t# print(bestAttrName, bestAttrId)\n",
+    "\t\t\t\t# print(attributeIds)\n",
+    "\t\t\t\tif len(attributeIds) > 0 and bestAttrId in attributeIds:\n",
+    "\t\t\t\t\ttoRemove = attributeIds.index(bestAttrId)\n",
+    "\t\t\t\t\tattributeIds.pop(toRemove)\n",
+    "\t\t\t\tchild.next = self.id3Recv(\n",
+    "\t\t\t\t\tchildSampleIds, attributeIds, child.next)\n",
+    "\t\treturn root\n",
+    "\n",
+    "\tdef printTree(self):\n",
+    "\t\tif self.root:\n",
+    "\t\t\troots = deque()\n",
+    "\t\t\troots.append(self.root)\n",
+    "\t\t\twhile len(roots) > 0:\n",
+    "\t\t\t\troot = roots.popleft()\n",
+    "\t\t\t\tprint(root.value)\n",
+    "\t\t\t\tif root.childs:\n",
+    "\t\t\t\t\tfor child in root.childs:\n",
+    "\t\t\t\t\t\tprint('({})'.format(child.value))\n",
+    "\t\t\t\t\t\troots.append(child.next)\n",
+    "\t\t\t\telif root.next:\n",
+    "\t\t\t\t\tprint(root.next)\n",
+    "\n",
+    "\n",
+    "def test():\n",
+    "\tf = open('DataFiles/rideclass.csv')\n",
+    "\tattributes = f.readline().split(',')\n",
+    "\tattributes = attributes[1:len(attributes)-1]\n",
+    "\tprint(attributes)\n",
+    "\tsample = f.readlines()\n",
+    "\tf.close()\n",
+    "\tfor i in range(len(sample)):\n",
+    "\t\tsample[i] = re.sub('\\d+,', '', sample[i])\n",
+    "\t\tsample[i] = sample[i].strip().split(',')\n",
+    "\tlabels = []\n",
+    "\tfor s in sample:\n",
+    "\t\tlabels.append(s.pop())\n",
+    "\t# print(sample)\n",
+    "\t# print(labels)\n",
+    "\tdecisionTree = DecisionTree(sample, attributes, labels)\n",
+    "\tprint(\"System entropy {}\".format(decisionTree.entropy))\n",
+    "\tdecisionTree.id3()\n",
+    "\tdecisionTree.printTree()\n",
+    "\n",
+    "\n",
+    "if __name__ == '__main__':\n",
+    "\ttest()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Cancer Data again now with Decision Trees and other Methods"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 7,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "import matplotlib.pyplot as plt\n",
+    "import numpy as np\n",
+    "from sklearn.model_selection import  train_test_split \n",
+    "from sklearn.datasets import load_breast_cancer\n",
+    "from sklearn.svm import SVC\n",
+    "from sklearn.linear_model import LogisticRegression\n",
+    "from sklearn.tree import DecisionTreeClassifier\n",
+    "\n",
+    "# Load the data\n",
+    "cancer = load_breast_cancer()\n",
+    "\n",
+    "X_train, X_test, y_train, y_test = train_test_split(cancer.data,cancer.target,random_state=0)\n",
+    "print(X_train.shape)\n",
+    "print(X_test.shape)\n",
+    "# Logistic Regression\n",
+    "logreg = LogisticRegression(solver='lbfgs')\n",
+    "logreg.fit(X_train, y_train)\n",
+    "print(\"Test set accuracy with Logistic Regression: {:.2f}\".format(logreg.score(X_test,y_test)))\n",
+    "# Support vector machine\n",
+    "svm = SVC(gamma='auto', C=100)\n",
+    "svm.fit(X_train, y_train)\n",
+    "print(\"Test set accuracy with SVM: {:.2f}\".format(svm.score(X_test,y_test)))\n",
+    "# Decision Trees\n",
+    "deep_tree_clf = DecisionTreeClassifier(max_depth=None)\n",
+    "deep_tree_clf.fit(X_train, y_train)\n",
+    "print(\"Test set accuracy with Decision Trees: {:.2f}\".format(deep_tree_clf.score(X_test,y_test)))\n",
+    "#now scale the data\n",
+    "from sklearn.preprocessing import StandardScaler\n",
+    "scaler = StandardScaler()\n",
+    "scaler.fit(X_train)\n",
+    "X_train_scaled = scaler.transform(X_train)\n",
+    "X_test_scaled = scaler.transform(X_test)\n",
+    "# Logistic Regression\n",
+    "logreg.fit(X_train_scaled, y_train)\n",
+    "print(\"Test set accuracy Logistic Regression with scaled data: {:.2f}\".format(logreg.score(X_test_scaled,y_test)))\n",
+    "# Support Vector Machine\n",
+    "svm.fit(X_train_scaled, y_train)\n",
+    "print(\"Test set accuracy SVM with scaled data: {:.2f}\".format(logreg.score(X_test_scaled,y_test)))\n",
+    "# Decision Trees\n",
+    "deep_tree_clf.fit(X_train_scaled, y_train)\n",
+    "print(\"Test set accuracy with Decision Trees and scaled data: {:.2f}\".format(deep_tree_clf.score(X_test_scaled,y_test)))"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Another example, the moons again"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "from __future__ import division, print_function, unicode_literals\n",
+    "\n",
+    "# Common imports\n",
+    "import numpy as np\n",
+    "import os\n",
+    "\n",
+    "# to make this notebook's output stable across runs\n",
+    "np.random.seed(42)\n",
+    "\n",
+    "# To plot pretty figures\n",
+    "import matplotlib\n",
+    "import matplotlib.pyplot as plt\n",
+    "from matplotlib.colors import ListedColormap\n",
+    "plt.rcParams['axes.labelsize'] = 14\n",
+    "plt.rcParams['xtick.labelsize'] = 12\n",
+    "plt.rcParams['ytick.labelsize'] = 12\n",
+    "\n",
+    "\n",
+    "from sklearn.svm import SVC\n",
+    "from sklearn import datasets\n",
+    "from sklearn.tree import DecisionTreeClassifier\n",
+    "from sklearn.datasets import make_moons\n",
+    "from sklearn.tree import export_graphviz\n",
+    "\n",
+    "Xm, ym = make_moons(n_samples=100, noise=0.25, random_state=53)\n",
+    "\n",
+    "deep_tree_clf1 = DecisionTreeClassifier(random_state=42)\n",
+    "deep_tree_clf2 = DecisionTreeClassifier(min_samples_leaf=4, random_state=42)\n",
+    "deep_tree_clf1.fit(Xm, ym)\n",
+    "deep_tree_clf2.fit(Xm, ym)\n",
+    "\n",
+    "\n",
+    "def plot_decision_boundary(clf, X, y, axes=[0, 7.5, 0, 3], iris=True, legend=False, plot_training=True):\n",
+    "    x1s = np.linspace(axes[0], axes[1], 100)\n",
+    "    x2s = np.linspace(axes[2], axes[3], 100)\n",
+    "    x1, x2 = np.meshgrid(x1s, x2s)\n",
+    "    X_new = np.c_[x1.ravel(), x2.ravel()]\n",
+    "    y_pred = clf.predict(X_new).reshape(x1.shape)\n",
+    "    custom_cmap = ListedColormap(['#fafab0','#9898ff','#a0faa0'])\n",
+    "    plt.contourf(x1, x2, y_pred, alpha=0.3, cmap=custom_cmap)\n",
+    "    if not iris:\n",
+    "        custom_cmap2 = ListedColormap(['#7d7d58','#4c4c7f','#507d50'])\n",
+    "        plt.contour(x1, x2, y_pred, cmap=custom_cmap2, alpha=0.8)\n",
+    "    if plot_training:\n",
+    "        plt.plot(X[:, 0][y==0], X[:, 1][y==0], \"yo\", label=\"Iris-Setosa\")\n",
+    "        plt.plot(X[:, 0][y==1], X[:, 1][y==1], \"bs\", label=\"Iris-Versicolor\")\n",
+    "        plt.plot(X[:, 0][y==2], X[:, 1][y==2], \"g^\", label=\"Iris-Virginica\")\n",
+    "        plt.axis(axes)\n",
+    "    if iris:\n",
+    "        plt.xlabel(\"Petal length\", fontsize=14)\n",
+    "        plt.ylabel(\"Petal width\", fontsize=14)\n",
+    "    else:\n",
+    "        plt.xlabel(r\"$x_1$\", fontsize=18)\n",
+    "        plt.ylabel(r\"$x_2$\", fontsize=18, rotation=0)\n",
+    "    if legend:\n",
+    "        plt.legend(loc=\"lower right\", fontsize=14)\n",
+    "plt.figure(figsize=(11, 4))\n",
+    "plt.subplot(121)\n",
+    "plot_decision_boundary(deep_tree_clf1, Xm, ym, axes=[-1.5, 2.5, -1, 1.5], iris=False)\n",
+    "plt.title(\"No restrictions\", fontsize=16)\n",
+    "plt.subplot(122)\n",
+    "plot_decision_boundary(deep_tree_clf2, Xm, ym, axes=[-1.5, 2.5, -1, 1.5], iris=False)\n",
+    "plt.title(\"min_samples_leaf = {}\".format(deep_tree_clf2.min_samples_leaf), fontsize=14)\n",
+    "plt.show()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Playing around with regions"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "np.random.seed(6)\n",
+    "Xs = np.random.rand(100, 2) - 0.5\n",
+    "ys = (Xs[:, 0] > 0).astype(np.float32) * 2\n",
+    "\n",
+    "angle = np.pi/4\n",
+    "rotation_matrix = np.array([[np.cos(angle), -np.sin(angle)], [np.sin(angle), np.cos(angle)]])\n",
+    "Xsr = Xs.dot(rotation_matrix)\n",
+    "\n",
+    "tree_clf_s = DecisionTreeClassifier(random_state=42)\n",
+    "tree_clf_s.fit(Xs, ys)\n",
+    "tree_clf_sr = DecisionTreeClassifier(random_state=42)\n",
+    "tree_clf_sr.fit(Xsr, ys)\n",
+    "\n",
+    "plt.figure(figsize=(11, 4))\n",
+    "plt.subplot(121)\n",
+    "plot_decision_boundary(tree_clf_s, Xs, ys, axes=[-0.7, 0.7, -0.7, 0.7], iris=False)\n",
+    "plt.subplot(122)\n",
+    "plot_decision_boundary(tree_clf_sr, Xsr, ys, axes=[-0.7, 0.7, -0.7, 0.7], iris=False)\n",
+    "\n",
+    "plt.show()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Regression trees"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 10,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "# Quadratic training set + noise\n",
+    "np.random.seed(42)\n",
+    "m = 200\n",
+    "X = np.random.rand(m, 1)\n",
+    "y = 4 * (X - 0.5) ** 2\n",
+    "y = y + np.random.randn(m, 1) / 10"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 11,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "from sklearn.tree import DecisionTreeRegressor\n",
+    "\n",
+    "tree_reg = DecisionTreeRegressor(max_depth=2, random_state=42)\n",
+    "tree_reg.fit(X, y)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Final regressor code"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 12,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "from sklearn.tree import DecisionTreeRegressor\n",
+    "\n",
+    "tree_reg1 = DecisionTreeRegressor(random_state=42, max_depth=2)\n",
+    "tree_reg2 = DecisionTreeRegressor(random_state=42, max_depth=3)\n",
+    "tree_reg1.fit(X, y)\n",
+    "tree_reg2.fit(X, y)\n",
+    "\n",
+    "def plot_regression_predictions(tree_reg, X, y, axes=[0, 1, -0.2, 1], ylabel=\"$y$\"):\n",
+    "    x1 = np.linspace(axes[0], axes[1], 500).reshape(-1, 1)\n",
+    "    y_pred = tree_reg.predict(x1)\n",
+    "    plt.axis(axes)\n",
+    "    plt.xlabel(\"$x_1$\", fontsize=18)\n",
+    "    if ylabel:\n",
+    "        plt.ylabel(ylabel, fontsize=18, rotation=0)\n",
+    "    plt.plot(X, y, \"b.\")\n",
+    "    plt.plot(x1, y_pred, \"r.-\", linewidth=2, label=r\"$\\hat{y}$\")\n",
+    "\n",
+    "plt.figure(figsize=(11, 4))\n",
+    "plt.subplot(121)\n",
+    "plot_regression_predictions(tree_reg1, X, y)\n",
+    "for split, style in ((0.1973, \"k-\"), (0.0917, \"k--\"), (0.7718, \"k--\")):\n",
+    "    plt.plot([split, split], [-0.2, 1], style, linewidth=2)\n",
+    "plt.text(0.21, 0.65, \"Depth=0\", fontsize=15)\n",
+    "plt.text(0.01, 0.2, \"Depth=1\", fontsize=13)\n",
+    "plt.text(0.65, 0.8, \"Depth=1\", fontsize=13)\n",
+    "plt.legend(loc=\"upper center\", fontsize=18)\n",
+    "plt.title(\"max_depth=2\", fontsize=14)\n",
+    "\n",
+    "plt.subplot(122)\n",
+    "plot_regression_predictions(tree_reg2, X, y, ylabel=None)\n",
+    "for split, style in ((0.1973, \"k-\"), (0.0917, \"k--\"), (0.7718, \"k--\")):\n",
+    "    plt.plot([split, split], [-0.2, 1], style, linewidth=2)\n",
+    "for split in (0.0458, 0.1298, 0.2873, 0.9040):\n",
+    "    plt.plot([split, split], [-0.2, 1], \"k:\", linewidth=1)\n",
+    "plt.text(0.3, 0.5, \"Depth=2\", fontsize=13)\n",
+    "plt.title(\"max_depth=3\", fontsize=14)\n",
+    "\n",
+    "plt.show()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 13,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "tree_reg1 = DecisionTreeRegressor(random_state=42)\n",
+    "tree_reg2 = DecisionTreeRegressor(random_state=42, min_samples_leaf=10)\n",
+    "tree_reg1.fit(X, y)\n",
+    "tree_reg2.fit(X, y)\n",
+    "\n",
+    "x1 = np.linspace(0, 1, 500).reshape(-1, 1)\n",
+    "y_pred1 = tree_reg1.predict(x1)\n",
+    "y_pred2 = tree_reg2.predict(x1)\n",
+    "\n",
+    "plt.figure(figsize=(11, 4))\n",
+    "\n",
+    "plt.subplot(121)\n",
+    "plt.plot(X, y, \"b.\")\n",
+    "plt.plot(x1, y_pred1, \"r.-\", linewidth=2, label=r\"$\\hat{y}$\")\n",
+    "plt.axis([0, 1, -0.2, 1.1])\n",
+    "plt.xlabel(\"$x_1$\", fontsize=18)\n",
+    "plt.ylabel(\"$y$\", fontsize=18, rotation=0)\n",
+    "plt.legend(loc=\"upper center\", fontsize=18)\n",
+    "plt.title(\"No restrictions\", fontsize=14)\n",
+    "\n",
+    "plt.subplot(122)\n",
+    "plt.plot(X, y, \"b.\")\n",
+    "plt.plot(x1, y_pred2, \"r.-\", linewidth=2, label=r\"$\\hat{y}$\")\n",
+    "plt.axis([0, 1, -0.2, 1.1])\n",
+    "plt.xlabel(\"$x_1$\", fontsize=18)\n",
+    "plt.title(\"min_samples_leaf={}\".format(tree_reg2.min_samples_leaf), fontsize=14)\n",
+    "\n",
+    "plt.show()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Pros and cons of trees, pros\n",
+    "\n",
+    "* White box, easy to interpret model. Some people believe that decision trees more closely mirror human decision-making than do the regression and classification approaches discussed earlier (think of support vector machines)\n",
+    "\n",
+    "* Trees are very easy to explain to people. In fact, they are even easier to explain than linear regression!\n",
+    "\n",
+    "* No feature normalization needed\n",
+    "\n",
+    "* Tree models can handle both continuous and categorical data (Classification and Regression Trees)\n",
+    "\n",
+    "* Can model nonlinear relationships\n",
+    "\n",
+    "* Can model interactions between the different descriptive features\n",
+    "\n",
+    "* Trees can be displayed graphically, and are easily interpreted even by a non-expert (especially if they are small)\n",
+    "\n",
+    "## Disadvantages\n",
+    "\n",
+    "* Unfortunately, trees generally do not have the same level of predictive accuracy as some of the other regression and classification approaches\n",
+    "\n",
+    "* If continuous features are used the tree may become quite large and hence less interpretable\n",
+    "\n",
+    "* Decision trees are prone to overfit the training data and hence do not well generalize the data if no stopping criteria or improvements like pruning, boosting or bagging are implemented\n",
+    "\n",
+    "* Small changes in the data may lead to a completely different tree. This issue can be addressed by using ensemble methods like bagging, boosting or random forests\n",
+    "\n",
+    "* Unbalanced datasets where some target feature values occur much more frequently than others may lead to biased trees since the frequently occurring feature values are preferred over the less frequently occurring ones. \n",
+    "\n",
+    "* If the number of features is relatively large (high dimensional) and the number of instances is relatively low, the tree might overfit the data\n",
+    "\n",
+    "* Features with many levels may be preferred over features with less levels since for them it is *more easy* to split the dataset such that the sub datasets only contain pure target feature values. This issue can be addressed by preferring for instance the information gain ratio as splitting criteria over information gain\n",
+    "\n",
+    "However, by aggregating many decision trees, using methods like bagging, random forests, and boosting, the predictive performance of trees can be substantially improved. \n",
+    "\n",
+    "\n",
+    "## Ensemble Methods: From a Single Tree to Many Trees and Extreme Boosting, Meet the Jungle of Methods\n",
+    "\n",
+    "As stated above and seen in many of the examples discussed here about\n",
+    "a single decision tree, we often end up overfitting our training\n",
+    "data. This normally means that we have a high variance. Can we reduce\n",
+    "the variance of a statistical learning method?\n",
+    "\n",
+    "This leads us to a set of different methods that can combine different\n",
+    "machine learning algorithms or just use one of them to construct forests and jungles of trees, homogeneous ones or heterogenous ones. These methods are recognized by different names which we will try to explain here. These are\n",
+    "1. Voting classifiers\n",
+    "\n",
+    "2. Bagging and Pasting\n",
+    "\n",
+    "3. Random forests\n",
+    "\n",
+    "4. Boosting methods, from adaptive to Extreme Gradient Boosting (XGBoost)\n",
+    "\n",
+    "We discuss these methods here.\n",
+    "\n",
+    "\n",
+    "## An Overview of Ensemble Methods\n",
+    "\n",
+    "\n",
+    "\n",
+    "\n",
+    "
    \n",
+    "\n",
+    "\n",
+    "\n",
+    "\n",
+    "\n",
+    "\n",
+    "\n",
+    "## Bagging\n",
+    "\n",
+    "The **plain** decision trees suffer from high\n",
+    "variance. This means that if we split the training data into two parts\n",
+    "at random, and fit a decision tree to both halves, the results that we\n",
+    "get could be quite different. In contrast, a procedure with low\n",
+    "variance will yield similar results if applied repeatedly to distinct\n",
+    "data sets; linear regression tends to have low variance, if the ratio\n",
+    "of $n$ to $p$ is moderately large. \n",
+    "\n",
+    "**Bootstrap aggregation**, or just **bagging**, is a\n",
+    "general-purpose procedure for reducing the variance of a statistical\n",
+    "learning method. \n",
+    "\n",
+    "\n",
+    "## More bagging\n",
+    "\n",
+    "Bagging typically results in improved accuracy\n",
+    "over prediction using a single tree. Unfortunately, however, it can be\n",
+    "difficult to interpret the resulting model. Recall that one of the\n",
+    "advantages of decision trees is the attractive and easily interpreted\n",
+    "diagram that results.\n",
+    "\n",
+    "However, when we bag a large number of trees, it is no longer\n",
+    "possible to represent the resulting statistical learning procedure\n",
+    "using a single tree, and it is no longer clear which variables are\n",
+    "most important to the procedure. Thus, bagging improves prediction\n",
+    "accuracy at the expense of interpretability.  Although the collection\n",
+    "of bagged trees is much more difficult to interpret than a single\n",
+    "tree, one can obtain an overall summary of the importance of each\n",
+    "predictor using the MSE (for bagging regression trees) or the Gini\n",
+    "index (for bagging classification trees). In the case of bagging\n",
+    "regression trees, we can record the total amount that the MSE is\n",
+    "decreased due to splits over a given predictor, averaged over all $B$ possible\n",
+    "trees. A large value indicates an important predictor. Similarly, in\n",
+    "the context of bagging classification trees, we can add up the total\n",
+    "amount that the Gini index  is decreased by splits over a given\n",
+    "predictor, averaged over all $B$ trees.\n",
+    "\n",
+    "## Simple Voting Example, head or tail"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 14,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "heads_proba = 0.51\n",
+    "coin_tosses = (np.random.rand(10000, 10) < heads_proba).astype(np.int32)\n",
+    "cumulative_heads_ratio = np.cumsum(coin_tosses, axis=0) / np.arange(1, 10001).reshape(-1, 1)\n",
+    "plt.figure(figsize=(8,3.5))\n",
+    "plt.plot(cumulative_heads_ratio)\n",
+    "plt.plot([0, 10000], [0.51, 0.51], \"k--\", linewidth=2, label=\"51%\")\n",
+    "plt.plot([0, 10000], [0.5, 0.5], \"k-\", label=\"50%\")\n",
+    "plt.xlabel(\"Number of coin tosses\")\n",
+    "plt.ylabel(\"Heads ratio\")\n",
+    "plt.legend(loc=\"lower right\")\n",
+    "plt.axis([0, 10000, 0.42, 0.58])\n",
+    "plt.show()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Using the Voting Classifier"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 15,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "from sklearn.model_selection import train_test_split\n",
+    "from sklearn.datasets import make_moons\n",
+    "\n",
+    "X, y = make_moons(n_samples=500, noise=0.30, random_state=42)\n",
+    "X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=42)\n",
+    "\n",
+    "from sklearn.ensemble import RandomForestClassifier\n",
+    "from sklearn.ensemble import VotingClassifier\n",
+    "from sklearn.linear_model import LogisticRegression\n",
+    "from sklearn.svm import SVC\n",
+    "\n",
+    "log_clf = LogisticRegression(solver=\"liblinear\", random_state=42)\n",
+    "rnd_clf = RandomForestClassifier(n_estimators=10, random_state=42)\n",
+    "svm_clf = SVC(gamma=\"auto\", random_state=42)\n",
+    "\n",
+    "voting_clf = VotingClassifier(\n",
+    "    estimators=[('lr', log_clf), ('rf', rnd_clf), ('svc', svm_clf)],\n",
+    "    voting='hard')\n",
+    "\n",
+    "voting_clf.fit(X_train, y_train)\n",
+    "\n",
+    "from sklearn.metrics import accuracy_score\n",
+    "\n",
+    "for clf in (log_clf, rnd_clf, svm_clf, voting_clf):\n",
+    "    clf.fit(X_train, y_train)\n",
+    "    y_pred = clf.predict(X_test)\n",
+    "    print(clf.__class__.__name__, accuracy_score(y_test, y_pred))\n",
+    "\n",
+    "log_clf = LogisticRegression(solver=\"liblinear\", random_state=42)\n",
+    "rnd_clf = RandomForestClassifier(n_estimators=10, random_state=42)\n",
+    "svm_clf = SVC(gamma=\"auto\", probability=True, random_state=42)\n",
+    "\n",
+    "voting_clf = VotingClassifier(\n",
+    "    estimators=[('lr', log_clf), ('rf', rnd_clf), ('svc', svm_clf)],\n",
+    "    voting='soft')\n",
+    "voting_clf.fit(X_train, y_train)\n",
+    "\n",
+    "from sklearn.metrics import accuracy_score\n",
+    "\n",
+    "for clf in (log_clf, rnd_clf, svm_clf, voting_clf):\n",
+    "    clf.fit(X_train, y_train)\n",
+    "    y_pred = clf.predict(X_test)\n",
+    "    print(clf.__class__.__name__, accuracy_score(y_test, y_pred))"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Please, not the moons again! Voting and Bagging"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 16,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "from sklearn.model_selection import train_test_split\n",
+    "from sklearn.datasets import make_moons\n",
+    "\n",
+    "X, y = make_moons(n_samples=500, noise=0.30, random_state=42)\n",
+    "X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=42)\n",
+    "from sklearn.ensemble import RandomForestClassifier\n",
+    "from sklearn.ensemble import VotingClassifier\n",
+    "from sklearn.linear_model import LogisticRegression\n",
+    "from sklearn.svm import SVC\n",
+    "\n",
+    "log_clf = LogisticRegression(random_state=42)\n",
+    "rnd_clf = RandomForestClassifier(random_state=42)\n",
+    "svm_clf = SVC(random_state=42)\n",
+    "\n",
+    "voting_clf = VotingClassifier(\n",
+    "    estimators=[('lr', log_clf), ('rf', rnd_clf), ('svc', svm_clf)],\n",
+    "    voting='hard')\n",
+    "voting_clf.fit(X_train, y_train)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 17,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "from sklearn.metrics import accuracy_score\n",
+    "\n",
+    "for clf in (log_clf, rnd_clf, svm_clf, voting_clf):\n",
+    "    clf.fit(X_train, y_train)\n",
+    "    y_pred = clf.predict(X_test)\n",
+    "    print(clf.__class__.__name__, accuracy_score(y_test, y_pred))"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 18,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "log_clf = LogisticRegression(random_state=42)\n",
+    "rnd_clf = RandomForestClassifier(random_state=42)\n",
+    "svm_clf = SVC(probability=True, random_state=42)\n",
+    "\n",
+    "voting_clf = VotingClassifier(\n",
+    "    estimators=[('lr', log_clf), ('rf', rnd_clf), ('svc', svm_clf)],\n",
+    "    voting='soft')\n",
+    "voting_clf.fit(X_train, y_train)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 19,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "from sklearn.metrics import accuracy_score\n",
+    "\n",
+    "for clf in (log_clf, rnd_clf, svm_clf, voting_clf):\n",
+    "    clf.fit(X_train, y_train)\n",
+    "    y_pred = clf.predict(X_test)\n",
+    "    print(clf.__class__.__name__, accuracy_score(y_test, y_pred))"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Now Bagging"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 20,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "from sklearn.ensemble import BaggingClassifier\n",
+    "from sklearn.tree import DecisionTreeClassifier\n",
+    "\n",
+    "bag_clf = BaggingClassifier(\n",
+    "    DecisionTreeClassifier(random_state=42), n_estimators=500,\n",
+    "    max_samples=100, bootstrap=True, n_jobs=-1, random_state=42)\n",
+    "bag_clf.fit(X_train, y_train)\n",
+    "y_pred = bag_clf.predict(X_test)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 21,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "from sklearn.metrics import accuracy_score\n",
+    "print(accuracy_score(y_test, y_pred))"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 22,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "tree_clf = DecisionTreeClassifier(random_state=42)\n",
+    "tree_clf.fit(X_train, y_train)\n",
+    "y_pred_tree = tree_clf.predict(X_test)\n",
+    "print(accuracy_score(y_test, y_pred_tree))"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 23,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "from matplotlib.colors import ListedColormap\n",
+    "\n",
+    "def plot_decision_boundary(clf, X, y, axes=[-1.5, 2.5, -1, 1.5], alpha=0.5, contour=True):\n",
+    "    x1s = np.linspace(axes[0], axes[1], 100)\n",
+    "    x2s = np.linspace(axes[2], axes[3], 100)\n",
+    "    x1, x2 = np.meshgrid(x1s, x2s)\n",
+    "    X_new = np.c_[x1.ravel(), x2.ravel()]\n",
+    "    y_pred = clf.predict(X_new).reshape(x1.shape)\n",
+    "    custom_cmap = ListedColormap(['#fafab0','#9898ff','#a0faa0'])\n",
+    "    plt.contourf(x1, x2, y_pred, alpha=0.3, cmap=custom_cmap)\n",
+    "    if contour:\n",
+    "        custom_cmap2 = ListedColormap(['#7d7d58','#4c4c7f','#507d50'])\n",
+    "        plt.contour(x1, x2, y_pred, cmap=custom_cmap2, alpha=0.8)\n",
+    "    plt.plot(X[:, 0][y==0], X[:, 1][y==0], \"yo\", alpha=alpha)\n",
+    "    plt.plot(X[:, 0][y==1], X[:, 1][y==1], \"bs\", alpha=alpha)\n",
+    "    plt.axis(axes)\n",
+    "    plt.xlabel(r\"$x_1$\", fontsize=18)\n",
+    "    plt.ylabel(r\"$x_2$\", fontsize=18, rotation=0)\n",
+    "plt.figure(figsize=(11,4))\n",
+    "plt.subplot(121)\n",
+    "plot_decision_boundary(tree_clf, X, y)\n",
+    "plt.title(\"Decision Tree\", fontsize=14)\n",
+    "plt.subplot(122)\n",
+    "plot_decision_boundary(bag_clf, X, y)\n",
+    "plt.title(\"Decision Trees with Bagging\", fontsize=14)\n",
+    "plt.show()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Making our own Bagging with Bootstrap"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 24,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "import matplotlib.pyplot as plt\n",
+    "import numpy as np\n",
+    "from sklearn.model_selection import train_test_split\n",
+    "from sklearn.pipeline import make_pipeline\n",
+    "from sklearn.utils import resample\n",
+    "from sklearn.tree import DecisionTreeRegressor\n",
+    "\n",
+    "\n",
+    "np.random.seed(2018)\n",
+    "\n",
+    "n = 40\n",
+    "n_boostraps = 100\n",
+    "maxdegree = 14\n",
+    "\n",
+    "# Make data set.\n",
+    "x = np.linspace(-3, 3, n).reshape(-1, 1)\n",
+    "y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)\n",
+    "error = np.zeros(maxdegree)\n",
+    "bias = np.zeros(maxdegree)\n",
+    "variance = np.zeros(maxdegree)\n",
+    "polydegree = np.zeros(maxdegree)\n",
+    "X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)\n",
+    "\n",
+    "from sklearn.preprocessing import StandardScaler\n",
+    "scaler = StandardScaler()\n",
+    "scaler.fit(X_train)\n",
+    "X_train_scaled = scaler.transform(X_train)\n",
+    "X_test_scaled = scaler.transform(X_test)\n",
+    "\n",
+    "for degree in range(maxdegree):\n",
+    "    model = DecisionTreeRegressor(max_depth=5) \n",
+    "    y_pred = np.empty((y_test.shape[0], n_boostraps))\n",
+    "    for i in range(n_boostraps):\n",
+    "        x_, y_ = resample(X_train_scaled, y_train)\n",
+    "        model.fit(x_, y_)\n",
+    "        y_pred[:, i] = model.predict(X_test_scaled).ravel()\n",
+    "\n",
+    "    polydegree[degree] = degree\n",
+    "    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )\n",
+    "    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )\n",
+    "    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )\n",
+    "    print('Polynomial degree:', degree)\n",
+    "    print('Error:', error[degree])\n",
+    "    print('Bias^2:', bias[degree])\n",
+    "    print('Var:', variance[degree])\n",
+    "    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))\n",
+    "\n",
+    "plt.plot(polydegree, error, label='Error')\n",
+    "plt.plot(polydegree, bias, label='bias')\n",
+    "plt.plot(polydegree, variance, label='Variance')\n",
+    "plt.legend()\n",
+    "plt.show()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Changing the Level of the Decision Tree"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 25,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "\n",
+    "import matplotlib.pyplot as plt\n",
+    "import numpy as np\n",
+    "from sklearn.model_selection import train_test_split\n",
+    "from sklearn.pipeline import make_pipeline\n",
+    "from sklearn.utils import resample\n",
+    "from sklearn.tree import DecisionTreeRegressor\n",
+    "\n",
+    "n = 100\n",
+    "n_boostraps = 100\n",
+    "maxdepth = 8\n",
+    "\n",
+    "# Make data set.\n",
+    "x = np.linspace(-3, 3, n).reshape(-1, 1)\n",
+    "y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)\n",
+    "error = np.zeros(maxdepth)\n",
+    "bias = np.zeros(maxdepth)\n",
+    "variance = np.zeros(maxdepth)\n",
+    "polydegree = np.zeros(maxdepth)\n",
+    "X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)\n",
+    "\n",
+    "from sklearn.preprocessing import StandardScaler\n",
+    "scaler = StandardScaler()\n",
+    "scaler.fit(X_train)\n",
+    "X_train_scaled = scaler.transform(X_train)\n",
+    "X_test_scaled = scaler.transform(X_test)\n",
+    "\n",
+    "for degree in range(1,maxdepth):\n",
+    "    model = DecisionTreeRegressor(max_depth=degree) \n",
+    "    y_pred = np.empty((y_test.shape[0], n_boostraps))\n",
+    "    for i in range(n_boostraps):\n",
+    "        x_, y_ = resample(X_train_scaled, y_train)\n",
+    "        model.fit(x_, y_)\n",
+    "        y_pred[:, i] = model.predict(X_test_scaled)#.ravel()\n",
+    "\n",
+    "    polydegree[degree] = degree\n",
+    "    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )\n",
+    "    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )\n",
+    "    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )\n",
+    "    print('Polynomial degree:', degree)\n",
+    "    print('Error:', error[degree])\n",
+    "    print('Bias^2:', bias[degree])\n",
+    "    print('Var:', variance[degree])\n",
+    "    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))\n",
+    "\n",
+    "plt.xlim(1,maxdepth)\n",
+    "plt.plot(polydegree, error, label='Error')\n",
+    "plt.plot(polydegree, bias, label='bias')\n",
+    "plt.plot(polydegree, variance, label='Variance')\n",
+    "plt.legend()\n",
+    "plt.show()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Random forests\n",
+    "\n",
+    "Random forests provide an improvement over bagged trees by way of a\n",
+    "small tweak that decorrelates the trees. \n",
+    "\n",
+    "As in bagging, we build a\n",
+    "number of decision trees on bootstrapped training samples. But when\n",
+    "building these decision trees, each time a split in a tree is\n",
+    "considered, a random sample of $m$ predictors is chosen as split\n",
+    "candidates from the full set of $p$ predictors. The split is allowed to\n",
+    "use only one of those $m$ predictors. \n",
+    "\n",
+    "A fresh sample of $m$ predictors is\n",
+    "taken at each split, and typically we choose"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "$$\n",
+    "m\\approx \\sqrt{p}.\n",
+    "$$"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "In building a random forest, at\n",
+    "each split in the tree, the algorithm is not even allowed to consider\n",
+    "a majority of the available predictors. \n",
+    "\n",
+    "The reason for this is rather clever. Suppose that there is one very\n",
+    "strong predictor in the data set, along with a number of other\n",
+    "moderately strong predictors. Then in the collection of bagged\n",
+    "variable importance random forest trees, most or all of the trees will\n",
+    "use this strong predictor in the top split. Consequently, all of the\n",
+    "bagged trees will look quite similar to each other. Hence the\n",
+    "predictions from the bagged trees will be highly correlated.\n",
+    "Unfortunately, averaging many highly correlated quantities does not\n",
+    "lead to as large of a reduction in variance as averaging many\n",
+    "uncorrelated quanti- ties. In particular, this means that bagging will\n",
+    "not lead to a substantial reduction in variance over a single tree in\n",
+    "this setting.\n",
+    "\n",
+    "\n",
+    "## Random Forest Algorithm\n",
+    "The algorithm described here can be applied to both classification and regression problems.\n",
+    "\n",
+    "We will grow of forest of say $M$ trees.\n",
+    "1. For $m=1:M$ we\n",
+    "\n",
+    "  * Draw a bootstrap sample of from the training data organized in our $\\boldsymbol{X}$ matrix.\n",
+    "\n",
+    "  * We grow then a random forest tree $T_m$ based on the bootstrapped data by repeating the steps outlined till we reach the maximum node size is reached\n",
+    "\n",
+    "1. we select $m \\le p$ varibales at random from the $p$ predictors/features\n",
+    "\n",
+    "2. pick the best split point among the $m$ features using either the CART algorithm or the ID3 for classification and create a new node\n",
+    "\n",
+    "3. split the node into daughter nodes\n",
+    "\n",
+    "\n",
+    "\n",
+    "4. Output then the ensemble of trees $\\{T_m\\}_1^{M}$ and make predictions for either a regression type of problem or a classification type of problem. \n",
+    "\n",
+    "## Bootstrap with Random Forests Instead of a Single Tree, own Bagging"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 26,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "\n",
+    "import matplotlib.pyplot as plt\n",
+    "import numpy as np\n",
+    "from sklearn.model_selection import train_test_split\n",
+    "from sklearn.pipeline import make_pipeline\n",
+    "from sklearn.utils import resample\n",
+    "from sklearn.ensemble import RandomForestRegressor\n",
+    "\n",
+    "np.random.seed(2018)\n",
+    "\n",
+    "n = 100\n",
+    "n_boostraps = 100\n",
+    "maxdegree = 14\n",
+    "\n",
+    "# Make data set.\n",
+    "x = np.linspace(-3, 3, n).reshape(-1, 1)\n",
+    "y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)\n",
+    "error = np.zeros(maxdegree)\n",
+    "bias = np.zeros(maxdegree)\n",
+    "variance = np.zeros(maxdegree)\n",
+    "polydegree = np.zeros(maxdegree)\n",
+    "X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)\n",
+    "\n",
+    "from sklearn.preprocessing import StandardScaler\n",
+    "scaler = StandardScaler()\n",
+    "scaler.fit(X_train)\n",
+    "X_train_scaled = scaler.transform(X_train)\n",
+    "X_test_scaled = scaler.transform(X_test)\n",
+    "\n",
+    "for degree in range(maxdegree):\n",
+    "    model = RandomForestRegressor()\n",
+    "    y_pred = np.empty((y_test.shape[0], n_boostraps))\n",
+    "    for i in range(n_boostraps):\n",
+    "        x_, y_ = resample(X_train_scaled, y_train)\n",
+    "        model.fit(x_, y_.ravel())\n",
+    "        y_pred[:, i] = model.predict(X_test_scaled).ravel()\n",
+    "\n",
+    "    polydegree[degree] = degree\n",
+    "    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )\n",
+    "    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )\n",
+    "    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )\n",
+    "    print('Polynomial degree:', degree)\n",
+    "    print('Error:', error[degree])\n",
+    "    print('Bias^2:', bias[degree])\n",
+    "    print('Var:', variance[degree])\n",
+    "    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))\n",
+    "\n",
+    "plt.plot(polydegree, error, label='Error')\n",
+    "plt.plot(polydegree, bias, label='bias')\n",
+    "plt.plot(polydegree, variance, label='Variance')\n",
+    "plt.legend()\n",
+    "plt.show()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Random Forests Compared with other Methods on the Cancer Data"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 27,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "import matplotlib.pyplot as plt\n",
+    "import numpy as np\n",
+    "from sklearn.model_selection import  train_test_split \n",
+    "from sklearn.datasets import load_breast_cancer\n",
+    "from sklearn.svm import SVC\n",
+    "from sklearn.linear_model import LogisticRegression\n",
+    "from sklearn.tree import DecisionTreeClassifier\n",
+    "\n",
+    "# Load the data\n",
+    "cancer = load_breast_cancer()\n",
+    "\n",
+    "X_train, X_test, y_train, y_test = train_test_split(cancer.data,cancer.target,random_state=0)\n",
+    "print(X_train.shape)\n",
+    "print(X_test.shape)\n",
+    "# Logistic Regression\n",
+    "logreg = LogisticRegression(solver='lbfgs')\n",
+    "logreg.fit(X_train, y_train)\n",
+    "print(\"Test set accuracy with Logistic Regression: {:.2f}\".format(logreg.score(X_test,y_test)))\n",
+    "# Support vector machine\n",
+    "svm = SVC(gamma='auto', C=100)\n",
+    "svm.fit(X_train, y_train)\n",
+    "print(\"Test set accuracy with SVM: {:.2f}\".format(svm.score(X_test,y_test)))\n",
+    "# Decision Trees\n",
+    "deep_tree_clf = DecisionTreeClassifier(max_depth=None)\n",
+    "deep_tree_clf.fit(X_train, y_train)\n",
+    "print(\"Test set accuracy with Decision Trees: {:.2f}\".format(deep_tree_clf.score(X_test,y_test)))\n",
+    "#now scale the data\n",
+    "from sklearn.preprocessing import StandardScaler\n",
+    "scaler = StandardScaler()\n",
+    "scaler.fit(X_train)\n",
+    "X_train_scaled = scaler.transform(X_train)\n",
+    "X_test_scaled = scaler.transform(X_test)\n",
+    "# Logistic Regression\n",
+    "logreg.fit(X_train_scaled, y_train)\n",
+    "print(\"Test set accuracy Logistic Regression with scaled data: {:.2f}\".format(logreg.score(X_test_scaled,y_test)))\n",
+    "# Support Vector Machine\n",
+    "svm.fit(X_train_scaled, y_train)\n",
+    "print(\"Test set accuracy SVM with scaled data: {:.2f}\".format(logreg.score(X_test_scaled,y_test)))\n",
+    "# Decision Trees\n",
+    "deep_tree_clf.fit(X_train_scaled, y_train)\n",
+    "print(\"Test set accuracy with Decision Trees and scaled data: {:.2f}\".format(deep_tree_clf.score(X_test_scaled,y_test)))\n",
+    "\n",
+    "\n",
+    "from sklearn.ensemble import RandomForestClassifier\n",
+    "from sklearn.preprocessing import LabelEncoder\n",
+    "from sklearn.model_selection import cross_validate\n",
+    "# Data set not specificied\n",
+    "#Instantiate the model with 500 trees and entropy as splitting criteria\n",
+    "Random_Forest_model = RandomForestClassifier(n_estimators=500,criterion=\"entropy\")\n",
+    "Random_Forest_model.fit(X_train_scaled, y_train)\n",
+    "#Cross validation\n",
+    "accuracy = cross_validate(Random_Forest_model,X_test_scaled,y_test,cv=10)['test_score']\n",
+    "print(accuracy)\n",
+    "print(\"Test set accuracy with Random Forests and scaled data: {:.2f}\".format(Random_Forest_model.score(X_test_scaled,y_test)))\n",
+    "\n",
+    "\n",
+    "import scikitplot as skplt\n",
+    "y_pred = Random_Forest_model.predict(X_test_scaled)\n",
+    "skplt.metrics.plot_confusion_matrix(y_test, y_pred, normalize=True)\n",
+    "plt.show()\n",
+    "y_probas = Random_Forest_model.predict_proba(X_test_scaled)\n",
+    "skplt.metrics.plot_roc(y_test, y_probas)\n",
+    "plt.show()\n",
+    "skplt.metrics.plot_cumulative_gain(y_test, y_probas)\n",
+    "plt.show()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Compare  Bagging on Trees with Random Forests"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 28,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "bag_clf = BaggingClassifier(\n",
+    "    DecisionTreeClassifier(splitter=\"random\", max_leaf_nodes=16, random_state=42),\n",
+    "    n_estimators=500, max_samples=1.0, bootstrap=True, n_jobs=-1, random_state=42)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 29,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "bag_clf.fit(X_train, y_train)\n",
+    "y_pred = bag_clf.predict(X_test)\n",
+    "from sklearn.ensemble import RandomForestClassifier\n",
+    "rnd_clf = RandomForestClassifier(n_estimators=500, max_leaf_nodes=16, n_jobs=-1, random_state=42)\n",
+    "rnd_clf.fit(X_train, y_train)\n",
+    "y_pred_rf = rnd_clf.predict(X_test)\n",
+    "np.sum(y_pred == y_pred_rf) / len(y_pred)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Boosting, a Bird'e Eye\n",
+    "\n",
+    "The basic idea is to combine weak classifiers in order to create a good\n",
+    "classifier. With a weak classifier we often intend a classifier which\n",
+    "produces results which are only slightly better than we would get by\n",
+    "random guesses.\n",
+    "\n",
+    "This is done by applying in an iterative way a weak (or a standard\n",
+    "classifier like decision trees) to modify the data. In each iteration\n",
+    "we emphasize those observations which are misclassified by weighting\n",
+    "them with a factor.\n",
+    "\n",
+    "\n",
+    "## Adaptive boosting: AdaBoost, Basic Algorithm\n",
+    "\n",
+    "The algorithm here is rather straightforward. Assume that our weak\n",
+    "classifier is a decision tree and we consider a binary set of outputs\n",
+    "with $y_i \\in \\{-1,1\\}$ and $i=0,1,2,\\dots,n-1$ as our set of\n",
+    "observations. Our design matrix is given in terms of the\n",
+    "feature/predictor vectors\n",
+    "$\\boldsymbol{X}=[\\boldsymbol{x}_0\\boldsymbol{x}_1\\dots\\boldsymbol{x}_{p-1}$. Finally, we define also a\n",
+    "classifier determined by our data via a function $G(\\boldsymbol{X})$. This function tells us how well we are able to classify our outputs/targets $\\boldsymbol{y}$. \n",
+    "\n",
+    "We can then define the misclassification error $\\mathrm{err}$ as"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "$$\n",
+    "\\mathrm{err}=\\frac{1}{n}\\sum_{i=0}^{n-1}I(y_i\\ne G(\\boldsymbol{X}_{i*}),\n",
+    "$$"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "where the function $I()$ is one if we misclassify and zero if we classify correctly. \n",
+    "\n",
+    "## Basic Steps of AdaBoost\n",
+    "\n",
+    "With the above definitions we are now ready to set up the algorithm for AdaBoost.\n",
+    "The basic idea is to set up weights which will be used to scale the correctly classified and the misclassified cases.\n",
+    "1. We start by initializing all weights to $w_i = 1/n$, with $i=0,1,2,\\dots n-1$. It is to see then that $\\sum_{i=0}^{n-1}w_i = 1$.\n",
+    "\n",
+    "2. We rewrite the misclassification error as"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "$$\n",
+    "\\mathrm{err}=\\frac{\\sum_{i=0}^{n-1}w_iI(y_i\\ne G(\\boldsymbol{X}_{i*})}{\\sum_{i=0}^{n-1}w_i},\n",
+    "$$"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "1. Then we start looping over all attempts at classifying, namely we start an iterative process for $m=1:M$, where $M$ is the final number of classifications. Our given classifier could for example be a plain decision tree.\n",
+    "\n",
+    "a. Fit then a given classifier to the training using the weights $w_i$.\n",
+    "\n",
+    "b. Compute then $\\mathrm{err}$ and figure out which events are classified properly and which are classified wrongly.\n",
+    "\n",
+    "c. Define a quantity $\\alpha_{m} = \\log{(1-\\mathrm{err})/\\mathrm{err}}\n",
+    "\n",
+    "d. Set the new weights to $w_i = w_i\\times \\exp{(\\alpha_m I(y_i\\ne G(\\boldsymbol{X}_{i*})}.\n",
+    "\n",
+    "\n",
+    "5. Compute the new classifier $G(\\boldsymbol{X})= \\sum_{i=0}^{n-1}\\alpha_m I(y_i\\ne G(\\boldsymbol{X}_{i*}).\n",
+    "\n",
+    "For the iterations with $m \\le 2$ the weights are modified\n",
+    "individually at each steps. The obersvations which were misclassified\n",
+    "at iteration $m-1$ have a weight which is larger than those which were\n",
+    "classified properly. As this proceeds, the observations which were\n",
+    "difficult to classifiy correctly are given a larger influence. Each\n",
+    "new classification step $m$ is then forced to concentrate on those\n",
+    "observations that are missed in the previous iterations.\n",
+    "\n",
+    "\n",
+    "\n",
+    "## AdaBoost Examples\n",
+    "\n",
+    "Using **Scikit-Learn** it is easy to appply the adaptive boosting algorithm, as done here."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 30,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "from sklearn.ensemble import AdaBoostClassifier\n",
+    "\n",
+    "ada_clf = AdaBoostClassifier(\n",
+    "    DecisionTreeClassifier(max_depth=1), n_estimators=200,\n",
+    "    algorithm=\"SAMME.R\", learning_rate=0.5, random_state=42)\n",
+    "ada_clf.fit(X_train, y_train)\n",
+    "\n",
+    "plot_decision_boundary(ada_clf, X, y)\n",
+    "\n",
+    "m = len(X_train)\n",
+    "\n",
+    "plt.figure(figsize=(11, 4))\n",
+    "for subplot, learning_rate in ((121, 1), (122, 0.5)):\n",
+    "    sample_weights = np.ones(m)\n",
+    "    plt.subplot(subplot)\n",
+    "    for i in range(5):\n",
+    "        svm_clf = SVC(kernel=\"rbf\", C=0.05, gamma=\"auto\", random_state=42)\n",
+    "        svm_clf.fit(X_train, y_train, sample_weight=sample_weights)\n",
+    "        y_pred = svm_clf.predict(X_train)\n",
+    "        sample_weights[y_pred != y_train] *= (1 + learning_rate)\n",
+    "        plot_decision_boundary(svm_clf, X, y, alpha=0.2)\n",
+    "        plt.title(\"learning_rate = {}\".format(learning_rate), fontsize=16)\n",
+    "    if subplot == 121:\n",
+    "        plt.text(-0.7, -0.65, \"1\", fontsize=14)\n",
+    "        plt.text(-0.6, -0.10, \"2\", fontsize=14)\n",
+    "        plt.text(-0.5,  0.10, \"3\", fontsize=14)\n",
+    "        plt.text(-0.4,  0.55, \"4\", fontsize=14)\n",
+    "        plt.text(-0.3,  0.90, \"5\", fontsize=14)\n",
+    "\n",
+    "save_fig(\"boosting_plot\")\n",
+    "plt.show()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Gradient boosting: Basics\n",
+    "\n",
+    "Gradient boosting is again a similar technique to Adapative boosting,\n",
+    "it combines so-called weak classifiers or regressors into a strong\n",
+    "method via a series of iterations.\n",
+    "\n",
+    "In order to understand the method, let us illustrate its basics by\n",
+    "bringing back the essential steps in linear regression, where our cost\n",
+    "function was the least squares function.\n",
+    "\n",
+    "## Gradient Boosting, algorithm\n",
+    "\n",
+    "Suppose we have a cost function $C(f)=\\sum_{i=0}^{n-1}L(y_i, f(x_i))$ where $y_i$ is our target and $f(x_i)$ the function which is meant to model $y_i$. The above cost function could be our standard least squares  function"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "$$\n",
+    "C(\\boldsymbol{y},\\boldsymbol{f})=\\frac{1}{n}\\sum_{i=0}^{n-1}(y_i-f(x_i))^2.\n",
+    "$$"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "The way we proceed in an iterative fashion is to\n",
+    "1. Initialize our estimate by $f_0(x)=0$.\n",
+    "\n",
+    "2. For $m=1:M$, we\n",
+    "\n",
+    "a. compute the negative gradient vector $\\boldsymbol{u}_m = -\\partial C(\\boldsymbol{y},\\boldsymbol{f})/\\partial \\boldsymbol{f}(x)$ at $f(x) = f_{m-1}(x);\n",
+    "\n",
+    "b. fit the so-called base-learner to the negative gradient $h_m(u_m,x)$;\n",
+    "\n",
+    "c. update the estimate $f_m(x) = f_{m-1}(x)+\\nu h_m(u_m,x)$;\n",
+    "\n",
+    "\n",
+    "4. The final estimate is then $f_M(x) = \\sum_{m=1}^M\\nu h_m(u_m,x)$.\n",
+    "\n",
+    "## Gradient Boosting, Examples"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 31,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "np.random.seed(42)\n",
+    "X = np.random.rand(100, 1) - 0.5\n",
+    "y = 3*X[:, 0]**2 + 0.05 * np.random.randn(100)\n",
+    "\n",
+    "from sklearn.tree import DecisionTreeRegressor\n",
+    "\n",
+    "tree_reg1 = DecisionTreeRegressor(max_depth=2, random_state=42)\n",
+    "tree_reg1.fit(X, y)\n",
+    "\n",
+    "y2 = y - tree_reg1.predict(X)\n",
+    "tree_reg2 = DecisionTreeRegressor(max_depth=2, random_state=42)\n",
+    "tree_reg2.fit(X, y2)\n",
+    "\n",
+    "y3 = y2 - tree_reg2.predict(X)\n",
+    "tree_reg3 = DecisionTreeRegressor(max_depth=2, random_state=42)\n",
+    "tree_reg3.fit(X, y3)\n",
+    "\n",
+    "X_new = np.array([[0.8]])\n",
+    "y_pred = sum(tree.predict(X_new) for tree in (tree_reg1, tree_reg2, tree_reg3))\n",
+    "\n",
+    "def plot_predictions(regressors, X, y, axes, label=None, style=\"r-\", data_style=\"b.\", data_label=None):\n",
+    "    x1 = np.linspace(axes[0], axes[1], 500)\n",
+    "    y_pred = sum(regressor.predict(x1.reshape(-1, 1)) for regressor in regressors)\n",
+    "    plt.plot(X[:, 0], y, data_style, label=data_label)\n",
+    "    plt.plot(x1, y_pred, style, linewidth=2, label=label)\n",
+    "    if label or data_label:\n",
+    "        plt.legend(loc=\"upper center\", fontsize=16)\n",
+    "    plt.axis(axes)\n",
+    "\n",
+    "plt.figure(figsize=(11,11))\n",
+    "\n",
+    "plt.subplot(321)\n",
+    "plot_predictions([tree_reg1], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label=\"$h_1(x_1)$\", style=\"g-\", data_label=\"Training set\")\n",
+    "plt.ylabel(\"$y$\", fontsize=16, rotation=0)\n",
+    "plt.title(\"Residuals and tree predictions\", fontsize=16)\n",
+    "\n",
+    "plt.subplot(322)\n",
+    "plot_predictions([tree_reg1], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label=\"$h(x_1) = h_1(x_1)$\", data_label=\"Training set\")\n",
+    "plt.ylabel(\"$y$\", fontsize=16, rotation=0)\n",
+    "plt.title(\"Ensemble predictions\", fontsize=16)\n",
+    "\n",
+    "plt.subplot(323)\n",
+    "plot_predictions([tree_reg2], X, y2, axes=[-0.5, 0.5, -0.5, 0.5], label=\"$h_2(x_1)$\", style=\"g-\", data_style=\"k+\", data_label=\"Residuals\")\n",
+    "plt.ylabel(\"$y - h_1(x_1)$\", fontsize=16)\n",
+    "\n",
+    "plt.subplot(324)\n",
+    "plot_predictions([tree_reg1, tree_reg2], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label=\"$h(x_1) = h_1(x_1) + h_2(x_1)$\")\n",
+    "plt.ylabel(\"$y$\", fontsize=16, rotation=0)\n",
+    "\n",
+    "plt.subplot(325)\n",
+    "plot_predictions([tree_reg3], X, y3, axes=[-0.5, 0.5, -0.5, 0.5], label=\"$h_3(x_1)$\", style=\"g-\", data_style=\"k+\")\n",
+    "plt.ylabel(\"$y - h_1(x_1) - h_2(x_1)$\", fontsize=16)\n",
+    "plt.xlabel(\"$x_1$\", fontsize=16)\n",
+    "\n",
+    "plt.subplot(326)\n",
+    "plot_predictions([tree_reg1, tree_reg2, tree_reg3], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label=\"$h(x_1) = h_1(x_1) + h_2(x_1) + h_3(x_1)$\")\n",
+    "plt.xlabel(\"$x_1$\", fontsize=16)\n",
+    "plt.ylabel(\"$y$\", fontsize=16, rotation=0)\n",
+    "\n",
+    "save_fig(\"gradient_boosting_plot\")\n",
+    "plt.show()\n",
+    "\n",
+    "from sklearn.ensemble import GradientBoostingRegressor\n",
+    "\n",
+    "gbrt = GradientBoostingRegressor(max_depth=2, n_estimators=3, learning_rate=1.0, random_state=42)\n",
+    "gbrt.fit(X, y)\n",
+    "\n",
+    "gbrt_slow = GradientBoostingRegressor(max_depth=2, n_estimators=200, learning_rate=0.1, random_state=42)\n",
+    "gbrt_slow.fit(X, y)\n",
+    "\n",
+    "plt.figure(figsize=(11,4))\n",
+    "\n",
+    "plt.subplot(121)\n",
+    "plot_predictions([gbrt], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label=\"Ensemble predictions\")\n",
+    "plt.title(\"learning_rate={}, n_estimators={}\".format(gbrt.learning_rate, gbrt.n_estimators), fontsize=14)\n",
+    "\n",
+    "plt.subplot(122)\n",
+    "plot_predictions([gbrt_slow], X, y, axes=[-0.5, 0.5, -0.1, 0.8])\n",
+    "plt.title(\"learning_rate={}, n_estimators={}\".format(gbrt_slow.learning_rate, gbrt_slow.n_estimators), fontsize=14)\n",
+    "\n",
+    "save_fig(\"gbrt_learning_rate_plot\")\n",
+    "plt.show()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Gradient Boots with Early Stopping"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 32,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "\n",
+    "from sklearn.model_selection import train_test_split\n",
+    "from sklearn.metrics import mean_squared_error\n",
+    "\n",
+    "X_train, X_val, y_train, y_val = train_test_split(X, y, random_state=49)\n",
+    "\n",
+    "gbrt = GradientBoostingRegressor(max_depth=2, n_estimators=120, random_state=42)\n",
+    "gbrt.fit(X_train, y_train)\n",
+    "\n",
+    "errors = [mean_squared_error(y_val, y_pred)\n",
+    "          for y_pred in gbrt.staged_predict(X_val)]\n",
+    "bst_n_estimators = np.argmin(errors) + 1\n",
+    "\n",
+    "gbrt_best = GradientBoostingRegressor(max_depth=2,n_estimators=bst_n_estimators, random_state=42)\n",
+    "gbrt_best.fit(X_train, y_train)\n",
+    "\n",
+    "min_error = np.min(errors)\n",
+    "plt.figure(figsize=(11, 4))\n",
+    "\n",
+    "plt.subplot(121)\n",
+    "plt.plot(errors, \"b.-\")\n",
+    "plt.plot([bst_n_estimators, bst_n_estimators], [0, min_error], \"k--\")\n",
+    "plt.plot([0, 120], [min_error, min_error], \"k--\")\n",
+    "plt.plot(bst_n_estimators, min_error, \"ko\")\n",
+    "plt.text(bst_n_estimators, min_error*1.2, \"Minimum\", ha=\"center\", fontsize=14)\n",
+    "plt.axis([0, 120, 0, 0.01])\n",
+    "plt.xlabel(\"Number of trees\")\n",
+    "plt.title(\"Validation error\", fontsize=14)\n",
+    "\n",
+    "plt.subplot(122)\n",
+    "plot_predictions([gbrt_best], X, y, axes=[-0.5, 0.5, -0.1, 0.8])\n",
+    "plt.title(\"Best model (%d trees)\" % bst_n_estimators, fontsize=14)\n",
+    "\n",
+    "save_fig(\"early_stopping_gbrt_plot\")\n",
+    "plt.show()\n",
+    "\n",
+    "\n",
+    "gbrt = GradientBoostingRegressor(max_depth=2, warm_start=True, random_state=42)\n",
+    "\n",
+    "min_val_error = float(\"inf\")\n",
+    "error_going_up = 0\n",
+    "for n_estimators in range(1, 120):\n",
+    "    gbrt.n_estimators = n_estimators\n",
+    "    gbrt.fit(X_train, y_train)\n",
+    "    y_pred = gbrt.predict(X_val)\n",
+    "    val_error = mean_squared_error(y_val, y_pred)\n",
+    "    if val_error < min_val_error:\n",
+    "        min_val_error = val_error\n",
+    "        error_going_up = 0\n",
+    "    else:\n",
+    "        error_going_up += 1\n",
+    "        if error_going_up == 5:\n",
+    "            break  # early stopping\n",
+    "\n",
+    "\n",
+    "print(gbrt.n_estimators)\n",
+    "print(\"Minimum validation MSE:\", min_val_error)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## XGBoost: Extreme Gradient Boosting\n",
+    "\n",
+    "\n",
+    "[XGBoost](https://github.com/dmlc/xgboost) or Extreme Gradient\n",
+    "Boosting, is an optimized distributed gradient boosting library\n",
+    "designed to be highly efficient, flexible and portable. It implements\n",
+    "machine learning algorithms under the Gradient Boosting\n",
+    "framework. XGBoost provides a parallel tree boosting that solve many\n",
+    "data science problems in a fast and accurate way. See the [article by Chen and Guestrin](https://arxiv.org/abs/1603.02754).\n",
+    "\n",
+    "The authors design and build a highly scalable end-to-end tree\n",
+    "boosting system. It has  a theoretically justified weighted quantile\n",
+    "sketch for efficient proposal calculation. It introduces a novel sparsity-aware algorithm for parallel tree learning and an effective cache-aware block structure for out-of-core tree learning.\n",
+    "\n",
+    "It is now the algorithm which wins essentially all ML competitions!!!\n",
+    "\n",
+    "## Regression Case"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 33,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "import matplotlib.pyplot as plt\n",
+    "import numpy as np\n",
+    "from sklearn.model_selection import train_test_split\n",
+    "import xgboost as xgb\n",
+    "from sklearn.preprocessing import StandardScaler\n",
+    "import scikitplot as skplt\n",
+    "from sklearn.metrics import mean_squared_error\n",
+    "\n",
+    "n = 100\n",
+    "maxdegree = 6\n",
+    "\n",
+    "# Make data set.\n",
+    "x = np.linspace(-3, 3, n).reshape(-1, 1)\n",
+    "y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)\n",
+    "\n",
+    "error = np.zeros(maxdegree)\n",
+    "bias = np.zeros(maxdegree)\n",
+    "variance = np.zeros(maxdegree)\n",
+    "polydegree = np.zeros(maxdegree)\n",
+    "X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)\n",
+    "scaler = StandardScaler()\n",
+    "scaler.fit(X_train)\n",
+    "X_train_scaled = scaler.transform(X_train)\n",
+    "X_test_scaled = scaler.transform(X_test)\n",
+    "\n",
+    "for degree in range(maxdegree):\n",
+    "    model =  xgb.XGBRegressor(objective ='reg:squarederror', colsample_bytree = 0.3, learning_rate = 0.1,\n",
+    "                max_depth = degree, alpha = 10, n_estimators = 10)\n",
+    "    model.fit(X_train_scaled,y_train)\n",
+    "    y_pred = model.predict(X_test_scaled)\n",
+    "    polydegree[degree] = degree\n",
+    "    error[degree] = np.mean( np.mean((y_test - y_pred)**2) )\n",
+    "    bias[degree] = np.mean( (y_test - np.mean(y_pred))**2 )\n",
+    "    variance[degree] = np.mean( np.var(y_pred) )\n",
+    "    print('Max depth:', degree)\n",
+    "    print('Error:', error[degree])\n",
+    "    print('Bias^2:', bias[degree])\n",
+    "    print('Var:', variance[degree])\n",
+    "    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))\n",
+    "\n",
+    "plt.xlim(1,maxdegree-1)\n",
+    "plt.plot(polydegree, error, label='Error')\n",
+    "plt.plot(polydegree, bias, label='bias')\n",
+    "plt.plot(polydegree, variance, label='Variance')\n",
+    "plt.legend()\n",
+    "plt.show()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Xgboost on the Cancer Data"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 34,
+   "metadata": {
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": [
+    "import matplotlib.pyplot as plt\n",
+    "import numpy as np\n",
+    "from sklearn.model_selection import  train_test_split \n",
+    "from sklearn.datasets import load_breast_cancer\n",
+    "from sklearn.preprocessing import LabelEncoder\n",
+    "from sklearn.model_selection import cross_validate\n",
+    "import scikitplot as skplt\n",
+    "import xgboost as xgb\n",
+    "# Load the data\n",
+    "cancer = load_breast_cancer()\n",
+    "\n",
+    "X_train, X_test, y_train, y_test = train_test_split(cancer.data,cancer.target,random_state=0)\n",
+    "print(X_train.shape)\n",
+    "print(X_test.shape)\n",
+    "#now scale the data\n",
+    "from sklearn.preprocessing import StandardScaler\n",
+    "scaler = StandardScaler()\n",
+    "scaler.fit(X_train)\n",
+    "X_train_scaled = scaler.transform(X_train)\n",
+    "X_test_scaled = scaler.transform(X_test)\n",
+    "\n",
+    "xg_clf = xgb.XGBClassifier()\n",
+    "xg_clf.fit(X_train_scaled,y_train)\n",
+    "xgb.plot_tree(xg_clf,num_trees=0)\n",
+    "plt.rcParams['figure.figsize'] = [50, 10]\n",
+    "plt.show()\n",
+    "xgb.plot_importance(xg_clf)\n",
+    "plt.rcParams['figure.figsize'] = [5, 5]\n",
+    "plt.show()"
+   ]
+  }
+ ],
+ "metadata": {},
+ "nbformat": 4,
+ "nbformat_minor": 2
+}
diff --git a/doc/src/DecisionTrees/DecisionTrees.p.tex b/doc/src/DecisionTrees/DecisionTrees.p.tex
new file mode 100644
index 000000000..1005fdb09
--- /dev/null
+++ b/doc/src/DecisionTrees/DecisionTrees.p.tex
@@ -0,0 +1,2425 @@
+%%
+%% Automatically generated file from DocOnce source
+%% (https://github.com/hplgit/doconce/)
+%%
+%%
+% #ifdef PTEX2TEX_EXPLANATION
+%%
+%% The file follows the ptex2tex extended LaTeX format, see
+%% ptex2tex: http://code.google.com/p/ptex2tex/
+%%
+%% Run
+%%      ptex2tex myfile
+%% or
+%%      doconce ptex2tex myfile
+%%
+%% to turn myfile.p.tex into an ordinary LaTeX file myfile.tex.
+%% (The ptex2tex program: http://code.google.com/p/ptex2tex)
+%% Many preprocess options can be added to ptex2tex or doconce ptex2tex
+%%
+%%      ptex2tex -DMINTED myfile
+%%      doconce ptex2tex myfile envir=minted
+%%
+%% ptex2tex will typeset code environments according to a global or local
+%% .ptex2tex.cfg configure file. doconce ptex2tex will typeset code
+%% according to options on the command line (just type doconce ptex2tex to
+%% see examples). If doconce ptex2tex has envir=minted, it enables the
+%% minted style without needing -DMINTED.
+% #endif
+
+% #define PREAMBLE
+
+% #ifdef PREAMBLE
+%-------------------- begin preamble ----------------------
+
+\documentclass[%
+oneside,                 % oneside: electronic viewing, twoside: printing
+final,                   % draft: marks overfull hboxes, figures with paths
+10pt]{article}
+
+\listfiles               %  print all files needed to compile this document
+
+\usepackage{relsize,makeidx,color,setspace,amsmath,amsfonts,amssymb}
+\usepackage[table]{xcolor}
+\usepackage{bm,ltablex,microtype}
+
+\usepackage[pdftex]{graphicx}
+
+\usepackage{ptex2tex}
+% #ifdef MINTED
+\usepackage{minted}
+\usemintedstyle{default}
+% #endif
+
+\usepackage[T1]{fontenc}
+%\usepackage[latin1]{inputenc}
+\usepackage{ucs}
+\usepackage[utf8x]{inputenc}
+
+\usepackage{lmodern}         % Latin Modern fonts derived from Computer Modern
+
+% Hyperlinks in PDF:
+\definecolor{linkcolor}{rgb}{0,0,0.4}
+\usepackage{hyperref}
+\hypersetup{
+    breaklinks=true,
+    colorlinks=true,
+    linkcolor=linkcolor,
+    urlcolor=linkcolor,
+    citecolor=black,
+    filecolor=black,
+    %filecolor=blue,
+    pdfmenubar=true,
+    pdftoolbar=true,
+    bookmarksdepth=3   % Uncomment (and tweak) for PDF bookmarks with more levels than the TOC
+    }
+%\hyperbaseurl{}   % hyperlinks are relative to this root
+
+\setcounter{tocdepth}{2}  % levels in table of contents
+
+% Tricks for having figures close to where they are defined:
+% 1. define less restrictive rules for where to put figures
+\setcounter{topnumber}{2}
+\setcounter{bottomnumber}{2}
+\setcounter{totalnumber}{4}
+\renewcommand{\topfraction}{0.95}
+\renewcommand{\bottomfraction}{0.95}
+\renewcommand{\textfraction}{0}
+\renewcommand{\floatpagefraction}{0.75}
+% floatpagefraction must always be less than topfraction!
+% 2. ensure all figures are flushed before next section
+\usepackage[section]{placeins}
+% 3. enable begin{figure}[H] (often leads to ugly pagebreaks)
+%\usepackage{float}\restylefloat{figure}
+
+% --- fancyhdr package for fancy headers ---
+\usepackage{fancyhdr}
+\fancyhf{} % sets both header and footer to nothing
+\renewcommand{\headrulewidth}{0pt}
+\fancyfoot[LE,RO]{\thepage}
+% Ensure copyright on titlepage (article style) and chapter pages (book style)
+\fancypagestyle{plain}{
+  \fancyhf{}
+  \fancyfoot[C]{{\footnotesize \copyright\ 1999-2019, Morten Hjorth-Jensen. Released under CC Attribution-NonCommercial 4.0 license}}
+%  \renewcommand{\footrulewidth}{0mm}
+  \renewcommand{\headrulewidth}{0mm}
+}
+% Ensure copyright on titlepages with \thispagestyle{empty}
+\fancypagestyle{empty}{
+  \fancyhf{}
+  \fancyfoot[C]{{\footnotesize \copyright\ 1999-2019, Morten Hjorth-Jensen. Released under CC Attribution-NonCommercial 4.0 license}}
+  \renewcommand{\footrulewidth}{0mm}
+  \renewcommand{\headrulewidth}{0mm}
+}
+
+\pagestyle{fancy}
+
+
+\usepackage[framemethod=TikZ]{mdframed}
+
+% --- begin definitions of admonition environments ---
+
+% --- end of definitions of admonition environments ---
+
+% prevent orhpans and widows
+\clubpenalty = 10000
+\widowpenalty = 10000
+
+% --- end of standard preamble for documents ---
+
+
+% insert custom LaTeX commands...
+
+\raggedbottom
+\makeindex
+\usepackage[totoc]{idxlayout}   % for index in the toc
+\usepackage[nottoc]{tocbibind}  % for references/bibliography in the toc
+
+%-------------------- end preamble ----------------------
+
+\begin{document}
+
+% matching end for #ifdef PREAMBLE
+% #endif
+
+\newcommand{\exercisesection}[1]{\subsection*{#1}}
+
+
+% ------------------- main content ----------------------
+
+
+
+% ----------------- title -------------------------
+
+\thispagestyle{empty}
+
+\begin{center}
+{\LARGE\bf
+\begin{spacing}{1.25}
+Data Analysis and Machine Learning: From Decision Trees to Forests and all that
+\end{spacing}
+}
+\end{center}
+
+% ----------------- author(s) -------------------------
+
+\begin{center}
+{\bf Morten Hjorth-Jensen${}^{1, 2}$} \\ [0mm]
+\end{center}
+
+\begin{center}
+% List of all institutions:
+\centerline{{\small ${}^1$Department of Physics, University of Oslo}}
+\centerline{{\small ${}^2$Department of Physics and Astronomy and National Superconducting Cyclotron Laboratory, Michigan State University}}
+\end{center}
+    
+% ----------------- end author(s) -------------------------
+
+% --- begin date ---
+\begin{center}
+Nov 7, 2019
+\end{center}
+% --- end date ---
+
+\vspace{1cm}
+
+
+% !split
+\subsection{Decision trees, overarching aims}
+
+
+Decision trees are supervised learning algorithms used for both,
+classification and regression tasks.
+
+
+The main idea of decision trees
+is to find those descriptive features which contain the most
+\textbf{information} regarding the target feature and then split the dataset
+along the values of these features such that the target feature values
+for the resulting underlying datasets are as pure as possible.
+
+The descriptive features which reproduce best the target/output features are normally  said
+to be the most informative ones. The process of finding the \textbf{most
+informative} feature is done until we accomplish a stopping criteria
+where we then finally end up in so called \textbf{leaf nodes}. 
+
+A decision tree is typically divided into a \textbf{root node}, the \textbf{interior nodes},
+and the final \textbf{leaf nodes} or just \textbf{leaves}. These entities are then connected by so-called \textbf{branches}.
+
+The leaf nodes
+contain the predictions we will make for new query instances presented
+to our trained model. This is possible since the model has 
+learned the underlying structure of the training data and hence can,
+given some assumptions, make predictions about the target feature value
+(class) of unseen query instances.
+
+% !split
+\subsection{A typical Decision Tree with its pertinent Jargon, Classification Problem}
+
+
+
+\vspace{6mm}
+
+% inline figure
+\centerline{\includegraphics[width=0.8\linewidth]{DataFiles/cancer.png}}
+
+\vspace{6mm}
+
+
+
+This tree was produced using the Wisconsin cancer data (discussed here as well, see code examples below) using \textbf{Scikit-Learn}'s decision tree classifier. Here we have used the so-called \textbf{gini} index (see below) to split the various branches.
+
+
+
+% !split
+\subsection{General Features}
+
+The overarching approach to decision trees is a top-down approach.
+
+\begin{itemize}
+\item A leaf provides the classification of a given instance.
+
+\item A node specifies a test of some attribute of the instance.
+
+\item A branch corresponds to a possible values of an attribute.
+
+\item An instance is classified by starting at the root node of the tree, testing the attribute specified by this node, then moving down the tree branch corresponding to the value of the attribute in the given example.
+\end{itemize}
+
+\noindent
+This process is then repeated for the subtree rooted at the new
+node.
+
+
+% !split
+\subsection{How do we set it up?}
+
+
+In simplified terms, the process of training a decision tree and
+predicting the target features of query instances is as follows:
+
+\begin{enumerate}
+\item Present a dataset containing of a number of training instances characterized by a number of descriptive features and a target feature
+
+\item Train the decision tree model by continuously splitting the target feature along the values of the descriptive features using a measure of information gain during the training process
+
+\item Grow the tree until we accomplish a stopping criteria create leaf nodes which represent the \emph{predictions} we want to make for new query instances
+
+\item Show query instances to the tree and run down the tree until we arrive at leaf nodes
+\end{enumerate}
+
+\noindent
+Then we are essentially done!
+
+
+
+
+
+% !split
+\subsection{Decision trees and Regression}
+\bpycod
+import numpy as np
+import matplotlib.pyplot as plt
+from sklearn.preprocessing import PolynomialFeatures
+from sklearn.linear_model import LinearRegression
+
+steps=250
+
+distance=0
+x=0
+distance_list=[]
+steps_list=[]
+while x 0 and bestAttrId in attributeIds:
+					toRemove = attributeIds.index(bestAttrId)
+					attributeIds.pop(toRemove)
+				child.next = self.id3Recv(
+					childSampleIds, attributeIds, child.next)
+		return root
+
+	def printTree(self):
+		if self.root:
+			roots = deque()
+			roots.append(self.root)
+			while len(roots) > 0:
+				root = roots.popleft()
+				print(root.value)
+				if root.childs:
+					for child in root.childs:
+						print('({})'.format(child.value))
+						roots.append(child.next)
+				elif root.next:
+					print(root.next)
+
+
+def test():
+	f = open('DataFiles/rideclass.csv')
+	attributes = f.readline().split(',')
+	attributes = attributes[1:len(attributes)-1]
+	print(attributes)
+	sample = f.readlines()
+	f.close()
+	for i in range(len(sample)):
+		sample[i] = re.sub('\d+,', '', sample[i])
+		sample[i] = sample[i].strip().split(',')
+	labels = []
+	for s in sample:
+		labels.append(s.pop())
+	# print(sample)
+	# print(labels)
+	decisionTree = DecisionTree(sample, attributes, labels)
+	print("System entropy {}".format(decisionTree.entropy))
+	decisionTree.id3()
+	decisionTree.printTree()
+
+
+if __name__ == '__main__':
+	test()
+\epycod
+
+% !split
+\subsection{Cancer Data again now with Decision Trees and other Methods}
+\bpycod
+import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import  train_test_split 
+from sklearn.datasets import load_breast_cancer
+from sklearn.svm import SVC
+from sklearn.linear_model import LogisticRegression
+from sklearn.tree import DecisionTreeClassifier
+
+# Load the data
+cancer = load_breast_cancer()
+
+X_train, X_test, y_train, y_test = train_test_split(cancer.data,cancer.target,random_state=0)
+print(X_train.shape)
+print(X_test.shape)
+# Logistic Regression
+logreg = LogisticRegression(solver='lbfgs')
+logreg.fit(X_train, y_train)
+print("Test set accuracy with Logistic Regression: {:.2f}".format(logreg.score(X_test,y_test)))
+# Support vector machine
+svm = SVC(gamma='auto', C=100)
+svm.fit(X_train, y_train)
+print("Test set accuracy with SVM: {:.2f}".format(svm.score(X_test,y_test)))
+# Decision Trees
+deep_tree_clf = DecisionTreeClassifier(max_depth=None)
+deep_tree_clf.fit(X_train, y_train)
+print("Test set accuracy with Decision Trees: {:.2f}".format(deep_tree_clf.score(X_test,y_test)))
+#now scale the data
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+# Logistic Regression
+logreg.fit(X_train_scaled, y_train)
+print("Test set accuracy Logistic Regression with scaled data: {:.2f}".format(logreg.score(X_test_scaled,y_test)))
+# Support Vector Machine
+svm.fit(X_train_scaled, y_train)
+print("Test set accuracy SVM with scaled data: {:.2f}".format(logreg.score(X_test_scaled,y_test)))
+# Decision Trees
+deep_tree_clf.fit(X_train_scaled, y_train)
+print("Test set accuracy with Decision Trees and scaled data: {:.2f}".format(deep_tree_clf.score(X_test_scaled,y_test)))
+
+\epycod
+
+
+% !split
+\subsection{Another example, the moons again}
+\bpycod
+from __future__ import division, print_function, unicode_literals
+
+# Common imports
+import numpy as np
+import os
+
+# to make this notebook's output stable across runs
+np.random.seed(42)
+
+# To plot pretty figures
+import matplotlib
+import matplotlib.pyplot as plt
+from matplotlib.colors import ListedColormap
+plt.rcParams['axes.labelsize'] = 14
+plt.rcParams['xtick.labelsize'] = 12
+plt.rcParams['ytick.labelsize'] = 12
+
+
+from sklearn.svm import SVC
+from sklearn import datasets
+from sklearn.tree import DecisionTreeClassifier
+from sklearn.datasets import make_moons
+from sklearn.tree import export_graphviz
+
+Xm, ym = make_moons(n_samples=100, noise=0.25, random_state=53)
+
+deep_tree_clf1 = DecisionTreeClassifier(random_state=42)
+deep_tree_clf2 = DecisionTreeClassifier(min_samples_leaf=4, random_state=42)
+deep_tree_clf1.fit(Xm, ym)
+deep_tree_clf2.fit(Xm, ym)
+
+
+def plot_decision_boundary(clf, X, y, axes=[0, 7.5, 0, 3], iris=True, legend=False, plot_training=True):
+    x1s = np.linspace(axes[0], axes[1], 100)
+    x2s = np.linspace(axes[2], axes[3], 100)
+    x1, x2 = np.meshgrid(x1s, x2s)
+    X_new = np.c_[x1.ravel(), x2.ravel()]
+    y_pred = clf.predict(X_new).reshape(x1.shape)
+    custom_cmap = ListedColormap(['#fafab0','#9898ff','#a0faa0'])
+    plt.contourf(x1, x2, y_pred, alpha=0.3, cmap=custom_cmap)
+    if not iris:
+        custom_cmap2 = ListedColormap(['#7d7d58','#4c4c7f','#507d50'])
+        plt.contour(x1, x2, y_pred, cmap=custom_cmap2, alpha=0.8)
+    if plot_training:
+        plt.plot(X[:, 0][y==0], X[:, 1][y==0], "yo", label="Iris-Setosa")
+        plt.plot(X[:, 0][y==1], X[:, 1][y==1], "bs", label="Iris-Versicolor")
+        plt.plot(X[:, 0][y==2], X[:, 1][y==2], "g^", label="Iris-Virginica")
+        plt.axis(axes)
+    if iris:
+        plt.xlabel("Petal length", fontsize=14)
+        plt.ylabel("Petal width", fontsize=14)
+    else:
+        plt.xlabel(r"$x_1$", fontsize=18)
+        plt.ylabel(r"$x_2$", fontsize=18, rotation=0)
+    if legend:
+        plt.legend(loc="lower right", fontsize=14)
+plt.figure(figsize=(11, 4))
+plt.subplot(121)
+plot_decision_boundary(deep_tree_clf1, Xm, ym, axes=[-1.5, 2.5, -1, 1.5], iris=False)
+plt.title("No restrictions", fontsize=16)
+plt.subplot(122)
+plot_decision_boundary(deep_tree_clf2, Xm, ym, axes=[-1.5, 2.5, -1, 1.5], iris=False)
+plt.title("min_samples_leaf = {}".format(deep_tree_clf2.min_samples_leaf), fontsize=14)
+plt.show()
+
+\epycod
+
+% !split
+\subsection{Playing around with regions}
+\bpycod
+np.random.seed(6)
+Xs = np.random.rand(100, 2) - 0.5
+ys = (Xs[:, 0] > 0).astype(np.float32) * 2
+
+angle = np.pi/4
+rotation_matrix = np.array([[np.cos(angle), -np.sin(angle)], [np.sin(angle), np.cos(angle)]])
+Xsr = Xs.dot(rotation_matrix)
+
+tree_clf_s = DecisionTreeClassifier(random_state=42)
+tree_clf_s.fit(Xs, ys)
+tree_clf_sr = DecisionTreeClassifier(random_state=42)
+tree_clf_sr.fit(Xsr, ys)
+
+plt.figure(figsize=(11, 4))
+plt.subplot(121)
+plot_decision_boundary(tree_clf_s, Xs, ys, axes=[-0.7, 0.7, -0.7, 0.7], iris=False)
+plt.subplot(122)
+plot_decision_boundary(tree_clf_sr, Xsr, ys, axes=[-0.7, 0.7, -0.7, 0.7], iris=False)
+
+plt.show()
+\epycod
+
+% !split
+\subsection{Regression trees}
+\bpycod
+# Quadratic training set + noise
+np.random.seed(42)
+m = 200
+X = np.random.rand(m, 1)
+y = 4 * (X - 0.5) ** 2
+y = y + np.random.randn(m, 1) / 10
+\epycod
+
+\bpycod
+from sklearn.tree import DecisionTreeRegressor
+
+tree_reg = DecisionTreeRegressor(max_depth=2, random_state=42)
+tree_reg.fit(X, y)
+\epycod
+
+% !split
+\subsection{Final regressor code}
+\bpycod
+from sklearn.tree import DecisionTreeRegressor
+
+tree_reg1 = DecisionTreeRegressor(random_state=42, max_depth=2)
+tree_reg2 = DecisionTreeRegressor(random_state=42, max_depth=3)
+tree_reg1.fit(X, y)
+tree_reg2.fit(X, y)
+
+def plot_regression_predictions(tree_reg, X, y, axes=[0, 1, -0.2, 1], ylabel="$y$"):
+    x1 = np.linspace(axes[0], axes[1], 500).reshape(-1, 1)
+    y_pred = tree_reg.predict(x1)
+    plt.axis(axes)
+    plt.xlabel("$x_1$", fontsize=18)
+    if ylabel:
+        plt.ylabel(ylabel, fontsize=18, rotation=0)
+    plt.plot(X, y, "b.")
+    plt.plot(x1, y_pred, "r.-", linewidth=2, label=r"$\hat{y}$")
+
+plt.figure(figsize=(11, 4))
+plt.subplot(121)
+plot_regression_predictions(tree_reg1, X, y)
+for split, style in ((0.1973, "k-"), (0.0917, "k--"), (0.7718, "k--")):
+    plt.plot([split, split], [-0.2, 1], style, linewidth=2)
+plt.text(0.21, 0.65, "Depth=0", fontsize=15)
+plt.text(0.01, 0.2, "Depth=1", fontsize=13)
+plt.text(0.65, 0.8, "Depth=1", fontsize=13)
+plt.legend(loc="upper center", fontsize=18)
+plt.title("max_depth=2", fontsize=14)
+
+plt.subplot(122)
+plot_regression_predictions(tree_reg2, X, y, ylabel=None)
+for split, style in ((0.1973, "k-"), (0.0917, "k--"), (0.7718, "k--")):
+    plt.plot([split, split], [-0.2, 1], style, linewidth=2)
+for split in (0.0458, 0.1298, 0.2873, 0.9040):
+    plt.plot([split, split], [-0.2, 1], "k:", linewidth=1)
+plt.text(0.3, 0.5, "Depth=2", fontsize=13)
+plt.title("max_depth=3", fontsize=14)
+
+plt.show()
+\epycod
+
+\bpycod
+tree_reg1 = DecisionTreeRegressor(random_state=42)
+tree_reg2 = DecisionTreeRegressor(random_state=42, min_samples_leaf=10)
+tree_reg1.fit(X, y)
+tree_reg2.fit(X, y)
+
+x1 = np.linspace(0, 1, 500).reshape(-1, 1)
+y_pred1 = tree_reg1.predict(x1)
+y_pred2 = tree_reg2.predict(x1)
+
+plt.figure(figsize=(11, 4))
+
+plt.subplot(121)
+plt.plot(X, y, "b.")
+plt.plot(x1, y_pred1, "r.-", linewidth=2, label=r"$\hat{y}$")
+plt.axis([0, 1, -0.2, 1.1])
+plt.xlabel("$x_1$", fontsize=18)
+plt.ylabel("$y$", fontsize=18, rotation=0)
+plt.legend(loc="upper center", fontsize=18)
+plt.title("No restrictions", fontsize=14)
+
+plt.subplot(122)
+plt.plot(X, y, "b.")
+plt.plot(x1, y_pred2, "r.-", linewidth=2, label=r"$\hat{y}$")
+plt.axis([0, 1, -0.2, 1.1])
+plt.xlabel("$x_1$", fontsize=18)
+plt.title("min_samples_leaf={}".format(tree_reg2.min_samples_leaf), fontsize=14)
+
+plt.show()
+\epycod
+
+
+
+% !split
+\subsection{Pros and cons of trees, pros}
+
+\begin{itemize}
+\item White box, easy to interpret model. Some people believe that decision trees more closely mirror human decision-making than do the regression and classification approaches discussed earlier (think of support vector machines)
+
+\item Trees are very easy to explain to people. In fact, they are even easier to explain than linear regression!
+
+\item No feature normalization needed
+
+\item Tree models can handle both continuous and categorical data (Classification and Regression Trees)
+
+\item Can model nonlinear relationships
+
+\item Can model interactions between the different descriptive features
+
+\item Trees can be displayed graphically, and are easily interpreted even by a non-expert (especially if they are small)
+\end{itemize}
+
+\noindent
+% !split
+\subsection{Disadvantages}
+
+\begin{itemize}
+\item Unfortunately, trees generally do not have the same level of predictive accuracy as some of the other regression and classification approaches
+
+\item If continuous features are used the tree may become quite large and hence less interpretable
+
+\item Decision trees are prone to overfit the training data and hence do not well generalize the data if no stopping criteria or improvements like pruning, boosting or bagging are implemented
+
+\item Small changes in the data may lead to a completely different tree. This issue can be addressed by using ensemble methods like bagging, boosting or random forests
+
+\item Unbalanced datasets where some target feature values occur much more frequently than others may lead to biased trees since the frequently occurring feature values are preferred over the less frequently occurring ones. 
+
+\item If the number of features is relatively large (high dimensional) and the number of instances is relatively low, the tree might overfit the data
+
+\item Features with many levels may be preferred over features with less levels since for them it is \emph{more easy} to split the dataset such that the sub datasets only contain pure target feature values. This issue can be addressed by preferring for instance the information gain ratio as splitting criteria over information gain
+\end{itemize}
+
+\noindent
+However, by aggregating many decision trees, using methods like bagging, random forests, and boosting, the predictive performance of trees can be substantially improved. 
+
+
+% !split
+\subsection{Ensemble Methods: From a Single Tree to Many Trees and Extreme Boosting, Meet the Jungle of Methods}
+
+As stated above and seen in many of the examples discussed here about
+a single decision tree, we often end up overfitting our training
+data. This normally means that we have a high variance. Can we reduce
+the variance of a statistical learning method?
+
+This leads us to a set of different methods that can combine different
+machine learning algorithms or just use one of them to construct forests and jungles of trees, homogeneous ones or heterogenous ones. These methods are recognized by different names which we will try to explain here. These are
+\begin{enumerate}
+\item Voting classifiers
+
+\item Bagging and Pasting
+
+\item Random forests
+
+\item Boosting methods, from adaptive to Extreme Gradient Boosting (XGBoost)
+\end{enumerate}
+
+\noindent
+We discuss these methods here.
+
+
+% !split
+\subsection{An Overview of Ensemble Methods}
+
+
+
+\vspace{6mm}
+
+% inline figure
+\centerline{\includegraphics[width=0.8\linewidth]{DataFiles/ensembleoverview.png}}
+
+\vspace{6mm}
+
+
+
+
+
+% !split
+\subsection{Bagging}
+
+The \textbf{plain} decision trees suffer from high
+variance. This means that if we split the training data into two parts
+at random, and fit a decision tree to both halves, the results that we
+get could be quite different. In contrast, a procedure with low
+variance will yield similar results if applied repeatedly to distinct
+data sets; linear regression tends to have low variance, if the ratio
+of $n$ to $p$ is moderately large. 
+
+\textbf{Bootstrap aggregation}, or just \textbf{bagging}, is a
+general-purpose procedure for reducing the variance of a statistical
+learning method. 
+
+
+% !split
+\subsection{More bagging}
+
+Bagging typically results in improved accuracy
+over prediction using a single tree. Unfortunately, however, it can be
+difficult to interpret the resulting model. Recall that one of the
+advantages of decision trees is the attractive and easily interpreted
+diagram that results.
+
+However, when we bag a large number of trees, it is no longer
+possible to represent the resulting statistical learning procedure
+using a single tree, and it is no longer clear which variables are
+most important to the procedure. Thus, bagging improves prediction
+accuracy at the expense of interpretability.  Although the collection
+of bagged trees is much more difficult to interpret than a single
+tree, one can obtain an overall summary of the importance of each
+predictor using the MSE (for bagging regression trees) or the Gini
+index (for bagging classification trees). In the case of bagging
+regression trees, we can record the total amount that the MSE is
+decreased due to splits over a given predictor, averaged over all $B$ possible
+trees. A large value indicates an important predictor. Similarly, in
+the context of bagging classification trees, we can add up the total
+amount that the Gini index  is decreased by splits over a given
+predictor, averaged over all $B$ trees.
+
+% !split
+\subsection{Simple Voting Example, head or tail}
+\bpycod
+heads_proba = 0.51
+coin_tosses = (np.random.rand(10000, 10) < heads_proba).astype(np.int32)
+cumulative_heads_ratio = np.cumsum(coin_tosses, axis=0) / np.arange(1, 10001).reshape(-1, 1)
+plt.figure(figsize=(8,3.5))
+plt.plot(cumulative_heads_ratio)
+plt.plot([0, 10000], [0.51, 0.51], "k--", linewidth=2, label="51%")
+plt.plot([0, 10000], [0.5, 0.5], "k-", label="50%")
+plt.xlabel("Number of coin tosses")
+plt.ylabel("Heads ratio")
+plt.legend(loc="lower right")
+plt.axis([0, 10000, 0.42, 0.58])
+plt.show()
+
+\epycod
+
+% !split
+\subsection{Using the Voting Classifier}
+\bpycod
+from sklearn.model_selection import train_test_split
+from sklearn.datasets import make_moons
+
+X, y = make_moons(n_samples=500, noise=0.30, random_state=42)
+X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=42)
+
+from sklearn.ensemble import RandomForestClassifier
+from sklearn.ensemble import VotingClassifier
+from sklearn.linear_model import LogisticRegression
+from sklearn.svm import SVC
+
+log_clf = LogisticRegression(solver="liblinear", random_state=42)
+rnd_clf = RandomForestClassifier(n_estimators=10, random_state=42)
+svm_clf = SVC(gamma="auto", random_state=42)
+
+voting_clf = VotingClassifier(
+    estimators=[('lr', log_clf), ('rf', rnd_clf), ('svc', svm_clf)],
+    voting='hard')
+
+voting_clf.fit(X_train, y_train)
+
+from sklearn.metrics import accuracy_score
+
+for clf in (log_clf, rnd_clf, svm_clf, voting_clf):
+    clf.fit(X_train, y_train)
+    y_pred = clf.predict(X_test)
+    print(clf.__class__.__name__, accuracy_score(y_test, y_pred))
+
+log_clf = LogisticRegression(solver="liblinear", random_state=42)
+rnd_clf = RandomForestClassifier(n_estimators=10, random_state=42)
+svm_clf = SVC(gamma="auto", probability=True, random_state=42)
+
+voting_clf = VotingClassifier(
+    estimators=[('lr', log_clf), ('rf', rnd_clf), ('svc', svm_clf)],
+    voting='soft')
+voting_clf.fit(X_train, y_train)
+
+from sklearn.metrics import accuracy_score
+
+for clf in (log_clf, rnd_clf, svm_clf, voting_clf):
+    clf.fit(X_train, y_train)
+    y_pred = clf.predict(X_test)
+    print(clf.__class__.__name__, accuracy_score(y_test, y_pred))
+
+\epycod
+
+% !split
+\subsection{Please, not the moons again! Voting and Bagging}
+\bpycod
+from sklearn.model_selection import train_test_split
+from sklearn.datasets import make_moons
+
+X, y = make_moons(n_samples=500, noise=0.30, random_state=42)
+X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=42)
+from sklearn.ensemble import RandomForestClassifier
+from sklearn.ensemble import VotingClassifier
+from sklearn.linear_model import LogisticRegression
+from sklearn.svm import SVC
+
+log_clf = LogisticRegression(random_state=42)
+rnd_clf = RandomForestClassifier(random_state=42)
+svm_clf = SVC(random_state=42)
+
+voting_clf = VotingClassifier(
+    estimators=[('lr', log_clf), ('rf', rnd_clf), ('svc', svm_clf)],
+    voting='hard')
+voting_clf.fit(X_train, y_train)
+\epycod
+
+\bpycod
+from sklearn.metrics import accuracy_score
+
+for clf in (log_clf, rnd_clf, svm_clf, voting_clf):
+    clf.fit(X_train, y_train)
+    y_pred = clf.predict(X_test)
+    print(clf.__class__.__name__, accuracy_score(y_test, y_pred))
+\epycod
+
+\bpycod
+log_clf = LogisticRegression(random_state=42)
+rnd_clf = RandomForestClassifier(random_state=42)
+svm_clf = SVC(probability=True, random_state=42)
+
+voting_clf = VotingClassifier(
+    estimators=[('lr', log_clf), ('rf', rnd_clf), ('svc', svm_clf)],
+    voting='soft')
+voting_clf.fit(X_train, y_train)
+\epycod
+
+\bpycod
+from sklearn.metrics import accuracy_score
+
+for clf in (log_clf, rnd_clf, svm_clf, voting_clf):
+    clf.fit(X_train, y_train)
+    y_pred = clf.predict(X_test)
+    print(clf.__class__.__name__, accuracy_score(y_test, y_pred))
+\epycod
+
+% !split
+\subsection{Now Bagging}
+
+\bpycod
+from sklearn.ensemble import BaggingClassifier
+from sklearn.tree import DecisionTreeClassifier
+
+bag_clf = BaggingClassifier(
+    DecisionTreeClassifier(random_state=42), n_estimators=500,
+    max_samples=100, bootstrap=True, n_jobs=-1, random_state=42)
+bag_clf.fit(X_train, y_train)
+y_pred = bag_clf.predict(X_test)
+\epycod
+
+
+\bpycod
+from sklearn.metrics import accuracy_score
+print(accuracy_score(y_test, y_pred))
+\epycod
+
+\bpycod
+tree_clf = DecisionTreeClassifier(random_state=42)
+tree_clf.fit(X_train, y_train)
+y_pred_tree = tree_clf.predict(X_test)
+print(accuracy_score(y_test, y_pred_tree))
+\epycod
+
+\bpycod
+from matplotlib.colors import ListedColormap
+
+def plot_decision_boundary(clf, X, y, axes=[-1.5, 2.5, -1, 1.5], alpha=0.5, contour=True):
+    x1s = np.linspace(axes[0], axes[1], 100)
+    x2s = np.linspace(axes[2], axes[3], 100)
+    x1, x2 = np.meshgrid(x1s, x2s)
+    X_new = np.c_[x1.ravel(), x2.ravel()]
+    y_pred = clf.predict(X_new).reshape(x1.shape)
+    custom_cmap = ListedColormap(['#fafab0','#9898ff','#a0faa0'])
+    plt.contourf(x1, x2, y_pred, alpha=0.3, cmap=custom_cmap)
+    if contour:
+        custom_cmap2 = ListedColormap(['#7d7d58','#4c4c7f','#507d50'])
+        plt.contour(x1, x2, y_pred, cmap=custom_cmap2, alpha=0.8)
+    plt.plot(X[:, 0][y==0], X[:, 1][y==0], "yo", alpha=alpha)
+    plt.plot(X[:, 0][y==1], X[:, 1][y==1], "bs", alpha=alpha)
+    plt.axis(axes)
+    plt.xlabel(r"$x_1$", fontsize=18)
+    plt.ylabel(r"$x_2$", fontsize=18, rotation=0)
+plt.figure(figsize=(11,4))
+plt.subplot(121)
+plot_decision_boundary(tree_clf, X, y)
+plt.title("Decision Tree", fontsize=14)
+plt.subplot(122)
+plot_decision_boundary(bag_clf, X, y)
+plt.title("Decision Trees with Bagging", fontsize=14)
+plt.show()
+\epycod
+
+
+% !split
+\subsection{Making our own Bagging with Bootstrap}
+\bpycod
+import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import train_test_split
+from sklearn.pipeline import make_pipeline
+from sklearn.utils import resample
+from sklearn.tree import DecisionTreeRegressor
+
+
+np.random.seed(2018)
+
+n = 40
+n_boostraps = 100
+maxdegree = 14
+
+# Make data set.
+x = np.linspace(-3, 3, n).reshape(-1, 1)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
+error = np.zeros(maxdegree)
+bias = np.zeros(maxdegree)
+variance = np.zeros(maxdegree)
+polydegree = np.zeros(maxdegree)
+X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
+
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+for degree in range(maxdegree):
+    model = DecisionTreeRegressor(max_depth=5) 
+    y_pred = np.empty((y_test.shape[0], n_boostraps))
+    for i in range(n_boostraps):
+        x_, y_ = resample(X_train_scaled, y_train)
+        model.fit(x_, y_)
+        y_pred[:, i] = model.predict(X_test_scaled).ravel()
+
+    polydegree[degree] = degree
+    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )
+    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )
+    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )
+    print('Polynomial degree:', degree)
+    print('Error:', error[degree])
+    print('Bias^2:', bias[degree])
+    print('Var:', variance[degree])
+    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
+
+plt.plot(polydegree, error, label='Error')
+plt.plot(polydegree, bias, label='bias')
+plt.plot(polydegree, variance, label='Variance')
+plt.legend()
+plt.show()
+
+\epycod
+
+
+% !split
+\subsection{Changing the Level of the Decision Tree}
+
+\bpycod
+
+import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import train_test_split
+from sklearn.pipeline import make_pipeline
+from sklearn.utils import resample
+from sklearn.tree import DecisionTreeRegressor
+
+n = 100
+n_boostraps = 100
+maxdepth = 8
+
+# Make data set.
+x = np.linspace(-3, 3, n).reshape(-1, 1)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
+error = np.zeros(maxdepth)
+bias = np.zeros(maxdepth)
+variance = np.zeros(maxdepth)
+polydegree = np.zeros(maxdepth)
+X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
+
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+for degree in range(1,maxdepth):
+    model = DecisionTreeRegressor(max_depth=degree) 
+    y_pred = np.empty((y_test.shape[0], n_boostraps))
+    for i in range(n_boostraps):
+        x_, y_ = resample(X_train_scaled, y_train)
+        model.fit(x_, y_)
+        y_pred[:, i] = model.predict(X_test_scaled)#.ravel()
+
+    polydegree[degree] = degree
+    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )
+    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )
+    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )
+    print('Polynomial degree:', degree)
+    print('Error:', error[degree])
+    print('Bias^2:', bias[degree])
+    print('Var:', variance[degree])
+    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
+
+plt.xlim(1,maxdepth)
+plt.plot(polydegree, error, label='Error')
+plt.plot(polydegree, bias, label='bias')
+plt.plot(polydegree, variance, label='Variance')
+plt.legend()
+plt.show()
+
+
+
+
+\epycod
+
+
+
+
+% !split
+\subsection{Random forests}
+
+Random forests provide an improvement over bagged trees by way of a
+small tweak that decorrelates the trees. 
+
+As in bagging, we build a
+number of decision trees on bootstrapped training samples. But when
+building these decision trees, each time a split in a tree is
+considered, a random sample of $m$ predictors is chosen as split
+candidates from the full set of $p$ predictors. The split is allowed to
+use only one of those $m$ predictors. 
+
+A fresh sample of $m$ predictors is
+taken at each split, and typically we choose 
+
+\[
+m\approx \sqrt{p}.
+\]
+
+In building a random forest, at
+each split in the tree, the algorithm is not even allowed to consider
+a majority of the available predictors. 
+
+The reason for this is rather clever. Suppose that there is one very
+strong predictor in the data set, along with a number of other
+moderately strong predictors. Then in the collection of bagged
+variable importance random forest trees, most or all of the trees will
+use this strong predictor in the top split. Consequently, all of the
+bagged trees will look quite similar to each other. Hence the
+predictions from the bagged trees will be highly correlated.
+Unfortunately, averaging many highly correlated quantities does not
+lead to as large of a reduction in variance as averaging many
+uncorrelated quanti- ties. In particular, this means that bagging will
+not lead to a substantial reduction in variance over a single tree in
+this setting.
+
+
+% !split
+\subsection{Random Forest Algorithm}
+The algorithm described here can be applied to both classification and regression problems.
+
+We will grow of forest of say $M$ trees.
+\begin{enumerate}
+\item For $m=1:M$ we
+\begin{itemize}
+
+  \item Draw a bootstrap sample of from the training data organized in our $\bm{X}$ matrix.
+
+  \item We grow then a random forest tree $T_m$ based on the bootstrapped data by repeating the steps outlined till we reach the maximum node size is reached
+\begin{enumerate}
+
+   \item we select $m \le p$ varibales at random from the $p$ predictors/features
+
+   \item pick the best split point among the $m$ features using either the CART algorithm or the ID3 for classification and create a new node
+
+   \item split the node into daughter nodes
+
+\end{enumerate}
+
+\noindent
+\end{itemize}
+
+\noindent
+\item Output then the ensemble of trees $\{T_m\}_1^{M}$ and make predictions for either a regression type of problem or a classification type of problem. 
+\end{enumerate}
+
+\noindent
+% !split
+\subsection{Bootstrap with Random Forests Instead of a Single Tree, own Bagging}
+
+\bpycod
+
+import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import train_test_split
+from sklearn.pipeline import make_pipeline
+from sklearn.utils import resample
+from sklearn.ensemble import RandomForestRegressor
+
+np.random.seed(2018)
+
+n = 100
+n_boostraps = 100
+maxdegree = 14
+
+# Make data set.
+x = np.linspace(-3, 3, n).reshape(-1, 1)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
+error = np.zeros(maxdegree)
+bias = np.zeros(maxdegree)
+variance = np.zeros(maxdegree)
+polydegree = np.zeros(maxdegree)
+X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
+
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+for degree in range(maxdegree):
+    model = RandomForestRegressor()
+    y_pred = np.empty((y_test.shape[0], n_boostraps))
+    for i in range(n_boostraps):
+        x_, y_ = resample(X_train_scaled, y_train)
+        model.fit(x_, y_.ravel())
+        y_pred[:, i] = model.predict(X_test_scaled).ravel()
+
+    polydegree[degree] = degree
+    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )
+    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )
+    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )
+    print('Polynomial degree:', degree)
+    print('Error:', error[degree])
+    print('Bias^2:', bias[degree])
+    print('Var:', variance[degree])
+    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
+
+plt.plot(polydegree, error, label='Error')
+plt.plot(polydegree, bias, label='bias')
+plt.plot(polydegree, variance, label='Variance')
+plt.legend()
+plt.show()
+
+
+
+\epycod
+
+
+
+
+% !split
+\subsection{Random Forests Compared with other Methods on the Cancer Data}
+\bpycod
+import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import  train_test_split 
+from sklearn.datasets import load_breast_cancer
+from sklearn.svm import SVC
+from sklearn.linear_model import LogisticRegression
+from sklearn.tree import DecisionTreeClassifier
+
+# Load the data
+cancer = load_breast_cancer()
+
+X_train, X_test, y_train, y_test = train_test_split(cancer.data,cancer.target,random_state=0)
+print(X_train.shape)
+print(X_test.shape)
+# Logistic Regression
+logreg = LogisticRegression(solver='lbfgs')
+logreg.fit(X_train, y_train)
+print("Test set accuracy with Logistic Regression: {:.2f}".format(logreg.score(X_test,y_test)))
+# Support vector machine
+svm = SVC(gamma='auto', C=100)
+svm.fit(X_train, y_train)
+print("Test set accuracy with SVM: {:.2f}".format(svm.score(X_test,y_test)))
+# Decision Trees
+deep_tree_clf = DecisionTreeClassifier(max_depth=None)
+deep_tree_clf.fit(X_train, y_train)
+print("Test set accuracy with Decision Trees: {:.2f}".format(deep_tree_clf.score(X_test,y_test)))
+#now scale the data
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+# Logistic Regression
+logreg.fit(X_train_scaled, y_train)
+print("Test set accuracy Logistic Regression with scaled data: {:.2f}".format(logreg.score(X_test_scaled,y_test)))
+# Support Vector Machine
+svm.fit(X_train_scaled, y_train)
+print("Test set accuracy SVM with scaled data: {:.2f}".format(logreg.score(X_test_scaled,y_test)))
+# Decision Trees
+deep_tree_clf.fit(X_train_scaled, y_train)
+print("Test set accuracy with Decision Trees and scaled data: {:.2f}".format(deep_tree_clf.score(X_test_scaled,y_test)))
+
+
+from sklearn.ensemble import RandomForestClassifier
+from sklearn.preprocessing import LabelEncoder
+from sklearn.model_selection import cross_validate
+# Data set not specificied
+#Instantiate the model with 500 trees and entropy as splitting criteria
+Random_Forest_model = RandomForestClassifier(n_estimators=500,criterion="entropy")
+Random_Forest_model.fit(X_train_scaled, y_train)
+#Cross validation
+accuracy = cross_validate(Random_Forest_model,X_test_scaled,y_test,cv=10)['test_score']
+print(accuracy)
+print("Test set accuracy with Random Forests and scaled data: {:.2f}".format(Random_Forest_model.score(X_test_scaled,y_test)))
+
+
+import scikitplot as skplt
+y_pred = Random_Forest_model.predict(X_test_scaled)
+skplt.metrics.plot_confusion_matrix(y_test, y_pred, normalize=True)
+plt.show()
+y_probas = Random_Forest_model.predict_proba(X_test_scaled)
+skplt.metrics.plot_roc(y_test, y_probas)
+plt.show()
+skplt.metrics.plot_cumulative_gain(y_test, y_probas)
+plt.show()
+
+\epycod
+
+
+% !split
+\subsection{Compare  Bagging on Trees with Random Forests}
+\bpycod
+bag_clf = BaggingClassifier(
+    DecisionTreeClassifier(splitter="random", max_leaf_nodes=16, random_state=42),
+    n_estimators=500, max_samples=1.0, bootstrap=True, n_jobs=-1, random_state=42)
+\epycod
+
+
+
+\bpycod
+bag_clf.fit(X_train, y_train)
+y_pred = bag_clf.predict(X_test)
+from sklearn.ensemble import RandomForestClassifier
+rnd_clf = RandomForestClassifier(n_estimators=500, max_leaf_nodes=16, n_jobs=-1, random_state=42)
+rnd_clf.fit(X_train, y_train)
+y_pred_rf = rnd_clf.predict(X_test)
+np.sum(y_pred == y_pred_rf) / len(y_pred) 
+\epycod
+
+
+
+
+
+% !split
+\subsection{Boosting, a Bird'e Eye}
+
+The basic idea is to combine weak classifiers in order to create a good
+classifier. With a weak classifier we often intend a classifier which
+produces results which are only slightly better than we would get by
+random guesses.
+
+This is done by applying in an iterative way a weak (or a standard
+classifier like decision trees) to modify the data. In each iteration
+we emphasize those observations which are misclassified by weighting
+them with a factor.
+
+
+% !split
+\subsection{Adaptive boosting: AdaBoost, Basic Algorithm}
+
+The algorithm here is rather straightforward. Assume that our weak
+classifier is a decision tree and we consider a binary set of outputs
+with $y_i \in \{-1,1\}$ and $i=0,1,2,\dots,n-1$ as our set of
+observations. Our design matrix is given in terms of the
+feature/predictor vectors
+$\bm{X}=[\bm{x}_0\bm{x}_1\dots\bm{x}_{p-1}$. Finally, we define also a
+classifier determined by our data via a function $G(\bm{X})$. This function tells us how well we are able to classify our outputs/targets $\bm{y}$. 
+
+We can then define the misclassification error $\mathrm{err}$ as
+\[
+\mathrm{err}=\frac{1}{n}\sum_{i=0}^{n-1}I(y_i\ne G(\bm{X}_{i*}),
+\]
+where the function $I()$ is one if we misclassify and zero if we classify correctly. 
+
+% !split
+\subsection{Basic Steps of AdaBoost}
+
+With the above definitions we are now ready to set up the algorithm for AdaBoost.
+The basic idea is to set up weights which will be used to scale the correctly classified and the misclassified cases.
+\begin{enumerate}
+\item We start by initializing all weights to $w_i = 1/n$, with $i=0,1,2,\dots n-1$. It is to see then that $\sum_{i=0}^{n-1}w_i = 1$.
+
+\item We rewrite the misclassification error as 
+\end{enumerate}
+
+\noindent
+\[
+\mathrm{err}=\frac{\sum_{i=0}^{n-1}w_iI(y_i\ne G(\bm{X}_{i*})}{\sum_{i=0}^{n-1}w_i},
+\]
+\begin{enumerate}
+\item Then we start looping over all attempts at classifying, namely we start an iterative process for $m=1:M$, where $M$ is the final number of classifications. Our given classifier could for example be a plain decision tree.
+\begin{enumerate}
+
+ \item Fit then a given classifier to the training using the weights $w_i$.
+
+ \item Compute then $\mathrm{err}$ and figure out which events are classified properly and which are classified wrongly.
+
+ \item Define a quantity $\alpha_{m} = \log{(1-\mathrm{err})/\mathrm{err}}
+
+ \item Set the new weights to $w_i = w_i\times \exp{(\alpha_m I(y_i\ne G(\bm{X}_{i*})}.
+
+\end{enumerate}
+
+\noindent
+\item Compute the new classifier $G(\bm{X})= \sum_{i=0}^{n-1}\alpha_m I(y_i\ne G(\bm{X}_{i*}).
+\end{enumerate}
+
+\noindent
+For the iterations with $m \le 2$ the weights are modified
+individually at each steps. The obersvations which were misclassified
+at iteration $m-1$ have a weight which is larger than those which were
+classified properly. As this proceeds, the observations which were
+difficult to classifiy correctly are given a larger influence. Each
+new classification step $m$ is then forced to concentrate on those
+observations that are missed in the previous iterations.
+
+
+
+% !split
+\subsection{AdaBoost Examples}
+
+Using \textbf{Scikit-Learn} it is easy to appply the adaptive boosting algorithm, as done here.
+
+\bpycod
+from sklearn.ensemble import AdaBoostClassifier
+
+ada_clf = AdaBoostClassifier(
+    DecisionTreeClassifier(max_depth=1), n_estimators=200,
+    algorithm="SAMME.R", learning_rate=0.5, random_state=42)
+ada_clf.fit(X_train, y_train)
+
+plot_decision_boundary(ada_clf, X, y)
+
+m = len(X_train)
+
+plt.figure(figsize=(11, 4))
+for subplot, learning_rate in ((121, 1), (122, 0.5)):
+    sample_weights = np.ones(m)
+    plt.subplot(subplot)
+    for i in range(5):
+        svm_clf = SVC(kernel="rbf", C=0.05, gamma="auto", random_state=42)
+        svm_clf.fit(X_train, y_train, sample_weight=sample_weights)
+        y_pred = svm_clf.predict(X_train)
+        sample_weights[y_pred != y_train] *= (1 + learning_rate)
+        plot_decision_boundary(svm_clf, X, y, alpha=0.2)
+        plt.title("learning_rate = {}".format(learning_rate), fontsize=16)
+    if subplot == 121:
+        plt.text(-0.7, -0.65, "1", fontsize=14)
+        plt.text(-0.6, -0.10, "2", fontsize=14)
+        plt.text(-0.5,  0.10, "3", fontsize=14)
+        plt.text(-0.4,  0.55, "4", fontsize=14)
+        plt.text(-0.3,  0.90, "5", fontsize=14)
+
+save_fig("boosting_plot")
+plt.show()
+\epycod
+
+
+% !split
+\subsection{Gradient boosting: Basics}
+
+Gradient boosting is again a similar technique to Adapative boosting,
+it combines so-called weak classifiers or regressors into a strong
+method via a series of iterations.
+
+In order to understand the method, let us illustrate its basics by
+bringing back the essential steps in linear regression, where our cost
+function was the least squares function.
+
+% !split
+\subsection{Gradient Boosting, algorithm}
+
+Suppose we have a cost function $C(f)=\sum_{i=0}^{n-1}L(y_i, f(x_i))$ where $y_i$ is our target and $f(x_i)$ the function which is meant to model $y_i$. The above cost function could be our standard least squares  function
+\[
+C(\bm{y},\bm{f})=\frac{1}{n}\sum_{i=0}^{n-1}(y_i-f(x_i))^2.
+\]
+
+The way we proceed in an iterative fashion is to
+\begin{enumerate}
+\item Initialize our estimate by $f_0(x)=0$.
+
+\item For $m=1:M$, we
+\begin{enumerate}
+
+ \item compute the negative gradient vector $\bm{u}_m = -\partial C(\bm{y},\bm{f})/\partial \bm{f}(x)$ at $f(x) = f_{m-1}(x);
+
+ \item fit the so-called base-learner to the negative gradient $h_m(u_m,x)$;
+
+ \item update the estimate $f_m(x) = f_{m-1}(x)+\nu h_m(u_m,x)$;
+
+\end{enumerate}
+
+\noindent
+\item The final estimate is then $f_M(x) = \sum_{m=1}^M\nu h_m(u_m,x)$.
+\end{enumerate}
+
+\noindent
+% !split
+\subsection{Gradient Boosting, Examples}
+\bpycod
+np.random.seed(42)
+X = np.random.rand(100, 1) - 0.5
+y = 3*X[:, 0]**2 + 0.05 * np.random.randn(100)
+
+from sklearn.tree import DecisionTreeRegressor
+
+tree_reg1 = DecisionTreeRegressor(max_depth=2, random_state=42)
+tree_reg1.fit(X, y)
+
+y2 = y - tree_reg1.predict(X)
+tree_reg2 = DecisionTreeRegressor(max_depth=2, random_state=42)
+tree_reg2.fit(X, y2)
+
+y3 = y2 - tree_reg2.predict(X)
+tree_reg3 = DecisionTreeRegressor(max_depth=2, random_state=42)
+tree_reg3.fit(X, y3)
+
+X_new = np.array([[0.8]])
+y_pred = sum(tree.predict(X_new) for tree in (tree_reg1, tree_reg2, tree_reg3))
+
+def plot_predictions(regressors, X, y, axes, label=None, style="r-", data_style="b.", data_label=None):
+    x1 = np.linspace(axes[0], axes[1], 500)
+    y_pred = sum(regressor.predict(x1.reshape(-1, 1)) for regressor in regressors)
+    plt.plot(X[:, 0], y, data_style, label=data_label)
+    plt.plot(x1, y_pred, style, linewidth=2, label=label)
+    if label or data_label:
+        plt.legend(loc="upper center", fontsize=16)
+    plt.axis(axes)
+
+plt.figure(figsize=(11,11))
+
+plt.subplot(321)
+plot_predictions([tree_reg1], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="$h_1(x_1)$", style="g-", data_label="Training set")
+plt.ylabel("$y$", fontsize=16, rotation=0)
+plt.title("Residuals and tree predictions", fontsize=16)
+
+plt.subplot(322)
+plot_predictions([tree_reg1], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="$h(x_1) = h_1(x_1)$", data_label="Training set")
+plt.ylabel("$y$", fontsize=16, rotation=0)
+plt.title("Ensemble predictions", fontsize=16)
+
+plt.subplot(323)
+plot_predictions([tree_reg2], X, y2, axes=[-0.5, 0.5, -0.5, 0.5], label="$h_2(x_1)$", style="g-", data_style="k+", data_label="Residuals")
+plt.ylabel("$y - h_1(x_1)$", fontsize=16)
+
+plt.subplot(324)
+plot_predictions([tree_reg1, tree_reg2], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="$h(x_1) = h_1(x_1) + h_2(x_1)$")
+plt.ylabel("$y$", fontsize=16, rotation=0)
+
+plt.subplot(325)
+plot_predictions([tree_reg3], X, y3, axes=[-0.5, 0.5, -0.5, 0.5], label="$h_3(x_1)$", style="g-", data_style="k+")
+plt.ylabel("$y - h_1(x_1) - h_2(x_1)$", fontsize=16)
+plt.xlabel("$x_1$", fontsize=16)
+
+plt.subplot(326)
+plot_predictions([tree_reg1, tree_reg2, tree_reg3], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="$h(x_1) = h_1(x_1) + h_2(x_1) + h_3(x_1)$")
+plt.xlabel("$x_1$", fontsize=16)
+plt.ylabel("$y$", fontsize=16, rotation=0)
+
+save_fig("gradient_boosting_plot")
+plt.show()
+
+from sklearn.ensemble import GradientBoostingRegressor
+
+gbrt = GradientBoostingRegressor(max_depth=2, n_estimators=3, learning_rate=1.0, random_state=42)
+gbrt.fit(X, y)
+
+gbrt_slow = GradientBoostingRegressor(max_depth=2, n_estimators=200, learning_rate=0.1, random_state=42)
+gbrt_slow.fit(X, y)
+
+plt.figure(figsize=(11,4))
+
+plt.subplot(121)
+plot_predictions([gbrt], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="Ensemble predictions")
+plt.title("learning_rate={}, n_estimators={}".format(gbrt.learning_rate, gbrt.n_estimators), fontsize=14)
+
+plt.subplot(122)
+plot_predictions([gbrt_slow], X, y, axes=[-0.5, 0.5, -0.1, 0.8])
+plt.title("learning_rate={}, n_estimators={}".format(gbrt_slow.learning_rate, gbrt_slow.n_estimators), fontsize=14)
+
+save_fig("gbrt_learning_rate_plot")
+plt.show()
+
+\epycod
+
+
+% !split
+\subsection{Gradient Boots with Early Stopping}
+\bpycod
+
+from sklearn.model_selection import train_test_split
+from sklearn.metrics import mean_squared_error
+
+X_train, X_val, y_train, y_val = train_test_split(X, y, random_state=49)
+
+gbrt = GradientBoostingRegressor(max_depth=2, n_estimators=120, random_state=42)
+gbrt.fit(X_train, y_train)
+
+errors = [mean_squared_error(y_val, y_pred)
+          for y_pred in gbrt.staged_predict(X_val)]
+bst_n_estimators = np.argmin(errors) + 1
+
+gbrt_best = GradientBoostingRegressor(max_depth=2,n_estimators=bst_n_estimators, random_state=42)
+gbrt_best.fit(X_train, y_train)
+
+min_error = np.min(errors)
+plt.figure(figsize=(11, 4))
+
+plt.subplot(121)
+plt.plot(errors, "b.-")
+plt.plot([bst_n_estimators, bst_n_estimators], [0, min_error], "k--")
+plt.plot([0, 120], [min_error, min_error], "k--")
+plt.plot(bst_n_estimators, min_error, "ko")
+plt.text(bst_n_estimators, min_error*1.2, "Minimum", ha="center", fontsize=14)
+plt.axis([0, 120, 0, 0.01])
+plt.xlabel("Number of trees")
+plt.title("Validation error", fontsize=14)
+
+plt.subplot(122)
+plot_predictions([gbrt_best], X, y, axes=[-0.5, 0.5, -0.1, 0.8])
+plt.title("Best model (%d trees)" % bst_n_estimators, fontsize=14)
+
+save_fig("early_stopping_gbrt_plot")
+plt.show()
+
+
+gbrt = GradientBoostingRegressor(max_depth=2, warm_start=True, random_state=42)
+
+min_val_error = float("inf")
+error_going_up = 0
+for n_estimators in range(1, 120):
+    gbrt.n_estimators = n_estimators
+    gbrt.fit(X_train, y_train)
+    y_pred = gbrt.predict(X_val)
+    val_error = mean_squared_error(y_val, y_pred)
+    if val_error < min_val_error:
+        min_val_error = val_error
+        error_going_up = 0
+    else:
+        error_going_up += 1
+        if error_going_up == 5:
+            break  # early stopping
+
+
+print(gbrt.n_estimators)
+print("Minimum validation MSE:", min_val_error)
+\epycod
+
+% !split
+\subsection{XGBoost: Extreme Gradient Boosting}
+
+
+\href{{https://github.com/dmlc/xgboost}}{XGBoost} or Extreme Gradient
+Boosting, is an optimized distributed gradient boosting library
+designed to be highly efficient, flexible and portable. It implements
+machine learning algorithms under the Gradient Boosting
+framework. XGBoost provides a parallel tree boosting that solve many
+data science problems in a fast and accurate way. See the \href{{https://arxiv.org/abs/1603.02754}}{article by Chen and Guestrin}.
+
+The authors design and build a highly scalable end-to-end tree
+boosting system. It has  a theoretically justified weighted quantile
+sketch for efficient proposal calculation. It introduces a novel sparsity-aware algorithm for parallel tree learning and an effective cache-aware block structure for out-of-core tree learning.
+
+It is now the algorithm which wins essentially all ML competitions!!!
+
+% !split
+\subsection{Regression Case}
+
+\bpycod
+import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import train_test_split
+import xgboost as xgb
+from sklearn.preprocessing import StandardScaler
+import scikitplot as skplt
+from sklearn.metrics import mean_squared_error
+
+n = 100
+maxdegree = 6
+
+# Make data set.
+x = np.linspace(-3, 3, n).reshape(-1, 1)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
+
+error = np.zeros(maxdegree)
+bias = np.zeros(maxdegree)
+variance = np.zeros(maxdegree)
+polydegree = np.zeros(maxdegree)
+X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+for degree in range(maxdegree):
+    model =  xgb.XGBRegressor(objective ='reg:squarederror', colsample_bytree = 0.3, learning_rate = 0.1,
+                max_depth = degree, alpha = 10, n_estimators = 10)
+    model.fit(X_train_scaled,y_train)
+    y_pred = model.predict(X_test_scaled)
+    polydegree[degree] = degree
+    error[degree] = np.mean( np.mean((y_test - y_pred)**2) )
+    bias[degree] = np.mean( (y_test - np.mean(y_pred))**2 )
+    variance[degree] = np.mean( np.var(y_pred) )
+    print('Max depth:', degree)
+    print('Error:', error[degree])
+    print('Bias^2:', bias[degree])
+    print('Var:', variance[degree])
+    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
+
+plt.xlim(1,maxdegree-1)
+plt.plot(polydegree, error, label='Error')
+plt.plot(polydegree, bias, label='bias')
+plt.plot(polydegree, variance, label='Variance')
+plt.legend()
+plt.show()
+
+
+\epycod
+
+
+
+% !split
+\subsection{Xgboost on the Cancer Data}
+\bpycod
+import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import  train_test_split 
+from sklearn.datasets import load_breast_cancer
+from sklearn.preprocessing import LabelEncoder
+from sklearn.model_selection import cross_validate
+import scikitplot as skplt
+import xgboost as xgb
+# Load the data
+cancer = load_breast_cancer()
+
+X_train, X_test, y_train, y_test = train_test_split(cancer.data,cancer.target,random_state=0)
+print(X_train.shape)
+print(X_test.shape)
+#now scale the data
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+xg_clf = xgb.XGBClassifier()
+xg_clf.fit(X_train_scaled,y_train)
+xgb.plot_tree(xg_clf,num_trees=0)
+plt.rcParams['figure.figsize'] = [50, 10]
+plt.show()
+xgb.plot_importance(xg_clf)
+plt.rcParams['figure.figsize'] = [5, 5]
+plt.show()
+\epycod
+
+% ------------------- end of main content ---------------
+
+% #ifdef PREAMBLE
+\end{document}
+% #endif
+
diff --git a/doc/src/DecisionTrees/DecisionTrees.tex b/doc/src/DecisionTrees/DecisionTrees.tex
new file mode 100644
index 000000000..5ab09d845
--- /dev/null
+++ b/doc/src/DecisionTrees/DecisionTrees.tex
@@ -0,0 +1,2395 @@
+%%
+%% Automatically generated file from DocOnce source
+%% (https://github.com/hplgit/doconce/)
+%%
+%%
+
+
+%-------------------- begin preamble ----------------------
+
+\documentclass[%
+oneside,                 % oneside: electronic viewing, twoside: printing
+final,                   % draft: marks overfull hboxes, figures with paths
+10pt]{article}
+
+\listfiles               %  print all files needed to compile this document
+
+\usepackage{relsize,makeidx,color,setspace,amsmath,amsfonts,amssymb}
+\usepackage[table]{xcolor}
+\usepackage{bm,ltablex,microtype}
+
+\usepackage[pdftex]{graphicx}
+
+\usepackage{fancyvrb} % packages needed for verbatim environments
+\usepackage{minted}
+\usemintedstyle{default}
+
+\usepackage[T1]{fontenc}
+%\usepackage[latin1]{inputenc}
+\usepackage{ucs}
+\usepackage[utf8x]{inputenc}
+
+\usepackage{lmodern}         % Latin Modern fonts derived from Computer Modern
+
+% Hyperlinks in PDF:
+\definecolor{linkcolor}{rgb}{0,0,0.4}
+\usepackage{hyperref}
+\hypersetup{
+    breaklinks=true,
+    colorlinks=true,
+    linkcolor=linkcolor,
+    urlcolor=linkcolor,
+    citecolor=black,
+    filecolor=black,
+    %filecolor=blue,
+    pdfmenubar=true,
+    pdftoolbar=true,
+    bookmarksdepth=3   % Uncomment (and tweak) for PDF bookmarks with more levels than the TOC
+    }
+%\hyperbaseurl{}   % hyperlinks are relative to this root
+
+\setcounter{tocdepth}{2}  % levels in table of contents
+
+% Tricks for having figures close to where they are defined:
+% 1. define less restrictive rules for where to put figures
+\setcounter{topnumber}{2}
+\setcounter{bottomnumber}{2}
+\setcounter{totalnumber}{4}
+\renewcommand{\topfraction}{0.95}
+\renewcommand{\bottomfraction}{0.95}
+\renewcommand{\textfraction}{0}
+\renewcommand{\floatpagefraction}{0.75}
+% floatpagefraction must always be less than topfraction!
+% 2. ensure all figures are flushed before next section
+\usepackage[section]{placeins}
+% 3. enable begin{figure}[H] (often leads to ugly pagebreaks)
+%\usepackage{float}\restylefloat{figure}
+
+% --- fancyhdr package for fancy headers ---
+\usepackage{fancyhdr}
+\fancyhf{} % sets both header and footer to nothing
+\renewcommand{\headrulewidth}{0pt}
+\fancyfoot[LE,RO]{\thepage}
+% Ensure copyright on titlepage (article style) and chapter pages (book style)
+\fancypagestyle{plain}{
+  \fancyhf{}
+  \fancyfoot[C]{{\footnotesize \copyright\ 1999-2019, Morten Hjorth-Jensen. Released under CC Attribution-NonCommercial 4.0 license}}
+%  \renewcommand{\footrulewidth}{0mm}
+  \renewcommand{\headrulewidth}{0mm}
+}
+% Ensure copyright on titlepages with \thispagestyle{empty}
+\fancypagestyle{empty}{
+  \fancyhf{}
+  \fancyfoot[C]{{\footnotesize \copyright\ 1999-2019, Morten Hjorth-Jensen. Released under CC Attribution-NonCommercial 4.0 license}}
+  \renewcommand{\footrulewidth}{0mm}
+  \renewcommand{\headrulewidth}{0mm}
+}
+
+\pagestyle{fancy}
+
+
+\usepackage[framemethod=TikZ]{mdframed}
+
+% --- begin definitions of admonition environments ---
+
+% --- end of definitions of admonition environments ---
+
+% prevent orhpans and widows
+\clubpenalty = 10000
+\widowpenalty = 10000
+
+% --- end of standard preamble for documents ---
+
+
+% insert custom LaTeX commands...
+
+\raggedbottom
+\makeindex
+\usepackage[totoc]{idxlayout}   % for index in the toc
+\usepackage[nottoc]{tocbibind}  % for references/bibliography in the toc
+
+%-------------------- end preamble ----------------------
+
+\begin{document}
+
+% matching end for #ifdef PREAMBLE
+
+\newcommand{\exercisesection}[1]{\subsection*{#1}}
+
+
+% ------------------- main content ----------------------
+
+
+
+% ----------------- title -------------------------
+
+\thispagestyle{empty}
+
+\begin{center}
+{\LARGE\bf
+\begin{spacing}{1.25}
+Data Analysis and Machine Learning: From Decision Trees to Forests and all that
+\end{spacing}
+}
+\end{center}
+
+% ----------------- author(s) -------------------------
+
+\begin{center}
+{\bf Morten Hjorth-Jensen${}^{1, 2}$} \\ [0mm]
+\end{center}
+
+\begin{center}
+% List of all institutions:
+\centerline{{\small ${}^1$Department of Physics, University of Oslo}}
+\centerline{{\small ${}^2$Department of Physics and Astronomy and National Superconducting Cyclotron Laboratory, Michigan State University}}
+\end{center}
+    
+% ----------------- end author(s) -------------------------
+
+% --- begin date ---
+\begin{center}
+Nov 7, 2019
+\end{center}
+% --- end date ---
+
+\vspace{1cm}
+
+
+% !split
+\subsection*{Decision trees, overarching aims}
+
+
+Decision trees are supervised learning algorithms used for both,
+classification and regression tasks.
+
+
+The main idea of decision trees
+is to find those descriptive features which contain the most
+\textbf{information} regarding the target feature and then split the dataset
+along the values of these features such that the target feature values
+for the resulting underlying datasets are as pure as possible.
+
+The descriptive features which reproduce best the target/output features are normally  said
+to be the most informative ones. The process of finding the \textbf{most
+informative} feature is done until we accomplish a stopping criteria
+where we then finally end up in so called \textbf{leaf nodes}. 
+
+A decision tree is typically divided into a \textbf{root node}, the \textbf{interior nodes},
+and the final \textbf{leaf nodes} or just \textbf{leaves}. These entities are then connected by so-called \textbf{branches}.
+
+The leaf nodes
+contain the predictions we will make for new query instances presented
+to our trained model. This is possible since the model has 
+learned the underlying structure of the training data and hence can,
+given some assumptions, make predictions about the target feature value
+(class) of unseen query instances.
+
+% !split
+\subsection*{A typical Decision Tree with its pertinent Jargon, Classification Problem}
+
+
+
+\vspace{6mm}
+
+% inline figure
+\centerline{\includegraphics[width=0.8\linewidth]{DataFiles/cancer.png}}
+
+\vspace{6mm}
+
+
+
+This tree was produced using the Wisconsin cancer data (discussed here as well, see code examples below) using \textbf{Scikit-Learn}'s decision tree classifier. Here we have used the so-called \textbf{gini} index (see below) to split the various branches.
+
+
+
+% !split
+\subsection*{General Features}
+
+The overarching approach to decision trees is a top-down approach.
+
+\begin{itemize}
+\item A leaf provides the classification of a given instance.
+
+\item A node specifies a test of some attribute of the instance.
+
+\item A branch corresponds to a possible values of an attribute.
+
+\item An instance is classified by starting at the root node of the tree, testing the attribute specified by this node, then moving down the tree branch corresponding to the value of the attribute in the given example.
+\end{itemize}
+
+\noindent
+This process is then repeated for the subtree rooted at the new
+node.
+
+
+% !split
+\subsection*{How do we set it up?}
+
+
+In simplified terms, the process of training a decision tree and
+predicting the target features of query instances is as follows:
+
+\begin{enumerate}
+\item Present a dataset containing of a number of training instances characterized by a number of descriptive features and a target feature
+
+\item Train the decision tree model by continuously splitting the target feature along the values of the descriptive features using a measure of information gain during the training process
+
+\item Grow the tree until we accomplish a stopping criteria create leaf nodes which represent the \emph{predictions} we want to make for new query instances
+
+\item Show query instances to the tree and run down the tree until we arrive at leaf nodes
+\end{enumerate}
+
+\noindent
+Then we are essentially done!
+
+
+
+
+
+% !split
+\subsection*{Decision trees and Regression}
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+import numpy as np
+import matplotlib.pyplot as plt
+from sklearn.preprocessing import PolynomialFeatures
+from sklearn.linear_model import LinearRegression
+
+steps=250
+
+distance=0
+x=0
+distance_list=[]
+steps_list=[]
+while x 0 and bestAttrId in attributeIds:
+					toRemove = attributeIds.index(bestAttrId)
+					attributeIds.pop(toRemove)
+				child.next = self.id3Recv(
+					childSampleIds, attributeIds, child.next)
+		return root
+
+	def printTree(self):
+		if self.root:
+			roots = deque()
+			roots.append(self.root)
+			while len(roots) > 0:
+				root = roots.popleft()
+				print(root.value)
+				if root.childs:
+					for child in root.childs:
+						print('({})'.format(child.value))
+						roots.append(child.next)
+				elif root.next:
+					print(root.next)
+
+
+def test():
+	f = open('DataFiles/rideclass.csv')
+	attributes = f.readline().split(',')
+	attributes = attributes[1:len(attributes)-1]
+	print(attributes)
+	sample = f.readlines()
+	f.close()
+	for i in range(len(sample)):
+		sample[i] = re.sub('\d+,', '', sample[i])
+		sample[i] = sample[i].strip().split(',')
+	labels = []
+	for s in sample:
+		labels.append(s.pop())
+	# print(sample)
+	# print(labels)
+	decisionTree = DecisionTree(sample, attributes, labels)
+	print("System entropy {}".format(decisionTree.entropy))
+	decisionTree.id3()
+	decisionTree.printTree()
+
+
+if __name__ == '__main__':
+	test()
+\end{minted}
+
+% !split
+\subsection*{Cancer Data again now with Decision Trees and other Methods}
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import  train_test_split 
+from sklearn.datasets import load_breast_cancer
+from sklearn.svm import SVC
+from sklearn.linear_model import LogisticRegression
+from sklearn.tree import DecisionTreeClassifier
+
+# Load the data
+cancer = load_breast_cancer()
+
+X_train, X_test, y_train, y_test = train_test_split(cancer.data,cancer.target,random_state=0)
+print(X_train.shape)
+print(X_test.shape)
+# Logistic Regression
+logreg = LogisticRegression(solver='lbfgs')
+logreg.fit(X_train, y_train)
+print("Test set accuracy with Logistic Regression: {:.2f}".format(logreg.score(X_test,y_test)))
+# Support vector machine
+svm = SVC(gamma='auto', C=100)
+svm.fit(X_train, y_train)
+print("Test set accuracy with SVM: {:.2f}".format(svm.score(X_test,y_test)))
+# Decision Trees
+deep_tree_clf = DecisionTreeClassifier(max_depth=None)
+deep_tree_clf.fit(X_train, y_train)
+print("Test set accuracy with Decision Trees: {:.2f}".format(deep_tree_clf.score(X_test,y_test)))
+#now scale the data
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+# Logistic Regression
+logreg.fit(X_train_scaled, y_train)
+print("Test set accuracy Logistic Regression with scaled data: {:.2f}".format(logreg.score(X_test_scaled,y_test)))
+# Support Vector Machine
+svm.fit(X_train_scaled, y_train)
+print("Test set accuracy SVM with scaled data: {:.2f}".format(logreg.score(X_test_scaled,y_test)))
+# Decision Trees
+deep_tree_clf.fit(X_train_scaled, y_train)
+print("Test set accuracy with Decision Trees and scaled data: {:.2f}".format(deep_tree_clf.score(X_test_scaled,y_test)))
+
+\end{minted}
+
+
+% !split
+\subsection*{Another example, the moons again}
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+from __future__ import division, print_function, unicode_literals
+
+# Common imports
+import numpy as np
+import os
+
+# to make this notebook's output stable across runs
+np.random.seed(42)
+
+# To plot pretty figures
+import matplotlib
+import matplotlib.pyplot as plt
+from matplotlib.colors import ListedColormap
+plt.rcParams['axes.labelsize'] = 14
+plt.rcParams['xtick.labelsize'] = 12
+plt.rcParams['ytick.labelsize'] = 12
+
+
+from sklearn.svm import SVC
+from sklearn import datasets
+from sklearn.tree import DecisionTreeClassifier
+from sklearn.datasets import make_moons
+from sklearn.tree import export_graphviz
+
+Xm, ym = make_moons(n_samples=100, noise=0.25, random_state=53)
+
+deep_tree_clf1 = DecisionTreeClassifier(random_state=42)
+deep_tree_clf2 = DecisionTreeClassifier(min_samples_leaf=4, random_state=42)
+deep_tree_clf1.fit(Xm, ym)
+deep_tree_clf2.fit(Xm, ym)
+
+
+def plot_decision_boundary(clf, X, y, axes=[0, 7.5, 0, 3], iris=True, legend=False, plot_training=True):
+    x1s = np.linspace(axes[0], axes[1], 100)
+    x2s = np.linspace(axes[2], axes[3], 100)
+    x1, x2 = np.meshgrid(x1s, x2s)
+    X_new = np.c_[x1.ravel(), x2.ravel()]
+    y_pred = clf.predict(X_new).reshape(x1.shape)
+    custom_cmap = ListedColormap(['#fafab0','#9898ff','#a0faa0'])
+    plt.contourf(x1, x2, y_pred, alpha=0.3, cmap=custom_cmap)
+    if not iris:
+        custom_cmap2 = ListedColormap(['#7d7d58','#4c4c7f','#507d50'])
+        plt.contour(x1, x2, y_pred, cmap=custom_cmap2, alpha=0.8)
+    if plot_training:
+        plt.plot(X[:, 0][y==0], X[:, 1][y==0], "yo", label="Iris-Setosa")
+        plt.plot(X[:, 0][y==1], X[:, 1][y==1], "bs", label="Iris-Versicolor")
+        plt.plot(X[:, 0][y==2], X[:, 1][y==2], "g^", label="Iris-Virginica")
+        plt.axis(axes)
+    if iris:
+        plt.xlabel("Petal length", fontsize=14)
+        plt.ylabel("Petal width", fontsize=14)
+    else:
+        plt.xlabel(r"$x_1$", fontsize=18)
+        plt.ylabel(r"$x_2$", fontsize=18, rotation=0)
+    if legend:
+        plt.legend(loc="lower right", fontsize=14)
+plt.figure(figsize=(11, 4))
+plt.subplot(121)
+plot_decision_boundary(deep_tree_clf1, Xm, ym, axes=[-1.5, 2.5, -1, 1.5], iris=False)
+plt.title("No restrictions", fontsize=16)
+plt.subplot(122)
+plot_decision_boundary(deep_tree_clf2, Xm, ym, axes=[-1.5, 2.5, -1, 1.5], iris=False)
+plt.title("min_samples_leaf = {}".format(deep_tree_clf2.min_samples_leaf), fontsize=14)
+plt.show()
+
+\end{minted}
+
+% !split
+\subsection*{Playing around with regions}
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+np.random.seed(6)
+Xs = np.random.rand(100, 2) - 0.5
+ys = (Xs[:, 0] > 0).astype(np.float32) * 2
+
+angle = np.pi/4
+rotation_matrix = np.array([[np.cos(angle), -np.sin(angle)], [np.sin(angle), np.cos(angle)]])
+Xsr = Xs.dot(rotation_matrix)
+
+tree_clf_s = DecisionTreeClassifier(random_state=42)
+tree_clf_s.fit(Xs, ys)
+tree_clf_sr = DecisionTreeClassifier(random_state=42)
+tree_clf_sr.fit(Xsr, ys)
+
+plt.figure(figsize=(11, 4))
+plt.subplot(121)
+plot_decision_boundary(tree_clf_s, Xs, ys, axes=[-0.7, 0.7, -0.7, 0.7], iris=False)
+plt.subplot(122)
+plot_decision_boundary(tree_clf_sr, Xsr, ys, axes=[-0.7, 0.7, -0.7, 0.7], iris=False)
+
+plt.show()
+\end{minted}
+
+% !split
+\subsection*{Regression trees}
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+# Quadratic training set + noise
+np.random.seed(42)
+m = 200
+X = np.random.rand(m, 1)
+y = 4 * (X - 0.5) ** 2
+y = y + np.random.randn(m, 1) / 10
+\end{minted}
+
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+from sklearn.tree import DecisionTreeRegressor
+
+tree_reg = DecisionTreeRegressor(max_depth=2, random_state=42)
+tree_reg.fit(X, y)
+\end{minted}
+
+% !split
+\subsection*{Final regressor code}
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+from sklearn.tree import DecisionTreeRegressor
+
+tree_reg1 = DecisionTreeRegressor(random_state=42, max_depth=2)
+tree_reg2 = DecisionTreeRegressor(random_state=42, max_depth=3)
+tree_reg1.fit(X, y)
+tree_reg2.fit(X, y)
+
+def plot_regression_predictions(tree_reg, X, y, axes=[0, 1, -0.2, 1], ylabel="$y$"):
+    x1 = np.linspace(axes[0], axes[1], 500).reshape(-1, 1)
+    y_pred = tree_reg.predict(x1)
+    plt.axis(axes)
+    plt.xlabel("$x_1$", fontsize=18)
+    if ylabel:
+        plt.ylabel(ylabel, fontsize=18, rotation=0)
+    plt.plot(X, y, "b.")
+    plt.plot(x1, y_pred, "r.-", linewidth=2, label=r"$\hat{y}$")
+
+plt.figure(figsize=(11, 4))
+plt.subplot(121)
+plot_regression_predictions(tree_reg1, X, y)
+for split, style in ((0.1973, "k-"), (0.0917, "k--"), (0.7718, "k--")):
+    plt.plot([split, split], [-0.2, 1], style, linewidth=2)
+plt.text(0.21, 0.65, "Depth=0", fontsize=15)
+plt.text(0.01, 0.2, "Depth=1", fontsize=13)
+plt.text(0.65, 0.8, "Depth=1", fontsize=13)
+plt.legend(loc="upper center", fontsize=18)
+plt.title("max_depth=2", fontsize=14)
+
+plt.subplot(122)
+plot_regression_predictions(tree_reg2, X, y, ylabel=None)
+for split, style in ((0.1973, "k-"), (0.0917, "k--"), (0.7718, "k--")):
+    plt.plot([split, split], [-0.2, 1], style, linewidth=2)
+for split in (0.0458, 0.1298, 0.2873, 0.9040):
+    plt.plot([split, split], [-0.2, 1], "k:", linewidth=1)
+plt.text(0.3, 0.5, "Depth=2", fontsize=13)
+plt.title("max_depth=3", fontsize=14)
+
+plt.show()
+\end{minted}
+
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+tree_reg1 = DecisionTreeRegressor(random_state=42)
+tree_reg2 = DecisionTreeRegressor(random_state=42, min_samples_leaf=10)
+tree_reg1.fit(X, y)
+tree_reg2.fit(X, y)
+
+x1 = np.linspace(0, 1, 500).reshape(-1, 1)
+y_pred1 = tree_reg1.predict(x1)
+y_pred2 = tree_reg2.predict(x1)
+
+plt.figure(figsize=(11, 4))
+
+plt.subplot(121)
+plt.plot(X, y, "b.")
+plt.plot(x1, y_pred1, "r.-", linewidth=2, label=r"$\hat{y}$")
+plt.axis([0, 1, -0.2, 1.1])
+plt.xlabel("$x_1$", fontsize=18)
+plt.ylabel("$y$", fontsize=18, rotation=0)
+plt.legend(loc="upper center", fontsize=18)
+plt.title("No restrictions", fontsize=14)
+
+plt.subplot(122)
+plt.plot(X, y, "b.")
+plt.plot(x1, y_pred2, "r.-", linewidth=2, label=r"$\hat{y}$")
+plt.axis([0, 1, -0.2, 1.1])
+plt.xlabel("$x_1$", fontsize=18)
+plt.title("min_samples_leaf={}".format(tree_reg2.min_samples_leaf), fontsize=14)
+
+plt.show()
+\end{minted}
+
+
+
+% !split
+\subsection*{Pros and cons of trees, pros}
+
+\begin{itemize}
+\item White box, easy to interpret model. Some people believe that decision trees more closely mirror human decision-making than do the regression and classification approaches discussed earlier (think of support vector machines)
+
+\item Trees are very easy to explain to people. In fact, they are even easier to explain than linear regression!
+
+\item No feature normalization needed
+
+\item Tree models can handle both continuous and categorical data (Classification and Regression Trees)
+
+\item Can model nonlinear relationships
+
+\item Can model interactions between the different descriptive features
+
+\item Trees can be displayed graphically, and are easily interpreted even by a non-expert (especially if they are small)
+\end{itemize}
+
+\noindent
+% !split
+\subsection*{Disadvantages}
+
+\begin{itemize}
+\item Unfortunately, trees generally do not have the same level of predictive accuracy as some of the other regression and classification approaches
+
+\item If continuous features are used the tree may become quite large and hence less interpretable
+
+\item Decision trees are prone to overfit the training data and hence do not well generalize the data if no stopping criteria or improvements like pruning, boosting or bagging are implemented
+
+\item Small changes in the data may lead to a completely different tree. This issue can be addressed by using ensemble methods like bagging, boosting or random forests
+
+\item Unbalanced datasets where some target feature values occur much more frequently than others may lead to biased trees since the frequently occurring feature values are preferred over the less frequently occurring ones. 
+
+\item If the number of features is relatively large (high dimensional) and the number of instances is relatively low, the tree might overfit the data
+
+\item Features with many levels may be preferred over features with less levels since for them it is \emph{more easy} to split the dataset such that the sub datasets only contain pure target feature values. This issue can be addressed by preferring for instance the information gain ratio as splitting criteria over information gain
+\end{itemize}
+
+\noindent
+However, by aggregating many decision trees, using methods like bagging, random forests, and boosting, the predictive performance of trees can be substantially improved. 
+
+
+% !split
+\subsection*{Ensemble Methods: From a Single Tree to Many Trees and Extreme Boosting, Meet the Jungle of Methods}
+
+As stated above and seen in many of the examples discussed here about
+a single decision tree, we often end up overfitting our training
+data. This normally means that we have a high variance. Can we reduce
+the variance of a statistical learning method?
+
+This leads us to a set of different methods that can combine different
+machine learning algorithms or just use one of them to construct forests and jungles of trees, homogeneous ones or heterogenous ones. These methods are recognized by different names which we will try to explain here. These are
+\begin{enumerate}
+\item Voting classifiers
+
+\item Bagging and Pasting
+
+\item Random forests
+
+\item Boosting methods, from adaptive to Extreme Gradient Boosting (XGBoost)
+\end{enumerate}
+
+\noindent
+We discuss these methods here.
+
+
+% !split
+\subsection*{An Overview of Ensemble Methods}
+
+
+
+\vspace{6mm}
+
+% inline figure
+\centerline{\includegraphics[width=0.8\linewidth]{DataFiles/ensembleoverview.png}}
+
+\vspace{6mm}
+
+
+
+
+
+% !split
+\subsection*{Bagging}
+
+The \textbf{plain} decision trees suffer from high
+variance. This means that if we split the training data into two parts
+at random, and fit a decision tree to both halves, the results that we
+get could be quite different. In contrast, a procedure with low
+variance will yield similar results if applied repeatedly to distinct
+data sets; linear regression tends to have low variance, if the ratio
+of $n$ to $p$ is moderately large. 
+
+\textbf{Bootstrap aggregation}, or just \textbf{bagging}, is a
+general-purpose procedure for reducing the variance of a statistical
+learning method. 
+
+
+% !split
+\subsection*{More bagging}
+
+Bagging typically results in improved accuracy
+over prediction using a single tree. Unfortunately, however, it can be
+difficult to interpret the resulting model. Recall that one of the
+advantages of decision trees is the attractive and easily interpreted
+diagram that results.
+
+However, when we bag a large number of trees, it is no longer
+possible to represent the resulting statistical learning procedure
+using a single tree, and it is no longer clear which variables are
+most important to the procedure. Thus, bagging improves prediction
+accuracy at the expense of interpretability.  Although the collection
+of bagged trees is much more difficult to interpret than a single
+tree, one can obtain an overall summary of the importance of each
+predictor using the MSE (for bagging regression trees) or the Gini
+index (for bagging classification trees). In the case of bagging
+regression trees, we can record the total amount that the MSE is
+decreased due to splits over a given predictor, averaged over all $B$ possible
+trees. A large value indicates an important predictor. Similarly, in
+the context of bagging classification trees, we can add up the total
+amount that the Gini index  is decreased by splits over a given
+predictor, averaged over all $B$ trees.
+
+% !split
+\subsection*{Simple Voting Example, head or tail}
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+heads_proba = 0.51
+coin_tosses = (np.random.rand(10000, 10) < heads_proba).astype(np.int32)
+cumulative_heads_ratio = np.cumsum(coin_tosses, axis=0) / np.arange(1, 10001).reshape(-1, 1)
+plt.figure(figsize=(8,3.5))
+plt.plot(cumulative_heads_ratio)
+plt.plot([0, 10000], [0.51, 0.51], "k--", linewidth=2, label="51%")
+plt.plot([0, 10000], [0.5, 0.5], "k-", label="50%")
+plt.xlabel("Number of coin tosses")
+plt.ylabel("Heads ratio")
+plt.legend(loc="lower right")
+plt.axis([0, 10000, 0.42, 0.58])
+plt.show()
+
+\end{minted}
+
+% !split
+\subsection*{Using the Voting Classifier}
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+from sklearn.model_selection import train_test_split
+from sklearn.datasets import make_moons
+
+X, y = make_moons(n_samples=500, noise=0.30, random_state=42)
+X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=42)
+
+from sklearn.ensemble import RandomForestClassifier
+from sklearn.ensemble import VotingClassifier
+from sklearn.linear_model import LogisticRegression
+from sklearn.svm import SVC
+
+log_clf = LogisticRegression(solver="liblinear", random_state=42)
+rnd_clf = RandomForestClassifier(n_estimators=10, random_state=42)
+svm_clf = SVC(gamma="auto", random_state=42)
+
+voting_clf = VotingClassifier(
+    estimators=[('lr', log_clf), ('rf', rnd_clf), ('svc', svm_clf)],
+    voting='hard')
+
+voting_clf.fit(X_train, y_train)
+
+from sklearn.metrics import accuracy_score
+
+for clf in (log_clf, rnd_clf, svm_clf, voting_clf):
+    clf.fit(X_train, y_train)
+    y_pred = clf.predict(X_test)
+    print(clf.__class__.__name__, accuracy_score(y_test, y_pred))
+
+log_clf = LogisticRegression(solver="liblinear", random_state=42)
+rnd_clf = RandomForestClassifier(n_estimators=10, random_state=42)
+svm_clf = SVC(gamma="auto", probability=True, random_state=42)
+
+voting_clf = VotingClassifier(
+    estimators=[('lr', log_clf), ('rf', rnd_clf), ('svc', svm_clf)],
+    voting='soft')
+voting_clf.fit(X_train, y_train)
+
+from sklearn.metrics import accuracy_score
+
+for clf in (log_clf, rnd_clf, svm_clf, voting_clf):
+    clf.fit(X_train, y_train)
+    y_pred = clf.predict(X_test)
+    print(clf.__class__.__name__, accuracy_score(y_test, y_pred))
+
+\end{minted}
+
+% !split
+\subsection*{Please, not the moons again! Voting and Bagging}
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+from sklearn.model_selection import train_test_split
+from sklearn.datasets import make_moons
+
+X, y = make_moons(n_samples=500, noise=0.30, random_state=42)
+X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=42)
+from sklearn.ensemble import RandomForestClassifier
+from sklearn.ensemble import VotingClassifier
+from sklearn.linear_model import LogisticRegression
+from sklearn.svm import SVC
+
+log_clf = LogisticRegression(random_state=42)
+rnd_clf = RandomForestClassifier(random_state=42)
+svm_clf = SVC(random_state=42)
+
+voting_clf = VotingClassifier(
+    estimators=[('lr', log_clf), ('rf', rnd_clf), ('svc', svm_clf)],
+    voting='hard')
+voting_clf.fit(X_train, y_train)
+\end{minted}
+
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+from sklearn.metrics import accuracy_score
+
+for clf in (log_clf, rnd_clf, svm_clf, voting_clf):
+    clf.fit(X_train, y_train)
+    y_pred = clf.predict(X_test)
+    print(clf.__class__.__name__, accuracy_score(y_test, y_pred))
+\end{minted}
+
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+log_clf = LogisticRegression(random_state=42)
+rnd_clf = RandomForestClassifier(random_state=42)
+svm_clf = SVC(probability=True, random_state=42)
+
+voting_clf = VotingClassifier(
+    estimators=[('lr', log_clf), ('rf', rnd_clf), ('svc', svm_clf)],
+    voting='soft')
+voting_clf.fit(X_train, y_train)
+\end{minted}
+
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+from sklearn.metrics import accuracy_score
+
+for clf in (log_clf, rnd_clf, svm_clf, voting_clf):
+    clf.fit(X_train, y_train)
+    y_pred = clf.predict(X_test)
+    print(clf.__class__.__name__, accuracy_score(y_test, y_pred))
+\end{minted}
+
+% !split
+\subsection*{Now Bagging}
+
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+from sklearn.ensemble import BaggingClassifier
+from sklearn.tree import DecisionTreeClassifier
+
+bag_clf = BaggingClassifier(
+    DecisionTreeClassifier(random_state=42), n_estimators=500,
+    max_samples=100, bootstrap=True, n_jobs=-1, random_state=42)
+bag_clf.fit(X_train, y_train)
+y_pred = bag_clf.predict(X_test)
+\end{minted}
+
+
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+from sklearn.metrics import accuracy_score
+print(accuracy_score(y_test, y_pred))
+\end{minted}
+
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+tree_clf = DecisionTreeClassifier(random_state=42)
+tree_clf.fit(X_train, y_train)
+y_pred_tree = tree_clf.predict(X_test)
+print(accuracy_score(y_test, y_pred_tree))
+\end{minted}
+
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+from matplotlib.colors import ListedColormap
+
+def plot_decision_boundary(clf, X, y, axes=[-1.5, 2.5, -1, 1.5], alpha=0.5, contour=True):
+    x1s = np.linspace(axes[0], axes[1], 100)
+    x2s = np.linspace(axes[2], axes[3], 100)
+    x1, x2 = np.meshgrid(x1s, x2s)
+    X_new = np.c_[x1.ravel(), x2.ravel()]
+    y_pred = clf.predict(X_new).reshape(x1.shape)
+    custom_cmap = ListedColormap(['#fafab0','#9898ff','#a0faa0'])
+    plt.contourf(x1, x2, y_pred, alpha=0.3, cmap=custom_cmap)
+    if contour:
+        custom_cmap2 = ListedColormap(['#7d7d58','#4c4c7f','#507d50'])
+        plt.contour(x1, x2, y_pred, cmap=custom_cmap2, alpha=0.8)
+    plt.plot(X[:, 0][y==0], X[:, 1][y==0], "yo", alpha=alpha)
+    plt.plot(X[:, 0][y==1], X[:, 1][y==1], "bs", alpha=alpha)
+    plt.axis(axes)
+    plt.xlabel(r"$x_1$", fontsize=18)
+    plt.ylabel(r"$x_2$", fontsize=18, rotation=0)
+plt.figure(figsize=(11,4))
+plt.subplot(121)
+plot_decision_boundary(tree_clf, X, y)
+plt.title("Decision Tree", fontsize=14)
+plt.subplot(122)
+plot_decision_boundary(bag_clf, X, y)
+plt.title("Decision Trees with Bagging", fontsize=14)
+plt.show()
+\end{minted}
+
+
+% !split
+\subsection*{Making our own Bagging with Bootstrap}
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import train_test_split
+from sklearn.pipeline import make_pipeline
+from sklearn.utils import resample
+from sklearn.tree import DecisionTreeRegressor
+
+
+np.random.seed(2018)
+
+n = 40
+n_boostraps = 100
+maxdegree = 14
+
+# Make data set.
+x = np.linspace(-3, 3, n).reshape(-1, 1)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
+error = np.zeros(maxdegree)
+bias = np.zeros(maxdegree)
+variance = np.zeros(maxdegree)
+polydegree = np.zeros(maxdegree)
+X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
+
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+for degree in range(maxdegree):
+    model = DecisionTreeRegressor(max_depth=5) 
+    y_pred = np.empty((y_test.shape[0], n_boostraps))
+    for i in range(n_boostraps):
+        x_, y_ = resample(X_train_scaled, y_train)
+        model.fit(x_, y_)
+        y_pred[:, i] = model.predict(X_test_scaled).ravel()
+
+    polydegree[degree] = degree
+    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )
+    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )
+    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )
+    print('Polynomial degree:', degree)
+    print('Error:', error[degree])
+    print('Bias^2:', bias[degree])
+    print('Var:', variance[degree])
+    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
+
+plt.plot(polydegree, error, label='Error')
+plt.plot(polydegree, bias, label='bias')
+plt.plot(polydegree, variance, label='Variance')
+plt.legend()
+plt.show()
+
+\end{minted}
+
+
+% !split
+\subsection*{Changing the Level of the Decision Tree}
+
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+
+import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import train_test_split
+from sklearn.pipeline import make_pipeline
+from sklearn.utils import resample
+from sklearn.tree import DecisionTreeRegressor
+
+n = 100
+n_boostraps = 100
+maxdepth = 8
+
+# Make data set.
+x = np.linspace(-3, 3, n).reshape(-1, 1)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
+error = np.zeros(maxdepth)
+bias = np.zeros(maxdepth)
+variance = np.zeros(maxdepth)
+polydegree = np.zeros(maxdepth)
+X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
+
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+for degree in range(1,maxdepth):
+    model = DecisionTreeRegressor(max_depth=degree) 
+    y_pred = np.empty((y_test.shape[0], n_boostraps))
+    for i in range(n_boostraps):
+        x_, y_ = resample(X_train_scaled, y_train)
+        model.fit(x_, y_)
+        y_pred[:, i] = model.predict(X_test_scaled)#.ravel()
+
+    polydegree[degree] = degree
+    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )
+    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )
+    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )
+    print('Polynomial degree:', degree)
+    print('Error:', error[degree])
+    print('Bias^2:', bias[degree])
+    print('Var:', variance[degree])
+    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
+
+plt.xlim(1,maxdepth)
+plt.plot(polydegree, error, label='Error')
+plt.plot(polydegree, bias, label='bias')
+plt.plot(polydegree, variance, label='Variance')
+plt.legend()
+plt.show()
+
+
+
+
+\end{minted}
+
+
+
+
+% !split
+\subsection*{Random forests}
+
+Random forests provide an improvement over bagged trees by way of a
+small tweak that decorrelates the trees. 
+
+As in bagging, we build a
+number of decision trees on bootstrapped training samples. But when
+building these decision trees, each time a split in a tree is
+considered, a random sample of $m$ predictors is chosen as split
+candidates from the full set of $p$ predictors. The split is allowed to
+use only one of those $m$ predictors. 
+
+A fresh sample of $m$ predictors is
+taken at each split, and typically we choose 
+
+\[
+m\approx \sqrt{p}.
+\]
+
+In building a random forest, at
+each split in the tree, the algorithm is not even allowed to consider
+a majority of the available predictors. 
+
+The reason for this is rather clever. Suppose that there is one very
+strong predictor in the data set, along with a number of other
+moderately strong predictors. Then in the collection of bagged
+variable importance random forest trees, most or all of the trees will
+use this strong predictor in the top split. Consequently, all of the
+bagged trees will look quite similar to each other. Hence the
+predictions from the bagged trees will be highly correlated.
+Unfortunately, averaging many highly correlated quantities does not
+lead to as large of a reduction in variance as averaging many
+uncorrelated quanti- ties. In particular, this means that bagging will
+not lead to a substantial reduction in variance over a single tree in
+this setting.
+
+
+% !split
+\subsection*{Random Forest Algorithm}
+The algorithm described here can be applied to both classification and regression problems.
+
+We will grow of forest of say $M$ trees.
+\begin{enumerate}
+\item For $m=1:M$ we
+\begin{itemize}
+
+  \item Draw a bootstrap sample of from the training data organized in our $\bm{X}$ matrix.
+
+  \item We grow then a random forest tree $T_m$ based on the bootstrapped data by repeating the steps outlined till we reach the maximum node size is reached
+\begin{enumerate}
+
+   \item we select $m \le p$ varibales at random from the $p$ predictors/features
+
+   \item pick the best split point among the $m$ features using either the CART algorithm or the ID3 for classification and create a new node
+
+   \item split the node into daughter nodes
+
+\end{enumerate}
+
+\noindent
+\end{itemize}
+
+\noindent
+\item Output then the ensemble of trees $\{T_m\}_1^{M}$ and make predictions for either a regression type of problem or a classification type of problem. 
+\end{enumerate}
+
+\noindent
+% !split
+\subsection*{Bootstrap with Random Forests Instead of a Single Tree, own Bagging}
+
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+
+import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import train_test_split
+from sklearn.pipeline import make_pipeline
+from sklearn.utils import resample
+from sklearn.ensemble import RandomForestRegressor
+
+np.random.seed(2018)
+
+n = 100
+n_boostraps = 100
+maxdegree = 14
+
+# Make data set.
+x = np.linspace(-3, 3, n).reshape(-1, 1)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
+error = np.zeros(maxdegree)
+bias = np.zeros(maxdegree)
+variance = np.zeros(maxdegree)
+polydegree = np.zeros(maxdegree)
+X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
+
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+for degree in range(maxdegree):
+    model = RandomForestRegressor()
+    y_pred = np.empty((y_test.shape[0], n_boostraps))
+    for i in range(n_boostraps):
+        x_, y_ = resample(X_train_scaled, y_train)
+        model.fit(x_, y_.ravel())
+        y_pred[:, i] = model.predict(X_test_scaled).ravel()
+
+    polydegree[degree] = degree
+    error[degree] = np.mean( np.mean((y_test - y_pred)**2, axis=1, keepdims=True) )
+    bias[degree] = np.mean( (y_test - np.mean(y_pred, axis=1, keepdims=True))**2 )
+    variance[degree] = np.mean( np.var(y_pred, axis=1, keepdims=True) )
+    print('Polynomial degree:', degree)
+    print('Error:', error[degree])
+    print('Bias^2:', bias[degree])
+    print('Var:', variance[degree])
+    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
+
+plt.plot(polydegree, error, label='Error')
+plt.plot(polydegree, bias, label='bias')
+plt.plot(polydegree, variance, label='Variance')
+plt.legend()
+plt.show()
+
+
+
+\end{minted}
+
+
+
+
+% !split
+\subsection*{Random Forests Compared with other Methods on the Cancer Data}
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import  train_test_split 
+from sklearn.datasets import load_breast_cancer
+from sklearn.svm import SVC
+from sklearn.linear_model import LogisticRegression
+from sklearn.tree import DecisionTreeClassifier
+
+# Load the data
+cancer = load_breast_cancer()
+
+X_train, X_test, y_train, y_test = train_test_split(cancer.data,cancer.target,random_state=0)
+print(X_train.shape)
+print(X_test.shape)
+# Logistic Regression
+logreg = LogisticRegression(solver='lbfgs')
+logreg.fit(X_train, y_train)
+print("Test set accuracy with Logistic Regression: {:.2f}".format(logreg.score(X_test,y_test)))
+# Support vector machine
+svm = SVC(gamma='auto', C=100)
+svm.fit(X_train, y_train)
+print("Test set accuracy with SVM: {:.2f}".format(svm.score(X_test,y_test)))
+# Decision Trees
+deep_tree_clf = DecisionTreeClassifier(max_depth=None)
+deep_tree_clf.fit(X_train, y_train)
+print("Test set accuracy with Decision Trees: {:.2f}".format(deep_tree_clf.score(X_test,y_test)))
+#now scale the data
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+# Logistic Regression
+logreg.fit(X_train_scaled, y_train)
+print("Test set accuracy Logistic Regression with scaled data: {:.2f}".format(logreg.score(X_test_scaled,y_test)))
+# Support Vector Machine
+svm.fit(X_train_scaled, y_train)
+print("Test set accuracy SVM with scaled data: {:.2f}".format(logreg.score(X_test_scaled,y_test)))
+# Decision Trees
+deep_tree_clf.fit(X_train_scaled, y_train)
+print("Test set accuracy with Decision Trees and scaled data: {:.2f}".format(deep_tree_clf.score(X_test_scaled,y_test)))
+
+
+from sklearn.ensemble import RandomForestClassifier
+from sklearn.preprocessing import LabelEncoder
+from sklearn.model_selection import cross_validate
+# Data set not specificied
+#Instantiate the model with 500 trees and entropy as splitting criteria
+Random_Forest_model = RandomForestClassifier(n_estimators=500,criterion="entropy")
+Random_Forest_model.fit(X_train_scaled, y_train)
+#Cross validation
+accuracy = cross_validate(Random_Forest_model,X_test_scaled,y_test,cv=10)['test_score']
+print(accuracy)
+print("Test set accuracy with Random Forests and scaled data: {:.2f}".format(Random_Forest_model.score(X_test_scaled,y_test)))
+
+
+import scikitplot as skplt
+y_pred = Random_Forest_model.predict(X_test_scaled)
+skplt.metrics.plot_confusion_matrix(y_test, y_pred, normalize=True)
+plt.show()
+y_probas = Random_Forest_model.predict_proba(X_test_scaled)
+skplt.metrics.plot_roc(y_test, y_probas)
+plt.show()
+skplt.metrics.plot_cumulative_gain(y_test, y_probas)
+plt.show()
+
+\end{minted}
+
+
+% !split
+\subsection*{Compare  Bagging on Trees with Random Forests}
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+bag_clf = BaggingClassifier(
+    DecisionTreeClassifier(splitter="random", max_leaf_nodes=16, random_state=42),
+    n_estimators=500, max_samples=1.0, bootstrap=True, n_jobs=-1, random_state=42)
+\end{minted}
+
+
+
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+bag_clf.fit(X_train, y_train)
+y_pred = bag_clf.predict(X_test)
+from sklearn.ensemble import RandomForestClassifier
+rnd_clf = RandomForestClassifier(n_estimators=500, max_leaf_nodes=16, n_jobs=-1, random_state=42)
+rnd_clf.fit(X_train, y_train)
+y_pred_rf = rnd_clf.predict(X_test)
+np.sum(y_pred == y_pred_rf) / len(y_pred) 
+\end{minted}
+
+
+
+
+
+% !split
+\subsection*{Boosting, a Bird'e Eye}
+
+The basic idea is to combine weak classifiers in order to create a good
+classifier. With a weak classifier we often intend a classifier which
+produces results which are only slightly better than we would get by
+random guesses.
+
+This is done by applying in an iterative way a weak (or a standard
+classifier like decision trees) to modify the data. In each iteration
+we emphasize those observations which are misclassified by weighting
+them with a factor.
+
+
+% !split
+\subsection*{Adaptive boosting: AdaBoost, Basic Algorithm}
+
+The algorithm here is rather straightforward. Assume that our weak
+classifier is a decision tree and we consider a binary set of outputs
+with $y_i \in \{-1,1\}$ and $i=0,1,2,\dots,n-1$ as our set of
+observations. Our design matrix is given in terms of the
+feature/predictor vectors
+$\bm{X}=[\bm{x}_0\bm{x}_1\dots\bm{x}_{p-1}$. Finally, we define also a
+classifier determined by our data via a function $G(\bm{X})$. This function tells us how well we are able to classify our outputs/targets $\bm{y}$. 
+
+We can then define the misclassification error $\mathrm{err}$ as
+\[
+\mathrm{err}=\frac{1}{n}\sum_{i=0}^{n-1}I(y_i\ne G(\bm{X}_{i*}),
+\]
+where the function $I()$ is one if we misclassify and zero if we classify correctly. 
+
+% !split
+\subsection*{Basic Steps of AdaBoost}
+
+With the above definitions we are now ready to set up the algorithm for AdaBoost.
+The basic idea is to set up weights which will be used to scale the correctly classified and the misclassified cases.
+\begin{enumerate}
+\item We start by initializing all weights to $w_i = 1/n$, with $i=0,1,2,\dots n-1$. It is to see then that $\sum_{i=0}^{n-1}w_i = 1$.
+
+\item We rewrite the misclassification error as 
+\end{enumerate}
+
+\noindent
+\[
+\mathrm{err}=\frac{\sum_{i=0}^{n-1}w_iI(y_i\ne G(\bm{X}_{i*})}{\sum_{i=0}^{n-1}w_i},
+\]
+\begin{enumerate}
+\item Then we start looping over all attempts at classifying, namely we start an iterative process for $m=1:M$, where $M$ is the final number of classifications. Our given classifier could for example be a plain decision tree.
+\begin{enumerate}
+
+ \item Fit then a given classifier to the training using the weights $w_i$.
+
+ \item Compute then $\mathrm{err}$ and figure out which events are classified properly and which are classified wrongly.
+
+ \item Define a quantity $\alpha_{m} = \log{(1-\mathrm{err})/\mathrm{err}}
+
+ \item Set the new weights to $w_i = w_i\times \exp{(\alpha_m I(y_i\ne G(\bm{X}_{i*})}.
+
+\end{enumerate}
+
+\noindent
+\item Compute the new classifier $G(\bm{X})= \sum_{i=0}^{n-1}\alpha_m I(y_i\ne G(\bm{X}_{i*}).
+\end{enumerate}
+
+\noindent
+For the iterations with $m \le 2$ the weights are modified
+individually at each steps. The obersvations which were misclassified
+at iteration $m-1$ have a weight which is larger than those which were
+classified properly. As this proceeds, the observations which were
+difficult to classifiy correctly are given a larger influence. Each
+new classification step $m$ is then forced to concentrate on those
+observations that are missed in the previous iterations.
+
+
+
+% !split
+\subsection*{AdaBoost Examples}
+
+Using \textbf{Scikit-Learn} it is easy to appply the adaptive boosting algorithm, as done here.
+
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+from sklearn.ensemble import AdaBoostClassifier
+
+ada_clf = AdaBoostClassifier(
+    DecisionTreeClassifier(max_depth=1), n_estimators=200,
+    algorithm="SAMME.R", learning_rate=0.5, random_state=42)
+ada_clf.fit(X_train, y_train)
+
+plot_decision_boundary(ada_clf, X, y)
+
+m = len(X_train)
+
+plt.figure(figsize=(11, 4))
+for subplot, learning_rate in ((121, 1), (122, 0.5)):
+    sample_weights = np.ones(m)
+    plt.subplot(subplot)
+    for i in range(5):
+        svm_clf = SVC(kernel="rbf", C=0.05, gamma="auto", random_state=42)
+        svm_clf.fit(X_train, y_train, sample_weight=sample_weights)
+        y_pred = svm_clf.predict(X_train)
+        sample_weights[y_pred != y_train] *= (1 + learning_rate)
+        plot_decision_boundary(svm_clf, X, y, alpha=0.2)
+        plt.title("learning_rate = {}".format(learning_rate), fontsize=16)
+    if subplot == 121:
+        plt.text(-0.7, -0.65, "1", fontsize=14)
+        plt.text(-0.6, -0.10, "2", fontsize=14)
+        plt.text(-0.5,  0.10, "3", fontsize=14)
+        plt.text(-0.4,  0.55, "4", fontsize=14)
+        plt.text(-0.3,  0.90, "5", fontsize=14)
+
+save_fig("boosting_plot")
+plt.show()
+\end{minted}
+
+
+% !split
+\subsection*{Gradient boosting: Basics}
+
+Gradient boosting is again a similar technique to Adapative boosting,
+it combines so-called weak classifiers or regressors into a strong
+method via a series of iterations.
+
+In order to understand the method, let us illustrate its basics by
+bringing back the essential steps in linear regression, where our cost
+function was the least squares function.
+
+% !split
+\subsection*{Gradient Boosting, algorithm}
+
+Suppose we have a cost function $C(f)=\sum_{i=0}^{n-1}L(y_i, f(x_i))$ where $y_i$ is our target and $f(x_i)$ the function which is meant to model $y_i$. The above cost function could be our standard least squares  function
+\[
+C(\bm{y},\bm{f})=\frac{1}{n}\sum_{i=0}^{n-1}(y_i-f(x_i))^2.
+\]
+
+The way we proceed in an iterative fashion is to
+\begin{enumerate}
+\item Initialize our estimate by $f_0(x)=0$.
+
+\item For $m=1:M$, we
+\begin{enumerate}
+
+ \item compute the negative gradient vector $\bm{u}_m = -\partial C(\bm{y},\bm{f})/\partial \bm{f}(x)$ at $f(x) = f_{m-1}(x);
+
+ \item fit the so-called base-learner to the negative gradient $h_m(u_m,x)$;
+
+ \item update the estimate $f_m(x) = f_{m-1}(x)+\nu h_m(u_m,x)$;
+
+\end{enumerate}
+
+\noindent
+\item The final estimate is then $f_M(x) = \sum_{m=1}^M\nu h_m(u_m,x)$.
+\end{enumerate}
+
+\noindent
+% !split
+\subsection*{Gradient Boosting, Examples}
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+np.random.seed(42)
+X = np.random.rand(100, 1) - 0.5
+y = 3*X[:, 0]**2 + 0.05 * np.random.randn(100)
+
+from sklearn.tree import DecisionTreeRegressor
+
+tree_reg1 = DecisionTreeRegressor(max_depth=2, random_state=42)
+tree_reg1.fit(X, y)
+
+y2 = y - tree_reg1.predict(X)
+tree_reg2 = DecisionTreeRegressor(max_depth=2, random_state=42)
+tree_reg2.fit(X, y2)
+
+y3 = y2 - tree_reg2.predict(X)
+tree_reg3 = DecisionTreeRegressor(max_depth=2, random_state=42)
+tree_reg3.fit(X, y3)
+
+X_new = np.array([[0.8]])
+y_pred = sum(tree.predict(X_new) for tree in (tree_reg1, tree_reg2, tree_reg3))
+
+def plot_predictions(regressors, X, y, axes, label=None, style="r-", data_style="b.", data_label=None):
+    x1 = np.linspace(axes[0], axes[1], 500)
+    y_pred = sum(regressor.predict(x1.reshape(-1, 1)) for regressor in regressors)
+    plt.plot(X[:, 0], y, data_style, label=data_label)
+    plt.plot(x1, y_pred, style, linewidth=2, label=label)
+    if label or data_label:
+        plt.legend(loc="upper center", fontsize=16)
+    plt.axis(axes)
+
+plt.figure(figsize=(11,11))
+
+plt.subplot(321)
+plot_predictions([tree_reg1], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="$h_1(x_1)$", style="g-", data_label="Training set")
+plt.ylabel("$y$", fontsize=16, rotation=0)
+plt.title("Residuals and tree predictions", fontsize=16)
+
+plt.subplot(322)
+plot_predictions([tree_reg1], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="$h(x_1) = h_1(x_1)$", data_label="Training set")
+plt.ylabel("$y$", fontsize=16, rotation=0)
+plt.title("Ensemble predictions", fontsize=16)
+
+plt.subplot(323)
+plot_predictions([tree_reg2], X, y2, axes=[-0.5, 0.5, -0.5, 0.5], label="$h_2(x_1)$", style="g-", data_style="k+", data_label="Residuals")
+plt.ylabel("$y - h_1(x_1)$", fontsize=16)
+
+plt.subplot(324)
+plot_predictions([tree_reg1, tree_reg2], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="$h(x_1) = h_1(x_1) + h_2(x_1)$")
+plt.ylabel("$y$", fontsize=16, rotation=0)
+
+plt.subplot(325)
+plot_predictions([tree_reg3], X, y3, axes=[-0.5, 0.5, -0.5, 0.5], label="$h_3(x_1)$", style="g-", data_style="k+")
+plt.ylabel("$y - h_1(x_1) - h_2(x_1)$", fontsize=16)
+plt.xlabel("$x_1$", fontsize=16)
+
+plt.subplot(326)
+plot_predictions([tree_reg1, tree_reg2, tree_reg3], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="$h(x_1) = h_1(x_1) + h_2(x_1) + h_3(x_1)$")
+plt.xlabel("$x_1$", fontsize=16)
+plt.ylabel("$y$", fontsize=16, rotation=0)
+
+save_fig("gradient_boosting_plot")
+plt.show()
+
+from sklearn.ensemble import GradientBoostingRegressor
+
+gbrt = GradientBoostingRegressor(max_depth=2, n_estimators=3, learning_rate=1.0, random_state=42)
+gbrt.fit(X, y)
+
+gbrt_slow = GradientBoostingRegressor(max_depth=2, n_estimators=200, learning_rate=0.1, random_state=42)
+gbrt_slow.fit(X, y)
+
+plt.figure(figsize=(11,4))
+
+plt.subplot(121)
+plot_predictions([gbrt], X, y, axes=[-0.5, 0.5, -0.1, 0.8], label="Ensemble predictions")
+plt.title("learning_rate={}, n_estimators={}".format(gbrt.learning_rate, gbrt.n_estimators), fontsize=14)
+
+plt.subplot(122)
+plot_predictions([gbrt_slow], X, y, axes=[-0.5, 0.5, -0.1, 0.8])
+plt.title("learning_rate={}, n_estimators={}".format(gbrt_slow.learning_rate, gbrt_slow.n_estimators), fontsize=14)
+
+save_fig("gbrt_learning_rate_plot")
+plt.show()
+
+\end{minted}
+
+
+% !split
+\subsection*{Gradient Boots with Early Stopping}
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+
+from sklearn.model_selection import train_test_split
+from sklearn.metrics import mean_squared_error
+
+X_train, X_val, y_train, y_val = train_test_split(X, y, random_state=49)
+
+gbrt = GradientBoostingRegressor(max_depth=2, n_estimators=120, random_state=42)
+gbrt.fit(X_train, y_train)
+
+errors = [mean_squared_error(y_val, y_pred)
+          for y_pred in gbrt.staged_predict(X_val)]
+bst_n_estimators = np.argmin(errors) + 1
+
+gbrt_best = GradientBoostingRegressor(max_depth=2,n_estimators=bst_n_estimators, random_state=42)
+gbrt_best.fit(X_train, y_train)
+
+min_error = np.min(errors)
+plt.figure(figsize=(11, 4))
+
+plt.subplot(121)
+plt.plot(errors, "b.-")
+plt.plot([bst_n_estimators, bst_n_estimators], [0, min_error], "k--")
+plt.plot([0, 120], [min_error, min_error], "k--")
+plt.plot(bst_n_estimators, min_error, "ko")
+plt.text(bst_n_estimators, min_error*1.2, "Minimum", ha="center", fontsize=14)
+plt.axis([0, 120, 0, 0.01])
+plt.xlabel("Number of trees")
+plt.title("Validation error", fontsize=14)
+
+plt.subplot(122)
+plot_predictions([gbrt_best], X, y, axes=[-0.5, 0.5, -0.1, 0.8])
+plt.title("Best model (%d trees)" % bst_n_estimators, fontsize=14)
+
+save_fig("early_stopping_gbrt_plot")
+plt.show()
+
+
+gbrt = GradientBoostingRegressor(max_depth=2, warm_start=True, random_state=42)
+
+min_val_error = float("inf")
+error_going_up = 0
+for n_estimators in range(1, 120):
+    gbrt.n_estimators = n_estimators
+    gbrt.fit(X_train, y_train)
+    y_pred = gbrt.predict(X_val)
+    val_error = mean_squared_error(y_val, y_pred)
+    if val_error < min_val_error:
+        min_val_error = val_error
+        error_going_up = 0
+    else:
+        error_going_up += 1
+        if error_going_up == 5:
+            break  # early stopping
+
+
+print(gbrt.n_estimators)
+print("Minimum validation MSE:", min_val_error)
+\end{minted}
+
+% !split
+\subsection*{XGBoost: Extreme Gradient Boosting}
+
+
+\href{{https://github.com/dmlc/xgboost}}{XGBoost} or Extreme Gradient
+Boosting, is an optimized distributed gradient boosting library
+designed to be highly efficient, flexible and portable. It implements
+machine learning algorithms under the Gradient Boosting
+framework. XGBoost provides a parallel tree boosting that solve many
+data science problems in a fast and accurate way. See the \href{{https://arxiv.org/abs/1603.02754}}{article by Chen and Guestrin}.
+
+The authors design and build a highly scalable end-to-end tree
+boosting system. It has  a theoretically justified weighted quantile
+sketch for efficient proposal calculation. It introduces a novel sparsity-aware algorithm for parallel tree learning and an effective cache-aware block structure for out-of-core tree learning.
+
+It is now the algorithm which wins essentially all ML competitions!!!
+
+% !split
+\subsection*{Regression Case}
+
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import train_test_split
+import xgboost as xgb
+from sklearn.preprocessing import StandardScaler
+import scikitplot as skplt
+from sklearn.metrics import mean_squared_error
+
+n = 100
+maxdegree = 6
+
+# Make data set.
+x = np.linspace(-3, 3, n).reshape(-1, 1)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.normal(0, 0.1, x.shape)
+
+error = np.zeros(maxdegree)
+bias = np.zeros(maxdegree)
+variance = np.zeros(maxdegree)
+polydegree = np.zeros(maxdegree)
+X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+for degree in range(maxdegree):
+    model =  xgb.XGBRegressor(objective ='reg:squarederror', colsample_bytree = 0.3, learning_rate = 0.1,
+                max_depth = degree, alpha = 10, n_estimators = 10)
+    model.fit(X_train_scaled,y_train)
+    y_pred = model.predict(X_test_scaled)
+    polydegree[degree] = degree
+    error[degree] = np.mean( np.mean((y_test - y_pred)**2) )
+    bias[degree] = np.mean( (y_test - np.mean(y_pred))**2 )
+    variance[degree] = np.mean( np.var(y_pred) )
+    print('Max depth:', degree)
+    print('Error:', error[degree])
+    print('Bias^2:', bias[degree])
+    print('Var:', variance[degree])
+    print('{} >= {} + {} = {}'.format(error[degree], bias[degree], variance[degree], bias[degree]+variance[degree]))
+
+plt.xlim(1,maxdegree-1)
+plt.plot(polydegree, error, label='Error')
+plt.plot(polydegree, bias, label='bias')
+plt.plot(polydegree, variance, label='Variance')
+plt.legend()
+plt.show()
+
+
+\end{minted}
+
+
+
+% !split
+\subsection*{Xgboost on the Cancer Data}
+\begin{minted}[fontsize=\fontsize{9pt}{9pt},linenos=false,mathescape,baselinestretch=1.0,fontfamily=tt,xleftmargin=7mm]{python}
+import matplotlib.pyplot as plt
+import numpy as np
+from sklearn.model_selection import  train_test_split 
+from sklearn.datasets import load_breast_cancer
+from sklearn.preprocessing import LabelEncoder
+from sklearn.model_selection import cross_validate
+import scikitplot as skplt
+import xgboost as xgb
+# Load the data
+cancer = load_breast_cancer()
+
+X_train, X_test, y_train, y_test = train_test_split(cancer.data,cancer.target,random_state=0)
+print(X_train.shape)
+print(X_test.shape)
+#now scale the data
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+scaler.fit(X_train)
+X_train_scaled = scaler.transform(X_train)
+X_test_scaled = scaler.transform(X_test)
+
+xg_clf = xgb.XGBClassifier()
+xg_clf.fit(X_train_scaled,y_train)
+xgb.plot_tree(xg_clf,num_trees=0)
+plt.rcParams['figure.figsize'] = [50, 10]
+plt.show()
+xgb.plot_importance(xg_clf)
+plt.rcParams['figure.figsize'] = [5, 5]
+plt.show()
+\end{minted}
+
+% ------------------- end of main content ---------------
+
+\end{document}
+
diff --git a/doc/src/DecisionTrees/_minted-DecisionTrees/default.pygstyle b/doc/src/DecisionTrees/_minted-DecisionTrees/default.pygstyle
new file mode 100644
index 000000000..e69de29bb
diff --git a/doc/src/DecisionTrees/ipynb-DecisionTrees-src.tar.gz b/doc/src/DecisionTrees/ipynb-DecisionTrees-src.tar.gz
new file mode 100644
index 0000000000000000000000000000000000000000..0ff487d36292b6af24323c88adb0fa4b5eb3d383
GIT binary patch
literal 294061
zcmV(?K-a$?iwFSK7sOov1Js;lR9s!N=W&7)f&_;^LPBs4*0=-+?(PI@B)A7la0%{#
z;KAK3KyZR<9tth9J){Mb9U{j-@mFhVuqg#B|ci4Ix-s@
z+8UcWFxlCf!~H*gu&}Ujad1#QK67(%{_B~A?LU48{=&t=$<0c|%FfBj1-xKo;i6(;
z<>X{xfumyie?5@@>EYz~$!|M7jm+sw@VFX#E!#jsHQ*Yp1;kFN?J
zB(cy*(1Dj&(o$kdaB#?HaBv8ppP&Fo5>hp!fhTxJB}oyuk|DBf;1@Ad4QVquIXHUY
z^%FSw;E!-fk8c5fNd9#~#7uZN#D5(Dj%6bJ`>)7nnTY>>jqv&LjbUxxl5lWBaMEJJ
zDz5N*3&^hc(j+KLI4(HxnSQvR8=uXS6S~ymJ~=oD^{@*@CUnsy``n1+qKOxS{CN|~
zT(kL#BEHw;WI`Zzy99W+dmPx-Hb6^XFKQt?Lw+p;kFMa0fS(EXKfHX!h&z#^-P5i3
ze|p2eUS=ZkWDZjz{NGMRh44fy63)it*&CQY;Fcf&!{Kq)MjHDkz
zT4^lDDNjw#Y_22lUZT@Bmfg!LoQ>GvUh`d~pG>wezS
zS96FzxsR652w&Y<*JMw;YSSEc&*(K?>*ce%N$exzc6UiLcQWU8bHS*T_aU6)NrtZm;py?P@x-dNv<>wmni=oKGN%@_4UB((@o>W`FhufWyBBy#*k0je^$ow
z9Fw}%YOK}^5>x!UX$f5el`;b6TFA6_fv
z*s%;5Jw*>FY1Xa;Wj2_coZmvQU=}fPiTAlGC}6bt~TgbZ|!cc&^T;Hl?KNtdF(Ppi)7epPyHlZ{eU<$@fiSHxJsW)_C{VmVCFL*eU9a
z)e~8x{XHnegD#uSPb`zb^lOg0G@8Y*!_FFwSciLJae4IzdjsTgfmy2m5I$e;ct=k+j6>sh
zb-cQI9%|mJor&wPG4QoewJ2yb&W-r5z55;EeId~!+Hit3b^~+-{T5H*DJI=A*HMI6
z^h(Fe16|DL9M8dvO=!<@>xWJif4GM$)K6|OYO{C@%_|h}<-FaW$Ya*3vfR&`UGTWb
zl=#V8CYwS*K(E?fve4-DO|!-5WnX;hmpU}Og2S{%=W8C@rqioqe9q4!NWru+@tFMyOhKYkC0A$LvB%W5Fb1EP1&4Yz
zc$8Zq1G6^Ti8RWMkl9R!0?^m{Q6F+xUb{WCBg!UT@h!G^k)o(0@>W@e?rU=e_flhqdkgO(R(h*yLkWh(-tZ6j3QdgK&ix18^
z9ZxNH7N%k@*4jYe)zspr>kP^N{E!LNa@twm9?KKli6$rhRI1$y|DmzYpu6hLRxVbT
zd@`G=u6D%3=`d)z)(1qdUKmjGQ=3$fD~Y||&_{V&=|_}+q%#3q9|U0_=dW90AqA#B
zW$>qoB#fiF5A{85aiE?fs6Wd#%)jbU$DBl;Ewpc!u}g#L~gE{c3cn0d-Er`%vEV
z9Z?ZTDxV7lc7(vV+c^nGSGjQy*3=lcNlo2=J9xV8D{rmE*BBnRlW*P{Uc*UzIM)~i
z5)ENVD)o|)6a}6SvlS0*!eQhfVRHR4yQeuB?B;_KL;P;4F9u#?Z)TYuP=mZEWtU`gyX=BMs`SOV~8N?
zqmy-+y2W~b2h)BOS2e&NjtiCJyWT;oxpjV+
zXxl_Nchr@;=3`|h&}-`rR$QK(d=Bj=wlQdaV{+l;cY9rKH)<3%igKpSVHH%8#<#pz
z?>H;uLX`794$rjzo0^l_*vfZptz5OaZl4GWz98ELkG46|El4d6;4*^IvvO;QPcK
z#h_M|%K1zJ!`IvE6T{npuRL}~22oRh8>C`0tx!W;Ip3Wao81k6uUcfd0L98zZj8Su
z2#3(gsZm?xuszJ?GU2sl+-NnL3JJ(T!z^Z-kaPAO!hHqrRb#1y5p4WsVs~QSs(%29
z+cq|Iw3wh#UUP9J*`LokIi)qa9DeMP5uq63_&BMvpiCTg2^K&gE{h~)e}zuOWRlMu
z8XitE(CYKx2NmcJA&{9+tag`McQ{^oDLUiJwFZ^14wkrYCcqteH<0&tfqn5sm<4UI
z!EytY68he3b$_Es;K?GS)f4@w(qvS`rk8AyIbHDL-tmrd5m*lw5rmA%wS-!~=Em;A
zXZ0#AfABJ0d)=IIjXUi^RhC>MJn(%II@fvRlo`WooW=meC
zbJ&YSO~vfif~O|=j+C(VvLto(0wxG$#s?!0&Y3{SNdsr_7OcG{hkx`voKC;yFxgnnDI
zpr!vlWm2xBZ)N(0ZuH7KBB!A}@~I;t6{$kk4(mskDXsQvTdWrTb(XASKtx#Xiz~Xi
z*e89TP{BQJo;}zUufSMq3z3NHi&OEA?Z$e_xf4lb>hb9iMH7y**~v_$R82JnOJ8$v
zk6Q(|OX7`(hlT%Dp-K=YaTKeu1ZU?Bu&9oXj@pW*N~yqDThbm&nX^hw8j5kI^MnA~
z{S?6UO-oCgm9vf3@lr3*Q1t2#yQs5Y?NAQAq!_P0gPt9pZZhW&-l1aW<8oJ=F13pi
zI7pi1?}GV|#<|@1zDo`-MwJPlm<3+2AXi&Ws8lYVTVay2O+)Os9r#^Jbll4gd}O33
z(DF=Dt58Yk_l8b4L5p9?#=RwphSU7G@=H7DrJv_sV-Ok%J$QHL6>#I89f?5U3uIz0
zeBsI6s@$%d?B|XY|N51ApKnpZ?H;h
zzh*kB>T>zOobxuz^=!OJQ+aJ5S`-p!uZ@}XU~2(}yBhlJ%srb3Nkm^Y?<-*&S6?b86`r?6aNDYysYz*5I_04+#j^IZn%ALYty(ZFW6gDd<#6z}zwaZG>nO)Oul;m%Pm3
zrl2705w+s55k}3NiRD&Da}H~uiKTGHTPcUlro%8{{4;Kp#s*#~e?Bc@p~uC+uRshU
z)#5X@A2J2|Hghid?P~8pqEO625={+Nd6%lI+k@F9pAFfhGX_dXc2AgUWtnODL=VnE
z!OS(E^T9Zw#5o`C-fEYRi|t~AX>WYe37eIGM~!t^%~1Y{3_de9V^;)pAcb8!bT9*b
zd+Y~$>N7l2kO!mHJ|lwkLZcIV)lcoIl97(t{0roX-12NmVGOSTl{d;+YH9A4?GWnD
zAS=WJ9L&7VJuU1o~McgcFI&RNI4;+WcW~gyIO@4yb7e8vm=m9E(r
zb@5-G@kQ`%cKue4iekS15m%&U+&}q~eqO%z-~;`)NIyg%1g(_@hb6h%IuHQgkL+j9
zkY+fD%i;q+tq%tA)!GUz^4g`h!cB73(0h(X(@2ii$XQ`h@^(Sw4KS!sKOGYerWC?^
z5L>i*Trxo~C^HXAo6x9Q{he**EJB}v+xOl%Qe#Xz+i&Hf+?hT#klyG0)RjMKJnkI6
zGchjUytAOaP&=qylYc{dDJSocp;oF{*m
ztsy^m=vf0lC6$7aX^RMKW4b6mLW$nyqS(Fk_RJDNo)9YXl!@_MAdpGO)bkJOHA>K3
zmzIiCeH&kpH0!=aDzMRvzqP#eMM5qmOJ&BP&~vux;{=1(J7%AI&k@1_@%~AhDg|z|
zD!-%T!m#5EK)PfZuI5H5@V!yWX)7ikQ+1_!Q`zQLVku+%(_30;^Q=HQN;E~435Sf9
zxK_O8-0OOB^|t8q{cZHBVzomILSsBx!a1VJh=6&z=hfLVc|4ptR#$U&N0R$TcLoFBIGD)oBdAuYi{VB}O{+Xe!kNj4kD0Upa_dt$
z7ZTCldX8r+9i#Jbu;8&Z*&IBZKzvhaKBR!-Lu(AVM+{PsJeixRaeG!cQ5-mbeNuFS
zXC}Q9#bG>)?|F43EZ{r3dSMlc-kD`ZwJfEod(Q(bpY|P2_1jyzmXN4gJZ;qqqOFI{Q=Kea%EYD!Bar6<&X{9q9HY%#__Ru
zoRq-5<$iyAyI_G&eR|pi`$C&wAUw8b>Ni{b`T_=NUS)d#?%hvJV{dj>;CAr4Qr
zYXTH@uY!H+6)qRlzGi-O|u7s{<}@M_j`#052+--o5wxlL-rI3!9%>mwy%$
zE(tg_z&+ayCgHE9#0FzND}^5&}F-Hk>YiF;uWMKLIbzpmg=xw$&l8)qMTqiJrKmE}@7h|~C-RHowk@LtLy-LRRKd-Y)7s+FdQ7Ub4-O)%5}03j-(Jy!oOZ^8f;mcD4yfc0=C?*#(w4~_7Y#7I99tUqNdL=h
zuBZ*tA1*o?Wnitr;7Mhyp8aN!bk6XK`;;#R^s4j8-_2?k!fLg69|W57K$}
zKWo>5i}phkn6kAmDe4?AnRs3HG6UP(B60_{m+o#ZT`vy=uqoett_4G&{W-(vmxptg
zXu}yGb*I?`216bb<5*1_{Rt!JO~P8cm@y-R{Hd9oVu>{|rMO93%eQ6$=Y!dBuKn50
z_HsV99QjBfa27h$f*U(J#}ENH5vnGsW-*zzb9HsS%I{2del7}0J$Q8^@?TeSOSI(M
z^1w9k14d~fA#CtwU$6HkXjFV6M!_NT-B!*sm>B8Z0VSTmR=8Z`#0Q5MkAGrp4os0-KCKe~~5Xh!-y=Pea%~kn(TBPU;EE+6SrE71%KI^W=>|341blq2VXWT7SY@^EzgN
zzJ#7HirH8eOlUDmdEe1GS`ThSL0+WMq-
z3?{?(qQtj;$ReJFJqC|f%AoFDg`Ks*8{7H1piXF83pI@|79Ss9FJ0Xm@e};Qtvgg4
zg~D&mv+>;Uw5jsg`&jUhkF#e8A9Zt%jN%7L9Z9iL${r`YMg$;hKn1kC!=P+$Hu
z7-zeJp12#faklL}PzG|oty|uxOMPcGja?9~S-DB2sD$AI3Jf7UbtL|pK*{H9E#*ok
zupK(T(R<+1z7T}9(^;PKm;f&nOe+34p^a3LL2pV=z^vba;RQXDF2%;76jTCZ=`_1z
znf9x9l`J%3^_ZZv@(V?Ye;#B_+C7{hJ0!Rf`Qp;2y>)u=p-3;u~&jPnR`?y)arHt0ZlO|JRswb-<0;S6>utfCTU2Xpe(
z6W!esap0*enm>6m!PHs<50gK&p!vq#1DZT;JNN*fWS{UkTSS5;f2tdPA|ofWI~t)r
z-<}vlBWBA9?oVRh8Y#fIIGhRGWY#m7{t|5|@c=&Gf#xUh!In0@yhb1!iG1x6vJVrp
zp4h{qDZZd|l)=k2+b@i>SSTUi;kGUoIqVf`=!%F0NO-5jzBRmdJ+r0MhjgT1!0!T`GTAhNTccrBwzeG1eXD2a$N~`4)8k=cf&c)z<KlONCE0U5D5qcijHAgAnOr}+3>#y7sxjdM{
zov+x49Ya^!2H5M-*|zdyCb`_WCrEUP>D9b8S2)Zpu%X)pd7Cut?yP^1>R!xn4x=5|`t9q$k0rZu85_%8Chv#h8JQdTX!E%$|CwmWpRcigd3N
zg>_!axImPCGml*32U`nm$9KKCU?kz5=;*Xy@9PO8`C4bcTG;LdEFRB0D8XYvpkRz?
zYhet{$!e+_P@t?n?INl-!8m%BRE@g95oJ}f;r#WE#v8>my>Z+iqwcV;TtBs!m$F1H
z;q(D%vP)<-nC>r^!ksOX7tx!*jRQT~jT8?}xcc}L^Nf$jS(2oP!*1u`i_ENl;EP`}
z9b;q4ar-0n7?0JIuClDxx_uO@bh2HA6IQq5G{kH?lp{TmXH^4eT3Z<&mxsoGx{d`y
zXy1RsRi={{i6EQ7x&}Zje?O7cSg1y3X3luCWWi#rq{g;%5^MFY&EAuB$*!q@=4Kb1O;CS
zRxjAkm>X4eUlbLmV}#m&BGauY;A?fem(1RLiz<1m`*||gbDtRnHA1b0+f5Xcp;XO5
zJD;;FR&>9;p(mv1oTg|}o0`*#^$%kX(s22C?!m`15L{e@WiRK0;J6S8qgrT%?
zH-{3Nc!vA#iv(1}laaR8Z=6`AAe
z{KHMjXG)w~N~Q5~_5pq>gn&Nt{rdaoY!*~XKi>bK*Qhokj%U<|-$cf+)7Hkk02sfl
zmU%n-#OPN@p%-y+ZCRPn`g;Pb9_3uwER_25uut263fetq5>8CtC#Y3I&sR^ph&`hmV<;lF-;x;Yj~(gNUC*ta*z
zBp2b_CN<56w95`8Rv5GI^U1ao>=P+hEBQcL|I_Ijc*)N$wjVylpkBIkxX1G+g??4hn?KW)_$TTrrW{#vtapY3F5_J0BBa5Mrck}eEMJ@f!;8WUHfM%riR54-SlU$nt{Dc3)>R@@zV21x7~iIPcFq3)4bma@ZWwJy)Qz9H;CD-#J~eRG1HMtRh+M
zhFP|X9j_GXR?rx_(qevaw+VXiN5x^8aubA}Z-+;a&#nTpC10(QFn2I<*<489>==iH
zZCvsW8u8PHur8H9>#*=p=bhvCQb2>_A6ca+Id{Gxy_jA{Zx&fG7TPwOvS|v6D!p{kr@hEHYV-!4>7dgJ`68jh(2rp5ML@ywH2dQY
zw}B*1cAb)kw?uK`dXyk<1mfqBjjfeFYqFPq)Z-^WQOj+pOxQqo;ljD0Ru&_$B}ssonB_;95PAR?sYSXI8!mxetAwy76CwbubFus;RRNH
z`VkP!p4u)p{Yd)3Szm)Zl?_0EN%b@Kz>|~tcFJXF}HkIxXz~BZ8k_jB|_aSh3
zO-+#JtxU_Qa*oupe!0QsZ&^>TTrUH<-^EwOnw2OTVuTPz^%`s6p3YOD?_yuaJEM=R
zJ4^Mnz5DMOKi?LepLoK2)!5Uk@`||as$j;ffAl)ZiV2}C#lDiZeDTwe{hvKhLnE8u
zfAmH@RS_a=Z}z+n;d;ABc)Z$!bF3$ubmmdE%kR9$(GO?`h
zzw8d?uUVQ39}+AV3ptayz+dnEQ3?=REBjt`n^UWUWFi0K+H)YJyPj!7;IiE!dju6@3xqjp3~mR9wQmaXk^#FsMoJwYJu
z?`mo8k2w;2dX*8oW?wD}0r!CBn~Ng9+Z1l_%4$!fV2ECWCvhZ&K=&6zQLnY0Fyd#|
zCmVs7-$?WJ`J7kAx5q5)At_TOnQr}RM`yP^by9WKrOHm_S*iV0p?1oZO$h=554Cv6S{bt@-fDN4#NKo{*1D-E
zN}+bOP4F*_4I1ql>z(Q18ArRbY1Z3vv+7xp&ppCp#Il+y?c|anJlk5tssCIQN0_`wp@p9dJQJenxMARy^-`GYZvOhzPOl6O5Fp33#k*;Iy|H}Uyj9@M
zP1gdYi>S!-5}o=GKmT9@zf=CT;SBlVjenp|hnO$5OGiKw5zGQ&5rcdx~D1s
zVQR@UtZ$IW$z-Uv@oreMJB*l`+GvNzrT_$Mgv&-O;i0d=ODWmI!Goketc{B!u?;&4u~-C$1&
zlyfIX%FXqy#c@-TCoPfHNC;*8<&X0Bh;E6aA{Y?Ii}`ntxidDI#f3Y@Von@^-<@Hu
z$1BT07q~P$1rZz08{^V@^EDHYm4D~)s<{zpQ-=X(y9_+?_AU=0qb1
znZ412l!M;frgs+XHZlCoCU77u-+-B;gObR9{Ns)g7sm_F9?7z?F!*1&%U-25It#D-
z<{mpSRd@&Bu^N+h$U(#ci^US$H%4WY6fW@f_Nz#85d7$OasU3`t6K|2bX>*Sp@6zK
z9MUldISII(AdMPzV;?!{rYt|Ozi~g?D!iMid~kJKin)jqUVIfLe67Z)ZBc%5Sg=E{
zQu5tw@o=UQXP@_zSpTGrig!usyYh!NVSX>D%KqVW{WBoWo*W;on9r^@7Vna)r}8>!
z1uob-x^yO%uU1u~g#h&N03wTa}R4paonszE~DR5dtO&&`kd>jqOMFORqdy=@^6SKfaZVd21r>Jw}Gbo1jn-nIbJ?}m$zY^{H{k`t3T1f|h4&!+$GY(SVG~y&S)6Yd4
zxO5-LOva54gT*ZWW)4-HPJ+aUf^F$B6{eWduo4n;sv9li=-Wf)z&vj-xp_oEnB{J7
z{$>+CAdrpV(7Ect7C~;2O`KZ>EU|}aaR^8_tus;fO(m`((i&MTyX+=%0D9=)(#EdV
zEf;Z9DO6a_TYT+$x)Gjbg;(<`h;VUZq0uF%>AN^OjpBE)ATKXbvmtJ)SuqG!G)Y(7rwfQ*1xlpUxgHlXHrJbUe-b?5!|vz{o<-C?X;
zZrfQhCi5W+YM&2O>LsZ`4JI^z6t#R<&JL!>jN0S(ypme~AsdaSoF}JVd2w^OTXdq=
z>W!tjq)q{MYuVs_PD19vlO=zczSHhEetC6l?0It`7rG5L&A*R4-;L(6H|o{K&rOzH
z4u7UxZP@tNa_b-Ui_QV<4Q$uO7(y;vwkFE1sjnwAEGb{708c4bwgCD9W8?jQ!fv_6
zFT+RTlD+k|?p-5}B;r&9STzDg%P=HsZJ=3QC>MrJBqN<$ue&SyD$CFN^0D>Ahv
z3bh`Jscb?4UMG=#@H)Yto}Ddhf9RBq$X5`48A2p^vPNjPGl5#5^h2Z*szkTP
zSJ;0v+k@K)tNCEbv%eF`ZDb@%nP?AEDHHe*?c9qP1KFA2RH06P)RduTBy4ilo%oA9
ztgLuk!GGB}?Xh?7AihErHVZ-!MGRdg7u5$_jdh1V7CSh=lPbd^Ka)9CAJN!5pfx&a
zzk=m}$Hs3C5DV?gkgJ0~!H=mD7A}Wki?1|4AK)w8!u?N^jC%eEM~vj
zEyp*3rB0eD$DL>gvJaolKR3VD<9
z-Kr<{)IMAt(@d6V=PkxR*%KDCxLByO?H=KtT>6`_V37w{{=fILBZ*Q~3|IgRfqejS
z%@ji?VE7{T6%Sc-b2Fq^v-(dDTFa#u8Rjy%0LSo5<>u1;r0#@Xi_W`LK9|Lw7W=*1
zC<~DoT4}qhCBLYAda<1^ufeGSQs$a}``o|eJQm2M!TgPKf87jRFXmko#Tuvh)HMcC
zmiyrB*5m>jfFV!|qf$Yy(eJ8Olx}RedXMGl3;_R^`r?YbPtGf_Bigm4T+WlLEteMq
zq+{A=y9xiGpy_Al^(T~ON>^@3rS~b3XK=Won{YMJe$9q7*zo1_CrXpY+Gr99{iz~m
zA1RdMd(6mYfP5}0OVN*U{!vRPgEQZfTuSN5g^`W74ZfH=l0o0N!4#>y-m*^Tz^a@E
zVEsQgAY?oIcrW;bNw+0TGLpi$t1L_K#JD@G;8&S^NH-QKwAbh2_MOQ{vsN6v;K|(_
z;NzH+t1Z4(nD!Zre!iYX44QjP{T4GYj*b-sla!VIZy9RY!0PSUAMBPBKZApDD6_za
zdQz!8tgqyfKKD%&u6B*mXw-WW*VsTJw48PdWNsEOO9eLKi8Q^3H=H5Ulz)S^4#4oJ
z>LKeempkx*ifv%%@-X+ewF~i+1bl!EK;k;pxB_ktn-akn`?qA3<0y+BpIkWBeD0tG
zzuwPgn}w4+=QSy#24#i=GDDQ`dZE>&HmetaWft%~0%s=3C7peKVZna?M7%4KQa?^Q
zjknr8W#wR@k;i_8_EiLBur0RkL^zr+t>-ELBQv^V!*K#U>$+CQVa+6YkOK
zzKGE1%f-sgzl-A!hLlFUe1@PnZ!i)X-qJupY$4>=zRdM@Yry*P(V5WP1^Lm+FPXLe
z1fpgS-fL11pV>2Fdd%UB-TNcuNFU-KdotfmzODp`LR!2!7d$R8{Vz{gvi`HUz2I>{
zWjkMIF`0uFg8wFr-F&F|U(N~AEg2=QI
z_d!lp<=|naRDNfCN3+3zqN2C)Oxg$8OC92&6DyT>Hy6jzl&NWq+O;n{O8BRN<+U-G
z^q8tE8cP@OC_&kMEK)l%<=GFP6RsJy_l_Dz0Wgb6%z?YheT&v#wQU&3CxaNru8_CJS{P8Qr+J3F`6SIxvn*Afg_Qhz;J?>DdU
zGUpwlSD1htxJpcCVQ%H+3
zUGGXMprpIo;P31Y7rmSs0O`GI`C{QBI$P2(?iE&X*nwh>s6p`GlmspAlCL*y-S52@
zFmbO9ft=5&L&e5WJ(m&zrf3WZjgtpn&*y0kyi|a)y85Z7i#P`0E-sA(=C?Pxvr&2A
zDvJq{>?+)M_b<`PV`-)FG_}V;|++2x=-6|xgS!n>(u}vdkVM|
zoknkR-9{&MQ(Eicqv-6B0%KHyh@RG6@%W}U$;v$u
znHwBoZ`@-9xxdR-Xz|S9jt=}hRigJ*STv}ENs?*(k(?Ox8;}5*!*)d`=zw4ExT)m7
z7c}6*?R+?c!OiUgnX!Br-yXAbaZwLlCLxLN=H7Ytbq|PLn=IGk#*&yH{T=P?e*V{H4)QZte+Tjnz-xcJ6?>R9
zE=4_3q|W^5a8Va4s;}7G@%>|qhU{ha?&L>|p%(P>k~GJ6;hxtgK~N`8{iU{c(K`Fn
zSG1V&0325RitswPVWNz_T$rDaVv%O!upEDK2R#q%)is=hX>;)w@uhb0MhF#$IX0kg
z0Nl|;PAft+3)WRrgT${0xv&1cCy?vK7S~Y@cGtqx==uQG=lwTvp^7EBooj
zkA|`^I~Ke4o@m>sAuutSv$*Z3qzdkTPMQ_XOXjpnR(+U;-E^0U%#%r3_4FS*{6yfH
znVAPNK$NE&L%(itC>5%to&q}m10!Av!$&3mUKcP}EF4dKe|PSLT)h7jd+lbf7JPJc
zB=#pq8q@R#-W~r;y!#I(>F#U#RS2lt^m8n;DFQEDRfKdid^M=5gjj>}aU196^8{`t
z%K%y&fH$J5E25QVYZ!+nQWQn47g_Ej>
zu)=C&w#WU56{JJ3%@DUuZToYz_6PIt_}$N*45tf-0%(jGx8;8klLhW~8P8WdzPoYh
z6skGCkEN3@b2~K~|EaCc2rZwqJE>K=vhndzKii%#V4_8
zcG@K@(Ql0);k118Row{~W&hedyWedJe2@)VtD^v!?7v=|I$hU}xxX2$1;Tx(N`Vr!
zTr$UNg-keW8ynSk_b}
zLqkJ6-nTsQ%=#24>Z>ndQ|`FXM`DI{H^FR77NRA*+RX47Db)vAFEkLGZVZZ^?@Z8?
zOH8}yb{tI4&52Uvn*cX?kF@RWKRQy7ka#+Y5OB}8W5#Q3mQ7-VcZ$9?q#bGT@&I^;0+55)Cm|IU$Du^bw~TqrO1p>cKJ4rB
z3C#L0Tn^^_2v+~m%W_*Kbe19LPxp_?P5_eGU~*u&aSw`yc9uw>9rSD)E;41s?|gTv
zb9D4+I4QSU`N?W`xK+qt*e9FhcaD!+!HPR5oOZ_FjNHqo@e(NK${yE-`THa0l(hX>
z`4u{$4PUC)j9&aa64rSyan~-XV?+2W6kILa{KPYN8@91l!_FWm(@+K|5}km)?av>jz{Z&h)6iuN3u^|=
z3Z!?A8!8#VOgONBIHE)|KUi5=4WtW@0!HN@94wj1DFKK;fS1lrqgfPU)P8CO_Qky|
zT8*Pu{d|jZE0ZDUr3B9E?W9MWQqKE8&hK$Cpukb=Qxt^Z2vb3(5x;8jEgUZWSsRYV>C
z)UNgPJXmZl`VbA#2PO#jtr|lZDL1xs46QK09d~J?t#->IVZ`hbaS6dnz#znfvB^~g
z46n7Zic9%niXFK2NY5SUR_5QGMGHkQ$AjxTr_je*4Y9q)af!qM3Yyo16TUBw&ZJBiU~}*o~f0Nn;&T9
z)5+5KT-0#j-hX@T2V8FoORrwiJ^gEIfb^Q<&f4C
z7~VP!--lElOZ|?G9GL`c(O~Srf!Ty$*^?zH!4~C63IT<3BXJ;f=?N$lYE{#}l1rznqj&;wmF<(3<$5FCe!}-Wx(+V-DNC9UEKnc
zqlS^$=R8~W(WeGIo?956SOWz3e*kQ;(ZZk2bJ)ysnDl)^t7!2LaEXDWbd6jxhf@fU
z;F(EMQ1oJYyHSnTg7xbf8eVXC7u{F;!FJ_u(6ou$4{qvJ(EVL`N>5CF$B1Kn?-R+c
z;3;5uPk(%JKAt1}Y`)$RwKE8-^X}%7)eHKGa~cXgw`!IuYkjPuw)>+NMZLkZ(lyg?
zWC2ZkKK-7FS?A&IF0wL%%B`uTwF0bG*ON6gAo!Muu>7u9ie$*ON;O&1^X8L`r1*7n
zc{uT|;!DZ+)0kG(RVxSBEgzo1^!_)Q{8XvUS$L)mHVRI-P;GI>#Cs~Ys|0SO`-0@UY
z$ZWd*a}A)|yYzkTevs_naBnu!33yznRPM8z^!n=5JGA%pp{o|EDDHCgHlRDS1nfoA
zvYxH)ykw>;i~Ey
z8ct-*ZWVC9KKX%e^&&K(kFHMIOtJrS16wbj$AvWjHr#c26*@?SJa$WP06(Ch6?|he
z?W6KI0T775LGrQ7wwczAM$FslB&v<)cdNtiIa(mfNu=|;h2a$JKR~t5m7l`aPK0&;5{wjuzg5?}j81_8i{rbakoSKZy7uI`481HV-QH
zC=$i$WUOgm4(6jbdEF2;yPto?+!_WL6!7}!6T9UOWB5^;mjrLs!~pbRwH$vvW=*Rt
z4zq>}ke`c!lpx;D>3{E#0T7^i(*^hd%=WjMDzRq);XTmO9^3(Uq0IHzXm6&H4o{|R
z7q;t&9$<}Jck^i`*it%y>m8JnjEu}~xdWNgR~v|g3e6O6BY6;dA}M>ot#?6{=EH@^
zV44s8GfJ=3E2e~|&EveD)>CMS{KT(Lf@x)Wjwz)X~
zHbBY0QckN$v3Lee7PphNFY^R7Fm3E@VUHW2!fanP{%y>Ez0~T9ImFAyM0ce2m|n_t7}kie^M^Ly>_KERSMv+F
zIprq5@up{H0%xl=%CFH<-#MZH7atfJia{&(>>vL`r$gD7{j5_tV+tno{DjjF3~4D)
zcoT?DAhJJOZNJimhP@G-Tyw3aQ)VEHNzSYCYwY1{J9cO5v0B{fskgvN}(#=Eb56lCM=mEY$(J#d9&UaKf6Tv!Z0|N=QLV)xSudIjBrohs91}yRX(F!A&
zV^M^U%Lx!d%LfOx`Mf~X*0xopybBibyHwd8BL#EZ*gfja;+c
zn|Hl1pbC#3PzEP@5Ww=f2sZmk?so`j&DN^6jMCWX{@ve?QER*KIWI5Io~!l^t`Pzy
z$Xg6hDo+6<1|eQPree3p@_8YU{1_J?B{(4U7oGB3c6bF#)=hx
z53C4_)L%B}0g1i=lf;~s_yB=Ly>m1J6zY^h?ZB*lqw`P04AOrXjvRc@Jer*DB&>da
zfxzdyCj`J5b>*=a9a`WELl={v7KSJ*bMwu{Czp?-2AsmLL{?)K?^{<-=o#AL;-Xaa
z3uL_(Phz7f1Mz=+gB0MJ;eWGaz$|!LqytE6B(V5M2x;S7H}Ifh9hrcUO9n1Si#wPu
z|JfSNP7zC)d2&h=6)XubcxSd`B-L^z;9CNLh@IloGb0K!dv*lCjMkl;uKXM4aO)g5
zv4_$HLW6NASAZpBY53>0;`jWdqCcWZ?ex@dX@9pR9J6}OUDke^I5p88N^nG+<6>f7
z54#K>zVwp42#-B>>%N9MEA)8j%YcTj`109=$rBp9tI6BzB@|3QTZ$U=cu}&q#J4aW
z|2E~S$^oKo`=38}niVF(fCQVU!CbnzY-B0XhFsh+I_*qFgQcN|$~8Zla=5{ZZ9voo
z%=Uj#@s&|+ZC$j*9a@SLphb#9fg-_+6fN#rq__sx0L8t*y|_bhw_?TJ-8HyF;Bnvk
z?rqMm{jCPosh1&AqT&M$1uH!?+o{^88AtonLiQB;=O
z8ZX`9c?lGBL3+Zyk21PK9&*B4umNtH{MJ%1>9y#{MB8Pf=dn!dqO#Ooiz?qd#cj*x
zNQOS_y84+SNkXVYyo3MTVBXGz_)Js?>b|G_Tj4D}{>d8kU*JsSGvK4@j_4BUf#dJO
z;mYMRzB73<{Z<%=T2ru&zF6ttGS3yf$-?Ooi|YUqpIl!uf4$_Nn?`Yje@-=p1`xR2
zRI<58|Ik)bbY(4$CcoW_iT-h2fohG
zg<=JveX&P0HeUo3>@WStuk8+R(D3?_h-=mQPU@OyOPFKKuO*Jct|*||NBPtnFp4K_
zs%ym~Xq4^BweCuS!=*J$`|RSmP~>}hFZpk$2U<^E%VD3gsJv4(_U#n5A`V9DE*mer
zV0=FsH(iB$;fol5cos7LFnVi`wZVH~4O^(Ea3{Smf`iI<$AX_IK0Q6*6iSS@?0~7<
zAY%2G7g%!oEn$=a7@00@aDhS0Lei(Kf6SQrf6SQr>qzGRnKAw;?ay+1ELW6oHa9q6
zm}Vkd;a!-WB?hak?@t0AhbOFUa<^c{k!7RdhX966d5yXbKsl}LwWGxf0e89X#|-3L
z#X1%8uF#H`(%Y`(okj8)ZJymdRD^79ov$<&j04Vpb?-yde>Mu1|FQW*M4x*&Gqa!f
zf0%5-^#1D^wQSOR5%9hksdipiL7aqREiT}Ty#>~jU;HEm?Qx9i
z3WXv-#{wt`~>?c7t1=$z3SX<`)L(Y=LvwmoF26=7ug*D=$7beAk0
z!T)ep%-NTCBH!T7UvYLSn;N$_f_M%%3q~hG
z_o%dXyQ6PE6zZegyU#nAc>6bbc#}Sy4sb7e9>Jx=qP=1zFp9m$zq?IUPp43{-+#rV
zU*#KSiyU(dU8wt=#=+N9SLk0PUWszfM3CKHj$s9KCjJ(G)Sb&X6aB$_T)#Y;^KW0}
z`NxS2&z1f=e0l<1;NGkr@ca0cS@EreMaVHx?;U!54%hrr@%Q``PaQ*t^E8`jz@
zx^KTeGMPQBUhaZY=wKVf{0(nFBO5=H;mzPnt5q(z#Fd*qtVEP*_GVM=)4IOGBPZ6=?UsLGK0s$OSChyPT&
zuz6Rn?E`z>Rl{y$uLo;f{@y2!_9(M&(a6s07Dv$B_8n2arTmIW{|EZyv0mD|+BXMY(2qT~95__&lXQu*
zbQDE9Qs3yCY`^@9EbdXgnPo{~#vi8+4t;LryWcyJ`o7P#ZZP&MfM~usYvtQoDu?O$
zt9l*qsx2dVt6gXHWzYjA;9i?2co1f4YuKGW@iCTtsS{;*N#uS-ol5U8F>xvzE?J$y
ze=a5R%UI;=qOrE}xu_7!t4z&qznc~rvPJMCa-7FVPw&X%ldaS-khH@6iTDG9_KT`C
zlJ&#Vpj}?4Gzy<14z7a6R!^7>&(%m!6p!N?SO3mPC_x-LsbErxry%D0)F0NEHfnH;
z@#XDdgskCzGeKFsH}iJPNuG>z!v8V;&p#aQxpq}}h}m-`S8YkPED(xZsyoKlx#%bD
ziS3r$A1e&rRi9@rBzm0G&VvuH$aNDMzO3G1Xygo4dhb8NiL-0J3FPu=Ji0)TUGhLU
z7V;ABrOE%!Xl*EImUy~Qo_Rg*py7PT^7y#(I@Cli5_oTGa9N1rdOD2(Uh)j-5ofycv3k_*%=eA%5i$fj9g<}=<-?>w9~EB~kb+%G{I;TPvPiz-cD
zUp$zO&swXWDq&bxHFyvkOOjJ;)g;}@L@oy~$!c>PKbbBcIFYx&(9dxi)+wyFk1e5f
z%A5)Km81e)v!JyQ2Hqk)@=dvfu9J`=to4E8&)(YX+HKJ_0q!wtfn3VqmH|!nmn#_#rG-pR$x
ztd23dlnB^iixOI^bF90hHwM1Ns1rs0&TYA-KAtBDZz+Q@Nye|AV*Q9-7Nx&NQ7%uH
zIRrm5AwENEmspE!^W@on%IIQ{@V18ccs-4F_k%fBpgDAD1B4K=-up4WntoFDJW{K5
zP;Dgrv{##YJpQ?LR>x$xOBP)|4T&ZA?&jg6XU!Glx@Hd7H$86PodORwRc!
zu1%Q!#=Pj#T7sot^R?t=P1q;8WA_-6nqH
zXZ76kvaC&;;TT7j*`7spYNYgrDR;ve)V(`erGoq(XilvZAHA0;wEoZ
z=$*#4{q}3#&CP1OAgY=3r2^#s?4|~Vy!A?~4{`;EZzT9f3<3W+%=On}PGu$9mzK!X
z%k(5P`7*w-_ArOP<{;rp_sXd>Zr2rAI~)~c|IJ{2-q0Id-fez;_khLu*t_i3F79My
z@DPTi5E?33K^cr1+H$?@5ym;KVf3MPm!qevKr#FIrDCRlt!LTexFsxWR{YhgEL2hr=JwQLSsSu>v6CE}|LEoN$
zkx5NuKg7}0pMhJq=0J4VtM+a6kM>Bt$ieN3#jJq9z&2=?5hz+HSKNPNGv`^Z(MmN=
zQqsQX3k-U2P^$S#dz?^(Ls{&lby2*975T+hK8&O!(>uJUAKPs}g`O>^HAr
zgugnmJq-QuWX5XYwuwi&lBGFu)kO%;iPoxDcYU-4WRp`@t1z~kGO|2YMA|}efsq4D
zZwIT)7i*E%yc~as(zO7vRbXR)a{(8rgPe`Gyz`owRyIo_pE*J4e7RLhO_c9!8hy6A
z9e#*5=&*1!t
z;>>BCBq>eL+lw?7TT<1GFNOz2Lq2A-MJnJuk!$huVwn8u38PUfR4zgyDpf>wf5(rb
z_cElv--CAzQqBsGfRJ58xh3}HG8C-XrZRaZuF3JW{jo~jepsJSpv8T~9MI&5;PG2?
zut(0$xv`1YR3-9}p7FW>^^}5lvg0#od4`+^sYBjzBqVL+DWx?Ta3`VI`h`Vx*?drh
zC%W_L-nk}vdK{kFJGC_H=pdmF2jlLsa-xAps*5Pu2wB@}eR;pTWmQCOi?NRU$EP+S
zyPzBg_h+JjS!+IK(#3DZ@2Dpp)|Q9V;;XFw=H&GCgE+RQo=SoKv_#*?remLErGWX^
z&k`HOjp`L^9a@RDXDJp`46NdK0y0aVKJbO>y;!Dc#~>J6(mtRna(VK=IYnR=cCCS-
zrS(1rXm>Fc9aAQxRIyV_b>Z?Rx8wyVMgEdcW2ZSn3K%sss0W={&McXC_(Oj!m=?vI
zH?3GNUDJe#@?s`ckTkyp`4K4GdFde$6e<@M?ZBKy>@C^|eo7fJJvqhAYb&*!%G74=
zU2~jX+RAkB8&x9bo{yW_yU9*%HoklbFCvVJ3itL6A}&7MQd=bMRJ8H?e?$JlhHvfG
zyU_6L97Kxh#=dX&sD`A%7=N2Ty}1C2TMDVjDMb1RiDB=cyQRiXyQz-@_1$OnfyLZ7
z1x1vZ4XlyPUupM$2J!|?lY#)SxmH9H{Ju!q5&a@T-dG2TgK^+gmqUP&)P}b&%LfrY>9w+KAk_t;H@Nv{^%8KR@C_
z&jH=}wC_@R9yjuIU-8|x#vv3hcckZx!v>E`u*t!^rpkwGDNe$+<5j3iK3XA?&s>^x
zVMB%N0KshRmBe*aX`k#T6T>@#)RZ(o1UiHg%#g&v`9NMC<
zv|3XY+Ny}|m!3&eEuWw8`VFFp$)t-sPJfA>Qp`*=mlp;g>JbG;)|l&WZy!Jev1aLq
z$M#Fj2oaA(X5Ej<=37QvOZb4JN=3gfKbsD+@}3Cr+#czKW<3QFu)Ahv2FjZ}M5UvF
zZT3gud)Pv!xlIE%#>(`Yo)D3$h-rUrnpOZ;%hYJ(ju%&uNc6P^V*3?GGH|^0wA)Gp
zn>8Kz#BQ32ne|Q#M|kIvI(jfD6^Vs9{eWq9hqG*x1&UNf78#zmArqzw)%yM%vv{a{
z%gT9@zTbAr6hG9xj?`%`ORc~E-M3O9XMUJV_YfMuPp2h_c`h~4N(h{|uPZ2OG_by;
z)Q?oeez3uE33jOrvVnEN^kTT&IiUrcG915_|PX`tl`S=v9sMdJynH
z0g}}WgkK(IkEv?#CZ-tF1HZ5?EM)|4Hbv=Y^)z@H+Hwj$F01WcGMXq|@LJZ%ADsgk
zJYkIPvNxnbtT)-n%!t)RTCz?ltfnS*{+@)PCQ+qRj9cG;-xl5=RfOy<=Xq6jdQE)C
zuy|lfFp_7|Qe!UKGr|C9{%2n+gqQ~w@{H)F8TC;jbmueCNTxok4DVevPGAcKg==h
zeKZ>LtwhkO7D{^WYOEMVw$n_C#d~y8cCV)W|c2EqeLBc6W-Cbnar_kO+gL&9f(2YA3
zQ(M9fpMAo?Uabirdr*<8bnO;2nT5%Uc#&ybnK5C57I`;Q0xoU7O9N;nUD?n@WXB1B
zFG*PG*XKr+z8l44=z5kYj#&|{?ndxdpZx0K`_u}%=Xn&|EO(9`oORQL)_*a7E7zX7
zUfRjZ2OSO`zTxX!A!9p76cH7TLzs%p0cEkb_Pjbn8PQiBr!PrZu5-huDLxAnL;$o1
zH7o1tWYKv=`fsE(IDldwDw|FB*j?T<++13eMx+->8cN~^(tcV$te8g`oI%}b>@AsrCQ
zk-^&KXxrJ-+-Y!9ty)iZLxhjv{^hE!R54#g(CdP4hHSF25%R!hl}~oQE~-*{)UH{B
zv8K4o!6~GNH|PBTDH?_t1@?q>2FiIuwO)u{R8?Gz?v`Q2I;0b#x=M7ea!sj%bXx^u
zomCR69;gYsnwZ@cE7vEhJN;s?c8=eNLJnkER*Nt4wzGTBZXA63q@8sr
z$@2fO|Nj>6Dgy4yphpyc<~rPeD|r&tKOkU@!$FVcTL@3COm_9&T+)c#19CtWeFG$o
z5I%OINx|6z(o>=EYPh6r_--?~#l7@g!wXqgQ!7+<=eN_)rw{aET6e@hlP)$j!@!j0
zhYswM+nYp@!{mw|I8$1PdpTtae^z*gXMT@b`8}>B_P{}J*u_g9UjX=BVzgpD4Y!$B
zaIbWPZ2{*?)|Gq;kIeVLo!S~>TLZUjLv#K_C3YL;w=uq5L<4WTY;zKjD#Gkwpj
zn25=K(vRNYdp8{jz2PakO`hIBN^SFqr=`y&Wnep{oQJkI;So4@#^t2iz&{xIC
z8$}FmgCTFor>tPRR`2~#Ikj$JcABD#t!(QkQ$V(+mh^g~y^j1~zJK*X)GZ|P5q*4G
zKcDi3bIGgJM)tTvKAj=x{lI{k+Ts9yU>H6VQfE(77?YCtLz&rSwmS}`fR{2ys)W^1
zX^oKkX&`8Ohi2)i468rOU6Hkgr9~~Df!V5S>QSQ=V`WGu+nGFkZ(aeawJHc0pDQTS
zATKyv2zFn7+s-bjSN5G|5iowRUzeGN(=2KJ^*af35uPeGY~muE&`L18Bx$^uAAa$)
z`v6YbyPvc72^m1-BfMiHVQojuy`jcnd(!1AJEo#M0=OGF2HT)m0q4f6UJy`Kw
z9C05{z`w}pi%1pC>tsPwjH?`kFD3YuR6NDO|9*EnAP)OBhdCj(vvGN0cgN~#S~@w%
z>5A-k#fbA+pqx#>BAF$Pl$~7XV#vR!2`K1N;TM=ZG?DbYZtXPEA=by}~fwRg+-dD;w_OOga=>KXTQ>0|0UB|9Vg
zGvwDz6Ar*xQ>a92x|bD^VNjXveAJ!y8q1TF5Zu7cieeKCb_=p^LO=Y62nQm6Y<)^xN}*e3O=!
zikj1soy#V$oRs<7*-l9q#CuY;84j~AxY%4N9W6^@G066;Ri
z+(Aq{biz3wG+DjRxHAq?nCIQar0(o|S)p%McRiQU+SxxMNXmn+esJA22BI9EfJH)F
zIh!OmLWd0->kJl-7iN?iVO29^!UGgL1!U-l{I=TJJRqJuZz|H!$bgMlz)b3;xf(G{
zi>PWX74dNJsltft!bYpFAjyLO1=OGaYp(beNQWt*YdcN!+Ka;~dP$0qY~&W_23U2<
zmck)LcFDGvL5R+CYlo)AFLkPkUQG(5NR3%VAqg`{jq-WD(_>sa%?2hg+7~Uu_zNOM
z|AEM$9Omo)AhHWpEDm$bktD7;^>>AblzE$jn1DoZRp->T9xB6kY7A0|BL}4geXi8b
zXm`PWxBlT}^jNVE9O9)ZBr;V@)U^B(;#Ef6G(b26sr)un@zSb_i3@VzPhbVTbX^K65{%jrL`ao9ksbmaHdqbbz>Kb9dC|8)58H*$AAW|io0W6`m
zz6vP~+i%(iMQ&5bfCQL6SA{rUgDE8%%<&pp;mDZyK%(a3J3j(~U#TR}7!<{P5|E&k
z=;ghcGwQU-64*xdsHKW+o=E~MrB1ps)3B0HRsTrl7~v#?Qm&?v^@GZJDg0Z~85
zgv9Jjz<=IYaVZ2KLg~5TX=%#trju;uYiG5A?>a$xzt{W+1_L-Bb`+{T6*PFNCBctO
z55j{@DuWRqY8YNx1^*bDR^ndWCQ-bXDjeL~5xjeqi-s+6x_D$wv0GHD-Kx%U;bx-h
z^L2rTi7FAtr6S8$GCXW*Fk{B9?Ye;MGfmzdCYqYBh4^}?t7gYF)TiUHQZ7M>)T+n(
z!J7NHi43wwDnct2G^G?ZI^%d&u>9<+v`9e1rR3`jr}
zX_bNN%><#nsgWf+1*kVsopEuaJ{UCpOe5f#J{u3u+S^$0t8WhhT*t(#KExW=+w6J9
z+P-NUKPVT*6#}M0*s#KjAhGsiTB$}*KD5N0!%XLWJ03JY?OlK6Q-M)7
zRMGdgObfM9yrF-`O^8B;@+ej`f-51;UsXKP8E|kN4!I(!<>rqxc%5vZ3MBXo(zD*Y
z!G?*Q5*fezHwNMI#kIkf*2?8+siMl*(&iDl$E5)<%N|?ZO7cDBze75u0_tei`VaBR
z@P=Dzijl=cMF7V46<`KIFrRt1Qx2LT-TJ3w21b#wtEJ*iwO-XTUZY1rAr+z;;<|TS
zA!d>iOa~kRkb~NcRva?ZBNRUK0i{aqWr6mMlm~G%RXxxR@>FCmUyUwW?y=2q=18O%
zIfeE}QF`nyC*lTIs(Lv;-m51*@u_ooMA|hN=;|XLS*;#$MmyF~b$DCRau|Xr{|=9q
zn4DG1Y845^IozSE(ZpJ&X>GCeOU+4p7>-p9{afhFR91Y)Z<&Y2L12Ngvxec$&US#x
z$|4u~EP^XT{xsY&lFh
zp1xzkuZxpmX}hV$E)2}IsSkx^A4?N)EXNIzXxUBnXlzC@zQ4LiS|1@HJ+A!XmUkCW
znFI~r$t8|5C!xW6OiKTp^omrKZR9wLm2QrHOZ>}$?vPAGoQqygZZl$P0e4;i{FK5Y
zH7cCkL;o;zt4ThSnx_0Kqxzc!(5~g#d%%=n3z)H*42ToJ7p3M@BBf;7)2`N?GyY&t&>+M;mj=aA88kTKc`#@rf&r(j4pm%PBSz3ZAZdYVk5UzbL|n3s>kkWT
z)vKuGl=1FnyR|37iYq1_NJGssYTYxWw+EzjHQO=<)JApg9#2_k`*b6WZMRE=f|Ue%
z);jjtq@Q5t234arlb%1Tu~{Vtb9E7zr1HF+v378mS|X6GLBIMufsnMA#j-gtGN9&Z
zWcaoH`1KA8P?1e$*YWo2_c1b_TV}qC-b*XYW#+N<-no5ViGgy!lq&0G)rLQ7v-!us
zh_st3vmKG%F^o~;T};(&>97~)LDkvaoJ(0*4B_h|iAPxGuQ&k$W^h|H%)>ux@=Ygl
z^^)kcBrE=MQ6MaEknXrd|D5|9~K!uVI?)zjrDfo~aUl(xG46T@qE+y)QncDNKZ%itL4
zGs_YnDFzWPwxNOCP@%Gt>_);Wcd_p$C37QTG=rCk;Al|6xyWbFL!>R7U`7Y1AU|t7
z={nhq!%49E#!e?%@Dji}*16RHTyAz`YL-cZ1HF+Hm4nUBwu48q90vq<$8%+@iohTz
zazW2ve)6OGvzbxE6h1zC)+I~@K7%u37bWcVMBW|AOKsl{I_XzSTAMA?+xD{@xh@$!
z>>zB6Rzew=9Y|fz@k`Sa9iI1Qvq_puByUvdsh39Q(z4t&-&rUKLnqt5!c*1?Fn0{H
z?PL?o1-wYrUiBh!Ffnb~ZjL%4=YNvnJ}PcT1oVwyaM`mBm##|mMK4L71-X3b>S>;j
z_;ot5vw8^7ToB@d=0^UtaDpG;LPnoPySBXlTgsPk&Wo*5t2JQf62ffB$5|4-?tvJl
zC?JQ^c*^uzfg1N~AbBY^t-#cWi<-*%W(V>Y%!ED$e9o>ZybYhz9s@5wM*+t(^eZ@>
z*{c?#fMJZ54qkS|RqTM*4$R3D&4O+!UWxI&Gf5+`ljaY5G!xz%4&+Pc276@ID8g3v
zYAd8#X1E;k!o9Qx0U=8H(uFgl+~Aqtxq!l9RtJJKj>O|`#ofT3oxHDsGugacEjWCq
z0+pXw3BS&(dmdZQj@*N`&jqp!tz5y-#GR2!;3=ovC!Jtw_jSW&)PW2sw*MH6Fe)`Z
zygi1SBtsR%9{oRNgC^obR#(E!9Djp4yP!drl-uR|Ikdrw}HAX
z@U!M}z=1o@oOrQ-?QGA*&B@RCJaJIEB#s=wMd~D=l)pE`O@&?I4=-bev)>(%_=x(ioDGKfgnr2L4>zo!
z8nxesQ#k|u`kfb$W!;WuZ~0J(ReY}^f+ih%U)F;?zQU#7)u>!2b?_Jf>AcNUgyYHQ
zD1Ym9Kc*a>pVQjJiWKb`fn>2YQi9tG%3)Y&Sm;Qs-lBBY#Z+*6@sG9Hd)A;rxzZXI
znM`K$Q96ASX-KU!q-EB*i{I|o65A5VCq9(61>b_n9d4_eTw3Xt
z&o`vj--uYwY}@_Y>WZoBKE5@e+m1t^;+^@1L6gO4>Rm^%3g%GTs6YN3Cz12MZ}MR=
z7Y_7wD5uC)!2K1+^igNuK;3FwC$Ru&oJ5VG-u`@8Liu~ED=jNNXEYahiq44YqkFFM
z@K0M9$xW(RtTD%ySzs}C;r|lX?(ZXcl=$Mi@_&&xelPe!Zotopd*31|Gc72-0oC08f`Z8y^&C)vXZu*
z>#ojQ!aS?;UmM5Kb5LJqKeq~{<|7WsyVv*-afSbp@uvQ3WT8b#*;Az}V-L%+
zAJQtT{GI2_r774*3@_8%LGZFT~0eqX3bt?RUw
zORcl@z4=eA!Xd!pX4<|G#r_j9f(JCvk=(tK)obXN#0;?Vje3mVPK(~cEX}}h{R#x7HX(^
zlSjJHsM3Vd^sL?6R;$zxwU1q7v!E-Mef)llNtoWsJ&)7*cP$0FE3OLYoOXwX<0?tu
zbT~2e@vCGwl%+Q#&t#0U(S;kec_Ka%@IfE^ce-mpV327
zLiRa+Y{YzUdHl)Yd9!$_0p3l8OquJJ55+6?3eqd2tCfWsJ1gVw7yeBPqkJkyi@e0M+0A-r_B{C
z+l{UGJ;|{K2u6zvUC8AXek+#{&EX?0Ru!;2KU}9Y1M>TlguE{GW*4+JqHG1#XBY^t
zx{6xz*({gfVh1uXw4?Myt^>E&N?PQuijDQuOZZL@wypVtj~8eJ=--cxa^JcB4*BE7
zBBK5Vn^XG+QH$dJFLY9YYPY#g8uDR!gQEI^$KO{?8-2-Wogu(U%aG8?9jiUAIQsgu9i)ZEa>PUwb@0OSdO$)1g-SnAnfcAOkIZDentLLG44y0FgZPrI
zfD2&@AvxR#Fr|f)gm9=#b*4?8mVoexUk&5+od{52sI5T)vPIR0xaugjaws6T1<
zN2MMDJJGytXIJ?bS)`JL(L>oEWj|a@QEe97b)&$>>hkIYNtTAjkT)d51Q`Z441?nI
z)e$9p2<(S~^T3CG$1N{4AQMasDV0!MZymS!!>>5B+J~^RyfIGa(JJ7q81=Mcb(PZC
zrLVjm(bU%KvJAn}0FoGIMeP@;?znOw0eQP^llQgJqARqP?y0mvFhJ;h#zeyh3&K(H
zb5N6lRydJ7yHFZ~>l>NCYAy9MIYd`=42jIJB-&^ds*7VG^zL6n67PSis(LOIH74tn
zF(tjfD<9gXmCh{@YO|4bxogc89^;=ITLn+p?R|OyL(Yk;9LfQ^819zTn3V6lyG`kw
zKqk>O67b{-a-6;c=~?}hG74$$Z45w$A@QCJ9_8q15fAJPYSmHTqi>3&9}a)n@qUZ5
zktT3BdfnniUZ|9pv-<$jQci7fM_Hm}%Is<-FjK85XyvCTll1pia@S&4`wzH%(1Jzv
z1yTM8vRFi(evJ#V7(P~bzWA1+$P
zSAE@^sonKj-M^I=Z?PRpM~1vOwpcT>c!qI^h5zH0_Vr%VIWj)R
zr_H_s6*`w{Y%e+2fnm>DfLx<48{0^)$#xrRZ|%oTkLEYKN4?oLOu{
z!1tJyYN4kloYNtSC&OMeeNVlH;PK@hURFu(ptgm;p2&xAWDI0adA;8^Mz<{W4Lk*#
zt#d2GBR_SOc;ZoEIVZ8k;s0gE=j}6FARUi3`ZETDeVzyQ>gc9@NhxLOJ~39-I+evI
zKyFQNa95Nyi{Ij~^QyYlXPHj}AYiV5VXNnP
z9h;rRn4W+lqNfxImmk-<>LP+<)fFFP0-eN$aF_1}lFZoF0;T-49X}
z!B<87XruX>G#Z*>4=nq9(v&E;nHgf_v6#yL5vS4DFN0di0|0;a1@iv+pMH{R9=iT8
zgErx~7YXHnHgY`0x4w}k&_+K+%aKGVGa+!18>_&=F=7i;BKd1)5&FgJ7H)jU)Toa_
z`FuE44?ivcu6xmuccEk_BW%-CNm-2bU`wA;^G)Q#OpdP&vH<5ir`^T8D1L*`)j2Ae
z^~Akr{|;^@1#CjT=+)gEfTJLUj34}&rhO~MsFgcA!-~HQq%@zgZT)_MU0qG$@TRnG
zjPsfads9I^po$utJtqjNt#aE5PPM!KXh*m2#fOgs`r46o%-0_O;S!5kUyd-f-V>vl
zE*j$f`UckCV0i<4_et}%7J(O+xf2MC$(
zvkGiRu`;?BH0sAXb@p#?pEA>Q!XOu?*Eh?(dhSl;fVUjKz8*fXZoGOcltVH=@isaf
zZSyVEEvh8x{oaZ$^;aWN486h=Xa?hJ&z|lUUABVHPF1k^g-^SimiQ&5Op0x?-v8{_
z;-kC2391qZ9M;M1DWd6xa@y0oel+}2Y%%LGkJG&1b6~@7oQjYGGLWw>P{J#|PV2yN
zn9ejxQ%Y=b7SecEV4*0lwd)aJ-OHNEzXx@I=yc{ZkypsUEf{i^>Id0RViRcOMJGxj7uawg3igz?PMebi0T9^|w_H+5%g
zn6zLGgmK=}H0LFs8euEbW6YF4W4!H_Jc*hfQ>Ebtp$7`=i-MfZHwwhZFMg-ckc|^b
zqMOfbrU{`y@VFaCd?V!@IdOM5r@vzN;0q$VaqiH|u
z@_Bq`?uRF~DYlS4&Yj4}PqHKt!n9SoZ`$e%aLqRzE}nMFuoQS)oB^$?@^^RT70_9l
zOLnV;(vNf0d!q@*#cRs3e-P~r{mYy1+GgpVvYeBgKiDS-@2gp5VsPKAe`0z5K(Tfd
zau8oa3f{GN!=HK>d&PgSK|y(&hTI1AUtz`S>v%fl%z+_ftZLp^www#{`tfHt)KPp(
zdx~`KfsignIw0#{qbh!f$?avyt&=VZ2u@id#HW2MBxo5c)vlAr|5$yFjfv@R%HpD<
z``}f5iM!CPUarr$M?7LB_r==N(!B!jGt^hv&S5TCa-5UG=pl9YYA?*L<@`~%=Y;2w
zm}R9o%F`w(ruvh&EGGX#WLF5s&gwd@W;$J7El3yj4@OaZd&X}t<~~UNPXhX0t)jx7
zS$OoMoHTyN@b?_;xTQ3%-@#jGS9alVp8^zMHX;RWj<*L>*(Sh@udv&Vufz^omS*gUgQ6%PZkj;O;Ky|;YT$)pkOEzh
zfZAnYj7vx>U@9gPQwtRAm1{*_0
z1G3voBahX&ums7^Ar(Lip-kv_#EaRS8Zm;P3$&%03)4X1P&~-_FH8b@-2bi?m+hbK
zT@RuX{dZsQqyo+E!yTzq-RnKJ<`LtcOWWik_RsjY)NGe?@BQ14kX3&pU)A*I#=5|w
z+XW_&WA2B`CBK;hPYOv7g*J;Jv~b~*D1$Vv097)9N)u7uadWshO4*wCZ-)~R$pq;J
zMYgCIMcAf$Rx7vdGWE<}%Xzb-C&Zbg{6qXPtLdkm3UH=>9b?a>*~4(j>nwy}GKI!o
zS4qSp3!x%n-=tywdab3FnG0Sc}n=r^TR|{hCs{!5&!36L1vgknP`uFrD5U
z;5>0{&^?2%I6I}MuaEuw#gfg4Hp^gtXs=g=;fU0Y5@>N}tvnog0o)I=;&H?lAXO7C
zD@i&TSZCh7Jy5Fomg(Hr%m6H(_93+Id1}F$WB;n)CRq(~BmDpBl<#@*def4s|1vY*
z`bL-@t(L>o?Qix)H{S$ZuDahU^bKeZT|7_ohlV9W56|OH-_+S$(AQtlcY;5q({LYg
z(tq4Hi<+rVlMO^I`REJNS!|Dlv-Z~;{jPjn?|RMVS^2g7*gEuWOeubzRg=H=NBIsZ
zQ+4t-sqdLTYHfBEyC;~YfKBDAtf7wJn@yz9?ZIFJEQ~kBEpm~)$KC~?bla?o^D^H+
zJOQB}&U!i2Xdx~&P{N~38F3s=`J>|c6a=nRy@-IWUGSx+UEI-9ejuS%TtD%inhcJ<
zdgK!Y3BJM%gM#tnowCdOux+K;8rPgVrRUm3@V!JCF?j3|<#Bx^;b}M=WwpqZezkOc
zc38ug*zC#eoYovJyJxMyn?V}rRDi`8Hd*h?{W7hWMWlB=_ovkknoa}k#o_3YLKAgs
zS|jdG`HCoI$bjL()b~U;9#LWia53rqh)#%cDCp$QBNy
z%JM{kE}J9qP9^A
zgxawxJ7-~^+ftsp{b62!&-Zig!n;Gvj8+qh=_8?U!%i}crca}qmTcgOOi8oZ+ZAv>
zy0Nol5P3xZ#^DrS!QId?s@t8;6clzZw7>jqVN$r^;poO`&zWh-o?l5w=jGIO#=wwf
z%edg|tv1C2N3%oDGmS)0)g&O9i6$-{b6Pa~hqA&1%W?Y*M#8YOpXS4##)x#;^TH+*
z@cx<;!l+#lKK>G25ebt7e|I+-A^b@=;QY|jG$gS{ywYehnFH1J0h8F8d2Rnk*eq+J
zE-j%$G+^#9`Uyfg%Xy$aKnmOI9%h^IW?YROMh|OBP)-0!XJkKcZt!=WH^wS9Ug$DB
z1}#L@8eOF`n(6XV&Znhqx7jrMG*L0n*Cn8nioZ+&c_%Y9w>1U2XU=l^gI~)kQUN
zBmkMZSe9~&zle*jm1F!)h9THZtVpXiCmi+Bm?QBHE~jrUkj8UsMjK;Y-bNRoS_wHz
z(KK3r9==v`bI6kF79kRJL4fxYiHi~1SqO%SG&ljNuQFPsv2Uw(r-iS{cKV(*K1oYa
z@yd&YPv6$8GO!1hc{^EeRyv78&t}43Be34JA}q4={qXqqWEVzxWeSOT3L~MZW?&fi
zT7;CEDZCTf#<~@Wn_)VNpwy4-NcJ*S$RHrV9<}isc9Jf3UgT1bmo8l_arh
zaSmYbIxp;ykj2#Ge4W`iZTH);d#A6;?x6)nbd}oZw}+rbTtmtRMyMQ*^mFEn$i9kT
z=nMI4{`d=*JPUrI>4KcBKemewn##4kZ%i6~vg~{=Zk^hx^QX4tWxE{q+Cr$z>apky
z8~9lD=itysP)^lhWF|fp5WwX;Ij61t<4n7OoW-b+?5X{|ph`TasMglt&g(M$z|?iX
zl&#qeJ!QHJDq^5u?-QN0K1cjSgNV`33hB=v?lOk?r_m{4l}qQ3PDtsv
zlvLD&0nMxYj;=Ji2N-V+5h%_$WIUSxmF005ns1ln4-+(7Q2RpWoEAGIE|!bLG!?SCmh{~2Mu)`TKkV(`pL!Ghmg8z!gQRS
zY0eq*+k%REh>7F8^HM&y
zlG%gZ>$LsVx=LOhwnK3-_I{IYP&;FlT&2~=6lX;hK~~oj!`{e_
zsh@W@%!taHHpILc!cABpzHVzwt*0{<#AR;ZYBhpcp0a#M%bfS1df%d_kR{phAX%1k
zi=7HTDc^lD-c?r$w=E@B2-#tvNled;hmGq{dX5)ngwxI(y^UFieEPn8XL`^u2Doc1h^|3D?GuM6!nk8?4l6Ziw6QmHRwdKYsF
z9OSP0S@p-=%h>=qK*qmK!lYje3_nWbptU)DQH@)S&u@f2w74X*9gMYQs`}#D#>^Zw
zc~Yda@Y1$SSLA&h!4>9z$L0L+RQU*~RP+~;=o?u&2+~tMFYs?Z^s$BQ=*ec&xhV%d
zGKD@;)Ley6*Nl0vK=$T7{3uc+&TvfeP}oAAooz!uGm)}&OuIKCPBX7|_!u+RAh%&)
z6hbEVX3z+hOaMO%FJ-ecI3!^v4*6s#@`ep})xVybF8Sf!xn}C%nZw84TflEy^P@Un
z388z1>c+Hwgu7<4xY3T(bImu`X?k5ifTW~QEwHdW;eFD-W;K~#TxByO-`8AbA?x{_
z>`5qwT!6NFYxjEY;3`|bsz27J4(}f$`TW?_H*Gf$+bi<%$$S|(`>`fQN0Y2#qi%WL
zykxnGZ_gRt9;81^+}_h`m4xXG&$Qwdn&W%M7`UJH@ak_G>T3F2^Hi`CA
z9~Ok4A04@A%CiTBO325FCZ`>;_OT81K!C%WC3=plB-k-spW#BocQz|d+2Y4ICq
z3pahi*!sPi<1?lq6rx5A8I=uWag$Ll+ZPNBU`p|Za(?gE5n(3kAYlx8HdfjnP~}bx
zdV1;jEw|A6{ZyYoFMNlAV7Qty5`w)bP4eH98Wuq=3@0$wWh_P3Qf$^2JPLZxBlKtu
zJRtW4m6^*b7b&tK7U4&5v4GU!E~N_)1&;NTadk!4=oamiD=ER)UX-O*U(-uTCPbyt
z>z$?~%jjp(_<6vvgDx3ptf=GVf`l3;KA4n=Z?Hi9@k}ksRg@Vi#J3vdk;~N
zK4)7Ea&V07kh!$;0LepO4=!*^D%6xs3byGmgn67gOjKt~8hd&Pnz-O
zFT_uLxELfj2CTIlUe>}EWTG4F`r50fD@{kFF#ceW|4e!5CzG~!1`yvNaelLkgC>n9%BweXQ1jgx#Kl~YuxLEvACQV!83BYiZ`B=_!~mQw~TBvfm3$a#1Z&nMvXzxV60OQt+Z09^D|
zQgKgyja+}qwzhXF={x@L=tZkHCg;yoo&DvEu-m|z{^{a_F@0}H&RK`wA87L3hSv2MBoGg_>L_Z2u@cFq|M6q|a~CElg%|c;g;-&kb}4;GKCS2|9Iw&`
zuL2%*Y7)E=l(G~X;CC0AD~rg-&~IP|CT4sfa^6VIf-=ns8(%i+stuSL$W-BApNI{d
z4LWoP$V#6(Roy@@DqEw7-&&h#N?f~{3lXwZAaw+^y!aeAtgzI`yI}Mf?Hn7cPCHa{
z9iG9!UDtlXvJqYTS2NvAqIq6X%;jT$+-Cg`-x6uxdt6x^CSGIoaEFb`aO-op#ooo>
z4W~(rC?oUi#0vRAYVIq5Nr-4D8l!Od#!
znjb$2Lrq&tdLROMmym_qzmAW9QndUAch)+Ces8QrGJ>%_M?d0Lz6ML{crd=#0QP7S
zE|1=jNOV*!Vl8}IRnSl(<9iMD?Bv>xY##YaovAV!tZyR=!Io
zCkK7IQ3}l7jO8*8(bZ$|GuRof;Zt$fahU$7*W3PDJJ-uGT+>w0g`!7SN{3~Q-FH?%!6kP@2JN{Hdf3~TK0X$(?z@4(*kF|ZN
zg^h1T>n6BT(I_`TP+$iBMeqjBbBTqjpIXA{2N4i}l-m4lmr=iq)JZ1p%r0=>t%dig
zA5k&t6ABgRbFpT>C;(mH{32@cZ|*oa#fR$>9dr>r@q4T{g;1)bk~xlLL{4168#|#$
zy4{y)iFQxBH@w?u7Nq>UL|s#-v;ES_!el9!KX+qY$UG!0YpB;l0JSZ4@im}!0|5My
z5S1Dn#&G&;L3SEg7s<5hcJ66p5^I3yq0SV-^u<1Y;^Nq9_tCP(mk_kI9lJ`=dYMM0
z0JJxKy_$jo;legobi}%{qnyu9~fPKPE=I?0d9Pe!bzV^Ft25fj7
z2%+EVBx^r5eO<+}UFvQ@46JZpHJ5?63WWZAk)x8PY&bx?)4W?lbTY9a-BC1gnR2@R
z#H!X-Ug>_M{5SI`B?cw1)E$+U98@t0$uETW(
z2^AnP>3*V6ibKvq-(WY!-FN8{+o=7N-
zM@5Z9_+p!7mo_vkC?5{|^u@6U%
zVmc1+Oljtd6==a!q>x@)-jIhm2Je60)`b-l&)f&Gu1g1hmIUBbR-5#PO*h;#K~=-l
zcFtS6FuC2@9Pzl}Buy@6Itp;UJ^Zjp-IR83BIyi*lV!S1+3Am6)>^i?J7}F#jJRy{
zUo5(a)v#F)oIm{Ho4v+g#gpyPZC|MJKBMWn?U`=T4WiG})(5>cc_C!EAPya%oUzk?
z@~NC5w>Z*41=Gbna;U
zBJ!zdAWxxeGsK*$o#C->^u=Fcmc`VFrTiRyWi3ZkSW{26t8nhb$it(6`4-9dWbPx^vU$yQ8Vto{}+GK@v$g}
zrSUmm@B0ZAwJpgQPA3|fC!14N$MRGiitg<~(1s`YIAR?5rHw-MN+Zc-=wHel0(qjz-mZmpOSlXUtm
z;5_afwek;Yx8v;AQug%mR2!9_o^tBtHIWDR8S7~|B|K*t1gGtIUs2Nzai?k4(#0zc
z&C~pAZYn=5Djo4S;lv~Cauk7>6>h4jS7#%I^-unK*3U#AvAw>;)S5q=6>unqNnf|9
z8>e#3oGac5$c}!gxv6sisoKxy_t+DE%{9{{C}D2U7nn?)6|W?0Ie&GixZInF=`^uZ
z>2--dpp_nzDxLzZsy4}aR9MX8OuO?Tt~p<&{K_ft-D`NSb_L$>%eH%O_BCmq>&n`f
z`hi4~3v*$=oWr4QXVu3Nem4Y0{nk#2m^M4Aq_oeHWyS*}-Vck#ehblG!ttNpRPXPU
zU#%w$K&L>SSQIF=T_`{y$oy(PpF>p5T84jD<5O|7#;ztv2cHx{%`58%W=jq`rt~K1
zud11`gl*cKyW%*CRk0B`hdOx%F5+?qSt_pmyPK
z>V7mW|9JdVy*HDon+nbJ1hl?PQ}jJX_voe!&%47{eL1pQwiuaA(Gq_5k;;9AQ35+2
zCmsKxx*dS!W8~%;d7-3ikT9TCxr;H<93&$$Yx~6?96CO!4fEXc<0i}&dDgf*(SZYG&Uzx=Uzk68IZ)s#6ifNX(J&zf
z*hShM%jDLm9}x>|X#<~d?QQNq$(Or)SF|>O@Fr@Cxb-Oa%Q-^6d-I;kMK^<2@~}E}
zq<`AHekW`9z9EwZi$F*Gb9ZAuF#t?&A$;3*vRlb0^N{;6Njo=JeJ
zN`}?`nV@#fnjS2E<(R#wcwlMYUae_mks<{jl~#CSO{THgkA5X{MHW8hB>H2dJ=uka
zB`n5d&8RrZ8=WL1g@u$QdD6=q?=B)Tk@k{EJYALrBLeRI`K)#ep!t_y4U{RYv206N
z;BZU+xoimuiuZCT>PSI1{Ij{$l=jD)rK`LA5+(2L8L!7i=cpA*#QZDUA88WqE_IP}
zdg%Kf`QUL?AQoQ=!44aG2G2(AMHdu;j}&;c%3K|blc53y#0!|cvm|rQsn?H-oawQz
zdnN@|M|AzcQe&dg#(c>gC2WA!d9yfJwh~~4++9TUph
z#@~pwB}d#Su~!Wl7~A?JkGgNfsBeTo;Ypy%S?(SeDb0|7c^ZDuke!JBKH9kjhx>q<
zi1%>DvU9#u(crx)_o`@hb{S5mJy(A3fMp8in348_!fxu^wS$b8$7@kIqMT4ls?}eoB3by?bFyx*R(nr9ODx%E
zz~(S##P4j<90e~GBRLP)sCO+~^SL4<_6tP9a~D{1o;MKRJhNn>@E*2sLOi)85R)kk
zi!iQp37~pIwI)WWgHq>{D@zb9Cjz5$BTaf*b&V1^&i(KuD;;tl0$B)W^X&|bnCW%D
zF!<!wcZD2vrFG_?+x{TYYF
zv{b#9+oHJ=7JFXhAZVF?Y1gRdHxm#02-QQA_Lr&?`G6^{;HyuK0lz0ZS4K#Dl2V2K
zOTu!JOt39?NNUaw9_Bz9;nZaJ>{(FioeqrpG*#@f^Hch~Y|*R5_MV&JMPf1=`1ktv+yhYBUCswwgBMxtho85_-~2_^u$g52;rl`iB5
zI>&n5Pa^wi+ASwJ-&M
zjv!Ps$Yc=1*$=p1(=o#TVu6E>x*5I$MUs#(2_oJh!}bhkoYPP8A;(hCp=ok
zsusHAYYeT!Ze<#?Vbtu*iITMM`hUqPqBs0iMPC~aHWmdNshd&>I!fYgtv~xN7nD)=
z)$)orSF;yu!oHk^D}0kOUZkyL#pO!aZBj1Y{!VS>8s)5alFxrZHIkZ%+g>WM;#_qe
zpY;QwFIn!fZGf$gM)&2kN9AJ!qk6Q;()MlXYv1Ty(2PW72%{~fH
zzP8)*mbKlj)?FNGMWOt_e_0hC!B*|{s8F8crwxB
zA4))SiX-0tfR*uWQsS>defr-e9LE=c&himKM9GqM^zBbosZbontqP=m6jmit{tLQa
zZ;vGvqi0AlS;V55t^xfHx*n%cS}eKKAXC@!8$v9CIwXRD*JHufiYu@p;RbcIw}&)z%1IKPeH>H2Q@cvWoNZb(}s{QpdT@h9B`!Xcd
z=xaV^1vI?)${Yb6V{Rh(NvbCIv&LR~AslPE_5JKvL@oC@qR)KzN?m$XuhCR8T7;CC
zn!VztxqNL=WwtCW+NUDEI&Uw@hmxgPZnCsgtY2DNMUpc58nx!S)1kIc2y$OFXInc2
zw;jN=9vpmp;2sl;zE#4ZYGhB`Q!gH~7f8;HR8RN-gc&msm5A9iS5r$&>AvL4zt5FP
znv^1xEcU+Ka01JFso5%Xap3lHAH72^V3vSO)(wF{;{7wqo9wJ-66zuTzuCX9EkO@5XdVf
z%D_8rbR=p6iNw{QQZ32b^seH~L*ZSisI}9^ebS;kfeP6f5Qg*2FggaL#U<>!EqR2Q
zDfN@#afII=4j~=4Ix%8Bfb3w1oxSXaT~qikr=yEG5yCyNC02Q7-N>T6-WYZ`E$vAEbOCZ
zPE5$(%n>Qc%Q4||;!oIrQkP_}otLD101xTI4_+F~#=Du!C1ztx{T$$
z$|dc?p$dL1Yf~)Ab1hpz?g0%c(v*54*)V+PhXiDk*PcZ1XhrSBy7*t6AeP$7Gk)37
z=SXNsYP}B$Ss4>O=UaZiEfL?sFz0gXD|2ut`S?6h*HAPh4XC{^>2WAzxkdMYI68@g
zeVn_HVPN02;R*s%YvRCI{(NLJb8<`8PrVjU0!ZC8PxK=SF8^P4O}}Wg2M@$#+f7ok
z;*z}OaWtN31#m1KG{vqUonV-wtdKMjI;Z70(UnJkQ+QLR>eN^0*9^(6*V*BaeaUZ1~)W;O#iT>Qdgx`3F(v7<#xeHHdApj;r*;6Hug98TL
zRQ54a-O?FNXRyM#@$LU5#TfX#Z*f-{_#>+8{jL-`2-WPnd*4^pj1B8)M0GAepz9pQ
zn5mQZJtw2rGr2@&$Yff&GI9f4AmA9K&QFYGtoS|Q+v&%;4K_gAyqI@v@1oliJ$=&-
z^?KqZHr*tB9hB>crxCv0G6;UY21!Z6`X1iYhiX%tOU&Eax117Ue2pQaOp5)xa~kyW+9_22w>TlcWt${~QT0;tx
zbd82urJbIL@R8c*Xyg```K(QYzvY
z-r$GV1E+uLFXbROI5=D(gE;}Srct4eCt8yx+m(2Rra(NjP6N-=mK9fFM8&h7#
zAA7HzSE?%o1SMw^hmK(tC%a8s@3G2P3^$W_rP4^>BoH)49JE{Y_|hV)K3ak@h;zbi
zUJGN66V=OcxZZ39IKKzNNv#RqjcV!~;^)H6Sy$~3c6CAgP?px6SVwreaa*U%Qvnm=
zhsPSJGWA^d@{a{+H_f$Td(
zkEi`d$}cR2va-fcz_(gF|KAo3Iq`E0Ao5F|Jx}i
zP;|oRaQadG>c`jJnxPE}?&G)gnA}gt(%&TM*5OWFzWwjDMso`beiSYu=>maz_~vij$G1==hp*^z?=J-pNJfdLyg!$;%nx;6ve}L%BNjhFhlLJq}?b#xQcvFor?t
z7PyxDWh|=lH7EYG64_
zy=5CrCk8&B0Z;eZIcRqQGjhNkpOe+>_aR%OWuH8)rZlq}T#$a@T~eKYatUFZ+0y-T
zG%3F|c3ArEd&~vjXPdv4e@+Uv{DDKaL90JgL+$$(hoyrmgR{<{Au;n2V3Ed;;2qu;
z8u9)t8+dS^iOdZuSIO3`TT)==y(Ic7AW3Hbv`gZaM&MtB!wqaiRw%olDw^^vtjF8x
zFzpja;WP&tkhXd{b5#FN@SoZTk8f#UZH@7w&w0_jN^f-N?(AGoyO$XyU^837D89~6
z990Q(w)szeQ}1p@py`)UZ$-;bgNV8D88u8RM!iBXGDuqGwY>yKubWmwGEp{rxV-{2
z3HdzE&a4btt%cTMzd2y)~aPx6N
zpe2`DS7p)r_O-|lJ#IxRD;^K5XTsJjZ*!g-Z>!ea!7_eVKVV{pDx7Uf6dkh81zK6d
z)!q;LyZfoDby`I-YC8P*)--urid~?TPdn{$+reW*EBW(|b^8rO@pXD{cOP&kRiP*s
zUd8bNnc#5G08(3BO~qWL6?;LQF=Fy|Db57&{A0-O{G9$s!0C;O#)+@21g{r{*Q{MA
zv-3m(7nziC5-wI5Oy+9bHY~+_juXu2M>g|YmlC~$*cc&qu`0tl^A$}hw#LA6mC31@v1AXLEEz);O(_3y!0iFfO3)j*USu`X>2T%@xP@-Mzqc2p
zkj86#<2Q11Tr82%Uy_q<0*{nPHMMe>I1ou@6rJUo`pIMQ?U5uEWJU)MXOm6~yx8_jF`X
zzv%pIWc~aEC&F{%H6&g`yKAF~v1^`!?}*v+1;h~_PPOqup&co|+bK0h^~|xP&<82M@bNx+*7IglX8Rq7U9w$UvC#M8
ztNQ|ORlJh+I~kz%;+@P#$_nza{BEq3eSzQ~XKk#2z1#bC&1xISqvK=rsqs&A{}#Ms
zz8r3Vk5&Dvn%Yj5VOFn!q0rt(FuX-Q*;H!>hT(KLlOKC3@mvkp#v_ZSGkI>u)7h>B
zYxF4Oy$FrrIAK0jRr~GV1`;YNHyar4A-KNP4X#UT&arKOlD57f@@jFr*y;J9QZQ~c
zN0+z#Gy!vo1lhkIgQ`Wqy0)rEWhZzuzO4Lmtek3{;GdtmPGcJUsZr^F)Xb7)(L
z?mtJ4grI!kHv5bo_M&fE#b3)z9$?!EM||y|lJJSe^sc=jZ#x0eLj!pC&{Mp+6hPxy
zqGrD`37dW~M3m!c2o1)exx`g`6+B*9X>s!xR6pvBV~43QntF#phu!*hzCd2qYa91Kjvz*g*MU3-KuhI47Ei_wsP%kFDL`>4{&!~RK)o7ucAq~y8d-7h!
z#qiWdK4z~O{H^}z?!sBi*|IEF3bT{haw3cAqJXg#N@|WG@_&yxNQpuH%*qHN{PtG-
z&r1eLPd9oHhLgF`z2)#D*Cd6h(}z&Cj+NKJs-mL4jwu-)2=mL6eBmsG^Lbt31GX-K
z74>?~Nz*#7ONVs>oj~72EpuzWjp1qEX&ZFUUAIf?C;`PcfseSNB4Mme9^ds@`8fGF
z2m;q@8}KciYF)YkjbZZlZx2Qty}50_=80P@!Mg<*&q5d(Cld0zUkUFrwu>dRnZGR_
zw1YsItFy+Vd|-k003AyPdp52mQIMU$yz!^7ei~a*^O1~Zn{a91g@Pmb#tY@w+>C$n
zrJuZe5
z3|1%G8lRFwX6uUsYI6Yi+NsXJ4p;>}UoOMk4jS
z?sYD0D?M-RB`5FzK186XIFfOT@mDrLeXm*R*0yC-ueTq6HLbk(Mml$0(qmy@Ml
zHJDTwHOUYAx=J71{JOf7o9vR?=)L9|i0;@NjJ%e$=WYti6Az8nEyKF++%x#x{MgJV
z3lFDzdftG$rs1|tYlZ51eT0KVBoKBuTfW&9fbhlHS^c8h;U`hL;=9MjYn_~toCL%Y8uWfvMd>DC1s|DgA7{VJrgKKk%BVmy6Uv6goa=E8eW4lU7!0Y@qJX}6%
zo815R9KL)AqPCJb@FgVE`gRDSdN_h{DePz{^HP{|~y#RaC$
zj#gUnQaG$do*r)7D|D8whgo@NCSK>To$9VZYoqd^BIJb5{7-?7#2L
zgb3A`O~mEsUT-s5FR+yuwCmHWBja(rjU*DRn=u+{_}dO~Dk(Ai;};(0$-nE8O|o`)
z7|Nj6L~Ps}=6ZWhF=w>-aC_0?YbKT=wO{_ePQ+ZS!hy*p0~E0{Z}B
znSQH9T^RYlvxSlj9*H2^L|oJFc73ISUOBw2w^_!0xvQ^4vocN@+!Rhlb#>@hvxf$f
z=kp=PAT)8pZ2iJl^~v(R121Li
zKSnT3Mr`Y=F4)5&^k-C7C=HJLXy6D3p`oFXtB|J6jf~G}gKuSJ)s-*&gB=Cp`F{2J
zbk+O$O-L*{_*X%Eu>H0|H{TCG4aBU2R=qX&?rSZkMG6$sqF7B8UM__u67Ym2N{iF~
zXRK-@#PD9l$qG1Oe+GXI`*Mrq%k`E&l=CncpGHzRnH~2&R9pz5{^=VF>B|j*HCFRz
ziHru&j5drqb#lk5cz`S-mGF6{bva%LSZ;Qql_7+IfpOd$e@#e8xT;%#fdTc8
z^@$2$+~K0x|MM&s_GqaQT|D%i_{GkMQoW5{g*ae$G!4_BZ2$s|=C7X*)(?YYkKOSC
zI}^KNmPo)aq-~R6-EtS_Z;uligNJ>E}joVFlkT0gZ*G5Uxfa*$Ci@d
zLCgqrvk3f|XDC?DyG!Y_CzFw6%;Ec?h^M9H<%vvTNU>njNJJ7Jmf`9ru;sB(8AVjF
zfg3Y`+TtXB7!7Ys?+L+_rMW2AZ!N-jxY!*FY;A4bSOS4SoEmZ)Z~k#w*sne+3w`kf0s>+@Wt}ncC;n_v<*dk`f%$IM#!&iePz1q2@ZK(b=He<
z-EBT{XefV{!k5TTs7Un(#x%jjQ6ildntTd}xG%(;3im4uCW7>!e_;X_Y#FmrQ8|Cy
z2M*=s4rOR$!na^^)ot_C(%LN7sYh}<-wKAxNhJ9b8`trCx$HKmB`;^o4J6wEk4(nB
zZ?7`=%_qrQT+c{7UG&|l{sAMX7cfHkp7+NE*LF$209W|Irl(@JCFpsBQlOa8JDDdN
zO{0>ZvB+$$hw^7QZ6ZDd6SIX%kS(D$-QN%oycb7Dk@(#&eIg^rG;S$G{(h4b89bUu
zJ_mX-%>R27^c8lg$qC_jv0gHPPAd*_S~iivXf&03aDfr6i$wI@pC95wyzqChU9X*-
zM3_xv`?`S+r;7t)sT3s|Dy0Jdft`C)lz=CYF*W<2!L1|tz7SF{tQb-Vfzf7decg8;
zmWs^j;MaOt`6S6dYmfU{M@J_Nk5fTw59_@uELaAMuXhD>{&IVM1R0EDN+JFIZUsDe
z@oh1usQv#8?=5UYV`CPWI#Vm>!m@nof5phi7>3V{HV{LB9`rYubiktoVKEXmIUg52
zZ6^5z-d?4proP}(i|6emyacuAUqg8<3r31nG@h704ljlR6+yuJZ7`lT2P_K>i}}H^
z?S&|Skvxo;EXwB3`uhn>e`P$9?t$t=YT0Wdr%NEOAj{DVop
zFH8?AnQp)6PYep{SZVc8s`@0&W;Xt6x=0PuVaElyLSVP?lX>+0?my_C2iwrp^zrIo
zn#<#wP5|%_gwJiCsgxseup|>m*NygP73p7=zea@-66?R^?}xRTuM*eVsP0|_`*hdT
z6uR|dop=~FGwFFnF7R&)rv$JS0Rk$grx4!hd~P4AP2}UKl^{|$ZF|3WCJ6j~7r^rq
zs2y0S74pvL^=Inf
z3%fH+0|Tv()97I&E~dj#_0?0SKg@Jl8`!Q$)QVIEi`>j5|GJb29N2RM(V~>Y|E%N!
zY>Nj72^^LCUWg=(3gz}zLj6dnm_8KXb{i|yABz_;^i(RF7-bz5%pza~1J*yj(%092
zVUplQ$bSHyMhLDvQjz`*52%o@nb$+GFUVW}1HcMWC_+A0LpwG`gSMYHXPfJ5Yd(6-
z&PI=SS97f%^&1_vnjg?g34PbQ1IbK=zZ-TO|-(^+})9W9*i4&
zKia?`C@A=rfCuCH`WkG6CCVGNOAV;E=i97cOO3|x?G4B4jtgLzI${I+Sxh!*_q&&1
zbdHDOwsR*L#-@}lwzpVcZlAf>;^y!cn@Nb~28`vQ^|q_YM_piwsgx=7TKv|8#b4|n
z+xkuhW{OOKnz;W2fC`pXumYLL78eDF{ud~S?_lhE5INr3f+8X!!WRVK(`uB7&6H@p
z;SmOpsb#KvHn?(be!g>lex6Hv6l{APgJ5p=iZ*5Z3ri~4Y;ugczr9=x{ZcbQuT$sq
ztnuwF4g^4ALKe&v*TKOMm)&9w+`~I^xW9C-16c%Y_JXix<`Vx|vOQTQ%>@y^JCw8i
z5R3DXrsK&H4i3ww4zM(`dwOuA)2tA}sE)o@hO@tH!N#CcNcHi2xj_DKrWA6N;(ru6
z+~)yW6ZG(OG*MZe4hAe=g1su141)4{i)nZ;N$Uki)E63hHC$ynMkqdIJR0+C8Y#Mf
z^?Pnga9-PmSdLWF8vjR0V6kBL;%_qqyx{lHFVbC(=1{@bBN~E11F->aQB%YRlMc+E
zd%$EtjbrDRDdCpt4`r?Ql;$rULh<4u@p7M_aC*TI@N#IXr*SOK7mGm`c5OZqllC3B
z6A`T*AX5DnSJY2plk$IDdV`Gk-5sho(r2jO=z#6Z>vBS!Oi(vGOMB5c)Y8K*pUjp$
zjNEsDMcCXMDhNhaxIOJUt9h{`BA;(B)+_m9z1|VsQ?Rj=@*%#EFz-Vj!OIFoebE?c
zl{N+s0269HuRAtAC!M=LUZlwm4u16q`|+J{|4AP?D9xlXBx`w^i$()5lj;gWC4-FHo$VjpYlDbtDcu!zS^!a`oLh)AP&>7vl*GdOYWT6lBQE
z5T48AvI%lD8m^txV1FPf6NP`V@}J`1(UP7liT_N$xJqC$5((9dd`F*rXTSq?vm@N<
zpg8hBKCv1pR0Av*vG~8KK0-Bw(FN0nZ}s(@Hmhx=`q<$3llV$D>-iDr9fa0XvA+jq
zdj10P2zW$|voF~JzQkILnce!h$p0@n&}Q*pa=_i}l8nn5DxQj(jxG{(5Kt<7CcqOFbi
zbY}z)pWj`Td)ru6s5c7e(W}wCTf4LX^nF)8mJ`2D7Ru4Suw?FZ%G_Y5O>kF*&w*uI
zq%V6LemFgv)D_&v<#zsw{(XN=qd0=u?xs&{A?jIe4B!_RP@8DJ5H
z3udj#GwZ38XdJMox|L%YqJc<1JWhH-u|jMDcVp~!9TWN@iHf6`lPfN*l@m_p0>o*r?2oi8o_M`73DURRDID^`t{;IT0fqH`7SoE2^)v?!dHA8g6L@)dI7Ho9}MX>&D@J2
zO^o!Kv2&s{OGXp?F)m9+D{cMJ?(DIGo*z(m8&#dbQwQUN=8_I)`3Ku9JLi^HNj%BL
zL#p}_tg{kkKU0djRiA4-7nD{R=_4Zfa_$;`ZEy+pIV>mhU0zV-DLd+3Hichk*ko$g
zTI@~LXV}jaZbuvcl=MmKR7hrn3&o=MlRwY>GLVEg5KGzhyr;O5ISzCCo|*38;;Ji~
z=JgH|E2m#0-rM*zetrB|8RP2es^=_D`V9cPaH866#Ma#hvI01RophkX)vi7E0
z&~&(CNsIM#|50blK_(*Zte+w3-B`B7tfyE*WfP7DR1dxm{PL?qTi=rAGVDmiYNS%#{#X8e@c-%_fLVoMj;<)3Rhr4-M~zz+9>W
zme1o1UX}yc^J`ltPTU-n-a37PEll;1tbhfTkR`oElB)$}^=Ja^#M1>1JnHhAUav#M=s8fGhipKUu5Kv^=
z^}&j8
zIZHKwu??e3a?R-t?fKQ9Y65|-^uG3ZS64&
zHgZn3yeJ#z7lt=q{tBo4CGY&5$pph4L=<%4A3`SX=>~^jIz_g83kS>Rc15VRn<`p*
zC(mp&Sd%gCPqerNhI$NB^HP?8ichMMM2ihICh@Rx`c0^0yQRJ0zxB9%fl65i0av@{
z=e2G}-MkK4HVz^TMWarhprBk)3vE(txSXC?JiWXiZUM(Hd9};WA;mFMsVNFj%MVi*3bO7
zDBf6&{L18c5%qlI#l%sj|4nBKH;CmqMJdj2CPb)3bE2};)VxHu^_@TTW&MoN;7h)f
z7=jo~UtyV~C&h({S%@oaR|UdwuX#Bh63~M;#3J8{)*UccvC`D@J`GAIt07;)ATkIa
z?KKEBtM^6{M(}%FKV+(VTd`UxY3zJMJxU~BK{c+rK+;fYkg74Z0oDLIMjsu8%_
zeUHuIXhtqhyw$x@VFdYZEG1ZeS%y^Mt-UAhM@-ly?`vEvTfy3p%CCXKKSK!Bj`g)F
zT?y68^$7@@$i4nAq3ZMDM663Xkpux^9q~25cO;#L9~wjYD@#&0vgGcyEVA9|GrcxO
z6@qP4!2OeMBM2fdAHQ4h_`Gwvvr2SRM^WGSc7WH+V3<(4ANSF8XvH$Ja<34X;YL7G
zBE{8AhmAnSw8~Qt@Nk)dYg^8uP?*5_xdY`pgXw90F_+sR&PTCuKjoyDtTtJTPqJx3
z+qVwS@>Y#E#KeLyx<_b6uC;Q$8#U-Bi&0*YcDDiVcOJ0{y)`zOU@()rXjd#k0*ZH(Quqn>imXYycJlve2>*8H^ys8Q-*Q@&?l(-TV
zV}FyXHfMM9*613;X%JHwl+V6qcE8;Q*DvSW?tKnaDU@a!@-r4qw7rv2FG3oNPVT4E0Rq;NL2la_4@BVZE+
zb*iA^_ZqB)ctSpB_;8*}zrISnAaM}&&@VUH3im0FSoy2{IoxU5IHWlB94#0=cMI*a
z28U^ppP_hE#Lw-45n1ll*P6|SJn!S^a8b&Um%MMV-3Dj6f@7)Hx3+Q>V=7P~%FWNH
zbE+XtI7h=D6xMg!SP$}30gwhnu0h=vynY-V8eA_uHE>h4;Zc;
zW>&LD$4A)KwHWD3@9nMQP6%c}S#|)O%ks4_m|JR)FYaWZ>Sh_
z5#pNw5)}17!LAMO>rD0f=mzH@4VvYv$2|cXfdt$6Na*+w<7;5rmxCCX1>|Ry3<~0w
z-S2qXj{=x>#(iDG!UW?b+U9Zwm2WO)j+Ncq!!jZReyQ6f9(@wiP-Va@3=qd
z_TjmFoxhGf-G~YxVng|uS1w8U5Fil1OoQYAwE^b_c>Zutjq&wlq3K
z2Zq6mCE0q-3QcA2S7xgW7$Oeup2^BBy1_l!uE5ZGYns@D+(O<}BNJ;qXntR5L76O&
z3+iPBt@qd7Q>LWVj1ZvL>kD;@jAE+@3ffF;h{d1rcJc;ZRqcmyKJje%A~)?`b3f~M
zuCW&=vBF6w72X=`#(iwi&@xqXE6%gI2@ttR*;3whKz7@V1F;TUrlQwN
zck--d&~j~VM$S#-WZFC9UD9Tx*C9&w9I(1n?hcYwC_6%HNhp@ZhSPQB7_S)F9<@9K
zolA;4Rh6Bn&A?S7uXKT9?CJS3Ri4kXP{K~{c{(A4qz$OU1au!H5t7Z4w4Ybm0I5tB
z5`g@vlHX(((W>p#2V)=!w}(BV=)hShh;>h$_*VHnrr_!t&2qyyo#%IW
zx;XdOilB4Fps
z+1Vns*o7fo*zV+}%V{T6Kp%p3CYRd~vWl5erUIeAQ{?I)vPHxRpGR@YGe1$gM7?r^
zna~B6uK8v%go>%0c<4KNKOOR=U0%-{rjJp-BT_lz>8!n(73AzbuIcU(NBSY%Uiq@DZsWmnC(dXf>
zcppH-P$>kFHdY&&^+^O_vO>SJo(u2ZW$0#?lo&Q>Mjtb1eVs@&o71C#fE1hf09iNg
zeRhX9SWgg^=I~Y~r_s_SD(2aM@7km60B1fAZZcmk0uYO}`AZy`j$j$`hYESXSgWKN
zw=4X6tf9cig5`C^$DJymv~*9xv31~RM@c6^)I4+P0SJs-cE658&Y3Gd9s;bWWI^_9
zJJQ~d{#xxWLcaJ(s%|RQDQ{+}*Po09r$I8}dQtrTtx84PQw|y}6+Wl;r)P(6nhsam
zH0wSSC4}z>;O4;>hNSiv0Tw%_*jA8zf`L$MgBP`ij|z_Fs+i4@
zX`+31@zX7K-ypbIKMwh-Yl_J{3GQh#QHYy?AbDMv+Pwv3;@^LMANuq0)Bm>BYb|Ft
z&O1vfJ-T?~7A|qaCLles&0b^awyp=PHX(~`8Wd;gv38G!YSZ&%iTubmo605RI9>{k
zRxw%cmAB26O@a2}dH8n6PFvihKoQsFi$K&Rb^B&nDY;IF6s
z_IMd*TLwcozSp|*-KE2mb^xe_H&-%H=&@N;A@Q(&KZ5XB$MC6&y+4|QtoQ9fK>T67
z;qBeXmu6RSn%zV7;vkbNi=rx`y(S_i3pD2w-mss+XxRWn(RQvg0h$t+$fOXOZWv
    x`(*7{pvA}h?Mc~oh1{O$;=o=4j`
zZMtbVU6f0nfyMc*oc;jG?jG7%X4fOtVM1OExBo}QH+EOLb-{M*q@x|%cE{+B-LY-k
zw#|-qY}>YW$F^{GXqE1y3{?&tI2>Itfu3uN&~2Js2BveQkmIEYPPD59Q*xzH6&>
z+lFtS0XtnPA)Z`<$z9Mi*(@06g-Is_`uyc6p#wL`IElr
zWq{iR+~h`^C2i*`8c)-!AHnK)&vy-F#TvapOu(xs`k-;~4mn@w6{A%sb>*#QU1@Q5
zgVtEs1xYPLp7&&MN2+@lM;;fsF{rJjy|&f>@g|rxaE{w&THmQKmkrAuMv#H
zkxnzOsBoG8^yb^kXolBK-tttJkqI;$6|}D{wai-d4#xGLO`RvoSxD_kcGspMy%F1jr*>&tBD#?lv@S^$9{D7$RK>$5T{j#v
z$(Oav_8x8BKuHwWJ5F6={|)3Z_f1F8%R$
zUvJgKesztAdzRe)s6egs$6cXfVhe+1&G2Gk@^Obn!a#BsaYj1WI+ds3m3DtX<+xzp
zWhYxX(NBXZ4-eGduPO1Uy?+{e=twR+j02PKMM{&vboHP2yh=}=G!*Vpfff!q$%kF$
zIe}=po$9W;$m4SSp;CT1zqxLud)ng4UONf9#K6cRiN#lK>@Pm0l~xpJGLc-JY5Wfu
z?z~e7qX5Z|u1vwSz#5`%SFD)_b96yU*1!G8Ot3edbx>b+>?k&=*=kX8dMd^k3iLjT
z+8vhxBXLt&h&Y&VPbHb0HX;vbp;**T*hw(Igearo_;>dED^N#SO?tw&C?_2^=6OgJ33U6TsWSr*wpMpAjospqbq{&20%Pb$-^OFtElKAwr^
z@8Z{PFeAP^=j4(FoY!o2s(8FaVSQuN?kSu0Arj~t_M!S~Fi5XSFA-V@hu*8W29}Zz
z%WV24-&Wwz12?MP#4yaHTa$cycu5q?NfG^&*E4K5kQ|^#;q{8`n>qVE=c9yv*iN{3
z2j5_(L$s%jvpQipT!GP?54`>-FWc%L_&8
z={zjk#`2us8{ftE47|kIg2iT^*x1okq=C^V~@-XIv
z^*69`(I}DBQ>4dMLgdLj%vX44&#D!R2L=D9>X)>|PA>o_>wLy&^w{51ugfzgXhePo
zhk>;AI`G?41>9VLy_el)ciplbLr)pm*2r6ImQYah#?HPbbS0lz)KmQCQ}jh$m0y9@
z8q-rn*`vFeOoxDQyr3E~d#TEha;lKSQkSj4;QpWIF0&D#4vt}vwtuF88?&kXmC(yn
zHuELa?P0Sf5&PjVL83bo6BFOllaiE;)4}Fnr^OYmxEY@l2T_AW1{=fL{-)<^hWieD
zsLD!ynbbL>AK7Xu?uoN)x2*)N;_nw16UN0y)7a($W<844j?l$)C#}!0hJSy(5i3?x
z72fMk#O_Vcsg=JE@%kS3oMLoJC~W1sLo=gzcZ7tUc-#j`E&1KB5!2b)33vx=ra1L7
zYVaWzg#=?3(jWOYq%>m{edw~qDGoR6c2hMC^wpP#P!a1vON*^~*w7D5
zIVwXR`Pni1$01@^%V5&ADK{5)(TAVnHN9Yo==}IDhH~Rx*OoQ&mhJd)%z5lC
zqOL(n`GLFcqMPE}ko0^Y=Ks*#b)oL&4Eq5cPPMw2m6hNDsi&gB$7I8Wtj&(pemG
z>j?v%pWe3&#E-y7Ev&Ywze22EMlRa}w&={17aDGsE1_2u$kb@-yT8KGjel-rB@ZKw
zd1eb*Hxb5EIxOSzN{>HFovhYhFX6vh2Xvp_>nzxA-AFm&l+glzh>>8y-lre``wAcV0g
z5{f&zQqFam%`dqBYODcEy=jm56L$rojg_0UOUJVD(}?FXp7k>y=3^qiM4t=QCb@-ZziN#g`eK*I7OzO@`)`H-)dV?6AJ;`9Zi#f>C|I-%&1o4rJC%zh~%r7id2yv)gD?xkISp-
zf+J=(393S=97qPM)S)Z3)I9xZ8EPG3tZCiuJiBq@bWGzYF=r+1d+P&2<+b6IXMVr3
z%0sN`EhH3R$2h5
zBvNG!Wq12TI$f}GLXE;>DUg!MJWV8R#%cat?qW&oqEY|MTh@K1@S+wcGN;5LB~i0+
z%;@v{gvD~qHOT_)2kg>`5cAL(|;k=?a)EZ1Abq4
z3P7+UG5&tgKv}*O6WUZxEBztafR%5_J>Y^SsH;8nG7Y%3`e8!_^`gaIp`~=Q$fkm*
zm9*vc4BNv|9u!bORiNL!POa5hyYjg}Eaz?;N0alp(hajPkv3f320hZvL6lQX
z1HiXbrjMei`1-k4{IFmMp_aZmlT&$Hl3ul@-G2!oKKaON5Lm-7yK-`w$bn};)n{2d
zn}tO~iF6SL%~7FGav?{pT+}^Q?EZ8Ay<3)YE)cyf;Cfp$d>~U5nu_-mn-j^kOzEdx
zlV28fT&VmjddDa7WC@XSNOfDo&g98lsYZk8Zw)dNo-H?zZ~=0zeyIyx701$}QD+@h
z+M=uS=^~XGx^7UiVTOFMwVRsiI`5mxuTJ$DmCxCNX2YETE9$g?RE&GO-x2gXE0$Tn
zsPg@~0kQgEGb_uf{*07eJg;P>)LRdN#JtC3eyP5U%$%V6{>6n#{jG%_H1)o5$G0&$
zd=iRhanI#l(U6Vg)H^5K9-`>Y{ZK(mKbBGTwJOB=Ldj^kgL+b@8V*bRo_4eazznF?
z&|RTpH}vTC+~T+AzG_kB(UF^z(RMbX^+xx`l=ua#e$iV_RQItuBLaVkpCaH(#bOd+
z6+`WVA#DQy6>6>geuWgyI-Go8m4Ti;f#*wvGbcD|xfPYX8CJcCvWWfQl0jHgu)5NV
zK-eiXpv*~S`&aIl=5GsS@~&5|HFEa1&BurwN!K~KZ;U;0fM1D!F9X`4@MqG;Z#%+d
z-%MT^A>oN$yJ%Twz0X3h3>vxTc0Cwfb)H{Wa_;KMW3Z$r;B|AiT|HnRCrbUMX=wBM
znr=%sP{D9$dY$~@T?5w(j_#BjN9*5q3HUs3R2kS$?*71caIKUc-6G;lUfHv#$nE}b
z(MJfG?;~$OL-(H+=IK8nV_~@2OG&teyt;Y-2$*l2_nusN7ncOZMd$rRhQQXWNhprN
zS*Z-F-m?MO*|{?tpE(1HcMK{lQ{C_VN2K;c#O*3PrIQ9`BQ9(LFMN)N#t
z&lx~slJd@>RIrX-J1DEX0d09qD~w0xpfs76^rF=(nnn3tuI13hjg4>BmiBt=jYO80
z+H)}>I7w|jN5_y*Uc3DaBR{2xGHQ!+ZM5o*as?ZFjI%_w+Fa46KODBos^OUVwxDMZ
z%~*g^9uhh$@;1yq+w5K{2ZlIN&9MG5wE#@>11k_+EC2}k*
zNyWO|2R_`OTdcVBJ4(vJ)N3B4`KInmqJ3A912399y2Fq}
zdg7t;-Kh3-zgq@M0vi1buC#T#&!YoD-HZF&9S
zMQ`&B4$x>nwYuH&@_96d8X!w-eDL@qf<9|xy%wUlc}sMJq+DzQg6>44a53*fT^F(7
zFb+k-<6)IvYT3&)rDm||l-8yT3(fGM8_TQ~SuRr{sAinr!LN4*8b56o`ok6(mslcQ
z6=k%#ab&g{ye{pRXS6jElZ}0RuDbg&wiE><-?(AdA09+q!;%4bW+AWsGfBzdU&muk
z$8%*IzEneR8c!2kLhv8?)r{e?iw~C~SBhoPKQ%@%Z@aXK?HDW9>e+HXBa5PF+VNsb
zwVNRoSIEU*UKtdc)eVxFG$0|`n<4IMVbo4hZIJ2iG?vS^7-eY@vzZ}sHU+r6J`*F4
zkQh68{oS&Uv0%nA7_hBFx_y&mBwjact|jspHEJ&HbZgc5Y!Gh`-|PMRl|_GyA;Bvw
zXl$qjKHoT-k)X)sbirauV+wb#uLRM1pEWbNgXv!e?|{i|dyoXW{dWGSqJeMUM_xYp
zZfufFq7u2!Se6^;6A?-!@IdCZ%4G2(7`HR*Bdmv(L;^?Wtl_TREV;K!gM(IA>%ys3
zVFfyfH74TO=g!t8BG2~G{*YACGFP73+GF5W^cD!<|o(TO2qFM1qP(C
zZDsti`uBRr$%rO{Mnx#uR1%~toaXZph;JnH6J8y+Mt}&UH_$I5WwTzcMcrNP4T8esQ8AtbcWJL$J(hmrSoMis(j7B)EXXbTgArxO}QIv8)?VcvM!I
zMZP$NmC0Igq`)B#KQ*YfblkFoK!C)gzl;_psO57PbfKmAd#P{(mR_5wC%Q8~jYjqsjmQ^V6Ox5$Y!~Vh
z#!khgJl)M0jtFCM*d+t6#zEi_v$kACcQwTUR@CPmOl>(i=m0V99
zt{;Y@kRV(TFyGOh=o~JH?%m6e6X+lLa%M^@B8$)txz
z-~ZhCb7*(;G+B<~M(z)`2;M+H{7d|((pPU3{5i9#69!IJ^Mj2IuX9FeC~zp
z@cG~sGlY6e)7!|;pRxGc#q#7Zw$5m%H>o6a2+GNW{PT!p`|2zDnAe)YHPby<+$7e;RUr$#bsxvtLcLvbjtC#rI@
z0~t5slZ>iw@0pQY`ql(YZ^G`hw8rkESx3Zd4&u)oZ$2yotFzq_cs2~8+E7jp^Sf%{
zECb-5Vz56LYcK=T5T3oN>Ui2?n1Nu4^BV!gU(*yzT+rT
zi2PJscO&n96QDzUztH?#-iMGz|GZ+S@|_1bjb*;pn9^eEnlnH6<{p??Vs;&Cd5R&x
zcG??A1>z3Wybxe8pgP?<=@Rw`-#(AG>s83dp(ynke|Fa97M;ae-&BuzX?{2MkO%Ge
z;S{LG#-$EdfJSx__N9qSbQI4Ev(cR%wezyEuu_1*o>1A~Cj7$pzs`+_r87j#b!d|m
zKhHcKSK@>SaP=gDkEX)jiH?Wj4#}KOYy9FoIKdE_-_sDxLSw1K!E&PkoOq`>EAtBwy(TT}R-
zX85}%Eac}Ae3fPq;eQ+IzzCuH;~7+?lI0(2(N*_fU5^qPt#RcEIOq1s1Z%8m9m5LK
zfPHAooF@sfNFM;_Zn;^VJes(|VrddcuO%Wgr%ZDhx7{M3aon=_rr#;@O@KN?fH&NP
zTl(RMvZRWBa!cSV9Hn@EtbF>!eq94r^6D5aJiDDxc2(BQo@LpaR-Z#~G{s^WJ5_r3^
zLpSo_v#K^TZ@-*eJD@AuImY_MsDR57gUAH0-k*Vm(yc=Ibq1%V=-*(ZdAOS8at#Qz
zBZJY8(?i8nSogPF0rmB4z#GC%?QrF*;I@6VI|i<&mnONpfg359U0SqmuzY}n?xs~g
zJjW7g3Ein6w4Ru|z4t{S4sTJ)YJc^RC|p=u0`)YGoL>}E{(Y~C%hl@s8RC+g&
zhhWmRusHv|
z^cKq$`fyP$pvK3S(9>U|UTetkI?L!E4^LN=eLlB4d5P309Y)VJc4IBR``2vevmuMi
z3yiT15pWhJkBuQq+jksA;Mqg2>mysO{WKpqP9Q7=HZG-8ZE3+snld5?l
zzZ3Mm<(4Av5JLZ6_kv$N1*^a;eSEL6T|OtGs?x>OR^cUcugP4if^^54wz@51?g
zf28CbISi0JigOGMpy=pYP!Oy0OFhydx|VLyH5b1VT#R}G2_9H4Vv%mT8yeHQfTc>i
zzofln?rAb+pYUJ6^$n!YA=H%ro?~4h>LYu|rXzG;&humIj!=xU?2}{=6YhDQ;wix)
zk(9L1p9pCAbj42%DtCWd8SmbS?n-ZXf{(_0&>;7G
zQYH$qQ;Rgy28|)&GzNUDq$UCQIyukR@T5J=%c^-Q6`Q7kE$w=`UK_6@gyRI(?PfUq
zUcYR}2pmNCzeZYhUrpqx)I!GpfJ=tB7k+J9Vk}xR#gLOxhP&D%
zgl!I@uIW7~4LXX|$LH1lV!aF#>kASW#%bS4L-pRVt5Yk0`_;p`MzT!6d4fYF8H5IlytSj1X?bK&2
zHU%J7Cxe}hIov8LUM^8m2)q%Si$$>`-Ie_OGg(}9e>ZapX=Fg>=h`|M1<^S=;B+kB
z!P00IT@#Y0RZ}2~%3I&bDJz+~FouBaZ4NY!eh<^aFg>FB_(aGb5cZ@~D4&Kqq*G7B1GGMwqp5hh
ziCQ#Z;6ARBM%J$JdZYb(Z$F6o?&fQTVY7*~F)~fo4ygrtm`ffFZ?Z1P&xSS*7vG&%x8)GDx(?v
z&#{W$9Ej#ROReg?+;`hyBD5IO((_W1%!Zd+jq!LoUFq@~?Wj2BS?3l0H%3?wDE^tzQhA3B*?-fB{)RT{q$^83OcK5|l7bN6^G=_VID43+K_Av+u~d}wi>Uc8K8&r`J%7sKJM7Zh|$2yvYp>HGjg
zC4sHI+dyL%Ki?8!n%vonW4?OVLQf_i8Km!wzxDT3Y6`(&5M=T;k0Qb!K_GWZt^C#S
z%Aru^{I?@wzk4bE{S_OqJWqLcFF+@!e{RY-Pi1wPeRQS29NKWK+E?puB0O#8BRMvq
zx=}M0={Kr55V(lo$!IoYBT=X-;I-S{Q<^pWSdofeIl3VwXGF#B2z7TMzu}OOd~lR<
zyPk$|Tj|W0!;Q&Zi1hAlS3adTrf_LoKYw^DgU+5x9k^`Cho|ri$pNkUYiZmrE<$T#
zfpO*3W3G%+JwQ7+3-sx;j;>*Yo;2h^Dnki`_{w^OjvLB~UbaVWo;Py^bUgRE7#D6J
z$$%-xoPOl6oUR!aaZS)T=1+7ZrxHHqnk~B>>#}rsJ{gL(M8WwXSbt1EUmpQxzA^Cc
z2T+7nS2&~)SFR-*?F;OC^<^EmT{ZDd_JEWH1}xUkVck)==}gCq6g7|#^66qXe@_Rj
zpg>%GEB$42w+GVt=|a!RL;YS$ku5d*
zXDx02$4AmFTTCQPh`Kfuo_E!z;T`-oj=M1V>_K~$
z`rz6eR~W2b$}Ed@#t=dYqt39?(G{<1LVw~x+D%V?wnz7L04~M6FPBZDJQMx!k6p(*
z-zItu-<7BSkh5u|QC)D8DpA}-^6|(70=o?Q{J~XJxcqy`WZstT5Y&8@j}+zs0rX^^
zT_?Y~9@Aa%4~Hfdpi`q&VAZ2X-NCtkuHN;-!}YExr{;EgJ17be-`R(p9yZ4@w1kVm
zqkuHZY*`bcR&EXex-ZHSa2LnM$4-=AZ$?r1wu#Qbv8X3@3imlcbBg*t3^~q
zy~XtRC)A34KS-kvXa>O;7y4JpvL2W>?~Q6u&o;(M8Tz_RNixvAN9|7sQhrdCv#WMK
zS!m#)1MGgoA3WderdMfg#R3@BsN;5bL0h>n@3GbW(E%4TwO8&YKc_v?0cPGHfK;2_
zVM{=`D16n-Mh4wTH5-lH4d7$OZrpJ4t)Rit_U)*=s{8|Lpq&=?4_wqzf$;>`=x#c|
zdI1z^2k0KY#{kDSE3=W~q|lZhd*_}%0%P!HuKCRObEO;U&c;*aP$6G-&~BlWsc3W;
zabqW&c*WHbFExLgr>QlxwhI|z%O_s(eU4rw>5ZUb(pBqx=|TEatdbXd&SXH5Vjxw-
zKe9eA4$Om?fq^ReyL6OhKPV!kbc{u}oe%vx8vmf4nIMj`i>FbM@6|=iZ6W_}u_V8v
zrF;{#@tI#gD=r}x!m0+&7D|K12M0ikEt;Y-dbQD)9n{ZOtLm5~J~z`S
z0H5qicDvlsWO5z2s71N_C|APg>xDe?Wv{dS
z8*9x%F^2>gAt-aN%wPOh_}i{Qr%!3C-NCKEqDYM2z&8{k^qNl|`odRB6~?gYs!ZQ8
z?7;XRt{0`4J~9|;odvlT1Q3^+d_=$(!I4G&HwcBZS|cNX+O3cbO7RcZLWeRlUs?+l)Rmn`Taw#0bknvjkUmdCf!DGC55Bks
zu2dZFk1Lb;Qw9JeUR;K=)QbrbF_29FV&#~!qhQT<*ttLTbA;5GYk*i0Erc{Ppm-e=
zgHldETd1bMG`L?#$l7O(6VL+Pg`H>>&s9Tg!hN5|R_E5{1qSu|dp!RENF+(dAmkoo
zL720B#YgU%tt32X(<@$HH`vb@WUBZ+!ZBamUBV#4a14KY55#8T%{$Nm&jt|n>0L-s
zQ}BVt>IXQLJI=FO`De&9(7aYtdLwVU8(ZsBFmA06dUP0Sv6)+xRe`DjS(JCb`jZJR
z)7#$a>wAe5mP8q5?YphmnPi{_dG-FNI$H<5>jS}K%r2D{3iCzDTt6#HK4+F{P?qwB
zf!S{4g+BI`MjxQDY>>cej#@<3Du|5IZlfQ#r5;NzuV&+43n{p-`j14qDO9NdH|
zV9sk-Dn*LPmeKLLj;vfUB4*%rSm-fNx7?@DfT}~|5{`5q2iFCLof#7PI>{-#P~B51
z%#f3rX~GH26`LQt>?G1yEjtQ8&>XD%5+uJ(r+-=TNF|toItgj(X4>-mInV2mY@{c5
zm&Z_WPM?Ow8<;3A(y=Ex?yH)-h9+=GJ`mq^thOuzLv|&tyt%Ww$#XF7fmXBIAj!Em
zw=WI>6qCLr?7nTojW^=K6XmO-!ui2Anr|YCI+}_k@jz3!ncv5IEcrxO2^J)gB|;92
zuTQ)+*H?moyfB77-6&1I&vk{9rT;Gr0G&y23+vpOldJy;jrYh!CJk`2H>4tJ`?{_$
z?fA0l_jK9i7(cQ2ZlU-`mws{8;L}h3G!uQnB+_+Cv8RcLo?yX8{;8x6v#W>|mBI#d
zhUG%wGPGzyO(dD|>#wg|78fd~+gAt)2Ln^Sw{!`ifL{TAbbaD-x?CtQfH6T>t+>bl
zrjmX#VEQGnAeiG4(}lyB}D^1)YMn^UWp?Ep~;?&QF0
zxYuN5HHUj`WmUk}8p5Qs+ZXw&!yD_N_#>*Jz!K`2LVv>B=w*||^5Mpb+i8J{%&h)h
z`}6h!CW|72^s+#@)yCr>?>GAVpkc$1EeV7|Opb6e!GmHIyUUz*+bqA|+`XBbO
z@xVGmej<%q7`5Gmz@}21R}3SHE{E)Q*4!hPrILoC@Zr3gI27CEZJ6yd7%$)2AF7Uc
zbRP^&30bai*Ho=&nXatN8*wynG>gTj+g!s^TU5g-1dqOP-W9CCwS9+!y0`AfrtSuC
zpc_d#dXBvGN%8oNA|IV@rKmf-fq`D{JUyevTxN*mm3|6ak%ZxLgh8`^NN_L^(?t`|
zjEw}>)2SL8H|%l)E-V4S=`OLUBGMK|Br^$5ee4oH1UHM!22y;Hjl?ZFf4$S=50r1z
zKJ)9x`g}sFslQAUDJ8}qyZ<8fW;&S6p!tmo!@K5xmoBnbP?7*A!HlX;;VVW3gH97l
zE&cIBOCl}z@(nv6CAWPamOvSlse0dovzlbkdiASARoF*{X@E$nA0#v!sbmz=R_1uP
zy9^tYeV&mBSx67MO72%ytldrb)7#r#!g|hOvDYx7#rD}K?oY)y-QpFi;`e1-0^u+<
zGsjFljkeI7ug>9=;CnLUyYxgVl3WjUJvhHXk`U?IpYaYb4QwRX;-4dhS>E
zB>U&Z0N?e8816Kqf(0QH2P5QEnQu8!KTShv2*Brff=2{G3*{?C;LD1%B~H^i1N{lX
zimkS-S&TktzJZq(Mag2t^F
zKis$S%@zJmYu68ci9jx+QPb0}`Yp!y_me)E#e~rKrq5DPdq}h6nT_>bRB+k_jnkgw
zMfPqp!dH%)d>g?wii)k(hQwO1Az9h)3KC@%P-sK=5h6
z4lN4lUJ)a62fsVrqOeTO72!|v5>#;J|9QZXrb``!R4rE-r961Xi1zt?d*cxG#gFct
zgq9!eAI7^)Oq|Ty_fzurzmNwK;#7B%pk_?&RfeC_p5n>~ojzNw0M~FWhgZ1wIYuIF
z0z_o>n7iQpEF#5{i;f5zETUbS8}Ih-1XJ;}IPt6j7RUx$W_BzQS+9hhH;`}^nYJO7
zG=B>k3wC$*T}v(=`iBvoZnj@|F&t$ABlojfOsjjfJOt7F%*tp+lJ|Bkxs^f-J5+&7
ztMvC(CvLe4TX57$%vjugHWoKLoepn=ksMcHI!<8)zpEz!chFP`R|>8<9S&P?LoHVs
z{A6M-Hj&C>N$K>yLC+s6s7y?>%wcvp^m5p1T~Mjq3pXG4Qcx+$S%(~U+Lb{UU6Aop3n)QS0K?oQq
zG(0S>1mdrzPX3bfKEuIx%)~3ha@bG9>q5%(Oc4ijOnmF=e|nuKz^V($nqmcYMHx9F
zR%K(+jDc0s-1xl0qc`jJ_EkK^$Cp~sCK~PLq&PzICD}RM^|dE)I_$4PU`WWzdEfEo
zS`tDq5IP+(@t+68lyo_4eR0qxB&|}4&iSEpc9nH&NM|+ir)f=ej&l8f!CJ&wWW6Xw
z@2GHu;0<^Ozu|9l9oLfX7N!Vp$(+XBf0`hvzV7>VCI14ZJX?&tTn_s2E7%;;;Gb(8
z(;X?FM>j8^adwi0yh~`hyWgHdK-;uOn(dpXVvzy}Z~7n=AD5=H;SVN?d#5K*W(B1bH!ir5_?7#r*AXtlI4=Pu#X^5;hW;yZdfTwlxsE4~;X?
zzNU%bTo>_hSzY@yqp3nZk8P*)#pQg(L3%-~RWLh~TT0)fxY{3X1Msa<#_`
z9X7=?+u}#-v;wNW(qliO#0sS9yxm0?2QseTWz-u9n`x2~*#4ShbJ!sPl74-~AtPb7
ze9fC=*1j!sXB9Uu&)B$Qy)M<2zfpDW5v;tnSRX<{hM8@RZ6d>bYJ^(AA^f)U5|wt$;Y`n{-5q)`
z|5bh<4fPMh`QkPz0f}nf3il!haBtZ-vcG$QX1DS_)Ein-LT&mB$jP{wiuJl%gNZJb7uFD|iPD-dQWV=9s#REDnxcLd9Q
zK4j#0T{A(OaZmU3QwS*Xx}5{P3nk_x+g&_8?r$_jL_%Tp8q5W4Y_8g^F{(}|hT>b1
zqkHi(4EXe9A)mdtZJ2Q%zz(JzJ;%RZ?h+gi>xAaH3D1M8ITK_z%Hw~b0$@Xgp8d`u
zf7?r$L4o}wE`+TWjkiDc`*{ID(1T()bpX+X3u
zl_{jK)2QQ12sQfp*?HxJPY&g7xZ8Q9Ig4sN^%+jMEUv&`P;U0CIC<+5Tk32D2xFV4
z5zq*R!FOXMDjc&?Qu{VeraXnZ?D@&8ao_}W|JirYV6+zH1~eLy)I%(dC-kb}7mGj~
zV{ViCq2lA~sNZ>ducJ-DL!x@Dms>9#x77!;f?Lz~wUEFNP#FQiN9u6g!0J5;q^rI7
zOH(MBfDu!XI(SMD#?t9zba3o4Twb*wzJz1rTjg&GL7Dgaa+1q!2*FD}M?5k3W+oHk
zxM>0~97F}W9+|{U|J$Kbs@^QG+m2dgm8N!a3pF6;Jkf*`eGvxAH9ZLos3y4tp?_g5
zvl}rxnmQ|6*oA|azO{O^+FumcH!or7Pxz%aCB4#_Y>wN&{2u3k$ylN;*xmAZ_Foh(
z-YZ`sFb3(@?DpTK*>8j%t+2c|ha3@L5FL9G@ww7ApRb0oms$U<)RnDVt;H;Wz87kq
z#9_AiMBAT8A-Krk5Gp|K`c3Yqo8>}9vD{}Qk;sMm>S7zCSE1VbAs)x#KB+8&Bk)aw
z?T_0g!TL%DmsLQ0?UAbn+HhQ)VpFE!Sr5$b4#zg3g>znSaNx0r(cSDyW35^3=|-&$
zCu2gBfOJJ@5A--F^@|=4@DYK{k{{%TRa>yNg}VTk%)|#QJ(At|%gHO+pZCSV{w^bM
z3v^m``23+r6&nQ|l9f#HcGYxCES)i!{&t1A-z*=R`cKFA&>1w|{t@`3bl&4A&#iKy
zN`hD3vAnl)$61~;Q}`PsbI`43dfpy26_T{tJf8mjqaYi%l%N<
zTzl{u@8NQf%30ePaw*)FT1#_|_ghml*_=WS9qF9)ggR8fb&$XPaTu^IJ_i;&0S&Af
z)^PJDDg^k)S6hDaBSnNNmFPS6as6w)qW@dhXD(~jjW0Av9J+hs-?2{ba7C-=Jx`^z
z*=q)J#X6?lXN4QkesTik(@r2GShYc!&4rz@_o3VO3l0jur!$yq7WrGM6^bsd8O+?r
zVE8M-ZJ{i7{9!UZ<7Rxy--{AE=5jGgnaXU@B90v&jLkBO@i
z;@_BBJ6Kz+lWtZFXbD{R9dVg)e+nC!w3s=q6DfaX3Yt=%-Io*+n{glb5)f{>sXX^T
zXdQF=f^}Z5tBpUt#l8Q%eJYFO{GxC;BI5Ca`()BB>m1fAzLaH{=DPYA@hzY3Z$o;*
zEa5GZEvjcJ>8Y!&_p^=zjPBl>xcO5le5(_oLh$aoRu`MniV3hegZlfF>^LQffLl@*
zaqNSf)g7);(M%$m7(ALPP)qt#Pu?9KVadilBc|pJZPO2>9T7j^1R*s~JA$-U^5=M`
zEkv!;mjR#0qYTpF{tAMKq?2?AlXiQgZIF2XT`I~ddsUbh3Sqv*QTO=k$3n#tm-{hl)##M3bu*Vg@yNFBoPgJ!ue{wUJfe)XT3~`|>z`u$m#`CSp6hjD+zgq-*qibMr=HeTmO6v%=|kwwR}6s7OLQ6B|2y$Wd(Id>wJIqfQ9t6dh_^*PyH9(
z6G^`(&^rxyMlP4yCFevL<=l5eAxZE3c#P3guDL%$?z3A|E}ie{)gB;PDd+c)X{F_uKE9A%HyqFol6Q@!Q_UKU;{dDNLA0)Wnu-CY8xRZlup2zx!u0U6>F+D@qC_hhO
zD)_=Im;0U^b{#L3BfoyBii4e6i@&!OUjLxdz@IZ5InT2c7wxvj`CYX4Tf55xhGlGg
zA&WU=s@FnVhSIE2K8UTN1X{{Sz&q7A0v&Sj&QO>yf1YJ11US|Qc%L!e3EfCZ-j`S9bGRjlzbq6~j4Z?r*o{9*T
zhwQ(*>Vv2+|FU`z#Qxhv0PSK|^G)VTzXoIT@b6^e435_#z3NhyRZdpv?;gx*?I&{D
z%JjoX{oAaeV@|`sv3r^6SgsyBrCj#rbTi4E`>4bzN)OgSGM*}N`FHSxgQUVnC4WI*
z(0c-<8cY4+SMYM3guzPrjhe^T{{+KO^{Y><2eOOrgpI`C@t74mZL!ANl=~hi$PaOBRY)PL^hRLp?6qh!YjL3~oLZkN
z$nk$2Qk*@mDm`w$^7{=&6E0Z)`P0F)7w2LoQ(Pkd=bQLvDIxXLS)RcTy+tRJK&2vi
zcuXrx!3^8Wnvlm!@Euhoy3&KBwDWW4x9em8rIe{0Qu8{n`7ZI*o=~l1A58m8XTnUW
z18rH)pK{==RV4k(s|sY`gMsi-E2e(>Z#J9KzX6jNv}pf^feIo9T-!!G4fyAV+VXjZ
z$nR&>r#sF1n?{w$zxq8^L(rEOSoQc-QtRhMNMU~GrFN}*LgR2XiegE{wmN6Q^@^z*
z+%)1G26$a{xz#C%en)_Hz56(t^i10@IU$~4J@R05yG&+dyE?es;e$cC{5n
zOEgAZLi2~sWR~^TD%Igt;WzB7tZuNdSxvr8P7?*ShysGVbpk4P_4BcQ`O}L%V_vYX68=`X~Gnd)NS8qRL
z@%n9Y$uc-g$;ufuo4|a;q6hTwTJqG=T&0|Xc%)H2TJ6H%i7uWxTeYy=P9SbOv99(-
z^DQV=wD|wIH<(V9ERAz|WzdNeyrvbNbgiaXt?LbVE$Nk?kD)Ev$0W_cxh$yPXbz)z
zVut;e&ZLv~F=;w>dJ>4$c$Z;1(3q{}@VTRuKkx9OB1}MfD&7E2CVqs7
z5NSNvcV=VxP9~(UiWXW={5k8H9KKh?!;C%Fnd?_cb9!*j+tVYmJj0ijXOJ_uy=v1g
z5t_`=057C-2o?g@_sg>#!c9M&X-t-gN9un1w`+5sq=fM6VJu;lv!o}?!>0!Eb*ktqyYNtkd
zv(A*e0Bez+!B8)259fe6;>RLQ=*RP=5bDye?SH=9ri}Ho0j%3{s6PMCGa_u9&`f_d
z%e;2Gs}f`=qf$ASrq6~DKA+s-c#qFjK>HQ9v}(eU?~g`vZ8hnUzd+x=?Y3XC==p$*
z@D3D0%LBaGko&)JZ?+OBJ~rN@o27tp*|GGh@S;$)&cqUrA9>jok>vMpn^Qlqo+V1v
z(Ke8x^p@$A9UD+Bek_O0A**ka;5H$=t;VZ4I)T;SI9bc
zGIcJc&R~W4qS#`#7I=c)47uCcOoELzKMnB3K=ChPS9o%d`@qs-dWI+K8BO*x9|?3A
zYEEoXCM8CdAcB<$@_lWUcT7eDD-i9C=fCM@KZX+YS_VG&(133{ex*T>Dtcp2$h{iP
zZl-6ouO~#Q$emxJUwDjqOs0V_1UQ3KLJ_BV?~b8>_t5JMQhi<;wgRytDKuEr5j&^vB<$9rM3;
zobIi>1_Sp7%o8|WLV2n}5wb9Hc@|Uff3=ghMZp0R>ofNPs)ImK=(t(q9Wj6S{a?=W
z4ʞU5}DE}zpw4!+Hrh{dQU?;v!@bu=^C1IxvEY}*yx8b?5D)&T%)M^(Sz$dqV4#NzGM(8~Smd`>?eWmd
zrhRTe=huLmNTM3FJbQKmuS62KsvNf$XU)=8eQQk#i;iu
znYl>#*$
z7Mm4(HrH;{O*~fipMn_u4cm(9*A=CLOwV>M2xD2??6_3+^)
zb4iVyn{1T^BYpCqO9WUk=*uIZ7`d)v8mR>CA+I~5LyhhE>R!?)IpQy-QZDHD}FPJ7{Nc_)tflkUKftU-slKrj~qrN)1Lp?s|OWcq|f~fC2^gYHmVWasH?lT{2
zVZ{s=DoKQVHKNs83)7JUDVGdujAMYJZhYwO`y;f6&O}w9?_}Cz{iD@1;qlp=L&wLz
zp3l#Cya`UxcZazkerWXKK(pgD2HVw2+$p2BX#H=K=tN(`D8lUT=dgXAG47ZjYl?-I
z2z#E`7O!>^H){PLcw(pl(RJNG#96Sv>`)L39;}5_KDWso`Z|DTD>%kiNb
z7KtXe0bR*mUd8U}Ycb&>uSXWEw^dr2rD98X$}ex{IOR}5M8rzaMvY$$nzvipn~VaG
zr^>DWL`;eSodrR_aIc4YxLpK|UhwM2-)gkwN-epdPQ(Fqmzz#=Eb1^ewyv8(h}~xz
z^C}2kRx1D>fYWLXWJ<@&B8zw-WP~R}2w;NlQ9irFVWpcrOMwq(J#I
zhkWcM9epKT2HNes88NThJ16R>v>_Gj)?lNK=(5X^8*N)G4?1`s$7+sVs$n9W&Tfh&
zfk&U5Mc7M^=DBJ&Xf-n-pex{bxF4q6XqCRV)M>L@Rda1UCT$oGJ>HJO8s{q7pO`J|
z_|wi1?r$u2XfR5p7?4*m`SXy*6f;`iLTNq2_{W;ikXYDIEF^+w%Vt>bU8g^mWijWdlc^XT#_Tp!*3^zsU^KcPRtN^F%
ztC(rg-LcQ?w9}?i;|b4am+1y%un7Tkdgxfa?W?n0G5tE=P)dV3n0)!UQBhwrLYuu^2pMv9Ery
zDA67ozqV>5z7v|}wBWWMb4;lO$sJm7zV7id72+ImYw_$t*<^c^!xCs?2e^lNE_P$u
ze2OkiN-iOV7N*C!jFw6^flsG^G7r9uKE7Eo_`TNs3v&5F&5u~0|AJ&FW`wUwiW}J?
z6fH9@xG!o;CnzA<&x^%gZu8DzMa@*ni=C`e0-}4BuXRW(P#=`};WM1OWf#fD4=hP}
zOU0FsRDcdhH21SRqvja~bVYW;$L_&Cqics%h?&!`;Kom~xk}mNmmGc1gWAe-5B+eP`tm7Z;0o*&(23pP?z!;uCnfs!i6b{|Al2SyySiUsV0jX5XumVl~
z4)Tef{&vP!#4(xld|xGFdU%gqx@3mqdw*(2kSLUlYr9G~+kicv0li*~wIQ_u{>r-7<4tGN||I`DEi^Yir_9-%_r1D0q#>Z?(I?TRZwnqWUV*%dK#|ZZbJEWK}`a
zmc$SHE-NHey*pQC{;TOiMM~NjC|c;Y`$hTg~rs
zG=#WHekY$QtvK=4dVh-OsfshokpIx%UOL4P+P$X~
z*3uv*7{5pE5{WauG&)=*L6q8v*TYtaNP8x91@+H*_d{eaV7Dz5YI1BydR~qutaf4t
z*Gx^E71brJJYAwXjwv=IDcPEie6W&)x25JW&w7mVzSM@xLl(ydxg7z!e_;sC|DAS@
zEv{6wTn`EErq|MyAs0Q{j0u+WG0V%TBF>)b2XF_H*@VpmA$$2lif_H
z>qnHA@BCvllMlfjarzoRIsU5LiS^t1-@K~ee4d$s;dot2&}qQ>>|a_L(#5EPL@%29
ze;tbYjs^FIQ;>0dSl&A1uaB_Qf5#SPdp+*yl-TLBve%cEXN6d$LUe6zlaQ4{1V~ic
zx&i<~mOg#}pj^`5%-dAY-3r5>so#A0i~eAnS;)E6?vj(AIU)mUJ8gn8>_H;1TU>$I
zU*i!?KYpXuz)OH}CnKM|N%ls-Xb&q0qVo1h>FHHr`}Ya7*hyn>83#pvnimW~Bfq0T
z{oD|(@be(J7&pL~5g9$eN&E58!e>!e)MD31#K0V!?;Enfh~Ak{GA8QdBINU8qwp#g
zUf0iLv9_E7V1TBna9z}~Ncxqj|EpYV2pUjKzA9@*@V}w|MTDam+L0t3qaW&gUr<$3
z#vobU9$z2d+8g|JRs$69g1zJQv>B;JLKV>+ia+yoTebD0=2IC4k$ZwzUpzUjf$VKP
zFwV6xCfMq}AY#e$3{{2>RPfW?n2-5u+&${aYO$gxd7{)D?Zv0vHwJ{cKhc+L%*Mc{B
z5X4cXvx_mW+-s_laM^V`oXKV$taAjJQNvjMEp5dErZaIO0_{Hle-EoyC-O6D<{Qf5
z?61@aoSR_SeL+*{q?nh@YIs2M{Bo~r;@&sDjD~NJK0ii&_kbuC;~&$yp*=Guc#rYr}(~7oZqFJOxpleuz+YVcU(2;w3Hw=bH@1BT;H!YrKUAHw^+DKyhoNYIw{b(1uT1Cdjfp6SbUg|O{0AW}pa(>X>wTPHX9t2m1oK}JyZicf0)+yhr0k=S
zM!mw=3Uglap4W2^0rv(Z6{5P#ksP{to4vElb8Oae@q6gzyMIJmWC=tlguHZ4%xn23
z!=HxzaUZuauP$tQ|5T_UY=2Z
zM2d`-;dt6az6oN?ud`n%R%{nmQk)6t1#R)1i$$d%K}c*zV^d^I59)68>h55szO>1J15YIb5J6VaQGB!C
zZci@xB?4c5=vOq;(;f30EZz{^`BZWstIAOR^}D`y%4zwAkM%U_R}(~>!lkcwp+)4h
z5}LY-MRzLyqm;M6S#fDk6R|DUVZt|+fqQnC3*`;w4IPh$PFGqZ&Xgw_Nt}v!!YZSu
zq{1P8N<#VNh48xWMKZn!3DqR#>>UNg!)@J5uOG7lXHS|h%=ZgZh5WbKE7@=Q+B=Ac
zuur(e7SgNNe*{FR<=YR|>($>oUx{IyUQtT;zkDY}*WH}2P-=8qK-npk_;kbB83x+1
z;o5x6M0aUq`)*ehkuO6E;rT{X`6XiSM3DY;89jwd6D?U0DNkC}#~^_eM8!
zy4`X;TLx}-yTF^cuPZHtCqz>{f|w`K)n9TcdUkxEc@p^bW56G~+gz(96mWjkNTzpD
zV4(*Z0<5%Ry2~o_$S%@HICQMFzB*AgQ@b=uLZd34dLO$h)z`x#wp#;VxZH|1^Kl=O
zHS8!F`esTIzjpJ1@SB#KM~07|H`j!7*?Py;5N_@_WHuX=GH{`<3o%{!>2CHX;nXJt
z^P-tnis$B&qu{>rT=C5y2X!RGoAN_%s?R3q*r1_22iQ#`_{lLLbEV{;x2k#mzEc%}
zz}o8!Y>8u^KR)@`GW6W}(x{C!r+5}?7bB07UK=V&&OiC@JV}NR^{oH5cQE3mt|cM7
zLf_q<{|5+YQGRqkeK&TfNgDPc!l1X{@4oaj_94Yh0j?T$n|L4+LyXai{Q9>A^a7&;
zZF4XQ$~QWoKgMCVNLr2<9VZ$J^a6R=8L;oNf^uXQCQML6A5aZYGU^|KOM@i&(8fz4
z)!DR%8Z%f7BTJH=CD;o;3%=+sJxZFNP%iC{#h+V#>Arg;B%2hE(RG%LE)RvahJAY7
zB2iX&`51O>iZ!$OB5Qp2{G&N*$n$XFK==ku79BI1!nnoMj0^lAH{2hCXG7H~CD>OC
zKCW2uzN#~r9I!rr>oB@5KNY{lto96!-+6W?4zSUP0+*
z35?c}R8#P_OyQmg68$DJW%58P8=%Qor4EY@1%+B!j@Y20p!VO>+gyRiI7$s$`3fV`
zw!!KRJT5DET8#bOC4RK(o{ww%>QDmG>!7?pT4HMOqJoNZC(U#)&Pq$@>eWvX#0W?~
zOZ;IS(6?BE)`yAbyUF7tM;0R;CbC3QzBvMGx}0{gU$g=6h?O_yj}wH40=@#A8-d?@
zHv;J6p3w5IbE7gntay9A
ze&|Yunu4bQr~v^1(c&ZzK10A6m29pun%SW7WX+5DIW9*k9QvNwsF0+1$|yvZX)5xW
zY05|rYQmIQcM(Lg2KJNd3XDTK3+zdr=6>t_0k#2-sW_sO(+{SXJ43H6h=-`$G@bMp
zi-oy@U-&l@?%B)u3fs3+n1>rh4TUlRUCR-GG!a3G`8fZ!gyT`FS_zU*
zFp}k{*ia8n_mf5i{2`wadq}YT;UPI6TW}#ROpPEG(|6=YYwFm>Arg*RqTMarQqLlc
z<7y}oO*DdE)H#v(e}!pITFUpZ1gl!3;y)HYEwdR3$$O
z+!>QRQASAyz5#A;f(eF|Tn4UJBSG1js_0MyV9!d}B`@X6&vUWseD3p_BLjn!2AgM<
z>ucQ~p~cmQyQoa#6pvd$_b9$TSmSEQ`2npV9|_I)06v!mm$oKr1t}vnngeL5CAkBf
z+tto;bNpgh#sr(5{si@7Q4_u$H$GL*ygU
z=#)+!*ZqLlr_*9P3f5~eh#E>&f3%#eqKElrGiWHbNq`eOHOfa4Dtb3UEsxmV7d{5
zWc$6(9X`&LbP12v{A}fSbxbdFR@pxy#Qeq@_)51C6Vd`|A(#5s1XE-u8Sd)eRaCaZ
zf=#TUzcYWU5*NdC5DQLen7D77C9-zOCj*9}pKj;|pVx<7PWx2YNmq3%syh-ZOeVt+4OVDWtvYgB@~CCp3i!aZOJ2>xp^tC+B=)W0FmI
z_<-#!AO%y9LwODJf*7<`3HQQL@XWqcLx`HgQ*7}5Q(z0#CB=@1(5;0w9s2*6lpM_<
zZXyzdAVQ1ve*l;V*HpF@(+94}z#n%03&-c#bv$@`(f+J!z5PENv|`h`xu^rOg;{36
z+i1)D1%l#VW>1pOR*C6i5W+U9sfPEHXx%`nbAKZmFHf_H1HMI@A|by!hv?Npgl#RZ>%RU{=Mu){n
zORPQTjeskPoCn=UABd$XN*6TBZt~}Kp$z9AebUan6~yPrczQ&hc{CUlw)TMkI9HSl
zzd%%uC_m}1dO`*jz8F1Nhoj7%8<-2;*rFsQ11lRudbG!z%RMY0QIn(}(8$x_5W~eC
zYeiJW5gs?4MT=;&6&c5|lFZaFUF%nrUh7O{hG+*l_iytD?SN!Wa)kVFL4+N1tjpi0
zhWv?)VNBHzeJ5&O-vjhzQF=fiL|NeWbPzfHv
z(LyfKsgBCbH^kj(Zcf`fIu;8W9hJ75qGcEPh8%EAsJlZ{f~U`n`@gco5-)YClSXjy
z@&7ghDKYH5*v6}ces2K~^sMa8`}IIPhx(jD3{;^k$;%hwr;xb~t
z$Ikl{!~J>V0UPHRuM2H0^0l-u77qe5Y;XB4rx}F7^>y1MntPs;Y)B7qZ*3<@0H^bL^(blzp|Tc7TK
zY7wdv?o18}i!N%8?kXpq!5dAC8)Z;hw5(zx3
zF>yVF_|(T-RSV(Mh;0j(a*}U~vbVE`>>iWdrYAi>e|7RaN$HoM&)lA#7Znn0I>`H0
zcXFD;yXdPMIoS4nyC&uP@jn*
z$oU6PG*Dbz^t*=(9?ypFw_9$QaZ*Jv@TA?Am!M)|cucw~@@0Kfh2sGwoJA{AAQy|R=S}9yW*|PT3h;4rz)Jq3}la|@QMnn-l268-k_>y
z1KurrwHQ0L5?kw;YS>h-LTGXc^VUA|7##DYd9JR82`bGl;9DFFvilu6M&D8aWWW^+
zf={W~z@xYnu))F?3nGngXorR?UeTH7`C(J5E%*zUR?;b(;FG){$gekg<72wh<#~!PjRY+4hanI
zW!9fF(z?U;XP*)P#6et}(yj1JRI@gYE3q7D^7yCxuh7@rB3TQcto-P0n4-Qd)!v*Ii<8QL=C|jt0gspcZ
zoiz00Bt#6RZKM3Z6WriNX(_VXStPIJe^@%P7w0irOm7%xSj^eONiO^_Ki&E2br$+r
z2vD}-o)_A;vOlKK>kx`CsKSC~J>F}POsD*t*LT?H8&lqMH<$Pj=Jq1B-lx^j;poG_
zjOd8a$$IJa+@bd;)3T23&j=gdyP>K2IxFfpDEGS;y0fI2wpeDj^d*e}$~Dwn1_UHp
ziiX|q&&5RA5#K)O8UOgB0PPO=0`+I^6N9A5ZMbK6zJIAd!%~mp_
z+%&Uz-5>8QJ^L!}|IE0gYSFbxtlIJN7yfgvdY&c`Y+$(J#4~#9q}%oh*&goWyV%$R
z>ae6-AUig^tm!8Lq97;9$XVTso(3gaTy8-YdNYY<1?|osGl6q}3OmY7rWB-XjS?x*4;7)C
z<{$?yk4&N4uw##0=1Bth(`q(iH4nIozKxpGnSRySAN9Y9JDK~=OL`WRc^l!qPrg`P
z{d;U^3~!z-wZB|t^>bk)d32KQ>#i{R)x*8`KMmJ7@xLBw$Pp8EibbN8G=Kfm@@IOP
z_E+;yRqOt&`Oh=}%%Rs;rg|&w0LBTuT8c0`*YCru@^q%mD3=dUhbqKJxngWWB#tz$
zD-E*f8l71s-WRzWVKOV3Fp5XqetS;iz>v^p+b*P2{4Wz93WKqk$-Sy
zP44p*_q$qW=xA&SxerO-E8
zfeH6FA0W^&Zp`rS6ee6sW!u+=l@jY^#hKnzX72susI74X#$Q0Bam3Sgdrd3jXBK;h
z=vPZAK)_>x9)hjOpYO(vOKe8CNwHl7{`Y>ix;=Q;-+)<1&`Xa|e>sPpcjXRz{TCe3
z+hA?o8TYDkXAlQ>o~`Gaq~uP9OV6}Sg9Wm7o_4~C3f+G0{VW;SrZ;K!Ag+89Gu(vW-KdFV9FOX>H
zu4bznG2`O}ThjI9JT8ar66RW(ob3<8`zKX6v{TU#rm>*g*1%-i2ywRr!zk+f?0O;U)B(olR%MCC4e_I>h};v{`zZE?
z%kv)bO26(ZqfOPT9MgQR@D7|~PNNRc$-_g|ymT@Bf+fQS8a1k#ak}D}5hh<;{x0DF
z#EdPLYS_?OfaaAwv0wWA8DbNzi~{C)A+j@0Jym9Q%gnxyKh!n3lc>;lRfs@4Y`bJ`
z<5EAn-&%*mU-2$u@mf*><~>xXtr1&Tg!BMRX>4MN_+seiH#0D|iSPS@TW
z%k32BcufgT)u;9=WRt!hvzlQg^QAf{Vq<(RJ-{*VIDzA^TP_U(VR5Yj*rkQ^O;ch5
zAS2UE-ipof#nTA*z3XbE;$<@~2$#6PWS#%fZEX;BzQ!=^A*d6aelDW@?C$D0us|xi
zd7CN2s4en=_P25nJ_%lU{oofw=Z<0n&T(Z}yR+*uv3(gp1idI}*W_qEfMnF92tQn-
zXZ@Vxq<5H0iR4=N(8{rIZT*4S^y^~v6NcoiF&8%aEd9?A_VOWM%f7$XVX
z9Nq$_*>dC0BfN9x!bq=jC3J(AIFHqzY_wRkNhl5Z`6GK#pR(Xpf^2hX;^4>P#BJ-g@s$&bc1_8T^Q!wa@
zh0e2b)y;4Mwd(;$)^h46X|*br>lA&{P`=v5Glys$U>i_r(4b@D8V<~s2oQO}s_rYb
z{{0?cRrHX-6T2!wG{J-}
zC@<%2T>^aHIL_7I2>g;jv+dr+Gp_41AztftW<8h86>nxHFmmS)`N@G)Q5OIFrU;KHdHE6s)0KW)*h9)Vel*?EbAx+#ew^
zGug~HFFcV*G}S#>GMP{rl^t1S@X3>Ky`$D8P2dm!1{uGY`ckdv
z=WP8*-}Zp#F1Eq(<~$ng{!UD3Zs-_%@-HZJVTw_6Ybw*9QAkz}f*V39AE%kxo=)`v6g&l#0%%M@Uwj`ky}Y6V
z6jk1q`;+s>lq<<~;P~%^;|=l*q13`ksKv9}iJ>ib{%g=eV0X*zedo8`0h7Fv?#T47s))m-`P@wN
z0y^Yh8UbMx!T(gb7#Q+R_gPL&mj8Pb4X%>iKcnDBde*xS|Lt8%I-KWY%DgV$1oNk4
zB3paq2Eg;3C5de5zxq+8|dc{p=&OUP5m@DpZfl=jvY+xYJ~7vPb=;Gzvb0F
zQ6+lGlXIUHIfkv+QPzA2PoY6?M{Ynapp!Y5BT
z_>_$ygOe=yx#H9>9lo*%XM3*>KDsFz<8~hOmfcY@H1A&G@>f~DrVFZ{rHt(u*daIQ
zUMIUrxJwc^Zv_34y3PC?GUUFsrCAVe7yM=W@>uPo3Cr?+^agK7tS^(64x2d}0Fr{B
zd;If%9)Y(KUEtiE(;-K5hHENY!S~00Bf{e#wD$1__1D<)0;5DET}h{wc)<(F-Xz^!r#i}^o?)?w`nUYggn6bBk#
z$72h%s)Tni&fgQ!0)as~Zq87N8a0wxcMC@V{02k>pYT(g#ZcncD2T4=%=1HJP`4ex
z)g|AwiIRK~2Kj!zgKD8FwXJ)LS9kL$28Eq;RyUW}WczJ1A4*!e#@e?N$M)8J66E$c
zk{mdLlEe((O1C4;o6mi!z`$a5R0gUUY-Jcgiid%rgyaGiQ$Mv)+RD+aGYysoN){s?
zAIJk%DdRm6r^`It=W5jDPoC3X%T1dq50;K@2!43soacf{2+6;q`%=8akYBP?m)HO
z4stT?S^e>5=`7}qypAP^s5;_SXAMi9e)w@o1}_pUjzyJ*F#h3)SZKiR1N9M{u@b@|
zPp`Z6>Rylt!#K-g`*m79zgL6=zyj$u%+I~FZwYq}&rBPkH#o1y+LTSv!{0EbIc#23
z*8<>-DNr+JjoFnBwR7Tm>CV_2{L>z&uTcLR%Ear<8WYOw~{PD5zrL`h)mK)~tCx
zvQL>j^uENktCtii79zPp^0?rbr7lO#of=oH7_giKHgA%e-3+VV%eI(e1z7k7V!+O(
zP=vmSDM@63p~kduclEkrF>aGb$p@TU6N79S-R2x;tlq_U?5Cn`aJgVC`M9yYI@@O<
zm(BH!;z{XOe}Th|PF_K^#2;dqvS4G$o&Jb~V{bw>8@1keNW@H`0|4Pj>FDD%l=Z8e
z_Tzw&Vp_2{T);8maDqp4NGC}D&+8Xmgmdvdd_0Q+-;E}zFa87y?e|bxLLF?
zxnbc2$A165I}Jqf(x3WgR(ae;@ozuHBdxfwzZ+i^Uo^(|X$~>@r-LwyUojs~iNvWx
zLA3VSY6f>t=1+D(X1fdQhtG?Il@b+&^_wpa(zv#rk8mI3Btx8u@`m%W3$o@lIRfH3
zt^wVz->oRcf7k?&*T5a>&*P@U_qp!4{%G6K=F^#)#q0n-44Qf`Wmx{-QOtLS+roht$OuKhvgz9dYKlBu^1u8;&_8-o$(z-G
ze-J$t@zxSQEnTcpuhK_mH*Ei~TeaE&A18V%C7n>-cM`A{&T-t96@|#QSg`l7qHwYx
zdE5UN=-a}%$g+`7h?ZKxfwA!_-J8X#!Qc7M_PtV
z!R`m&OEpFB)q2k(#TmFVJ^4Mg2llIj8Z6fct;*y!(P^hWr=)8aTZ=C}9vI3g`32!k
z@`2IG%>1X+b6K_c^n!VdVAZbE5u?zDh8GU!77iX44Zxv
z8@PE%6k~j+$+ss{z#;=t_6&w~z@F
zeBktWGaY2pj-V^g3rmd-R1nczgc81*`t72-YJ+BkYqgHjAoty7U#!^mFkWZ8pM<8z
znW9;ZyvQfU`Ni>ErWynW;OWAkooMYn_79iBE!UB3*>wTG{hGwkK)E&aIE{nB^v9b>
zu9uOUAT9dl$8s|5);e|MHz*`EBXjRfLV_W%!dCyV%aGFUzSO>s%HVENXv!!5=UY^4$V(E#r#?Xswd`hsgco?8r^Q}-@$kgW
z3zes};Xo#>M-pQ3g7>Tbt?rv#@7vhod26B~0{3{Rh4iwby{@a#6Sm(V)B_{%VJPq3
zv>#HwQ~t;sjU|CB7Bk=hJ;>5;QCJKHYH#>q%g{~R9#bK7|GF4lkw458V{2m$cGr5$P6^>JE*xr*=qJVGpR)g%^048ckO#nm=fy4Wb&KLk-z?EY7;x>}
zBgSB7Gq=QBw%1Bna{ElN_k(0(9(sm*-d^Pe`Y9do`Rf2JfFv{E5x1+bd1PJAwf`ma
zWhejqlvmGH1jZv;&#uf@IR?gxQJ$Q+3GHZmq~{~EBfeA4BFAEZ{Qi2o=)QufBGwkO+`G5Pih4#d
zf|0-RTiq2jeL2W(kA*ml5`A-fNPldPSXhz<8p8W>-{D$_x~TYH0a_xr%7FbC#oGmYCtj4bLes_w)`4eZ;>~y<9Yy$%E^4+vwTpb
zO_xu<$kOcv)v1UJg1Bh&C+@^bzrUwpkKrpdhK~`pi~rMtm2PLsJa{>b9ABsf0^4-*
z+_uiQV=E$>+0t3uzG?_Q^G3j*z6zZRB8UZsNHWEJObjf(CKa^cym!(=n$4L(zelyt
z8B2Zi)0B9q3$5%s2&&Un8fM@a7dAT^1rI=8N$x(bZ1vpH_%cBerVdEd;&qBiKy&z1
z0P$tu8g{ShbtSaGI!SKQcmMokL@s8~FM=O|e1vq?0N=}ed8WP8us%PndYG$llgGb7
zCvSi7N?ql%P|#mbz&Xuv&ZjQ3CEmv1;xWtL%d0^v@fAyA`B(u5HcN$Xm|lu=S5QZ!~d$=|qfh2{Q$>9|e}W*(_QNTE_}A9BBFh}-59mbr*=2z+achB!rQ{0e#Lt$Wb7gG1HpMz@lY&Ff
z-(xZAcNH38T#>^X=S`zJRSL-RUm}$pef?obqm`)gXqe#a)7FUqKN?2hRB)
zj|xKqZI&K5S0;ObS$pVLuhY4rmJG@zS&D<=NJS
zl$;W;(#H-A>NRe5dWMGyTuI}6f$?AD)jsAL&R2xsVm-|?Dj*iPjY|5-C$7zlVihS?
zr0~A`HRnwz{#VvSx{n)ZcV|3Z?PI}i+=&~ZmM1@tX1o+|NuuD7MYmch?ew;#t)Btj
zQdB4D)gXdRhU=tZSt!g8#SWRnh656A`@&}=%dr$xifJ9)
zv*QS$<*vT%mP5F?R@G=!9t3vpuBgne>e@F#{E+Gcb}kS=tgym{q$H-RH=<`8M)~6X
zR%?}6IHp96K7Slb^&RrJK41+D|Mfp1fh+2*KnHzjv@-d}voR3<_Th64(P+?ywr*yp
zjxojFcQifpubvZF-8Y+Giwad
zHluv5_`-HTq0d`nWe1V2P7#R`=Y^=X|0a%NW#WB3-Yx?XUPgP;USfs8wKf!;;Spj#
ztuv{kc)ub|?gRQ)E4Ut-Lb-C_E
zFSwwbA*qX9Jqz7=09atC!;&`CZVA;5S;!9c<47X2rnwO-fP)h9`m?<*24}KjU)WEp
zHq4g3wTKeRSpf?V7YoX`Etv>*4z`lEUpr$XwFIO5s|shfF|Zg7D-%Q8ls~ux`IAIr
z9AC#Rd|nYKjRx46^$iQ|oBz-m$?xfk3hkQyr-fQ*3=_I?
z$ORBZ>)ax0@Df)IOuNob*Xr=UNLXiV&Z#LicfX=gcHf$N01Gl!oOb&DdhD0pI
zDt1`Z{T`c(m*49ZjQO5@kdPfd}@EjJ{8M76`>@wd+{so
zY2-lcPWN|PcjeKs{5OcH9)43JLS+GDs)ad6p4G^mGNNeJ^T}y&;DB8}OeWs5r6nlFdp}e*uTaqS@B}|_
zpPnVZn$B@I-nVK+e&4djAhF`hAHE^1Gp2HGOzhdYC1S0+-;Fyr{|Tq%fvg<=(@i_A
z?<`9eUmA=^ZkXapA_0CLDk)uf@@w2mMU!j>wJk1tNNA#>1$t(Saz`KI#(~>SCsOexC%i#G*`pvwLrFOte?Fm0849R
zGE56`@>c#DO7=+{vR9-A5nT;Ilb~nVJ7T9zU9=FB4uw6Q)zop4gU)`k9)7?pr4J9Q
zgQa8pCi?wrO^NJPkvhZ7Xd&WFzS_F^v1d{vv)l3m=@-Wd^&g(X1T2Z7Q*Iuz^r|Bd
zX!KpJZuWSv@Cr05&*nX$DYD6LI53tNb_+LUN&Msp0ZU+^-QZwg8~?Rbe+f^=n)M?&
zR>t`DCoTRzt@UPStu#3|0Z|%S0Qh}-H21k~{~KIQTlIWtR|n<4`3qaVlXf$vBs1&E
zt6tPj9oH5YRV5_dJJc9Jtg(rA(n6
zOF>}nUSL;PM1OyJ%%(iwO)k^IEK3^!*z|rpy120+IPB>WduA$}f$~$75lYsx_?cTU
zw!928^Cen2Fb1~C!yK8}Blc(_A(qF7tT2AIs#vu|YnzDlOc7`l3OXYz@?FL=
zu&1=`-RheJKiX}>oo0v{u-hx^HNs}v1sCa8MH_fvp2_(BC0(JMp;9vlq^gw;I$mRV
z^aFoo&!RdDc6N~E!yjOvCZV8ggBy{&v%AvLj!DF^=uHW&-LfLC$bvP*;?`mpX`Dp7em)mbX{v)NZ}}
zuXhC$1GE0g1oLFd%#29Dy@ho!iLejj^S6^ovefa4apA2!bs36}j9k9#FAm}&%}$&h
zkkpsP|9;kxq$}>_y7&S3+p*GX(k5O$;v9v`-@zT9hcf#Mt9`a2!D*_U&jQZk>oT&M
zdi3#0S28i9+HLyo3@`>RPV9_ZI@BN{wdpOj(K-Ol$GUZ=|CrOBc9KQb8FCf`1)N=|
zRgW8kWy#Wae;LG3%|TerexL?WB`*@4fe(bBk#9n;4X>8WSPT
zi%;8>O-j66)C-(LK4WXoS4Ewtv-02jXlNlm{(xrD+2CgUmWmam$Zj#guOl-b5Mrrd
z?=!ah_4dfI4V7sv{Ko&zZ~^aP^wGOs&I!2FewS9UJSW2j-fu;aG|VFFSgz@Cm+gE-
z#TIp!mcKbjQZzT~$&B5O{>H3}BS_ud7aNK_*!zSzdbv^3maHKkV2%Au1&TjD*=6K>
zjzA|AlQddhX<%p1sXw-MFye&(>->N-KRaF|Umql;f&COvina8;sbr$+)DWM`1Fp{Lz(c-o
zPkY!iKSD=U!B3)QYr5Z-P(?JoDJ>@-0W$NzP$jEp7~+@FdS^#&OF#`
zx+~Z1Wq!Zlm(E513RvpT3=3WJ^&uecbuO7SjDF@X@0n-Gl5u#Q(YclmG!12f_3POSal`
zEv8xcM2N(3g?t~#jvth>z892e5=!`@#WKfq-ZD9Lhb)k331W(!BA5g=4-O9~$0P|p
ze@9~A{->_K9?%KzSeAGx&qLoM|GE3t*h5Y5;yDE(FO?U&n_8HOIijp+*;MFpf9bku
z6VfxI|GUu~*yz;`$K-zYP?=i~yl)t3ZXu`OB*IS#Ou1>$b*{Vcr6EiIqg(&eUq^zM
z;LA6{LEUP@4U+1(v(7b=^QfFqIU0Hz8q8p-0nEQhQm)*T(F
zRCfCbUN`5rX)oiI{!JO|MIcI^hOV9ctrK!(C;(&t(1LL*Lh-kcXNaX2bu
z`jQV=yYpASygY0Bm)C>(4+?5f633|^Wu%(|LzWUYw@+R4kJBj;2iW$;LrS_8dRP=<
z8AlP@_@KB;&P@gX`7XUy+904m56g+Oq34G`P|gS!QHhrwOWOrG~!>*U{Nv1YoftM2=%
z?q-MRNugYg#ZgIRwOY!=y!2j~J*DbU`C|ZTgWCRCocw~NU7?WhKDmCZi={#|p$n#Ilr%HVuV%-@eNa{*ZD2f&CJ6vV_*BGj8
z*Vj<=g!NgwTo^5+TukhMdr~fXyBrjdQVBs@@5gLUdL+~|s6$I|b=cB&L_~w}v7zC=
zrzd9ApCg0!U;cg}POB5u%fE@LG*lg#NKz4GW~9KFu%gR&_djBVsT+^EQu0nNV3Nx0
zFkdlYSEp8SkpH~~;Gs@1TOoHj@*r?^}(?Tq;pB$du7*XN0
zZEF$G_2uFIHnqN{Fa8n`Kefn3>TT;u6z=)cbSOuoiKI+l>xu@=>iU}7@8VnRtwr>p
z()~vx?6pf@2;vB;7h$!E^!9~13?;wn
zu}N(B_pUCsg}w8F&ZXK~L)B|d^kQWEjRq?<|7N*!)n431%OV_d1KZ*>8YTf0I7yralF?C#cJ`PbkIBX=ZpGT;U)DU_c>NKJ^zy3=|zp;NvS}8i!8uL
z`|ug+-+Z#;2YJojNB1H8&WjCBh(U4=k6(HR_>@wyi3P)rw2PKZP^<>jX;%FyKfpSE
zEDo{kSE1#-`FvnFlq3Vqh6Z_6O+~(D&#M&`1VPqABs#`VGr)pMW0tGu^Z=xS;#Ip-
zCm|P_-PK2kmbryMu09Ewz)ZHB5)OFBJr9@q^p!k%($>w?E=R;>_@}+wI{_X#TX8m2
zs$k6KRBL66-vOyxEoa5C;twWOIM^St(hCd~;U79vO^5wN_>s8IgoEXX&g&*B*#!Qm
z-o{upL<;@46IDd*J_C%QJtHU+!w`W@$Q>*`;m-HXb{ntwVIharpTt<1czeBVY7*9civy(zU;^fJS$A5%>2sh^ysN!l)(J1?XBP@YNv0t
zyq^ZBwyB4IdP7+EQ&J&fCNgG_Iss;`bv@Z`HJCev?##oY5CZH2@l>GI+!<0D@l_XH-h@juIY7MUe+j)Cy}5)W7+qYqP0p5E2#S2BI47i
z&_JaXw2Q{kVokTxb#^!HfKQ|R-g!UTQ~88z{`A+;#J=4#)j&(Z1=u8nv_eYxf6VC7x_maj#5Lwbp2lPmCWG^`ZgFxrG^8&K$J{sA_3%;vKIEk|--4pDQTUo2tNf
z9w?xKfiz2Yb0^$?L>&{%Ru-&nOlzM?vVBcvXLvDWv{sx-HB=@C>Sqrw&k{nN61?6T
zZ2dH=Hh!VIrq%W`=04YrWlBT;CS-46Zh0%xL4ZkkztTwwlNvv{+u)dFWYpC5dUA&ryIFHb8p+>>^MEqu+5ywIvW
zCY!2pDDJhLFdl_5T7569c1T*cT^N9Hxs{Eit4VKu$07=oFlLHNu^6m5IX=
z3xAdR->EX>^_!^FgAl&MpDAJMYsLp5-%ywMX?fNdrH8F%LZQQ>Gm@5DPkm(3wA8fv
zR%xBO-W=pe)aLjpM@qs0<9pc@t;IIDZO=D7>a
zcTX4%YC+7VZ=CzQ-d7rVhjKxL*Y5cd|7;xIJKEVv_gdH1Ny^RS^j&YbQv)q`msGot
ziTMuEmgGCM>gFnWGm+Vfr0}XQT5vP5(2)ia-sQ(NEaF1XyZ0N_M2FHldYk&5FvuG!
zb-eGt`cLkO=jR{0CRY8S#$Go{LfH9|Rk}BoD_|tKHm~#bEAfXV3Q!X#l?zByKL2-T
z2E=N^&|kRHz=wX3K{p@ehyvQGA1$s*KQH#K-bbEC9AN=`{_phnc5Nh20>V`VI3D@V
z;&)=d>O}xy$YZ=^0FegOxAnF9tmz$m1dMVvc{ms@*;!G8WxW1sQToCP5W4KW4nupS
zMuLtVjZq3fDs#Af4Kn`j<|?tc@4nq~{}i*{^&F&A*&HL5G8xQryZF5|Hs~dn{q#LF
zc9&*sV`bwOKmIC*vOX>LA}ZPaIUifht+PraHG2#RuU<>%>;b9rJ0D!Jf7
zyZG3lD0f|npT~bJ{fSppm8(IkeH}3qdMG>itA2CYu;KeC`<*m8N1U6O{_k;v(GoqP
zzCRvk%s$}Am-P7M0gbTJps8YP)!#tB_Q-%-k>{rAquC$qD_pkiO!`tPRF*H$U49Z|
zP%vUh^57#U6ds;|ZK
zKI#1OLo)O{7^W8;W^Z?Q-lb3uQ2a!VM-sYzm;U*-ydH3_*w+nm#cveOAc$Nu^s5f&4awq5bPm(AY4Non7!
zjcemiuKS_~C>xn5#IUg**8u_&D5Izoh>6_m^
zY?iOOw|$ueV*Dw{Wg7g77%K37L0fXx5-l7B5VE>sU@hTrKCuYv_D(SBrX?(7V8u=O
zZGfuI-)8t6HBEXln9xJLOkv?rvph}^Xs;vw-?eN+@SQI<9y#!}I}tS0x~3*c>!M39
zJZjVkb8cT*H{Q+#LvhdTI_8t~8{(>fNyEak+7jn-+$cfuALH7k`J7H|Jjjth3!@P#Ad%rpgbk*D)}a|9et@o~1X
zdS&)?K{`az*
zdeY6Kqo~R*`_k18y;thbu<#B{s|J$t^;AQ1F^pU;#j308YVd
zlyVKMWZv3Lj6`Y*ZlL?6L^hdb!Ej|Was9a;3jFQ_Q|xP&_q{9zWKK7#fAwn@ha4L4
zE&mO({#6&w5bafo@<1Tn`_Kfj=JjEaWOPS*6aBK@@ba_xPd3(OYrw`bi|zM2x|nNj
zV4-ICp-4>x@5~(Gd>+$=wfod>QH>V?e_B7k(8jAh$tx5d&sQQj-&FY=w7-1wC^QCb
z9G34wobxyM(c9JS$UVX|IMPbl;^SD^)@jKj>3pW!hy+5;A7`bI74N=SG3iF-@08@5
zRDXT&A7kk)&$TN)Qt$mc%01QTx#7n5G)K%O4ElWBcQM^#a@;^<{^6ksQCH>F&!QMN8>tQ;G;A$0mx
zi6^JDd*}bR|L=7Cw{zf%P<_SrLiWS?*kLl`%oA3JpIBVecx8Mhz!d!_mmW&GKesZV%Z40|9gB9!j?=$Gm9v%_iH3BUn>MVv6bdmAj
z&&7|`NWU~16Wx^SDSVcufNAPJeMV&B+NVht=8nb~ira>{t{AO-@DI;Bs8xwSk7;U+
z;uG%r^BbYdT0#Hqw2Ykc7NtcDBC+9sBQ#TQdpoUo{aew$w?h?@kO;4_k7WP2{tk!3
z0@?Y5)YBK$UnDAwdp|VGg#GY^f?}Ic69rZXf6g}WuF*>AMclKmuE?w$N6%3?$x7IF
z+qYfYc!SX%x%#UE7S7)<&y=@WKzSbUmU(W=0nrHRT0*#uCaS>V2m^h_kK{(I5kcAV
z>w<>-LtblrxpH2=mn-bWT<4Z4^EPHk~HpRxJ%5Kk74<&h_?pZJ}RunbX|oo|ol<1Cd1xls479Of%9
z_Zf|1j9}la?Pk>`0i%V`4`0G;BCy_?h_Y9>u!(N7-VM69oSa>Hm|uPA=F$At|8E}}
z-xfsWYHJgJDB3ghjIwQZqrB-C^L1t|S=iaxuYOWF`Uv{o_lDeF*Gu5tq`;aBR0|R^
z_C}JCZq?Smlp!x`d5C7C!KB*Vdb4^jUyH@`l?EPM%ExZF;0k_iRqN4zy@?U(-GYnr
zipdzD82^qpq%Oz6bM;}(7Z2#ELDx$)F9$Z#Ah3~s_@){THc|oTrmz`WJ!c72MIEh@Ur(i3a)ob6yymneeN$0x^DvPI+FNN~y
zZu77OD3=anMyPxjPA_+pe@^sdYWpFmw0gbI}
zqJm7jld#Q~%d1@nABLKMsTV|YD$)@NIegS3U8xgJ-Zd14S=va&Q-Loxye03gh|=*y
z8YqW_FwnYmof5D;(oBGh%o~%Zlu{6Gk1+v1bRsLO%~`
z@hZS(uWf=w*t-1|Dz;v3HU@`=v_oglHkZAK%oL5wk=ArMWqzUlD12<%T
z7|ihv@3l&^kA)qk{^?~w#5IsA)c@7FAgsS|x>z8#0SGL2y1Ap@s#=i9WOhk1+u#>?
zuIRh97r0UeL}OsdXIcVAk&P}d#;wd1a5OCDrSVBcbc-CAxa+(I@WMW!4@xyHD|60Q
zX!LiqHcST&x*$4YcVIDe#9RTxT|(Xshy{8}nT?WWXd(K--fkq
zLaO@jw{HQC>vV-?S4i^UD5l4;9H
z*tqQIr+NAe#cTQ90s2MC&dFxS4GYeagMjjtoRI~aHB8p+K|oM2)LFTn$66YlMn#_N
zD3{D%Z%Uy6_jaMvwtkPfRx6;;r{k_$PM`&!q>Ui!ZP$39`?
zftyt$exg{qcZMtst&9#>cl>6N@eIz;>7ko19FTayd10WWhAsX(<3A4v?Fzl@dHOA9
ztu61W!!?id7uE?J4^@@BO@wWzuqkx*W&INjJQ|Yj(@+OW~;*(E-jk
zbY235$??=_=Wdj|&@GP8H_1>e`>9sjW}{ENEEtTs90CC^IA&OHq0R`4ANx^#2vpBw
z%PGSro&*5`RN6fJbzX?SN)#PP=Ur|NYB%b901#ow*dK9K`L_1gq^|@57aG6evp?_=
zYJB4$=D-xBX6RXeg%y!L)eZRwS(LBc>Vla}A
z7A@XEC*1Dmj4yqyGMhoSKWycKqiFEzUsP@nhV-20A|~H4d7o-0S+?z=@@wticCD~<
zWWJ}howqhqiTPzlrC1ek;gr>hOhk@&
zT#puJJ%1&5tNxM9qS-R>Ilq;0kH63D?k#_1Q?Hr-&@UA`7i0|$j=G3ANm!lNCm|~4
z;E#n4+8}wUql82UB=ol=^|Q2omBYYL(rvPaCcq39Z#FFG&p0`ps>b%YIo)Gn5%7CJ
z=}uz`x1&WjRR=_nq;+7$1u3IGTpY1eG5rmSgPk^)sx^^Z
zhv5CI)M6zn92;C#>osIG7NPiA#5jxJZW+m};FO3QJhFe=52r;cP2
zeKNW4G}d!1tRCmWu}Gm*LFYuXRPqcAW1P8;C^-usPM%IYiCL`1xj4>+VU)_*O>Tv4
zQ8_E_PEd`}jdvo3a*aBD+&i8xP!(4^2KxcZNpw1@f4&io4>BGdhp<7a?;__4;1fQ5
z7Qg;c&0;fSu2{PV_{6AN#gj}SAIwF5usxQY<`j)fbXg#IV^-bPB1LjY>hyLlkLbah
zP~PrfJXU0LI90DUg&%faFC{{ba*8sA_0;PicN@e3*$W%YY}94^*5RGbXCtuQ;S;<%
zY)Y!KoA^Lcb!#eBr1quaag__#xfZqgK&|ESn`qO!Ot@mgSmpL=AL**Mk51Zhk2zxZ
z`xRkQ|Er6tjRJdw;>muiO6(clj>gi-jnS)`g+fBf>!l6f^>Y)k;cajA=M%HbHstR8
zp&GWpA=tlg*~hA!KTTH6gEeVx@KO@S*8~
zLdE9^>-7Q{G6|>CfvOm6X49LpBqwKQ(S45LB$Y3NFAr~@olPa4qb%SO4Mn#qS*WPH
zqpdRF+qd3NGT8Uky1koQo!d*jsFmjPN6HI}y>>W@udb|VyIB)#mQ)Q&9_Lz%x$ICK
zYmi)D)gI@P{MAf$l9sN!qsI`m1RD{UlvDdIZ^>T
z(`Q{hzhb$gTThmlovT*#`4_h1URfi?Z~N075T6?rUEQA?!!y56nac%FL*
z!%-eYc8Su|u6OhgyJBA+9T&X$_p#g;!@-m<$;
zC4@dCi6h`u0l!g{ts*vCz5sl$Cq40^RTmSwgwBSr`$L&8+=q;`XNat(qkuA#=0u)_RHL@P1XHC47SX78q{P8*^n$=
zS27U%v(>&`BLq+83h!+8hPJpJq1705$1l~JwNtmzdIz+^sNO?iP8RkV46q_gj)!LO
z-^)q%v^!UYy0%!1Vtr>eBD#SQQE?wd?)g5)DIi|HPXHv$Mgbtkj!|J%+k(11L
zJ5PPNf0c2PJ5m~d8Viq~fQWyPeDI=dB5g_GG%>W@{_}nf=Irx_>-2W!GB0wE!IOuzcF(02XL5K9nh(^}2{*!n
zeia?9JS9JE0lFO(lQ-7o0keJ!$8T@1?H5{gXqa^B7j>Bf`fCAEYG#`5+I&_5@A!cS
zU*B5HLQU33`@#_-p+BLK*e+BJooPvPfHNuwUr+bR)Zt
z@aQn%i?G4i_|KRie8olRZ_ezc+ycQ*?c$rq^OYf2MNxhb`YlfR>Ue(*s9*EN#Uqhm
zT4TKDz{I946*TgPfY&82@cpe!t2SpZHaM$q4as8kM=C;L@G9n7hZKnn*~QDF>+Wb8
zvOzeeSEXik$Sw|pdj><*J#YgcP|vOkPH1BzxI<5ulk48_)xQl^qmDlcBx6jM>P%#E
z1X^#!_%!}8IJ7f#z)voi%izOv+kXSrBkxJ;-|CMXO}Kx$Ju!QKdqpZ7Yv%mXCKM;)
zB$a2s@_UAEH7E7)PcD|F33L-WhDM=d3anOGEq8pU#$RR--`GB+BY
zs^rGZa8AM9QgL{yDiWiUn>4JZe?$wdVgFdi>E&KcrKh1
zG`G#D4@WI+Zbu>A?{E011*wAkeJIFSdxT{CjGL$!q
zMu9e!n!c?UZo=Tm8WYcd>(Wb>mL7#Qp!oeZZ?%hRbCX_O(@{977+32-hssk){8
z3#F(raJ_TrJ8jCUgHp{ApZhVilC)bTcxG$lB9XC;ON5+X)e~KQNWB526NrY9_1j@}
zJim||!@?!%<;dlO6b;t5w9#yb8iI4er>X`H1q#~CP%s#E5jb!E
z`Lt^Y6L4LG1COgJOa&Q8<%WEU8a$EBFE1GYirrbgFV^#)4Yy*sRF1QF$uPGNsZDH1
z*Mo^vtP{Q8b`a$R1frncIQ;SYVCVK^F`h+VP%2I&QADIW<=H-i&2sG9XM8Rhi?>)@
z_C!WKu2zH<$EM>M@cgAnoKUASGI|D#nrpzB!UpNrDxbzg_)y??G(!oJKcPflbK&uK
zVBq^K7Q-DoVH>qaSP!T&>;0cS#(uH`;(hjBYpj?%
z1*``gun$l*nX{nWI)w_hn#{=u1voGDEE*y$yXXUEFx8)aoK5qDF<$Eqp3&c
zJy)Y%a6{J^v`4^W(iwx)aVUvAvb)J_64Mp6(Cui3gfskqAs|2{fiWqRitF(WczbzN
zDf>d2D0SHz3ft@4@~a;c3eNd`)D}-5QNa7FXDlH+`n;jjK(QcGD&y=Pz+1Je8r8h5k0IEsj_n~YSBM*V09;;#-
zUH+^leb6AG9**}dZ!h^*x
z;Qa$XYHQVk(@6+&H@hfRXQ4``_;vtPq)CFlcRj%L|LVXTg(UYbgVVOoxHp7k*2_XUv|8>H}_c2D4LfiNQnD63Atm~nG2-5kvpiUdREnHm2VY?2mo6i|Z~
zi^B?u&20E*rP(owPJ=EMk5jIj8s;}5
zJywK_S4X=(Aou91rqceg<1ome^GhZWvezZZ=L=piNk~utHS$ALLHJvRMrBcVJ|?|p
zQ2UP-HMRfI9X6Qm3@yd(#lcUwkfqAGQWIIc^`1aO8=dLj_Ift<4hnm2(|Gf*K4&N0dS0Z$z_U>4xC7Q@Xuy8lW^9aq38-rJSqfIbXxT`G<
zNgNnYLM9&x2#di}_l15?Y_(lvQpo1R?2km2K3{F!3LZXwD8oAf)kQlipmLI<`$SDm4a8b}1j~P15G=IE`D#=+0`_)o_Z$3=m=L#%OaX6~
z)0O7Q{+8P)sU8)SypX|x0ZAsiY1RU)Nw`GmT%fA{#ngXx2I_uQi{*j?mcS-uIi4{L
z90#chA02_uD)`e0N%AnCJXk)pA%LGYP?x#`g+Y}Q#_xH)2?8zasl4C~`gu<9;v&WL
zq);pG`kNok6n{gfR;nAw2id-AKCT^X8Egq^T@YAIi~IY}R+`Bz$I`J}j;zTU3gaZVIoN@R)?C|HRD*u}r1YPyG*ZFibC0uUBh)T^{nZ5L}kFPP1R
z&cdaWiI9JoxwKzv)6mV$EIDyafcws%g9ggTg?spfxgNak>(Te!xslg_{HqTOvOPt@!EtrIHN5LpEr9y@S&>a0YX!J_nl
z6MQh4Tj;RXmPh#~nNps`_s$M%CxF{S$p179BNz%6Nd{mQze~~}mrA#MwQwUgb22ClOFC3)B
zvhNap6MnBxoHFs!p1R->;eicqDxG>w9`uEq)R3b;7{#A%j*@6p$c9siSC3e3{(zSf
zBC-e6CR1=9Zx6&Ifs`c3mM2lK;(w?#)HyY73;lu}yjW`_-?amxh0%@H_up7AU#UGg
zs&86DW&bPZDL5}F5}TQ#P5xaF$%wu)l&h6~^k=nsC(o&$Vf`8nk5_yBh0+O|$H&Jq
z{R?L9B+y-ckC3|QX|!q|L96YYm$c$hHeacgMlkJrcUo^h(8l$jy~UyorL&TAIjqun
zCM#!heb#BPoNvzznSMFW2RS|rMDOvaelZk0W+;Zy9bQt4`P0gF{Y;ye(X3
z!-INAA@#fr88xq)aOR9bw-K%=ZhD&=8dQYHUAcsUK68$kNB^@=QIt*o1rWwf*O`dV
zRq3elS(cI;{O{A5!TM)3HERJvXox`3Af-hRwTBCMUk#<@yguKP+b=i1&HSiQ2Q9gi
z9rSvKH#v|U#E9W3DJlISDg_dsvV{C~&sCkYQEqxd2_ZqtHV7h>-Sd<>lm3tGmoS4t
zEr)+AwV0JPpH{buB>!a&DEj0mbm}xvagV4ne#I9mjkR}me^)#3yWeoupN
zJV@`n3DiR%K<jNtta>KW
zGnjcp*_6fr^6qlayBU7K%meq5F)=Z(ia*Yiq2QeK3PsD5DE6>iH6o$j^nlREY}$D
zhanRSrN^lv|LdiYp&Ne|2j*3r%>7pjCyct@fe#nG;!hx%O5}C51Zyyee&$B}kPo3x
zhe2SC%B0Vu9Csz&32KaUw%xvye*k*K3yb`mHWYoxa>c{!bX{r5g8o
z>>z*r#So;&y9^_&ur`5h>^6fAok4O`MPEKqpy7#%F;V<`5`I;6``aJ>kqL
z2%gw@5QnMEHZ>iu+CTz6Pj!1F2U5wn4;!27{I0v=aM1%6h
zU7Cl!K@zzjd<_F}&po=IXpPhHT=`z}1&FWe%!kp;$I?}v0$Gv5;@uy!e}Jj(APelh
zJ@sH+1mpw?`oxI@k@QqAfrutqQy)<+U%3*@qB|hqZ1H05Ovf4V6I$k`BM!{dTUFY^Nv*K{0gKH
z-?9*DUK0=F_0du*XdUnV4ja!@f;=uCBr`Lm#Gv+tlo0lEwT8XS=Jq6J9CxJb$guY*5Q>B#Y&OX
zp@OP0SP&|BMX37nWT85yQma))h9ZqLfsH*Ld=AUM-*KX13k?Ykq1)?m
zx-^YGNrzu*whF4Hp80uBk*5OuOCiV`vqV8(k9){nr=or;M(|E3xb2uHW43#Gw4xN0B?
z&*T7HC_sf%fVPeT5s>PiD0Em@WFiSri^gI-MKVmO5MW|hs|$$*L3e8-$_HvOZ8|O=
zE_Z^?f*cgdH$Q-L1~%QCUVOJ9VXl-Tf}Pq=_++{;(vK_ILkY|?n>=t~C|HaRCARVG
zmV&DP^_-jb;Yid8g!~b;M%^O`f^-*;eW)O*#b7qnr+fZMZPFL^wu*1i>4@U>e8*@C
zh-bunn#>Utzd4-tSBqmRM1{kE_ChZBm9U@y`c+X7UKTefI4B}vGwJ8u46oxClj{e-
zq4BF0;D*EdtJ+Y=eILT9?#B!MAQ_~tqo26r3sbGR(qs!e^F+w&B0`xhJJpU6i~5@10T1Q+xD5=>;W+XHB^TxG+DY)L`*I)lrE`68uZ_;9gRAMok}
z0&bAf=_UNzUH=Fc(a(u)SSutVFukBqQ#$h5Xj;okN+K}&((E2troTTrZ3ki^;j$I^
z67aDRTL=q~Qq-yXl6)X!2Lt$3+m%DEQ4)N&S3
zoy)rPyWSsH{W;m;=9LL@l=7ZZabn*B2X7Fv-T^n5F*U3vvP7dXXm_>eaJ-9fCV7>1
zmc|WY#wPtLs*c<%Mn(~Z-*HB9cExy3%W*C|jNBQF&dx9WJ##)
z8#CHt1NbW%Chq1dbap=K?&V(Y6NMS#oPqGrbX=~xL&56hs`+2?#Z;ZEW^Sg1P{%)8~Q;%7|M>-cjJK1l^wiCxK&V{
zB;QHei=AXL<1Hya$mC<2p!prW(O{LJ&Xc5Gb|_Q5vcJ1X%K-f!pd%-S3KIqTq2;zSx5A#gB}oid4f
zr^R?$cA^oNB+bdiX_!-4$EO=@<{cixTI;zT=gwwFH^}FgJlL@?;zPa0gEjlRx-*5n
zw&_=Z{YvH4!l3;!^G>TL_}oLAFm8uS3;SBdtGdjMNP_@4@RCa*i}qMOVihHLb^~f%
z(Uh`yNH@9-yZmS|AN8jKo^EKpCdfC&%PtOTJmceb^35K+F1Nj(;%E?g0x&hD%?SFX
zVj05~M=$iF>J2eV;|HJAvo5U}*L3kt>K2%0t5chQVBuIO0!&YHi|guR;WI}oXJU6$
zAJe5&fM~H=lT>d%K>Sa3%kPJDiTj;Y{Nqof_fG_1dUkU1EYln=<@K0@+iJAH&DxO=
z9cnQbEQ`)g!3=j0Km740fM);7n|ngv
zziLI_+bpxhBzL#@`=WkF>)X$mYa5+Zqxk{XJ~=xLf&>hrxtEUTE5gZ*K=BtElgmBW
z3qc8-tiUVW=G;!$Y<_p(JnV2XWy&7p4l*qdRV;A=6fAO{NRGZt8XZiIU|=xDaPmkR
zu~60ADDktEfEF_MATF`rv8a{6uPPto3i2Iw-iTD3S7sXp6Jy&ed%ukJ~Bq)&6jdL*Mk*C^SlS
zEPq6E=$6~`tOBWniBmMjahq1<;wG#~IQz{j&1n6==1_ys2)SYdP
zp$!yb#B8?vONOllRcIt+lft2iP92-}MvL|*E{*_Kja=%x%(ej=bYJ9i@zlXHjAUnI
z**%g_%G~vkm$~6BgPQbhtXT$pCHYRQ>F9$zw|){=Yf8P({hgr1n?U^>R|;dXPIR-T
z)Yesl&V6Wcn;%~7v6i%%k$M&%n_{>Izd(hjV(MTaD|5e5w?Cknc;j;al{%}LZ9GTEp`zXl>EE=WS$9XxJo0v_}mV8;MR8W~i4uS6hBx31bw)wL*%1*l67IM|~5#aBq#D
zxhk4F+7XkK8tXWk?yTSoEsP5;0<&?qVX}RsE`g>aPnUdh%x8wnqyyetHXT
z41-(@;c*t~OZtYXmZzB`01o(C%t72*-oe1vOW~IXBRGLa49L>+&i;f%lc@?t-$qO+
zwZQ?u{rT;8aEJG9c3Q{i6z;Vd6p+qqPlvZ3OaBHC70BP5-)?~u@W00l%=U{22jrTx
z)Fs=m|4Mxx4G_ueONiI3e(TZ5$2nXnX8fv2S9-(^3He~W`jOE>v(>I|NR~ZG|MViA
z(^o0a9~#;$-Cux%nwt9J-hobC=K8OB1qR3X!O(#l!*aPAwerN+JZwF=Ss{SP3ocRE
zYpX9j2kS^Z;ro?x_R4IMKR&-D;$SRpoDZgaf%}@AOa@Jtpu5UFFrZaq=V
z+u!JOQ_^;;Nkg&xTQ&ThNwaNslf$X}bUt^wq099)cUuajtfgGN`FUID7i0+C_Q@<>
z^ajiEm
zke_(M`28hs1^R-+5&_A}>A_?!cB63d)bH{-^RFvpB1TiE&PZ;#b}=7>?EsVMk+iPZ
z+qr=YWJX!ouS@Mob7HUJ!sg|JJ|LYfv=zv$4dMuT_)QVYpq#X79>Ugt
zLV#b})2s*aCYjFZH=Ze0-!F&1&do?jr#wdu?7HrV?umtdykWdl;_u#2FmJ!mnt@+h
z^WUwUo|3b6l}=gPVe(8yJ+3$z5)u}VQ|+UBQQG&R21VEm+r^sj9CV-}oIi4rQZ`-@
zT)q=)*vbqDp&_Ay|29|^f_@>-)aX;W$B^$>V_!HzrFA=BvK1EyzPxUaG1s$Kn2z{P
zD?;1Yd*LDQusDucA^DMVzP~I-cdsvm?gCj{c89K(sr(H@&fw~9dEET*AZ+2BtT0;-
zkG!%Gw4;Saba6LI+<|jMJ5D
zL3F=rLeuXn^g6;c%{MZ#%)5hnzeSg*id%4>ePllB7{r=y
zE`jy#8)xN0hTY)v1kGvvBI?L4tT+*4l0*S-7f^Ss$E6tK_m<0M5q&dkRy!g=7fPp5
z6;vlIv|a1`3JJxvRuW2v!Yc0kommTu>4ba94>gLvY2-UCl$Dj$SSI(RDC$DL2L<%Q
zoA^5;2|Mbb!@8i2Q^;So2iEzrzDZz~=^6B%BwjNO$#odqSP9JL{a^Nq5c55A}2p
zh}4>_vsNGm{A3VRZ5P>meI$H%I;VWPYkF;xUT#aKln?3BYg0O{GD
zbuYVSS?50inZ<o#_Ia$;Xi&4vq%IQQ&kleO{`rjD=t`4qm|>&?-p8ArPABxj5C|b9)dobW
zPcIW$Ma_{d=7&pbg}-=@CQPd5q@aiUN_dJQvQYJQCXfX+>9x2(2jLVG{dvqz3
z37A|wS6t+Oa2=*|vzaBRHyylCYiG)A-s#<;wkCi12(6*hWhq!PX%rTtRYUGkg;S>g
z8zbW8m=5Fj*5&pd{S6KWTKhdQnx2VxzEhU~&U7v8_CVa(p3}sK4Zvu@9
zUd{Dr@K?!W?7^VlkG{bdx~i)RS0dUwf!foJ6l!IshYOR{7H6n-kF(^)FY1PB5BM58
z+pX}Gtav*L{euD}HWfsJYE^sgqg$7?HUVZ8HW^2)_4c+lbi|mEMFAs8jJILBNjy4e`1Ngj4u~u-?k%dZlxSp
zYc&7LJ;CR_X2ux%i5u=>ap`IhtkZiB6i%#^bG@-$ftw$B*EMSmljzjiJBdLrcYHuAz$I(`ixbty6eE`a|-!r+W*6&@|kF`OBV-=4?Y
zoNjo0e>?3lm)eX4$fEB_)lBsM(sKgdpkTq*IcCj`vRxOks;_GeJL$?b{57JWu(t&+
zI?b=lC$FvHN88*_E;f6i8tqpWIEGCon;q8RyArYlyumk{5`~h$C(jLD7rY>8mf4y;
zCVqSq>(<;EjJw`FF#)c3mT6RC=^GdnUxxf0k)SvV;|=ud-fbAd>bq~78%{A*(x5h1
zMw`ddST2fZOv{M@zA;-Tb0;YtU)RbhJnL9*vwoLBM}QhFxu~T&`?Vfe+xaJi9NFo7
z%eN>BA2kp|Ce{=Z*0pbA;CrWM{ob}~2S!^S2=`bw?=jBXNCrRubwP)9=#nZ+BrS3%
zbv)Yrc0G$tY9Pgm?NQIR3Z
z`=iV|HQJPpiS)q?{N6ypD!x~Y(zXLGUw;WWeE7APKjIrXYjw6j2#vVhc;#H7fsV&(
zr^hKJ;rSinaO%l7H5b+lwZr!H)&y2Ht9+Z+i~!9;$F)vuYNcF+U+}MI@8{zET-lUy
zi>tDQPG|e}hyq|v8^;(?W}i>x{R1_KlUAN
zpT^T0zA5bl_(ON2XKnEhG%eQJ`l`G$4=9i9PW-~Fof&W9EYh+n8rDPXe7+I&o3?t#
z2-u9psBiBcs?pjW?vHv52+4f)`YJqQQp=#Bt0y8&!P5W-E^I<~<
zqeH`LK~9;^@o(RM3@6WDwTVK1>f+T1EgWhGRfRqs&F(yO0&BMJ#noGP@%hVF_e&fz8lV`PoI|eZtl!lxNsphY}kN`
z6)Q>_oZo)?&C_HI8#W9hMvTDt@#B#+X;KUsG6V$+7R04Xm!#sJhA&)5pEhlpTvX>X
z*_uR&5(T{e2M!#FjT<*w%j)g4nKPy%Pm)M>rOapNsUSXc|DD|{^n*%2y?sIWA|sv9
zHd}fTm4#w`QnfVLapeUXExCY|O|l?=a?kggNkaECW$zsf6iWQW(ZgfHzjsh+&Kd0O
zm`fU9Hu>uU?!W$wHiexKEqrLq+_Vg;+F5DSTy46y-br
za}{e_WHsCS%!#6)Q1U34CJftp_w8#okB;;^0v_KW{9d1~!9fwYckiBj@tEDkvK3Cp
zHP?iNb|l<>@d<}+zmbaMOOHQb&cO#baO*WlEXc0&c%pJ=`S}&U5Re!@{GFu=mC*EU1^>eFmB>yNqI~W1wEn`1m5Yyk?7bHi;EU
zTASx^ih>d4f2~Nm7y21seYF0krFjmasP13d8HvP=k77EGWsNeT<%-K_T`)2JI{p|L
z;zvS%Vfnu8@(XdJyer208A)P1*ygAu{o(G9!A3_
zoRvJe$&w{Am&HeMix)3O^@`<$Rd-CY<)+SEh_xHHV&~dr(&}->pMPQL-)r&b)Nx3Y
zGC9s)5+Af(dt>4B@o-L(7z0L5!0HW~@%qhM%%B=k{;IhOIj>f9XmbI(8DtlR9BY?=C1+v@punYAme4-(l8*CHVUFD+YA$=-$7R
zLb21K_dt1Gvq>9F9yJ&R^W?;-a~IHi=xAILR_dum-#WHzf{MSF#O59U3e2XWZL>zG
zTfLGj$G*H0+e7o^%#L-!LZ6E0VwGxRx62CYthJc%aiJ{sFI~Nk7M*%x(ToWQFDyc9
zHEWBmZChf_;$_H^J{?96>TR)Y_fC5e;28fAIDh&qtUc=1@FKUzwdq
z*7#J$S13~)hmN1bsDZuEu6ZMw9#PP&a16Wl9Yp!k#ofyo1Gn$o6Z6_f&c*&C$J~>y
zi@Bq4XG9b;R@q*q%fcpqV#bJ$T9qqc!GZFRlYce
zbZYKi$L!@UShQvfp1pdFJXz9XWVb(%B%#~lGnKtl7OlpfBd5e7LJVmYKcsVW@mvop
z>feO@CxkzYtQjzO#d^H`@DUY?<;U1wZRNX``Ee5STbP%62*&yyFvlpxW{mmFPa~Fj-JMwcj6s?g$uq4|60|5FN&6ZMq}2n
z?nsx?S=Kvq=~^5-d0v)Z>arbM_8vj0g1Iq%{&LJ7)&qI7Wq_r&Wd+{&^XS8&sx}TBc~xqSA*ADS9X_;&(yC>hvjXQB@EuBAqe$1RX6K}zM|bQ8OIxjTwf
zeU&R$mWo?*4p3RgojZ3BEm}0JT)9&22BOnc)~CeU7&(^QxpSjUn>Nz_N1;N6q_sJ@
z*<;0ug+9Xgqh`&T7&B%J)~;PEt-^~GDI#&9WjK8n96Wds^eIA&Y3*-Ltc{7^MpQ`C
zvYoz#{`~V#=~$dtabEW9*#lAoah&biwUbJ3Qk_++Rt>|450|uOoZITvtD}4O?y{_+
z=QGzAV;tvj&6+iGzM3{|YE1GnJay_+iP8M|^JU&rrAlG->ecdEF{jWchOJw-mRPd@
z=ZtuB?Cd+Q<73B;x$pM!vI8PRwqm9@vYlB6?qRdgm3;X69aZL@F{`M(`TV2Vvb5-K
zwc?WOv#ZeG?799D852f9FQM02bo8OrH^ht>2E9uqLs8KeinV?-eXH0lK&A1m=bwZw
zB|O>{NrI|b;>mHC>*s5sZ)msbG7bv8QtZfK(79+*l+6$a>`z+}-@b?w<{r9_>rX!-
zozT~euaX9df9jT~zo2zG^-J_+P$qqBbSvg8-_NK=A~u^Yxb0q`W&4PVsibAQwKR^w
zJ97HoTdX_x6obp9^iFp7Jbxa3fEW=&%QeHp;{`$WC-63>(4W5luFwS~MB|^k
zFQN#gJ?D9+u(;>7M*U5HVffiBpWF2vLEzA0`6cWZI=19Ofxq(96QOrYk8&B^uG?h7
z;B&jscNujZ%+QflZq>p>(>~2ouWkC3RUS
z?*|Fx{NHDuqJUU6{Ql#A;wEt!r-VYdRe?k(D{h=rq%Re;I+TxshNas4^HSMgLMX7G
zzyBs~!Vld6zALQ46`wv=6E8ga;GV|7d9n5A!Rs$FZkbTd(@LBQW-7GLJ$MgVmYb7V
zLh;>q<4q)t77i_hbvhRv%FjLv-smZ&AH*
zKjcjs%{_?kv6$nYD5+<
z3`+PA+|@3-c$sk9J@G;;if~@zMF}f#_@($)`PGpx0%gRG3tC;+Le`59hpOdDS**$_
zQ9|rLavbL`UzK^NbUt?KEcP5YEYE4JedYQssZj3JcL*lWSs)bBH8HY(Puvw&;l(P~
zk;;8$K7Z*7x(yhPzgDe9QQ>%(O8g`tK1G*$gfh4G$3J+IYT<+
z%avWWrT)Z^8%LDSFHg&sEQ*Bj;!3OZ!sTn=p|H&A)VdjB3hSQgOItt(bziCC1VTD}bO=FEl^&Q6FW6p+;_mX&p$Ie!tZckij@geB-FIp&Wa
zKMRHXXBn>3qCL1vAV=0rR@-*3#6p0I^FBkzAZgYDXwbSd?mc*D_F7ffueoYSp)*D-
zDk%B!;|FfJ-W4Bnq4DbVTN#}~ES4xy6h)V>oBwUoX?J}wFZF9wL9xPaIIFT;j0q={
zA0sEuK>0dN@kf{5ICJ5WN|uW?*3#WGngps>C<9)L{vo95$eArO!reQJLv1slN$z-B
znQud&NgdZnV0xfUdPIJB3^~Pbf*-v8EqY%47vaE~SB{E`n<54Bp<1GM*w{FQc<+g2
zQ0DvZxE;B6E|`=_Cq8rq=5?Gu$z5Nzva&SAb?45V$e$w%qDApcyQ1>a(%MBR&d1MN
zj?#tm;JtVjH|aS7org`4%J}RVQe)iQ<)G4-GFd43nU*MiY_xA!U0Q1wsnbp>{z(T@
zr_&%DJ9`No#iBr^k_9nq`8v!O3k4LmwQ}Pwj1bf?bt{%cLP1G4bLl#)+O%7y8~^0Y
zlorX8Bt)#}(NMK?5t(GI?MBtgf>bbFhfkAvI}VwI%Qsw6yFv+6Dp5#Uo@Yvv0(rz@
z0>^k!UceHS8mGF2Tz`1-H!ceJz$))T8|YYs?~9;I?NsUUa{xMY1HdH7^#cb
z#u#A<&&3k;zG3cYt{#rzz^a*8J8Pt;C4xBiLeaFtCSYD6sb)E`}
zi0iKRBvntpc8%qHs61Q@=s9L4=8f!yD3LrGzua)WFBeE!^&5*St$Jhdlt0D!dsWoz
zk#!W#lT9kR&1UdH+3
zOQk=pxjS^|Akz{iOemGkpT#>a<>SYXm^Et_XgI`qqf(qa=Ik5)({D_R7A<66(?6u-
zV;`t&r&6AdQn?tQl+0d@ku->;E25ARjAJ4_7ZuN3VBok&!$PG$sluq_uTi50xWG@#
z^f+!iy;ozRGM+Rv&6_uelamu-$BvD2=gvv`8Y=sX37WD`VkZd~Mfko$
zrGLwoEv4d|M+(!A9z80lYUr#yXU?25-qIiD;bIIIU$}t7xiSskfQCqFlI95XI_fR&+COdAdHJKGJ};rY9YxM&kl~~^ZR$mk~oUR
z7O6~KclN1FfAQffNHIetHovPWEI8cg;T%Gmm1cRUrkJ=R^5`5p@T
zoXr=Xq3hbKNE9u+ROWYIdrcbTaj_u3pv;}N?=Jp1^90!vM-}ub;f0bo3@ZFo+R91{
ze=L<{7Wy_Ex&1b-3&X(jqWpoIuTXN@NvRj9ku4rRi@s8^PVuVW`$gHNVi6;$(4~wN
z%5$kpS#?F|Ql4AvCl$`FFF&E4z~J@A?`Xa9veYNBEgB%QzBF+oNo75CEB!XQeJ}Fb
zt3yll^UnnPLu~n<*6H2Wx&2<3op^*Pd+wsT7z3$@s1xb_6JKj-a-bIeP|vmG_#>1O
z_>`Me1CWI07e+9zxc5Za;c!Q1HQXv}-Ozj}TgNck!mtvqVy45nFO7N5x9>mP;?h%3wmQ
z7K(o=qW^B1Q9NwiZa$@FoW`x6Hxz5_n5OL(mRYgH!-*E^;e>U0?xa!0s`N|KY;uK>
zihnAj=hRLo)47;XX!1$PJx?BZig6sgnNmpqGplA=VM!7KwHBO5QnBJYt5zDhwYyrD
zIPyj~^S}eK@ZqNDHr634F!}TP(|4>A52htzf#arVgO>k81TS1wQ5Ss)l$~|fq=efk
zt>yQcSp2J-Fn2*HAs<>Kp@>&61}dJ%JK|wTslD)m+&S?<
zC?7(F3WbA5Pau8jlsJ0wwBRs}B$fPsG_EHs?T?_iQ2LUmeg2Y_=+e45`gC^VvMf>{
z57Or=Ar$!kpnZ$RvJF~w-#obwF~o=Ps&yO13-4z%s#8N+*w?OF5&edb#iJ)r(V}4;
z^cgk=`E%z$$=?cj=wG4)@j-?E>_tn_SXi?s7Yg~N?Yql<9^STA_KzrwBxk$2T`D&T{gRm|%C;dl^my0c2l$o*MFArHr=R>U4
zZ{2}ZDUxC7>UCm)WDT~jSt7XkbIZjIYv~>;Go0kgmIdd|pO-IbwxoIB2NJEcNQpJD
zdHv
zs&RT05FgN50tbOwf|4t3obbZpv?^Au{u|{hRI+JDPM$b{+?i5)tIkdH$4c(?{Mj?2
zM!Wv_qi!V(?c73I$Zy(n80UquyFm6#7&Utds+1~>r4!wD@Rj{7FEW>H1Q($Ir(&5(
z_<o>Hn
zS4He1{ARW7235+ScdG^%GIar(_a239g05%oh@MhOO-u6r6X#0Wm=XnY;^YN4#fY&z
z(=dG_EE?MvHOdzAR@rOJOd0@ITvhDpXY(S40RO4s9F!!rW5X!BZB9MT@v-TK$*R{zvi7*r4lB
zX@O6_1EzB0&I4)vA4yOJ(img(?4^ZDW%n@jg@9O>Wk
z=FMw0`>0W)q~9J|VJqb)Ex+m5dd7?y@|}g2?R*!iU%$Ro7*l~x4jn4<(&CyHi}TExGfQ6}RQj_%_PI%uCQ{K(9EKZ1fnc|D*+|VI8=Sv9}K-f3--&EKJjs3WI
z!krUboM_mvp_~_T(Ua4j*qakB25~-(!o>yCg##+-xyWKF%J6~8ovskbZ-z|=r6vBdnCh!Q^XLQ>F>}|
zP@k+m{Y3JaZ*7xZ7!|tdQ-mQuM(=c$*4C^`*a2cjt#ra4UR=EY@?F-^I8Q><&6z;{
zPazBsLy3iNT6L$7A6e)_{)eVNt+ZKQiM6sSPuh5q;LRavxPo}*IzYZlzDslCrJadu4Oi}{iOb3y-1@jhWd$`
zf1Z=p_+f=VKMF1OyO(g5VNs=|7vc#SA5Ps9yXQq)6*IZ93-BFr{H{9~By=xKr*gZI
zz)b3qoP{1qSf;D8meOPflRsC5@|+7UtHlBe7aB^XjV*UEaK0xWd
z6H_`g6-TT2>z6%$2YaOp;@pH4*MN(-=8C0`CBrL(K`e!OATE?pY6(oKPm;PJp{wI=
zh-o$5bY?PxnLpgo!23yh@eQTL_6UTweU-XKM?~Pky}KwVl(Ak(@zgP-GN1Xl?P^+$
zH27OkyO64eit(d&-ijM=R9Ow3dqfNy(gQbp;gm6arO#J
zk7Kl;bGdx|rhNIO5?@KGsSSpXn}TRjqhRK^;qLjo3J@bxsUsF2
z)XTY7Z5zw3-Lw^5`iX7U4))*dp1*jdPCXpCW5JTF-?9T8
z`wYfQVWl)<>@c|~!n}REw3D3r+)WWKY*?X$_$sS#c1rB2s%^rb0C!pJ-LT37cfQJ)
zEmH;z?b8ii1?3Ju1|JG%!YOkW!63oePr4$n(p5dC$4J6LjynoyVPz}27Fuu@77mps
zR8E0f+1y?{P21Jywnv~|_BaR?;(yYq8m;MUt<=}^YWq3!PL2LOuj~V(PG+W8Thxd^$mDdY?(V|3>%6zsP`={ev
z(#=q4SshVWLX#fn_WeiFGW)4`c6J>xRi^V#DDh%?O;~z2h+YN@#`M9`b=xI946VPZ
z&^IcDV+tQLROVCoKI$Z>dS1MKBhRh<@v3c>?b@t1)^6X6W$Sm~_ab>wuTm*%Ra#A>
zLY!9d6z{ZctgOrMsaUvXv*hfrS*{rV9Nou#frI(Tdu45awo#9q21)Wfy&r*#55Tk+_9O8@^Gvv7CBNSOM=QJ
z-SC;aay>c@nJgDDtT`}F5H}Vsow@CyF5jF`dF<8igx1fbxZyh#g-UO}k5foZQ@(t8
z58p*pj*>!032A|-%%|`=rKLZW@>JSVX;1nYCA5s^e+r$QvYt+zI>}uDyx&r|K*Tbn
zN1+m)3VqVF(6XM2B`!QrxQIYbbqbZ*+|9|wb#w1*v})(LDC{F?g=n=;p{2W5W0VvT2U}R^^Bpo?L^mCQrM8O*od<;q{9{5tQb?SW
z3i!xvK?x(2`~0M*USy0=+(!`xlNk+Eyt#hv$XB`xTKkg;ff6}fXw=COAAcTsDC;dI
z^c74~ivco|g+JaYOMM4_8c#wJqRd<{yZubT%gQ+XA)^DjxNo%4@5juqpLZ@I!
zsB@q)pF*Yu3LU-1kLq^qy*3x8#pMSWB(N63zpz*4+Sg4?7do3{!Mtx4#V$SY;GL~gH
zy)f)F73z1snbGYi%X(w}&p&lRKg{Ik9Fcy9iw0Z>+9z<}A_y0DC{8gV2tAnFB0+7@
z=ON`%NZn8mmuqrCi90klS^5+j1!
zjhx{(!UDbUq>~Ja2U2vg5*OHr^HTwhKZUnVk>3<&84H
z<@|HGrD=t*d@qqE1{T%N;GsHN#Vg@fC?(Yezk52vN4RkuDo=F7*tjSVU(nSU<>!6W
zJ<41iDshL{J-{1pZLuSuY2E}v0pRf_o=U7h^S^gRDebvnk?*HvAaCez1jgKYW-778
zjh({w`A`lYCWNFb;;xB{;-N?U_@Jz_=z`35RdCUC2$>j5CtM&zeF%j93}wwd7a%^5u(E!h2OuNx^A9ckyz`f1f>b2FbfVc+4cpJ^xmG
zgpl+8v{0y15(w)Kqwqt)C^3I3^BEU0qSz%N6vuxG4*l)_?nj4S1JP^9DBL`;&myP4
zu`XlymH6o9$I*L()p=gQk5A>jl5pYNt~I`_T5^!4@*0mXUwKxp#}K^lO1#sy%6s(0
z8L7=ejq!Vl-{j5$mCj@7V!uAdjPcgO6KBpxk3l2kg2}{T1HfGiO1NVpceX6}
z_rPI9jvNX31%*+Gs`ZdAbt;v`{Wy8{0^x`_6ncufDEGv|19wprtyoL?4It+}cRVPe
zW86|j3*q?bvoeZl*3wlub4OY#@tU-S@B^h%*)nLFEoy)%^GPel-#96=$j50-e3I8%
zJ@%SbTP&cw{PaImC|eqb#bTgRT==!WdS3mCpx?^-*SxY<^f6SUN+m3+l`K${`9Aqg
z`XsO45D)y)wB%dYEF)@HuPW%~&PWOlTl9*Hw_(F1@v^EY;+bVCQ6hVOuX`&=VSk*5`^(-aBm@%WnGjkaSwxQ0yAu?t%=dk{Oew6?>fuB|XIX<%Cp2G>YJ^
z2>J)$e~y3b+-10O^N#ysjcFUx7OxTc!n#u$F`YgW7X7mYiG|gE`Q|-jkxMsR;VKq$
zN;K#sdHTDztm~nwI^|1(ix}5#-}9PEq^SS^KmbWZK~xZLdcwKkHR)H8*R??*Yp%Xm
z;+@i*f|Yq$UMYfASyP2;B(Y1@DC#iy0;V>H%D#kvx6igA_3d|YrKhd!;{
z=_t2Hj~;UI!L&c#8540fD%aVL(Z>Sophdf}PYNG3C#GdoUwNeEJ)e8rWkBj9F7j|k
zv{A~GDPuhKGR*$Cy1IJE$HfnGzEHTcgTh4_wFt!baL$jp{2nUx;)LH@3S%#_T1_S;
zP>BEi(leOL98c&G(uwbXD)WyDC37WV;71ynJvUxTij%Cu;+@nuq)MSuSShkumonh{
zsaKag`9@llA>~s_B|ACv$)!&^naRSbYK`pi&2_~?oL6P3^fw9_3CxLC{iYtsRF`6^
zj5e*;*tYRnRrw+ce*sDqHasWyK84Eh??3fFF$Fh0nHkuBE=VkIl1W;x|0{e0(9a?b
zL>D*Agd!V`N5?hv^p
zl>OEe`l0Z{eR0>=R~@zu0v>;Hwy9Is;(!Ryk|<<|U&?UrKtSd2E1^K&+B&Pd)wZH%
zVR<^h-E{E4w6Y&Ieu|`saY~#}T7|QogU3z?-utw&${V*`
z1&8|^q)6s0&rSc(A+GDa`_kf@R^OMe-9XOb6);aw_w?%6+BCb>^W;gB2&$fEm^XQ}
zd%ndh*MQrzDObb}f}!Ik%l^|LFD=<=9dAxThYIasVc^o$8|JdP$9dJZdqxo$*0-DF
z?I+z%m4+?l4h1^FrQZQ%F<+{B1DrU0&REPn+`auDa;JfNgb2_d0G0V82lPZY;jGsv
zRL0YHK+Q_!F}z=QnY3Wp-}0C)bqZrv8LnBm9I|Fi?;aoK?0G@o^GNtH$boob=LDVU
zrf^P*5&hjhrkT<8?tQ^o9AB34D&4(eGXZl>cFkn-R@D6Y^AP&OWBifVqq4$Vg?*Xn
zw%BG_cFuXMZJI&eN3B)klf2gIQE7CfJ#o){?7#H}(Zuef(!yG&bCJXzK3r5@9{&{_
zlnakN7F&@&h{ei?f|4i!iiyQyTS}Y!f>(cqE9T6ZEk3N<*`#VcufcO(d2w{>)E*79
z$3%83T)8&6h%8j%li#FI@>*!lL$p2qevKW>N!#-BaSkax;kdVPf#%
zoH?p?%A2cz5(?!c?S^M#ir~rLw6~*0fGq)RS9X3AxhAQBV3BrYS3Ul}@Lm%GzDUnL8#E
zeD@UAO@4U_o#;-PGDZ5`A_qRL^~v{6VZ2eO@TWstQpS+~os=~c(&JD_os%`I+jkTB
z_gAb~Vb=PcdC7TCgC3>qCS^{B3>lc@r!mS3Ec~HETq?6kF~t9uFJG3{^TeQc@7^9t
z`DRQWDi9~m&d$=Rp7b=zqMdy66mt6$Lu)c(#0W{LL%JP0Hs)?dI>99^5$mRqMoHBv
zBMg7&?3k27M&aTBDLbf`=XJ;Ni`}9a-
zkZ?k=>y`W_4B|-lLhI;}6;nzU0rK5{5;_iY)Klmzmef4t*^l*8X-l00h5Ylh@HYy-
z*-81*R_vm0ARMGpe^P(ZdEuwzmzrhfnl%u?WRld*!R=--+Eh@V%naybK*p4
zo-eU*wUz<#W-f*ËFKSn6I$z?xshpRDJ
zI(gkJ76*owOD?yhJrXyiXhH$4wqU8uT2qR;ZK#7rPm&H)Ygesc
zwh6xI_ZT!1(Zx1*a^lk=EO|aNq;dP$4lgV`E(mA5PoF(U+&CUT#==K%TTD58{FGEC
zlYhTmPr)ZCb`Fpmf5Wz&=q;Sha$$fU6)igTl+;30;!m5m2t&qBMh)Qvx3XwAf>>yK
zAgE)=51%?kGNenB5~QywQKTSf9Y~rUYhtZ!({kJ$z&81xj%!I>G-vU0jF>oGw%1!&
z(>E0g{<0;CniiGkrVfG`$%PRvup|_gIm_3qcP|%FSn*X7T>qD@UPt?018{VQ3;x}I
zSdRUVM)lmOmRff11*gP`Kz{zl!k>Xrrp{d`eGV8E_+zKd2K^4?%AOgV?~7Ni38%cj
zJ}SIQH`Z?qd*@m-=C>5471@vh{ji{_=NC0gWo!}8t)VW>Ga{Y1&w3?3$?H{J+)CPf
z#Z|fGb7)y-@v-t91B7FB-cyE_P3EcidEpqn`HE|(oGBJ2RY{5MS6`v=vMX58Fuip8
z<*9%lBKjANg9d{K3O?8t;$GnPoAqsVdR~L)yz+V=AM&2vyKtnvo5IJp*N9)r`Q$h0
zle}K_$)10|#=f`88yESe?~`;HR2unJY<0g^{fHMkCW_?Ailr{wkxB5|)AD?@aPqo(
z%O2D!C+Pj+#FANW-F+yz>ldxrg2SgSfR1(PFxRINLEO(CK7B47x2AGVVzq5rhm${_
z6gTrm^g_W}tJqiAvUDw`x3chV6SjmdqOzLO_|iA
zfr#p_SL0Us)RBDlbPC*|aSiM_dK&XqZa}-i6VQFsbcs{PCNjv_imitehzd|Xoi^3@3
z!-d7LE-j2vICSbfQaB}&@$Wx;l32AHI02jHkHh0<;(KP=Lb>a(N~t0q%9)U#7}
zJ>;dQ!kPT%q;w${I<2Hh-BPPoEy=%5>YI`!OG=A+DqP9e?v?O&Nh%&L4A8M^)TmK0
zWXKStNs~tI2%yFJ)2C1IfA+2dAd2mKpWl1NQ%pcCBqUT+MC?}Vcy>R#ySqD{T|B!x
zpWWTvVj&)4IO?_QN!$1ykb(>&F!KEQ^Tm0_wShJU4<0PQebD^qlA=Y6MoI?%J}8n9To?3B
zh{*s8eaX^)$&w|4RtcUOxE|`%sU!Lf${#3%Fn-|fmxAL)+JTV?#@q)+Hbvni0tSnN
zR~_InV%{ORDBi08ZV{xk41@iUKXchJ6>cz1X1ruz1!sE9hA=D6Jr+Fs677t9C%`&Z
zF&+SmYXHQs?0?QwJ}@4H0Q3b;0wn@?)#VaZU3uB4)}!iob>{m=U|IwN2K~^E8_ziC
zE@LS*-9^)e`We-}weu0*aLk?uG?qbjj3JCkNAorh4GrNy!{DmldGi^KX3vJ)8B>b!
zQQ=5Q4d!X@2k{9e)Ntq61}?PaYM8)fU;z)q%Vx~a--u%xlw(RCT>_&?ZGnY60wts0
z%~(0Jl$8ZwsF1D=oWoJOA5dG4w^p3vE>y{BOUwD%z&Z?Pk>|?T4)%}h>Q2~48WeDk
z_H_2nK9EHn@G%L1LTJg^`H7CAw9ZCmO
zscb1h*Mgw3_<@24;J8HD(x6r~nmK=|m_#l2ohDTi%Y2@>8HCP6ExYs)-1%*qHl(XJ
z{n%?@miVDiO{V+k->riH@XuVhOv~nQ@GkzCMcT$~yQp*jp@PRA{g}UeHF=e)OdlWJ
zre2-eQjNdb2y1j}rq+Q%0>C@^FCBxl#C{Cz8~OSKv^W6DNA&MOTXycH`E0encJnsD
zf3L5Ou5<*@g(pPw&b>6U)3Xu;fIc|+p*XR2$s_U+ycTf}Bsp;OB;{g-j4?8O+;HLb
z(7Jm+HBTpwzjJHVhj>)^#_Yq~G6;6&Y@L*ql2+z66{$YFXC@qYC_r#L8^F!K<=S(;
zmmS211oFl(h>m-?%SKkDl*~*EPKWYG#V7IHm|Vj}C^((v%m4rkN{+n%J`1#O^tetpk{r>h$>c0p)+}4X94a5JO*EWlU}1l
z9zg&8?f+uxmo+qS(p*;FCs(&)Xn#^(g%SnX%6&WKui1v)T-i&8_U+T88v%@-*sr4i
zmwiX|p^^QBLXy
zmWK8%FDGBe*p?~mEq+g)y<{qzU9@<^PO4S0Bz0?U-m_XhKY|B9BMvG&Z2Cg+LmvRZ
zFP=D1&5eKX^hKsca#nZm*!insy^xn_3ceIp$!PiHAv9vfV)CliLbL(y{zW{-SsgN|
z+yB)oef#z$@aT_VAQ(SbLoZ#rR9L$MGzT|3fL=*DcI+tVZLlrw37(Vd63o?Cty(2`
z*r6zZb+eC;56zxEo9Px)09-+%z6dMzF=NJveF5wTzz#sVbm`Iru5t*h^kJdjw5e)N
zhqNqNvIr_31kja&zXI?DD6CJwTfcSdR+=$mhEV1M{FbcbjX6fCUHP?yG-+E}f+ru!
z4k$7Z5Eq^fEn2jo=FOW6z#3i#;Op0xsE$e3I7|DK+Jd*jv}w~sU}6C05b)Xqn1@K^
zrKL#J^9LRSI2TaBfLkAmfz_*5i>00_I+GDN8RriQ2E5tiSeXaRjK09W7^dD+_!z-wN%_=dou
zd?a(4k7jTdmiG`~DwArti*@mAmA!5-kMGTfk`vkZ5tstRo9EaaimMcnpnX|!{xSK?
zyg)&tRNum)5dVp)u534=3IK8GJtnmwEu-pp7q&?E<2irpRX3$70$gqwP79>nPhI;~
zSJbhBalt=GW@cfSQ8zpF-g1X3`KquB7(6a$>>=_2h<9A;N3|EIp5d8+BQemA^I<4Y
zRMznabT&;GfXD9&dmSu47fSO^1k=U-MRb(bpHwl2Eh}XlXy(y}H21_Kp^O<-)s^54
z00F)Aex_CWGu)eFe04Vw+_w3O8zNs|L9Ozrv7lZ)jtuhqll!QXA|HW;hgEhF>jE2U
z;KA6?E&7W+Iv~J?7BTerl^Z(?s#ZG8wf!Fm_RNAa7Y&bFA9bKfqQpyeo<+2k|7(H`XzQjL)-GeG4lG-
z`hRSZ4s9$r=H1!NNa~A}Z|oA(Kabyj71~Q(eKNKkc{%hA+<0;>Jb8Xp>lkmiRY?j8
zL*?Nq0V@C~Q}ANLWWrea@a2a(fH5?`czH!7;0T;{BMBWlb0Mmt2J>OHk)17fLT{Zj
zSk9D8OcMR>a)8|bevpT|t5H_Duv7J6o1
z%RGF@K*C@#7Z6AenKKE%U%QUJZR`i!m+(Y@VBMel<-(vltZ0mp=cblXy!Yk`Vkf|(0fCR@aSEh%qso`c$(Ig!iqfUaSM<7Y%gY*CZ{55pxxP3_
z%?oBQR4iI~eYg43`peI#S%LH%km);ZX`G$>!aj(6ee^trrX1QIREy4p(wKcg_t5WQD(`VmBc8u?YN{g)PJ(u!;xRu3gTUUB2Ocd}m4vSiM7{+P9%L
z(UzF&gwKJ%7yt70TQ#MQp}OYoTWTBqhqr>n&o8B0Vj0TT@1`HS#uNSrw8Ioy_g$7sv(YqVk0W^;Bcz+=#x0BDX%K-B_{aA#-dNJU{?
zj$o;^i@>O`*hdi4s04sH6cO;Y1As2g
zyS3}JZQGWvU%yVLPoE|L$N-69HLoq^9HYFgHyl$;%-?Ooy>R&WU@c#$P$B-Ni2Us!
zuWRfv$?IsZWB)i#Q06!}IjMns?Q*7Wqi@lrtwEm~ZBN&JKjMux$Mbm6tD}G1Vt{vE|F6(pc&%Bc_hs2?8Y`d^jmj`i`*V6?S
zOZnJWdOKTbQ!KHqou*Mh&xZDX0ODtJ&8nxg?&Rq1TizPH6k;r+C)-lakk5cy3tF+2O2OdsSnY&(}fXBX%
zkE$%xe21A1Eh?FN@-dBOfL@ot<@0D)P0;-TdlaQfmPlhl4N4FQf<9>r_`maH)_h*!
zI6%D<6XI>3f|eaHUumeESEBfMUO~<9^5rWA^POVi@>92|+`sCfi1rHT1V1p){QUWI
zy+J*VZ5-gS1s?a2b_b9#TULii=~4oeM{=>VQ7Q4|X~wRDyIn3(?OR?}dtJLNC>Tsp
zGMKY(sV`_FNs_2R4yn8;OHdw3v`41Yi*8vdwt%tq#eA5rNpnSYT^l!#0q?rRy{^4q
z2L}hjFFJznIDM2U2x1G6R}z4|#(?=K4{%xH=9lUlTB09@0Ah2FQNCYor3I|rC60V~
zIbG|f9+R#%uzy|jH@eE2k``SsANxR`pwKW@j2$my+q3BXWb96jEQeTj1=ynBSU~{U
z5=_0KOVE}fF;v%3JzaTG^&2q^^~Euas-L>{XDCg0K7veJZWS
z!x8xHf4_x0EycrwQx55;jR4P)HNd;*?ieptT$73@BM=)#07^Jqljcfj*i*`o($(B%
zBGvbQ{+T^bh)QsL4M|Ee)t4bazld{$auh+dbph;BThg+udm2iYBC)iov4rKnwtQx5
z6<7$#Me-8TWi|BWJ9{s&SCL#$URPOpndq)(U@NF6PoG7%y1JfnUPMUSnqNUm~a(A(^7?cL4dKJXV15r8^?GaEt-_4)Ti*};j3#Fnm0&!~uRQQg9%f!@K
zQvmrcNat~nO^zjCPHRs}j8T|r~y{rb#qNq5-&2vNSHqWzKAQ`Suz0hI<<6Oc2l
z1y$*ou+WcAKVqT|0=2^G-ab=CgUx?tm+Tw2?@$gmSLWL^-`m9C&Xi1pBVZ{D{TFZE
z8Rbf8uroR`-?G~+Bt_DQ$(D_k;pqXW%_K1vyu$MnKVu~wduM_E2QIMEbN3Z*Ct{FZ
z9wS9+59Nbje~~uD66MY1^4Dlkwvmi2K3~3hFH4Eus&Zb5;^TS!^5qLF%3|IkNx6Tc
zci(B(!gKo5r%&3+CU3`-g}%uhiQx*LzkH>e_W~$8dkt7c;elsL*%2f
zP-CuOVm#~EBih9Q8(%SX1R~21tK>|&xXQ3E#0Hpl_#qwVz_o~~j%e&XnHodlb~}9U
zm56M9E#y7(N?UQj+)Nx8*jjWGJa?;@>v=mzhYw>uf7ETl0ebcErkQ(e=F|r;XbDSZ
z)%6$`ugIuJ55ajWlqpHm#tt)B_}+s+DpIi)9p1K%iWkanu&gnexMnBJ_=o=GAh$1G
zy`sYTz3J}}1NHLOA3Xq&KySZ!hDJ@ENjLoNh?oNv%a*1NE&dW*_h@V8{3Wzx^#)<#
zTp({A>fWv;l`CDGuHU*t?R!N8ww3n{R5&vyjFczEiFHd|*C`n^!KBF2VjoUbb}s0h
z;EY83S00%URBQJW{PzPQOGeUk%>7jQ@^%AE;%OD?v5a9ocF
zg@V)|%3po_or*ceh@Oj#=Dl_)-*SGD1s&p`vuRSLpg(;
zx8orFvuX?7x*y0~_^GLSx#BdUTPsRtf#GYL0YPvUo;FqWw-b4-4j7t
zj~_oy#fuj=xKC5c8f+*$6Wsa;JPVF@@cP%RSyS*;NA*ktCw>0>`Kfg2(zIZ~0`;!`
z{r6uHi1^5nBUGY92@zDdPMtdH{Dit$uyjzDcFstZ8&L!yd>1FFsP`
ze=gCMwq8_-gSDC~;#-u1$?m@STzDDhXF44OfYpn|Fqi+d!$EZH{wqq&bfA?uAZ^bw
zj%qAM>d%N>{`BI*SDIBXhm=IuLf+AUZTHmM(#poxdDI3=Qd)dmTIxGEK8}qgj!6OZ
zEisBp8-gD?25_brA6eh}*#tly)sCDPY}(q^Qw?@SRX(cx=|_SY&<|!1KdT7h2(Z59
zf@^f7TfRsZxCoNcX0;!EWloIIRkQL-;u{Uy^?+V~`o`NXbe#b#Ww6;YqD(akuX0ze|QS1kncM{$Rh=P8EUq*+!dg}zmqUw80<^!z63)Wmg8~7I=
zW!3zd`~e$5ccSkYU7a22*p3a9+asI4A||DeojRKVeP5b1axgjCXQnYzW>Tp?>(lk)
z`yvJW4Gwusl^e9+V8D53>gb{LmV=}99y*4u-Snf?-2WN#mQdf}W2tAy)~swOK$~{#
zrdrM0(ZMZi*z+I*HK_HcM&Cl8Jf$hV3#f9rG8*Y|Wy4a}H6}zcAqa>CkbK^$$7IWt
zI7t{RAGSNt;=9Vvip$=m?RB=BHhJXfNQG=^`T6`ijo1@NlWSz9pqF20{gr3brm)I8
z1%WI_?+v0hMKV(Htf{H4S326|_mal#52B{5v|M)X2?xaeMrGV0sDyOx+>As^oD)m5
zq3aNo^Qvo9j7y8Db7I`yO)I6GXVb>ZP-%cH=8Bkq#Lwx6#?ZJpWTwNbW>QYKtQzUM
zHcTn2tG+z#%h#{eu={ZO^7R`{8PtWIzj{M`#?Pi#?7=pfx3vX8{TSb6)TmlH8q%dG
z?L2gfTJ#%B2q;^tLP^?x;yktNKaNTl%1d*{^yR?A!8CH_63WbRi@LR_r`=qf+g4qV
z=3kJOYA4POpa~szb#Fn9FCdC?1HyDprEOj4<
zcH+^vm~$xqH1H4rc~osU*(9TF?Y$U_F^#`nU%S&8dzpDOh=wzW503t+4A$3}e}#^A
zFQ5~PY0ZjO}N@?FWL*G*f
zD+KzMb&`_A65m@1K`iwhoDjw;I>#9K3pzTjD0LG22;d{gj5w_^Hhy@(PGE3n-|ZKa
zkt4V_@J>%#u05l{6`Vw{*M$sVdD*9=7KJiW$91=iLIK)i
z$j#6XwB}+M4Puaf$(hI8*F-d>mYc{c<(igu@=IkogINIdH(hy3B{{8eewCH#S5^>o
z+jN)iKK)2JGg{`QD0XatKe&c0@jXq^4qs>G-T{#+P?PyWoBBP77gptn$6p2
z{LuaaoS(^HcmW2=JGN{>joWpj;)M&)rK>mSz|j+IS(b(d^yozOYE%`L??c8-rWxag
zQ?4B5gZh(a&leW<*e@g%)BcpH-hg)QKSXu@tgLbN4;?>6KYsk69cz}T>3%lv*dwg<
zQE=t@%~T+-CtLUr67|ZJDo$H>?I+(w%W3JHDb&4vD|LI{zkjC^RqN89mCDi3KHb#m
zacRX;9Qqc}b?F>71UUMqo;&s0a-Z6*yUWx&zf)BYYYvqAOk+}O2Ig;vgeam=9qiVS>l{nq77XqRynV_M#Z?am^vrM?cKCe%6T?zybP5F$YQRj
zIF`+AC3e*R_U$_jo4$~C9zIQ(GuTqIT9s(sjzcu2cUwxq9%I9%FQTb~yE3TnOig=_
zrebWVfBDv3I&|tHS*J-w{o4ISbt;vnu;(u|%9?X@N6ucROE>QdWkMmZoT8~~x9`)^
z&AVw-&o-1inJRF)w2ytq&r$he1q4`+w4&a*X!({sbc^YN0O%h$b%C6-WTah-Cref6
z@R`fBVBPk30qBeQumwC%RM)lkn_fL+31$Uz^g)9L36R^~-kw^vY)PwEucnC;CsK+O
zDMa3^S+hjY+=dMs3hVZZ7cbKO{rkzr#zy4VuV0^@Jb9u~)|>%6bm$OWxNt!z6N(fm
zBAU8#S}-f{P}e5+&MuJB+p;5Vg)5jmQ2)5lqiv~-e0(IA#L5d
zHNgY=mb#zNj;?vK{5-^$bzm{$_XW^CKgMF%Be-F{47B;`QyNmynX#A@wBXbu$_p%}
zaAxYT_Li`|hebOq;eo|;EA7Boj2%6D_f@nr_D?s9U@>!!KV&Q>lh`lD8Wha67F=O0
z<^^Lh=`@D_r2i|nvVSiW3i%yW-T?sVS6>LFffbzTKb|}Mb3)~T{D+K^RB8lF55QcS
z@ct5_SPI`sh+}8WF)M(+F{7lEiE;$+5u_EnFv>bVqrjl;{wi>OM&0AgT4xA&`m2#9pdaw!Gh595kG%P1Cy9$~GWx{4{tsV$
zCVzfOXuF=Fz1z$)2LYJBosVnxt>+>N`XW{U*fQ6&M+U?Ej?%t<<`UPCp|Lc!FRJS?
z=DCZF?T;ztx(*ml%hqlZ;Q8~Huc&3$K4MJzc4;S?IeYQ4;Lyj$)f=~H!`7XGQl~xB
z*?`8TX~&*ag)Q0NfA~npPMxLKZ{O+My`gc|zGe}$^TWrFRHAS}bz5%EPGTLL{@Q?k
zwF+hF#>ssG%*TF?pE*yU;OX4DnW*!MU3@{=Bgr4^)^26R*X%uiDM`JY!N+u+X?xZ$
z_Eo3FwG~To=xaSnEH(@XB>RS~Ojb6{LCF*U&hDST391?Z`mp{5z#lSF^^u63%Ga^o
zsd#4b!+MVNR*~|N9$hKGO3WMm3(zfoo@GpxgoElPQ8$HWqIO^Um35#FOgVO+t@k~+
zFS;hy>IN*ZVu^ENi8gc{igI3cjf!z;mO3YvYR6EomGf+9bWF?Be)pMHTU9zfap?CL
zGlN$AyO-KEs7}vczo9k*CyMzkhbaSp87Q!$th0Z6{>`ZFrT7)diOzu;sDbzbksSd0yQ;CNc61Kx=56hms(
zrTh;(ir;k7~9Fqd*TVfQIHrP7t!!$QTkxTIGd)CuB{zg6ds4Ra)F+^@
z8ND}9XSQp)WOS5ibmRgpEyRXFbx8nHS+O-LT)4CaFOf9N)h6-fBOh*O5Cn__;GZXR
z1lW&s9Ou4gAF)cEOV|rdc$+o1#WwzosO)SJOL%pI7Fw~C1{WXYxF+M8ggt#uOV@0q
z!M(butp82Z0*QkL~%UTpzO|F}GsVPRpkWXTc%bPpRgOdJY;^ARIP=sm1l
zxpE2M9_OKO;lfn5Y+1T>>lRa0YI>5$8vvLNx*J`BfV-$K7XbR!*47&L1SowVP&#PK
zM>f!D(4YaeZ{MCumMp2AEr9!?MT^pe2@}K*1rQSeT>1%JEOj4<`!WFxCa{>bm!1f&
z`CdLw;)ns9PdOB%_vi%{(d?3SE49Aez{Xl1%u56fetVaQj
zJX5h4luyIn0&*-yGRAz#u@IvY)Q6$~22)T>%sBFpwzJ}5UzdCdf
zKtG|Lc?%!+fB*MitYa*PHk7@*sDw*|_O+yI8iRm-A-2w!9@bfyPYv>lEmt$7)c-M3
zmU+_TR-^Kvm8}2eY2d_%eAfbk67kbc4)RNNKC>mhcG=(YWMr{ssvWfV9S15CP@FAh;?_;^)7+uAgM(qux`pa(7
z@4O#x`xG>*j+^Cn5D#0zxxhIwSCsQ=?g(18zNOBqWqW0~HRb*_JU+(c;^}IM`#HW?
z=-&%`#PgB=$frnzQA*jO`3(a40Ls<#k<97Eu)lo$mNKQ+J>NiqifqAeXPZ{%DN4n2
z2yOZF`3ui^BhN)p4gqcT3HCEGtWu>4rBAQAf8^;{PpxJ6g0xScK8b!|T@p~`Kx}Oz-+SGUC($yyhu^9F;)|SuVSj-Pouo(7cmP-T{6Jek!&zG;)^t2-wc~;4(
zqVidO_OS?(yN;C|EtzU4zKf;so%lXZ(LFu|(2wrODfKi)AU0a)OY}G?*z#Qx8`iEG
z5^y$45{{Wmn|haSY}JpSu_A%?pvx)Ztor!C3k7`8+3X^JyrknGm9WJB$Qs?`;r)BX
z+F&REpyft>@M{&|D)j;R*|{z7
zFL87Fg^%*MXrcc$oE|5oRSNatr?ZK$idQGfR+1-=u=0^U8f4izS97?#5`M;t1lk8a
z45m^=RW7}7zkU7YEx|)TTP|L`F0A14dwEj*n$^S)ytWf&{6l{1J>cz`i!L4BNgKB9
zV%Ow1RH|4Z8aH(g_sK=w6h4|}Em%s`E0rf}jtvl(1=qACfFcWE$9jhmJoB4d)wp_w_T|vcx&DL>u~!SvjxzM#gZuCC-WJJ!IMY
zNMEmG;ylAG{26-<=8T2(eEP=0oODZ^bL=1QVN11T&heOA_szTaqKOnthbM`xwSK@H
zWuTas#Mt>rb{9v&&seduNh<>3eqb6OE9UWsKxqI219%X~1@n|cr!P?sS0}R1kd9i_
zt4b~FR-rsqThPA@>PvwB%%vM>z@)iUxm01AJg_r!>c?G=K>6`wT|9@vjXt#VOxd!~u{?9k28>c*i})XjE(p^1S%Cj?-xD^DjOge@q{f0_ex|
z2q%#CW21#WB9=qWJ$x$y;9~s>2U)SSASmsG13@~Q1=szq4qh60pu<71##_wUFUc9^
z=Z9oZA2IQh@^XZKOoQVeeA19J%pr%i6LWNQa8_UCKrMI?L0gRFxuCZq;4XNvA@Y|e
z`17HxFvdkpRQ`#7V8m<*3X8aEq3>X?Qm~vncR>We#n7L>bj{*tU=9bL9Xf7`&UUl8
zI@9UB|7zs9IXkg?a8kN@*cE3m#*KS7LDtv%b&h*nSw(e)11FYsng=>ilr>{V~UzZKCbeKv^?G1^+#{M
zP?h=DX=6(dD#F$`u|W_QrNjDrbhf8=w7B+Rt7D1hm2zG!b2a#$2%(*~UTQR!FG~s<
z#T0XOF+kex6wJ{$0!QhZdX;rRyBSBDToLJ%RLsKdNWecGKt*;s})Wy-K~y
z*i&`(K8%has2$Fkxe{M2^c^y(EQC*8x1(-j!cRT9Ys)5G7;=_ieEQ&4FPFx;SM@L7|{SzlnQ10Bh#X5q@
zT3Tk$o=t-X57yaiwrtr1xGvB1^z;;)SFc_rP~||59z9AZ3s7G!Ns=U?{{8#QlLVNb
zCQTYq7ogkAmoG(7UwNU788cE)knV3?dAaz$ZmIhqp;+kKhg;YKmp}kq$YqXSXlYq|
zI)o4i4JQp&i(oc^oZzHa+tMHxw;`(94+Wu?Kd}3`km5wPR
zVEPCkm?J>%dHj(N5IYHj-@|r3FukEDYUF2-y>~cAzjo2IOo^i{u(Gd`(}q@@52fN+
z(~v{jq=EtmeUj4*ENfgUY}=0WYwBnDZa5PoN)CBzBE}
zM>bc==9-nJF~_|fQ`sctiFeN-mIt(p6z~?bH@c4A#yawKVP%z^YFDd7^Omh5pAtox
z8podcjTkTb3XXqByPg9G3W*`TyHc&Hm1x>Oi)iA^xzwn3b*9|8EI@q6ESW`}14mEN
zkg=1fSfTs`4}{)B$4~~lbZp6ARkVdahmW5UA0nlT6;Y=p^c4URtcz2`%n`tc#f9K^
zwB+0qdd?Qe;LeVY;9TIGm@CS8HFpG6^-l)8<@a17s*pXxt0_}zl@*pLd$8^Jt;-^GfWOOM{uVxHfWkEm0aqXB;d=xf7jz$HqE2
zT$~7i`i!OPDIL!Xk|s$^ea6qRcvk_+50+8i$ai3qMrFhzPpeY82&GAtg8EJLCEroK
zMF88CTlZ1trnS|@Iu4yoP)H2y*i?Y|Yj+%=N~H>sQ|62!aPQeGx2Q|gI->l}!>2fY
zPfE&|&W6_QIH+zTS+b;5y=-yywz8s(fQN?%WzU{n1dO$_vlD@6yLazy@qqy-KVrlP
zo$cD#r~pBE=AVE5NomujrJg-|(!zxc$=~0fmMvRG^1iic(}tjs7&>&QnipSc0}2Xo
z?t?QQQmRxbTEBk12;y6=+$7bci~1=pa>5mdXeiH3IJGtb}Lo+!MhnEQSpl
zhwZp;R9#aF0gLHV)=6iTv>b;}TU1z#EiFIyn2NikWh~|o!NCtNVi;aRdUF6<7>D-s
zaisFuZ3GrGe3w5>spU?AjKyr=IE`(JM$L2p3g^p@-csLx@6x(fc|=g$Ne2T3(}H&J
zUG07<+sX(SGXjyXaARsIWt0)n8iCjdNPsfppZeLT7lY&N*4`F8@>M-}!p>G_f3RzA
zR3hOt^3QQya2g_&<7bMDtL8$19L%=GDpg>kjLZwSwnZ*sVUKo>@ds*-aBF?g04;0c
zK}}cOq;hjEi`|sVmX`YR3kFg#hYfWvm4#*=3#PdY+(Y9$vPxF+Vrm`8yoNbw>Y<0!
zX04xyD$jZi>f_@;MP-rQMZs9`2V-tIfJ}boV!^=~(FJf`u9`oS+H~tj4O(}mlqpgO
zAbr`|jp_he^6`b==ul<+c{Y1dqrE{i;AiJV4|2f7U
z0C{p|*KpyN^5WXkQr9&sgQ5dH@4fw!)?9o_=NZI?0#cE4E0%8Tc)J_
z%XDLx*{jK1V_cjIoD*|JIj@nGh|w8S1d#7m(w-W5+eN3kx{hVftLfJVK9@Ea7
zFX+iT+#4xqT-B^<{@Be|pV3DC-Qar{bT~EGBE4bwF?ZZ>SEL2-$L3MKcXM%m|8iMq
z`S~z9aOahXyEB0AZz%k)J$}z|GlIl;t>k@vEw_UE4hwyA1y5H?+|PLb
zO7ll^`}8a8gTiL{<>oB5J2d9`2J_}uYtv7i51@W%
z*A`;?+|^rxH^0?@@#=i6q-S3L_yFjuu1E9d!KHIl%r>a$}60G&n
zKGsK#8l|qbfB$~9SAjZF&&tf1GsUwIfq$VSfFPLfm@#AY)JQnV7+d-Ii|&z&cA{H3
z4%HJ_OhXUqx#E}iH6r47*UUqT{R~|&!NE|PKIgusqn>aoUG1mS6@#NL~S#XSXPiZ)=jDSz^
zLe$RfBbAf|p#9Z-#|PfIhi}M@E%lQ#9~&Zz!!on>qU&^LNMQ!}f4x2n*7o46*2UMR
zJ$^VIfNvSu;!sx*Uhvd`FW-Y-KH49>XbXI}cyS@xyS0vwb?yEAitW=(B{cm;Vmf|S
z^|Fj^AD%yd&aS5J6y|p!x=s9&df<&edG-QjcXMSPOy(NrN%{%i_Nmd3Uix1Z0RJ!P
zm~5vH0w2;x4!|XW_p+==tD#{}DS&zRZLO^-t0MzH{4pe#7cXB?@Z(VO%H<(Xj(6+!
zxr@-{D_00E%jQU$lvZ@AOHN9o!`eh&yP)0jo^p|D=CYwY%##l=8V}1NQ*g0Td=@P_3RF5JrM~GWd&OHrZr>2AZmO8Ij_?~Z4vjXjFV2Qtz
z002M$Nkl*$P`;MwFyg}vN(+W=n
zDF0!-j(ZHsu#z(`lRpFVxc|KP;}&%RwmY)*S#^E}>|k%c7azXU%>e}j;NN(KAA2wU
zk2)34#8iC$BVWER3pu2u)y+KENZ^s0atHSq_Ir=#7>)DU3Bv)Dt(DuBrqyyKLw(jh
z_~_oRXs5=)>%vM7VL|0@p){kwL@pE7g6-SoB*
zUOBq@1ggIW!H)&aj-!1BVU4e=TzpL1y!R+=K6aHhZTZ)%!$Y7}@TX&$Pz7^zShIuQ
z9$fYKVH+0h;MbRP>;H-|LjOR!V{dP-H>j>D{rd9d3!OiIUd&xxTwF}48gI)&G4l7{
ze-oDG$hdLi=*j;1G_bsrInBiSx&n&{Vl0MZ4N6$dS>CR_;40l6RxFlaF`pQV`OgF_
z<~?IENp)f|QJs*7jKd%xFBlm>o19Q&(t$uqU8*GwA3i*x9P&>+&d6WJKXv!Y&Xf^|
zxe0P}I=j@Bxke`7`g$K%O>oIU{{;;qLL7Io}W
zCJ{e<^_BErKLP-}2eU;xsCGsU?9JM2qF#eX32S;->&FJk>ZBRi$55X#rne*f^ofmi
zTEb5e<3=Dl7ClGUO7}BQR2$^Aqk?RGUy}j22?rjsg6|tyTm2sceJYE6fc+3y*8`*n
zpfAZf2JE32E$En%TCToBEtuMCP5r94f!8zyFa}$;rm6-aZ$c)Rv
zE2=7k{E`5CoxxyXRS#>L-$3Qn0k0zf{=0bJu>RH8XKmnDQVd9K%a?GSJ<1A#m=eA{
zEn;z^oY!xW0DV)>>qDkq0hJt-PDMFJ!=;DsXa`%k-w6FkBP%k&Co%VdD|z=V_bG&9T1e8@Ckq4mDz|YfwnBrVEr7ipQ5zI^#~_B}d@Ns}g};>8u(spyQIQb#ocz%It^dmw0!`>7Wbg(vlU-Ej&x-tTWMnD1d4ULYH7w1P{G`p#f+7m$a
zm)u}0)c*;ZH<)!6JRilV90+8s#{g9Bhb1aUYZ
zW!zL0DR3E0#-JX6eMw+#FM(oN*eck8t%hOMFBe$tzy16}@YfI76`;<7v5NomPXtiU
zP@mNmc%>{`VW&%>>X$qn>nXYKF)26~hOCW^-M4aHRo9|?yi7f>N%^C(Gh3NA&X=B?
zSW%AP@PD!O`NA`yB3-)P<@?aLZrSr{>K#kewn#>5wA_yhPQ64>bd}}%ZE$&&BDFG8
z>}6)lut|r4De%>2x)$<59ERN2)|;*KrF#(ZgKVr4sc8Z2_*npf7lS0loK?y=Y!?ex
z%z+E*}o+R0+uBZKN$qcbyx
z6D>u#6T{v7sm)Iw)P)94noE^i_hA|WRxai7pv4mh{?tAwJ5xp=HjM!2X!`W&BPeK~
zC@NgIFs)p5DP~v1$ayRCA?~yB|>f#aF|zm{c^oe)d?^N2N2$2*l+PPyqe7
zeD-2d-+%kxZ#sF-esveYBfq(gr=Yb-%pW7rtXf*`J^e`FN6W+?*@~Dq0?^!pD#2U<
zn1O3DE=qyCo^)vITKe?qGxN?pEqL`|JMkK))
z;Ccc84Q_Z-0DY;AU$q2v;hJVSXe-lW9bwD(wM<9QA(7v+2tW(@s>MnFCj6d-`}~jpRQIekD+(IqO(*6*u>RNA
zXYGB1zdt_9Z#>t;TqZt9{@mGV|MKbdiD|_Wv-K~m;$u?2e*H$*?>?Yx&JG+?N3mqI
z#8|7YNAoX8OSNN(WBQeCfRgLz(W5*^|0H5UX;VpBqHlhFe!}7()H#a8t`QL5m;8)3
zSI|y;c~3+&7GCzLXiHnK7{+2g@LVL3g2hC2GJa7$pq*d!9sG+9`KRty0rY?BaL3=B
z{r7+WnQS@wGe;vj{K&@kuk(gGRDh$akNwld*eWqn4m`PI_XRQl@QOm;ej#Ue>*~S*
z0;}gv7ioV3w%up2;2HT1E~3uca`h=KW^2Yftf0ULeq|0I+^?Ktw78|!jheUuE5rY^
z5k(W*Kyg^S;+l-hl8ght{VZ9sdIQazzl1_>se<+Vto<6=Pv7BV>Dlv_B3Q1x95_Pr
zmaam>`gW&!HLA+9V(mJvX-imgN9O=RLGp;{NDv4UzkAObv8RqUY~*`I#n{
zC1*ov!ogs9l6u|6<`3G{-c!9DpH>WwQ9?YgxDF27eNDdXagsVb#>B(dDl2mYI*Us(
z{pcgUH&W7+ny%ux@egy=kKO-}6%(HXr#(Cn4t7!nZUY5R|Lx`--F?1ip%8(b4tOh;
z?yNAtTtE`(o&^0p;42^s1$Xny3U*+3evZM)!_Y7)(V!FUT{@LY6v!j({+BGLE?P_f?mw=v??UXMFlk_C
zjkMUlp}HP5mFm&yv!&XJ&c0(~9eEI3OZTsFE4#t%NTE>NGcj*P=ZP|u!v@S3I^3Z&Z{6VctyipscKQ}aTZ)brL`Pu|n)
zqYnuId7+?a!FeqUXVw`TF*Y2hv3&rv^Q*ptjqPJxE~fzcarx{ieOE?6HUfj0lWxn^
zXVk#kj)N1YrS(_Bs1qw2^0JqL^s$cO+iyIl73ZI5yd0ne=(6!H6=!$*DYe|_0e|5C
zeK3eJq)JNdi>bJEWxG;Vl@W-M5%~7)8*SOSm(E_iLbDexql!MI)cbk(=rQd-a-247
z+ex?Y`O{|()|JpiSoD~9&bgun64w_%~xT4>&`seu%3qic+bcbJZmn6
z(RrrR@XTaIn2bh8HZfm!1?Kp6V9Yhe(3G9ydA9P=eQxgnZiStq}B2k&{f8qG}t^^!s0|ZuaRk1Og*nQ?70CC6i_$~
zWF-+4Q_IeUk+*#clVa82da&$yMQf%>0DBbdVoUtdRh%go2e}VrrC4gde>3nsir-`A
zzegIpTSoG|4SE0NyPz2|ZBMGwxKBaj15bu!JU0NCkAQ7*!96`@-$Pa`1koTK->~=J
zgr^Vg?;#Z&nR8!%pXKdfy^(i%rpo%Noh-KVvm)bH3EN9rFxP!Xg23hYug7bm`Jj_3G8tN1i21
z7WKNaqKrU78-Yh`NOF`hmF;Xad7llazHkgPNkTZU(b8)?FHJ?Gn1<*bdp-`|<F_hrK8?H4ubl^++hbO-=-+L
z-}7-o3NR1$Uca-n>cSHm5T5psxo-Pyy(g$O&T#NcNsh86*pbI@UCZ23k{G)c=R*}h
zUp*q0U-`nI^wfVA(dO;D=`l~13+MNySrbN5rVQ!D&Q@>SN-H;Pp({6T2|AnxwW?8z
zM)gEq<96LBPc9Gg4+x}vhmVt2ZVwtgun)~!x{7x0KSZ`1Y(AkM8ZeccT2EZ0`oJ$bBK1XDn%V|b9+t2;
zQO>KT!UyNU)bol!t*{thcloIZ0*f|sG9UYl+OAkNNi2I_O;?PyVGjLH8}3r=#WzL2
zGo?<#9Quv~&x>|NGxGiVjtWh?B$Q?KIoNd>*EGUQr7*{@(B?kG{TbCgjC-%loU3AA
zu)d#GQ{`=lJ5{O+z<)ZAWuLA7RB7IIQP!C~1?F>bUWlPSivqGgH9c)*Tb5mo?!Am&
zBy3D8YJijN6>e+_=
zNScHeZ`euuPn=_mdM{qzK}ixPrXDTp(>L~}o4#Z{y?y_I8dNJwqk6OvZPn>CShRcM
z!ZpIrDRTyz*uNtc^2#alwg1dpvz3-@+D*@1y(U=5kMGl-vSd=tkqq@&yB$-v?*%@h
z{_X#w=Cv!ET2(0@M@L}Wwr#X<;X*oj@+4)=nw1dvwqnJK;%Kg2yGB!|PNjnf50af7
z8%~uhN#nXs3X5Tn!_I6~->hIJ#$uk*{pTO4QVv^cS}-FG*?ym~m=BaY
zb4r>~CtElc6U=dYK67j=#$p&O$Go{2Td#L0u3Le$^h}7rVqO4?$--Dn<*e)hEMYMp
zX~D@yjK#bVECgUO=u_6oB3R6|kauh}`B)CvFB!ti2Mi`-&9hcViD^Evr3W(8S(
z_AwhYeWxSc3kWJhq=7CnzIpuNxD>JM2qdy>b4qK<2>hH8c*74oSfAJAhj4$Ug#j?%
zV%06WHlPqCPy9QzUE?Pn%pKUe{KMD(XzJmI0`zZPBolq&?aPne)76lBRM*Q+yjVar
z10cSXYg%f^LG_>v`1{~P`n!fJzhpiM#Q`i@oAZn2V(>c}y*E&}M>WcqLF0e{{v!|8
zY<2S3(Mo>FRhxH}j&;u`UP1uV5kRs}SqHkr?aV$N%$9I}P`$jWj}v)ghJYe%t58_e
z;#f;C_pT%X{IY~3Mx;p=aN~Abe_Qk`Gt*kg0e!q=3pqvv#zR0s1jV!B4}C>qZUh#s
zT2JGr&tV|GH@)MJ-i0gHQitAyY1^uW!b-hGm)=yhf)9-y+*eqqckDAn0Pzm?ndmBm
z{2R9Jq#~>YsPkuKTDWp8MGU(v30OEnG<*IktcEy(Y#KecXBUjX`(X3wV(c7Q!=83SFlk%0N?hyc&IJmb}^fzB9
zFHex6*n}YR$);90MP4E*m+-26-c_pyHnEPmy69eLuvhv{%&6%2lESOsIE4l&ecQNTif07Bvy7sb#RCDii@
z(0d@iTW|8m5fu4xHLeZlua0%~qDwqaNt)<4HN_3CQ*ZXTJKMvXZn5D3sCT3o7f{v%
zbU)TLug2vJgEXCGP+U#8t#NlJxVvj`2_9f@cL?qTcPF^J1$TG%0KuK$?(TZ`ch0#}
zHUH*EPu1+6?su*AEQmKsaN!;olKwH09i1h;LEONho;7xoJ9VdD6~fM|ps(s^8$alC
zDA#HT1!xkCH}Fmx8tvs6Tbd(S`Qlnq11ysZcN0n&PKTp=Pg^d_JT}=>Yn<@E%yeN_
zZF+>CxZ;bU#@drtn2wez#iGCIenhu(fIhc(X}+9xWD%ZofDWmf=+X{hbWpSMoHgD{
zw{!O`Ui%W~>pyv`A7S6s3LpeM9s;RZG{wyjydpP8BbO{6XE;GB(QR3@d$MJ@PQ)}^nX&JpUS6{J%uK0j=x)&+o`t;$sj*PJ
zVzc*8-+0P(%WHXaq_b5qc81_Oecv;_a9MprCg3itGa13?wb9HG_IXLFtgI{r;!(k*
zt#IM5AMajYzB*BYhhHvM5digfaoqMB-+|tnd}{fOX`q6SA*B=uU2z~FBV%v*t<;-I
zu&x8U;T*8Jl>)|9PQzX>@I`>9mvgb%qlnLL;j01A`a)hF&gDQ5dFLLt2N5EjE%uu@
zA|2gAfWB}oPm@imasfV{&S4%?ihz&}sE)D|;roD-mX4NInwAO(u25aX0~NIqy9>0X
zHiQ4r(KP?^372nMj)VqHHiiT|<1pAy--}8BPQ4!s@@%=6AaX<>(=(H?e9}I=>EhC^
zt+}4JBO((vw^C*kAcd~w!WW#vie@l2k@01JiZlt_j|8r;$GRQS+KmM=-
z7fvHTQb98zEhu>?O{hF!)F6$xhRUxS6U=8|cT~q(!vUVxsHkd7N&uhET2zBuMGJ0y
zJJ!r1g(@KaWM}y6;D{XoGf98eJO~Wfs2oJ5c)xsCdoFPE9U%N|)n>D(YCJ5lEfjm_
zY=ezj1a`80Xq*`ZR%jVnh&Qb;(#!x%*E{y-{)p9Xa!aVOK$)%~`6vqVZ0@HH+Z#v4
zabJs|b-lj@!Q74rzA&f$dP|6hs(yWM-oUkc$6WrU9-l@Kunrpo;v=v=nciWM&ETZN
zW=YCzz7(W&#T)DW89jz=@xJpl5GK#HNfx3v?A~W2MgCj5?$`R68KP5(kK0rhc%1i&
zgDFz`Dc!w;o$mnoCx0t$KJzgfB?O;H@ToTtvNx*@m@HZ#-ZU4r<&K>e&OH1!3
zQjrk`z2gl5JuEBQ6o-#->D0>!WuuF@NfO!=Lv`7ar_EroHLR=`?X(MgjM7eB>dkhM
zMU3HxNfap_hsB#qM841|eNEv|4mz*HO(&f9ze;r~`V1DHg_<4lff5#D)(ob*QrFWC_r7=L4D{sOlG0(E(iEORxt+nR-YvyPaH@uRn
zU)D<+-YXH01tXnz-KtsnsZzZt)=Rj!0S-{IXCkbfqHwg^U~(30Y>Xr6ypH*~AqI
zKZh(ppn(|$mQ3S4nb&We!=aDf=P*4!67TuU3(q)gHV7{}HTJAFq4vv*l^Sd+4odTH
z0(Bb{&aIyWeq{P%dP(7C|5vGr5Ajf+bR8=1qL
zIJ$qX93QukF`hz18U9FnW
z7y5zKHR2+oU~7yRN7pW!#sIy`K5SH>rv{`#L^qceuszfz_ZQN!`zx%)ggFecEg
zG{iX1s1x#jB>U2vv2Vm5XdEZ7zquSOXJ`ivsZcOCd8;ygeez$}-53KO1E~Lm(Xr4s
zc%_>k{7Ly{wq6AVLnH!kyDUbjCV&&B%M7_F;Q!$@+)M
zB5K+VYAj20JvoE5DUO_t2V(tSQ{fnLhRUKhRjHK+Nx5z{XEe#G;u#vwvDPw%-{7F-
z9?4T2v%b;{;h!pKg#B9#639|Rb8GSACsvv2cABki`Ez@YbAz1mdqQ1&1v9mOy-ZDvQ1-b@DqEn?ej&
zPC>%GUzPw^(Hx(Y$k!UaFsex!Y9kiH!F0o26A6qbI8{2mVD{pM{IRN6P}}e4GO>t2
z&OC9I94XDnc&(d9QtkUoA?-$A*mXoCYOt5Q^?1skvJBt{CjS<-_vy6rTgSW@iqxEB
zCndq{@t0GfezK*8^0M#8VCua7m+cwD8TXgUzdKlm^&E2!-`M?8cI$g6bFX(XAaaqL
zFI2&BB|DT5;vR#QzKUq3TvVHYi+w7SNQtun$#kk(REd%f<)BsYE&{-?Fz>C_L4wTj
z^z`tKu26>HxpXSj;H?^^pa394@?AU(MNQ{#jds&7K;frm+!N)3Y|jEZVasB9+
zYDTqd9BW{z8!4iUp>#3@40H*U##!{R99*46P2XqMqC9S6>z%gX8N`G*Vj9|PiHT_*
zicKfqGj409IUj6*G0RbmYkw&)u!G+f80jboPR8rCY&9T5BZlaYSs@TP^u$HG(4#6b
zz#7p_=<~fgp=fVz_9xM1R@eS_BVeJr4}pBObeRzy`1l>8wOUR&m1`011rA@0s{-um
zkt6KG8)n&@*k`b_(Jr3WZEZQ3NLU5x)uw4o(VfDDDLsTQwN@j@8ZoVfBnJ#|?1xjM
ztD+L4G~wtM#0+t8NW?{hAV_-%VQUH1h9qjZYUq5F$G#8;I2A?CUHlxkSnv5~d6{nl
zRw$y=^B;MjT)7*`RfE3%
z9KmQ9e48ax%lEKU0yM=$OF-w}qTY?%@|4Sj
zM-rip=FDqb6Pml8yWH8c_dV9xVIN7m-VwJ?yXP=1ozs2HyS>YgYmV(*!K1tpZJkAj
zF?XxyGgeS-Kd>#Jx$t|wA4(=O;a%#9_x|&6>ebhXq=%QP6C(ff$M3fkGP@`Hbg4S(
zz?BH9AsJBs_KHJ2&(z;7fHF_J&lfejTN*$i(D`?BvR+%j_Y#R3_T{5m`)Dpbo-V}#
zb7puj&t@C?ODzBQt%aq7WW{89cBas3fd51}ZLk{3C1k8Ry1!5X7lb$cCOz9|sWO(G
z>CuXOx3Z`{_ZwIJ=}AjAFIWT#T*e_b&*1g(l7?2Gu`H<2X^j0N3T)@a#`dtN-*{Qu
z56yWqcH`E6@}7y2H4H;#Dk6w!h!?3AuT3|>Kg^^{&=dPVca||r^{p#jGm4KMA1K5BL|_@&jFg|
zWbhZ5gzqF=R5qW^^ciECkkWUT;wR?3fCB2-E?96*=M4-eN2kJZDOdQ7$?4jX`Lav2
z20qurZsD(AOD8X9ROs!5W6$hdZWV&DgE*Y8#M_&#MfC|?aGjoCUbM6%q`*zxAXpg
zHg15i^k)c^)0y?q-fay-ax3l%1ESUG!Zz1mrHXE_)=+C6kOJ#73nCWIj~CRA+_e29
zrPnPHWZL#`8BP2YW!da3QXx1@&_G#mcD;1O24Sv)ciwL0w_Ax^nIL%FL;>{1c}q>l
zo!-tkAXP1qTZOfA-4Pm!_oJSj-*3h7)vfhyH*;^
zBf*+AmvCqq(B2lYM7{XkF0krgwz9#%kx>UXgp6aTF_MI@ouB~p;_|RR!sHGtsQv9M
zvhl>bH17=>r=g{R$ApbOL+GkMlcIJp)r74i-A_oTwXbl#x}a%n3)T^Qh6R|np+HL-
zvVAw1OGhTjbP~}wWRp+PGsg*vlO}wLOrP`X|M=r9+jwl8ndR!nJ`rv?zEIM5|Dw`9
z-vgI{W;{SD6kHq++wA`&6DMU;L_&fTIoL=0A7T(VKn{nhTk=v%ZSlA+
zgm8J#7v~04sKWUz`|g*J1r~V^sKi5yl5((%G#nVG<+Wb_O%gMocQko{r_V<3>}men
zhk8XI`sm3%6svpH%QQ?9o%b@@=l@?SI8cRTX4#r;MEC{>c~d~eAM5iWZsX-pytJbT
zX@i}2rEeOf_DbmMTwO`&f}edVEW=(Wj%^h|mXowG8dLQ{ql^Jt)6RIG_Z+1hB^Xnu
z17*NLYeHCbV;Lip0K_8+)GSW(U@o^E@BYNODA}86HuX>Xu|c~!e2!(n>Nav0J}1Ol
zW4^h)JGimg?ZvXzQ!tk-TJTxSn9eaH_z9NL;)3mA>P(?wP7_^Ru0byaDJhTkE2K{R
zWj3+a+?YUgdl>apG4!Q!2FN9Q=~Il+g}2DTr~h=rFQu2c?4_K@W2e)(UO1Mq-(jEtub$WZ>#<9Z_uxs
z4kr4x=e>e7#5=k~i2_=Pf!(Z+jN1%{i)CGb%as|0As<8$gYM{7ajAEMxn=^EU|6IB
z#ss{v#IYkC$)6L1rzxrJcI{GEFA=g=E=_~)gbTjLCPGna|CuQ)mrrXX+b{@^9_W1D
zAEKRD|HM%3^$m<9jZUgV)d~6>!+0FJXn0+xdw#ceCCn=qk{C@Fea%l`xB{dyDrt%U{;i6H_7SvLBHDi3ddEdouXd{-P8R)*X`}jua*vn2md)csO3$^
zt@tw=8$Sm&Q17!UusChH}f06Xv;NYfFF
zoy$u_U`~W1YNMDVAOn;`VY(#E;72DjX#C3S06Q1egG6+j{`dop^;JknDDY%tZcasW
zA95M?(RX(^8Xo8E<%Q86_EaJg2m2S;xr@!e<+P7%b8~YkuZOdIV91Xk=L0o-e}9h+
z?nNaLT;23|T3O6zR
zP$Xapn?XKUx*Wn1b^;EqD%X`gZ{I?Ot~yO&2-G2@Ym#ZYdX4iO{HgpNXJh(7Du!%)
z#yH@SqFcIp2`OxtD$sB5x62fUI68Yv#t$%{i@)Y?(3@nei@||L3?`D0E$h2!vFXQq
zK}$P;F87<1<7XwNw#Qt}a4@&n*!A!jI@qVTW%KG;-E+T*MNa-%kVlNFpjsUs0teH-%~HR
z6f|4E!7bjPK@~wyD7eBrl6CnIesfq|s5JF==SbV^*8EY&{27$zTMYFej)vVSASxW3
ztBVGwiL2Pt`IidjQ51jv=dPYdy~C$6m`^lW-H~L=1JiMh>-wJ(N%0xx-iK@P*9I<-
z)5%Oe7njw)WZm0TERv21f<-dWG7=F_I-NVhHwa>1z>)JiD<4I>&dV+*fnH&%v(0(7
zAKx!>2aJvK67$Yzv~qO5`cQ;kwLA0=>`{;13}KyJ>C|ukGsS_`h5uA2`EcN-F^()B
z1a61C_Tj%9v8?jdS5nN-XV&7G$>?~4*T8et`yStRZxdNv_F&K41lc!*Yj1036@uZo
z;Egd}r>R-rSxR&KGufbs^jkVtYR;@2Ewu-`Kx1G5*jCfO`aAA5n?rUpd`iBo_ijRy
zLi`a}OC=CC1;AvfTF~}RM5(3G;6l;2tK8B(aw6ky?p9U!T2me`;*C#RSu*c}4Iduc
zkQlVWB24lFD-U%QS9eanzeds0TV7B5q4P_-NyI^p_xBDDySmV1oGa#F!o5>8S?Wzs
zyB#8d1z@uc)TFJcFlFY`ss2Ts(&uNDG3>rvh_8Y2wSq08^J$6_{+3jze&gyK&m
z9{3#68D-9mYqe_K2*Lpq$mK3%*ms(e{SYPljeq=p!C%kAb;1^yagR?iTeYlXc+)!x
zu{Jf?XLApzb1Z<*9bVX^NTF!f<#|yMz0{3~frK{Zx8&&nLJTQ(FVvZ
zVw}sonE>e};l*t-|GZXYg7R)m#+F)CV@yL{g&)##jY>Vf
zKHGKvBHbVbkfu0!fJ)1=?Hh1(fr1AORsTA?SO{BmK
ze{Uam)-mC0T)2JkYA#Zi)b_*v{H_l(Ol1fwxH7-0ve#zRfuTbxuWgMzQ8
zYa(^{iZj0q!^(ndq1e(gF7B4L3beu%CTSO^um7W~X{WVt6S=?j4+!CTyFQy#nywR;
z0juUQygFmoN$Jl7>oa8}n0@6*a(*rBt
z6!_`?pO3}wEnZdLG02x<>--&2>Ug#4-}kQF-nV=DmP`LUL#HjnHXenGCEm^m3K$IE
z1pnMmowS~1BzPb46o$izyuk=e_((thy9A`8jrgm?Lc4qV{`!yM^Nu7=ON_B>ip&@B
zK~QedONu7(zcx{Gj&$}y4U*JJh&opErABXMTot{)c8{NDdeH7zqZri#L2_|heem!X
z7Zw!>;Vi0|S9qU~hek={?>yJic*YCr%*l
zolH*6>jk>Z(z@{Gdz__zN&ceK_h>&
z=`u&EZTt!^%ipHl|05K=7i_UF^B{(*cT>RhDU;kLf;cw!oixgP^Tv`K)%^t@?*eAN
z$qZ5nM#1d!Y(>9t+n~;y_-^Jgk2|KfVD6vGPPCR0N=^%8;GanBXA6=fE)~A)_S#V4
zCL;k{Cioussua0q2XHr0+L|2@tquzdvp{_=eqmhD
z*jepxv+tM6rT(5K0-R(hNJvNmhkRZ9SJ5w1y@i)$w{~r&;NRKAaT|U|byB97qM2_IN3i@#ES~*1T2)Z>OLvKF6!&l_Gypy}#YWA&q9>
z9StoyWTBI9-|Sb=AYO3Vn2SwD@8+Y3cJ&Y3$d4?Q8}y|tWd{WAN5a15E>Q9ySEYmm
z;+4Y*fwNON&#;yI-*o$2yST*qWc0z;3ceKIX)e^^P?B)CEay<+#A~`>1s~92LiPPPs?KIh+-_JbpK-f-4k|tF|i}&^93`#})2I
z#AZsH0;Raq5}Y>lApr-J{b_Lq2ZEF_CO=(rMp?h48Mp`GJk|qUS*YahL7K
z4@L%6;SI9UH@N$E!J;q}4u$w>WP*yOcDsCH&|e*fJ~%&sYUzIe7s5$#vixG_^miw)
zvQK?x!TSn}f6mf7W~U2_Uc-wI$O~Z&-z&|s;`2m!VCV{RQN}%W{y{nU1Xc6GKKJVL6a$
z_V@3Y+~mW$#c{n*aJ1=q`}-UPqQF%QCL}%?Zt*UPZPnV60$6^;4e<11c#&h!T;pBVnY)M&G|HcY&*Gan=V~
zr+Gl;qJ5lfzB_GucRln#Va0>X$IsCFr=ZF-b!cDzGHS!l`&`Rs<7C>WDISB3>{}a+
zf1ti^{#=wO;%tWxGO{?$`4@hC7$Wo5Y-iESZ%<@?vDbqBDjkK)8JLFTL7(?V%hjnQ
zJ#tFf6;fYzVG#MCDM*b^1t~7cW6l_^eNt{!2aKrV$nkX?>)8#c)F5WmAgZgYq0G8v
z%b`@^HK_xjY)a$j`p74_0k@Q;OIC*4
z+uLrVZ3ImKp4La3ClW7-K`Sa}toaFk{|a4=ofQeB?YhShd-%6_%}uJu
z0F^|rSjUYsPE4P;jh2i5H1dR!(ep1-z4Y+6%Es6MoSvE^4xQOumE7bv(yjQ<_6rm<
z^GF|&Nw65xX$3zHLh%NJ>lwAuJ_C-*h;O||sikcE#1j)}+7>-H5Zf?t>cYYwnp^4%1!7PWN1YE9D_PNWXdo`!rQ8}89|)^
zkqz%UN9`zOGMkH82?H4dc0VX&;;CK1svv+a2x*6vYFgf3IgT|dT^Lj{MT38
z;(>%B{%C;)V9BcR(ipXBxE%HB+!4tWOAU`&xVP^HzsA+2bIuyB$oWg4i{vn->VC2z
zq2IkU98fy0RF10ZMNPPg%^T&>-xW;iN}fnF*0DxaqEvU?CE;Wv6k7+m)UA*g2s`j0
zLgQuSMg91GdNA%o2Wg`5`8USXuO=@3P3xL<_Ft~|lf|2#Hx7&ND@<=FL}%{!F;h?0
zQre+7)mq{-mSCq}62YkcAvg(JL*0EOPv@0JhI7P=li!J_C961Y9pspb%{Ahi<>xjP
z{^w-lLVjn>VzAL!`QVQLV2e^Dslt!#8f(6|AxLJ_4*#KmXIVqdxa0MjBfRj`#$+~`
zmb1!j{T;x!%tfG{|4VE>-;k={u`MVlnN&XBktmfY@SLx8{TcZr?022uOTw3-kkcjN
z_?!g()MC4#*^5Rb=#erkpDFM524S&sRyJQR;}hvFDM?mzVJ-I(Q3Kzy1&QOt2=>aEsu(yU0#V7ofq
z6qo1#J5y9+xsRV))4^{C!P-e~E6+~wowjzrBSm36CX;LINXov46?KDE(S~3xXEZ{I
z6jR9GKei`1zaE9Pa~jk<1}k{*VQ+r=#*EZLQN;B1cG)m84)-YNjNj{3>&O!SnEzjZ
z4}4!57!=Lwzt#y=m@POz{R256;mQ{Yqm`Zu#$A20xp8H37K?E~lN$ZVd#QOnUr-0Q
zN!d4WMOV{*$w5+O!$%J2isB3M9Sq?q|
z3{Lk)GvcQ#y{d@7$qY@wAC*`@`icrns&5N=U(g!~fO|+9z7Me3Rj5~euX9=zDC&j{
zjGdG1dx@!OdrHH}Mk+oPm(@1V+WsQQvxaraF+CJzWZpKBs$YO~)=;X|$9d-^G4X;Z
zpM&BqpS@Cc+4eEi)D2#FJQlw0TN(LG?{Hh>IX&^#c1zsXi#eR_!1%_|5*ZCWt;U`A
zX7Z|6*80$1CoZ&TCRmvGQ8Uz`)BA_rb#BOlK#dhG{<06j^`E6?8G=tqI{nB+DHTp=
zj9_5+soo>|@k*wrXW#e=s}VLgXUvrjEC%~rk$0Ny)?q1q
zr;ol5nJ=Ez*n+1wzqZ<7S3drb((r|CsYH+QWO>rXCXd_E@U~apeb6gYT#A{uQ#Vd_6A_dAySID>jxYG#64{Mh}-kRM;N`<+)&5)45i`v!&YG+x1L
zs#*5BI*aQMqhHPW>dtE0@ex?OI9{GC^To9vnPw+;g^h|EN9qC@%j2@g8+S~_Rul!y
z?UPIANr4Y<5F|Ki`4&KG%guTy*G5x}eAbDHz~$|)5wnTXSfvPT-M_n$wvECQhBSQf
zAr6~&A7Pb7Fzi$_*v9suyzeuj-an^8XvkS8>=mMnzncDx9wDf@-jkRJ%_r~6T9!TO
zw4z+*^^@HWGFtl(vC^YN2JjWZ*#o}2=D>x#cfQlugiyfox|S}C1dY4c
zs?Gr3T6)E!>FlHT@yF^+7#9D5qUxcPA7(>O_EVH|EN-s`Nl!fBC&clq5v^@a-Sk|>
z3sUf-er*n)`XE(cZ$ue0fkaeI_A!rI!Tn~U)g0u3c|pgI96F#kAJ2{Wt&ik0qaj|j
zh|pkPX4Sz+2pTJX-uhP0HRepLhu{gB&236^PO*s8ZO~Iw&sUSeGjr@h=ONW$-s}3s
zwomGp-|ItmGZF8k`YVs+MLCmdlS{lq)9u}}oYjIkRshM{wtrsl2#uxhcob&;LakUb
z=VQS+Poc6m)|ZsA;b5pd-LR045X6kcHf2JDXTF~&AfUzEax&{H79rt(1_bOeZ`yl#
z^n)J2D2RoJCk+XMlm{4}XOrkuNR2cp3Hiu^lln31TwlNjQ2=DPy$bZ5~!eddSHsz^Xv2FOQe)
zS{Z~YzJ0U(Ra{i*?dkEdEu!sq46jG5DVCLT2e9$;Vvh~z8alrWZO=h9WF8dz@Z_p=
zBzFVxKQyHi!Dam6FZe#HAM$+clKIW}`&83bAvOLHc+5dP0oZ(u-w(Mbh89yuEXXe3
za<;PNMCh>&caMY0O;H*O+M}2xVmG`b9#cgL@p4ug>iXps
zK)qAK`5r4_4gA*d=brUPUwLQD3bZh^MrLzt#Ay<2Ldv?-nTga?wZ3Yn=a(+8pv`d!9fPnO%$
zrNz^WDi&4XbUFvp&j+`If+6&E`V<1x*8B|jV~AD_OeR$7d6~4@VqD!Xc`By;#-+j(
zb2_BueklA_m;Vw@-0lAStG!_@kXVR3Hf!a4-g?B$`WO+6R5s5vED^+8oqV*EU^nma
z&QH&Mo&eh8Vv74kZz@M$c^J^bq9NEnYWv{=AM)FGGD6DcK1;QN5UOYNXM=JH!1ewo
z8BGPB6kcG}N)Rm7cbR^dNqs+*qd_aOeT4hVOPsShrDGY*w&f+|kE;pdMU$;Suv%;W
z&G2mP;K=_yvzpP#lU?J*1?iQIIesT~x2#z*{~rjzE5lab-qDJx)IZ+|!Tm!&k&rW<
zCvR+Gw%5P!b0R8qrSXa@Wiqk;)@tP)T$7RZ?pjdp0g?vCKQtLuz0jK^Y4P0Xcwgc-gieLWbwfwz$EZG8NsI!-y_l(qj(M*=mzN->
zW$UE5EZz63Ys`fUP7%%a7JJ>u#S31M3b(To2A)f$4PC7=ndo2h#YUR@_kmphj73TlFk@uqATd<~2nF<wrOVxVtIA={(mKXm}Nrd1|T{YX?&X(hW
zoPxyX+bwFo4G@bc(`xw9kD%Xk>3ld@1Qg}X07ktgEx?0}wRn@w_Y=dCwgKRMBEZAv
zV^GOX?~Nvg{O|7#@CwfWXzV{rOQ{&t^O{f{ZEaeOG1s@Z1&K6@zq%jJR#fg-ecztu
zTOCdK9Jc(IX=_zl)S6Mq15(Co;Gh1YbBQB|7>0Z#DPm4qAM~}SJvMF)it}wh{@Qv|
zKQb==0bdbmB+)iK)Oa%=Q#n9ykpE_it(&0kkEzJnH^jHKKmQ
z>v_*`+x}jOHL@>g=Qfz+ZwPg6OKVO7w2>noBy|ushH0$VtkNil
z21+~7)mFi6w?KdQNGKpy!lL^dA5(IAB~oa?(tuk|5xmw(_}ORC-lSviUOMnItfjgUwc1&Yx4tGj3R(8n+XiC{aB2x1^K{-NN0
zd;6tnzKG}G>1u;6_xNWCO?Ww3$p~}m>dOhx*N;w1Fl!^tyIobKZ*$Et`bHUyqaf2>
zDxM#XvA!2adlCUG!icbiDaMVbJi!L^RK7Ly9>~dO{<8QX_OM$UUWE#Ix8xRf`W3Yh
z1SD%yS874evG2Pp_^W}>amu&l9w=H50^`HAo_!Q@!N-IOr%fyek^utLB4Xg$uxQ_S
z1rtMge((JR^MmFkG(0qUFHo#~tNBX+)A9t3FV*GDAxw2)@b3D3=>Ce{A7W~as*_ev;;uV`)&>+ME6*bvb
z!BaMn2d~=T?fB*1ONy@-;gy);^Ebi}VzAO6EQ~N&uFj*&i}&*|(O5Q9qI6~-!o8>p
z4yX!&s#Gg`tkLuxg?_0M_W4X(GGYZaPKd7ULwb<+lV))WQ(o1gOiARV#hh8_=e0J1
zA!J~5!w7Gcn86#$P_2((B^HS@du)`qqM{;1_4`;dLleR~;Ko24zXePdbHKYyK9gIu
z!`1eSux}@5+3jL2onEaROA;L*o%rP6rN-{g7Re>iC=$nW{yb4V+3e{7`49Z|=?jEb
z?R0nij}J_xK2wsW!G0YMFGa{KT=O}jS(+h`wCiw-uH!KIpxc-T#9csOt)
z%onG;8HD$i?Xooz+*dY8x9K-2_*=NVHL?j`;e~vo&Xg}d^7xOeN2OAG(x?bY&zqbI
zDl~$Rtd#oj_%XN(Tcub0+=A4lW`5Om>fYlU#e$A|`e{4pQ2F>6LTB8~aQwln6C6aP
zp$RER)kASK_tgF+s4j$t1cRHglI+O1vjZDL?S#3tQUz#27w_=W%A+NoPy{@)b`(P9PvzY2`WS_)>-SvKxou~Un^DE}QT!|7WXP;ep%BA#v-do#=nXI?1lZ&T#k*&FEq
zFtgsN*>VNP4+taVW^1qPpvr@*Jte}1AFiH_|DiJ+<}o06T#Ci-YGb9_LS41enB85E
zt(C>=XfYQf*niULuR*-pZq2e#p_IJ%DJVsY=l`|G=pXZH(Y4&esU7YI`W@9s%6B{NO6CW%fbzg>SL#DfmTDBGgTz{t3M3nS
z-#u5`oN2&=HkA7_2zhTydyEEPl#nx_fV@E7L?(|)hwHhErs(r#je&4BCtU)+n_au3
z_)?nG7>8>}vLH7u5Dk*R!@at5)CRLP@h*l#`3Z<4asb2(y9D`aqu!buXt^s@m8*q;sQp$MZ-{d&SNq7AorC(lckCkRa^#D6
zL8Qkdf}dVUcbn0a00yNCn^!x|{E2aSoqdm)IZ24pkvOi3`z(F|W9%I>BjaW?PIxa8
zURo~#$^0)xei7fix}Wuyo7P5u;CJP-JU@CpbMYA?iI&7H3}VHya8#1qhW#`#w{tW`
zg)l(aFAwq7jk{L%i0>Gcy~C4N8<}rtMj;=}!!$w4va1u&!>=%|5bzAXyeT1KFlZQy
z7>4E;*WwAE;f3Ze*y}b>nwgy42&!&q3~C4lW9hO)V&ki>vtX?M<#JL
zHy=p29x?m%p_E|#rmyn_y@DFdqWS~-1DNjEnh(whUAJ!*T%HCk>2-eNoeMdnk-2l?
zrQBPq6(AkY&4x$J{kil-)4Q7yq13Fvg86vMY`KJ6Jnum!F2lC`o5prL*#%~+ivT31
zbCgS5l7I6pQASj0Kt$)Tt7P$k1%vjTinQqblGv_z5aa;D{p!&b5BBqJOLaf|Rj4h0
z*<~%;e}3ft;jjj@q^edLTdKdpH4(oBz^w{3i@)vgu#&MDjyI^)3D_y*qe^@%qU3BR
zcz~Jc*qIoBN#N566yZ~m9`vtX7Uqwt9Nu#or-K^IRc()}Sa^04%AyYDbra&K5owqVm4C;94?K87_a
zwJ|a|o5%dcB)~Iy90b!c&uO)&eB*Z&&aBr%@nU>#uh%;;@;~-{0CavRR_lU6z3
z>iikN%?w?h%p_T+%Y{-^#CdG>C&OT@VdQ=x(`~+Z6YpfD86BG?E{s^f(<=4{d2q@9
z$}3~+U&UaEVkG~I*W4aXl>oP9cm`M6F7Q4`tLeM|fSG+eB@8d?XgC*n*qqGo_VMv?
z_i}IjUx*lh56JzK<4p?@*YPRWYAAN#t;mgWPoP{Ra252t3m5Wv`3A_81?*;H(?FKE
z6yOzbTFuioAoyev`wRHDp@y}CF8l!Bx68!(`Z@@=z%)a76VkkKnkD{g887~k&us|k
z=(3ZTe2bHcii>rFhn4tE_>cs?3%>Sfz{DS}l?abIYO`d(sZ%*oRXdC%j9C{lR_s^5hFA!AC8^N~`M%Tx9pIJ}8&&EF
zV4riw&p)^E-0*kbf
zQGMTnz3tcY5sKu1u!H7<$w0nz5nrS8rD%!r5rbJi3pY&vU+qDftX*W4r$;a>O?PgE
ziF>>&icG)j!kV>Vd^-|)`X)F$fpItPqhjSf`A3LOnb
z+j@pdrN?q#{f^y4mxEt^2h}gj#@OY@UMJHYjmo~NHttT#ybQ{=DdzS}Pc_j;f
zMQMKsB{Mqam?gas^A?13PY<~l8AFcIRMLz}!rJxZf!oBR27VO&H8}j%N_c9N7a35=NM$%G
zS5&7EK4ERGSt_R*j=d3TXwKP=hfLh1x~NnW`>Y4E_*m3HZ3dY_xuZfSg;36SNPgaP
zFGhQpf0T-;AdIx)k$Qo@{5LMu`L0&^g-D9T-1J^i?MNzRuRyaS16Sh}r5v2$4!3sq
zUj!@!^qPWCOc3{r0A^aA~+nFe&zN;vr^N$7A
zFmB{bxZ2sJl94Db1K$IxcT-z$jYIFw@h=7hJh<*39}Ha2uNim)vng%(n|ybt5hmK}
zV@JeVUcF=90&BGu75>*XZ5I;KLM|r5`X^P#{{QW1Lu(n_RZNjWVgWri6Ou*}2F<2y
zV{r<*w?+qnMM+%F)B$9u)-NN9xk@b`;>g?qySJCXm;B~{_qtsm>5)s7_`?WPu4(i|
zP}A84SISpPK1URv!(uhzWSh@?6yXcEUAce^!Q^ObN6BL--}CH`G`LPv8g~z;FyB-V
zr~FA0)}XrI!O?aBjcSFK7hk)xU^1h!G?trub=}76!|whnA*$3UE`knpccFS^^f}*M
zPUdCArV{VM+N9>4X5*BxRB?xWCsxl;2s>q=aK#Ut6oEw-)51p|&%Vf)bY!llWS~G{
z9a|*!nWv3&cW3Db(iQ$0?|iJ|t9gBpVwT}A8)ajKq^`u*D-H5T~dwIYs)!dwnxy72}#zsEbp2xC}Vc#W35hvdiNIGNx7#?<<;$p=;Aw0j&T|rhdb}Q#~CybJ&`ixP|ASLxFYQR#`rsZ&N_=Ah4Dy&!ta!H7ErE!
zTsV!TaaTh8*eGY>G#=i>V92&cc|>yYUf0d$^pX^CHq93z%COAz7aa_-Zh?Eh`ctiB
zXvsk&K)N`9V53TBc6?d76PN`6wxa%K^4JF_3t?^K9nX~@p1F(6hhGlvjzA&#h(kRQ
zx77;)!El191c`=t=!z#G$NuX?GcgvAP0WV+=oZpfIO!-94vqJ4(h{pQ^ai=eY%)u$
z_pY~|80R6cw^yW~1Z+>Bwl6>hTKMy=;flh)6UemdbT-hn9h2J6p%2LV(#Un5JgLyn0L1Xi_&w7M0rn-aIFYxQ?g)&8V{J&hN5!m4u@Fe^C%EW*1
z!iN)CTUswDOo(Zrb{-6skSIZ`BSBY?S*|lpj;fRh7sN)I1cAts`xEu?%-E9+KXg{q&i8^lU6~Kk0e40dtE+{
zIUnyl^Tyy2ZO*1NnmeeuSGWrmTFId?*}wM_`H~ru%@H55Pq`CQrI?M9g?(RiXNXr9
zX+&1uR(e`|>%&W=bLI`({~rK+K!d-jL~K$3`+vrQzlDR}VpI`XoOUJdh
z`2@`WXmriY?3nF4{kbPVtc^z3G_R}-e*J}x@+q0^*I$SeGvFr?PVZcJ@Q#-9M$U7_
z)R#^_B7b(H?0g@*r1q;d&=UG_iAQ8!p=7a@oWVjLV}wQiltpW3=MgpWjzc~fKI+U?Y)VGXS#VJXkiuVOu{fMvZ>mi~?`FDUfk^cFbrP*zKnHf>q~
zPxJ8ML;Ao5G=2N_6(*&c1856#02c62_8X!UDKrW{LwUy1unB>&;dB9%^MDt^M;ec>uP+T6G>Crt
z?KeU{<>cbxLdfsly}N)MLOt}YOP4MNo%zb^My0S)-zV^Kgbjad=LWDaFjic2Dwrzd
zNUqhzr1_z6*63p5(&`IO2B8z_;4cum7;FmE@NlIGe3cV62>>%Q`M`bpnRPLH
z*?N7fw?FAyucJ;PHc04KDYLw~PB_I6g}F|m4UyZdbRV}bfI6`w;L=$&MrM|>ga0Dv
zsiqvf&$#*7Sq9$-z8gUmS(X`+#-qy*-%|grcWB)&`GkS#WG)M1)Y_hrZ@^^fI1QCk
zO21=>x0*(zD)iO6KoJRuon}QinP$?EvBf5_vc0hk>ZTeyU<2$!XLcA1^5CQ1y;%W2
z>(~PtwKsrL@{c(j-GVCTOiuav;WC4NK<6BP$UmUTC`*cXA`kys7EVLHtcaL)gq^VQ
z2L#1`4?XL6q{D*#*HY;aRD`4>@fe>zZRzMmMH6UtbM^Bqjo)TxkNbF|;5+}S#*S&e
zmJ%P>cy<9*XJvQxESV)%w1DAmn*P#_T6Web
z>J!9|Pt(T!X%KR!U|vtcM=6@l?O@S5V&Zh__ov1h35xqsgETKg&tJTxUHcB{j9qFQ
zb2xt3KoD0w?tqO;=rZv~zQr?_N#t=zU
z>t_>1%kf6dE_PJE;`~#g&^K2sO{-ZKf?!g7j0`B{1GVNk%RDwD0Dh2vVO%|3nlP}8PODRbt`!a^U;Yirb~
zA(Zui;DKx;f|Fao(!lxc#EBC{IXJzQXwjlYf_JlKniC)i6Y@Lhj2VQs^{HUNf`pG!
zRK9$9aWY`|@Zs7G3pnVNP6333gwT#1J9Ne_ZyWXCq*uc4$Ji%No-8jC<-hUFUViw8MeogLEkVdz1}pM5)omgg$d59!gxLueJP{_ks--
z0FP5EkM5~NOWMPn52>Do`fs_zP|=!G1up#FGoofz0lxzt^P73l1l}-^@)Pp=)J*mf
z0EffQOpvAgchZ^1;)s1$75a`iD^A!avC|v@{q9{p1K}elSPO3Ike9G49E3D||
z^h7@vPcVF4;9U@uRsmyLIOGZK^;=@@TT;wLX>kMM{Z
zSqL9XXW}7cQR5UUhsng0v`z6{qqNRXv@M`?J4gq^dR^Vzbwb6
zw2xE^J74^5NuaBbeGF4d`CN%S4@99i8dVS)q*Iu$8(U%z8$18ITjC<1uOa5y+B
zTb!LcN#y=xlN5eF0Fz^iuvp#0PA@_A7GE<})FJaeGWN8st8
zWa9>jvhsHp$Jv9}`X2B|rbzNL@#O7SDm3+?PSJ6-HEWnfe3Ej6cc4z5qenh691N_G
zZ6&e(aE|2%onu%XeJYwaH{IH|imu)E7tlM25@_y+`|RWvxcjF!Eznj;k|-e^TRTTs
z^vCB7_vCt(^zIDpcWT3Yx_kd2z2*~!x!IyWq8;3#vxisB()L*jlr5T{n$@mMSu>h9
z%)x1ckG=<`s!ny)+>f^3kV^4!bm3jOawSX`ixn$I0RaI7$epZNvl8&*K}gG(F(X~N
zbV*wW%I>pg&kBqF1iW5No;5C~4B9gp&)=jgP2SN;(O$-jYR)DMRKQ
zy8y-J{g<$sC4WWK5wwos)o*cjL#0}!9R>PHo+z%fm*TFSYC(V>6GE64f&+>yKR9^f
zSWDz8=wmTl1ESC{Z;N2<1<(s)N*%X`cn9r;hVW7hJT2C{1i-+A!C+|a3YP|nR5J4O
z5IC)tmq#W=2VT&%dIxpCQPU;W`1LR{pGlihZYJ&D-%R5D^rBoFqxUEZ>||0*=-Eu7
z^U|Oxd$`TH&43$zba!Ci;HJ?}uYDHoI!^4l8ko?($@6dI$Lf&
zjjcFu|NZ0$&F2GTzELHYCTFI@IC-0u&JRzdOAN@xgcC`@csP|qrGggq;L850Dor8`
zp3Wy-qQchJM(Ci=f<4NBdEveVv3?Y3++E*$|LOEJexU;64h(=dl%n8vja0k^*1mIC
zINP@E#zgsk$AfQF0F7+CH?ce_IIqaG6e8|Sp=AWA(WBuA@osb#D<5Vfydd>Ky#d1r2FHD;0iB-MUa{-UP};Wm;pP;E^U*N7
zL%a|X=Yr?fi+<0MO(xd>=SJjZ(}8c!yxkHB2fRvg{o7#yOUIp*CcNkdgA1)^5B6oW
z7&?F>w`~_{R1#QAxA#Iw64@-3Fzl);E19YJ2j}5F0x&LYwcn~kh9-W{dbLwjO9QhisaDFjMCH8{587hwcaiPYH*h2>SbByC$Oe!}#B(
zt6EYb1)Ci(bkd(kfg9z-Eve+LbN!3PTKKQgWhu2G*4?bLh2#1RCEf)=Pz2o~8ydop
z-w(jgTYmbT29p`d%&Zo*VB=-*;y221%-@mS9RgyKPI6Zx?Ovh?--mYw6tMqPUPfMW
zBnuQZyjHaC*@zV92JQvakUKge_%!SGd=CEfe4vO}-IZ)*B8!Q{^ts^H(R4&4q@fa6
zZO}Pgt9oz!*jV4yh-(D%thPh)R?N=bUQ6L>G@589>nQEwr(-_>afJ>8)D$aI<6
zx3z+Ss6~sH$u;mEv@-5*J#J%=!FF*obnBzzNy}`R5pCW-f|AlTh0w_lHhOryhe<;4
za5F*)==S$1Ci*1>T>NKnzKkMqZ0!mJk^WMQTT3P;K2eNi25)Y`2WWho5sdm&%+8)~
zx}<8|{dP2h6FxF5Y5*ZO|lwLL_M7{YI2X`VGD1l$aXx^#Tp7
zi9qzn$Wwv--q*kw)>xI~TbXZ^yaA(s=On2|1bv>!6)Ht8tJwBN(-NYI_-W^No1i~^
zg7XHvRIRmns8T;|MJ}ZFtRa1Frd$xW6k=9~K8gq`L`qz!W2
zIxhDp2`9(@cmdNJT|z9
zh!Pa7e7`gqWg8Qf(LZA_#!N|f^?$Nyr&F8{mOL|>IQwj8wNt#5NBFep-CSLBwh&{5
zz+(i)(c#Jmwi$?CTRC#ta-TQ91}G*mri7EQ7g=dJIZp%{;tlJyp}z|JzH_>
z3=WO-n`Qm|!xFQ{K035Qx+Gz0quBXUqg+mHUCFoeEjH8fuRMTt$-Wb3d?|BD3U!ER
zQvdl%9o3O!hNBq4H&i{(%k59U?KbocmmqbBw~7^Wv{&^EpBl}leySA7(bPN^%cY6r
z2>KS?pKqv6Wb>CA^#(&pHb_PjpgC-A^m-Zq|D=E^jEnJTIP6A{$f0Ot`-<6U=oaVk
zp4Adj+_Xcq;Btkp0Z!8t2!*^{4crpK<#<4<&k1sJq4UUwP+S
zV{0Zu>XPRwcJ2DZAG}C%q<$vX{D|DpI#G6;!uqKKr4cP!-kd;ueb`Q;#=>-oS940V
zW){-n=ngTvG4L6-*X?=cyDQ`UlPa}>S;0%G81dFDav^cb1J8+i>9Z7la;7>g{9UVW
z%(%1nf<$-i5!2e?!;#Z$tqUtYYay}DTn;e&Hhtz9W#5RT>R}xd{&^H{dS?1t0u2^#
zTae@XDv+)Rt@^waQmxt)Yf52+@)qv$f?!46{t&rWrqrv}bSxe7;tD~}*Ba14YI#v}
zQF#=X`L?w&{2Cd9OmSAy?hmz>yGzixu;0pd`lmPaSS7ID>G}w!1)~T4mB+=}b^`5y
z9g>%%byUe7vf&va;zhMBj?X@%vu3;Mws)$w)WarIkb*boU1sMFCj4L})yLnv@A0$x
zpX^yvkEI0v`0H?h>)Cu(vVptp+*5b<4&-?mcY<=fJIj*AsZWo4yIn`&tpJG{Me#s9
z-o}<|x$P93@8EK1O^R;ESE#995fdI&p48KUDxLS?&XnZ=4(G~M9QqTX6K^xon
zHsXFqGvPermV=}m1#-Vz7irpwg!x02SN#Jfuj|58e<8LH;#~jXz@Cprk6~sDb_EQ79bWekV!EAm3?hx(
zq|x`QUbjRBU=;?QHPNUb8vSrz1fkE}TB6v2CPa&*oDk;Y{oNWbkH;A)LmGFOHv1Go
zgwN#_mJ2a+4vpqo&1K+I`-8zJKe~_>>G!@ZU!?a&UqQ}
zTjo7mWsw2|BBH@Dn34xJr_fsBDA3By-vqvg!g1S3^jed1E3vaz&pA>Q
zuba$FjqY@_z0Vpugd6{GadRav^0-@~W%ZiVbkv`cypEJf
zW|Z~ISNfUw)UKJ&6g>dRbVK--DkOgGIoW}6K!3Jw7JQr0;oMGrjy~}MM-qW28SqnP
zEk@=t44SmjJQM%`IA85br}KIIQW_{wd3m@jb=VmY8L}JKGF)%-xH+DqySlvv2L7FZ
zha&z`HZ2$0WBZQzN<{rQZPkb-$+%PAeXfObEj4(kd_2V~!No=V5M0DLEFOhm{=(7r
z^biGY+d2YGk%M>4>@OsHs3$}bS
zM36v4wLCDXOHyF_K1Zc!&S*(eMg|E*v*!S=7ZEevd%aA6Jo`z=F_~qUwBRZ3
zmxotzYrEOX1@&+qZ$I}L^rY?or145NO#YM*Ng7Vhx42H8ynvf25k9r7|9(o@Q1u)`
zo8pE{@QtiDJoX30c|S5C&-8QqR714>emIxd+qM(QjX)LM>ywpP|I<$BwRR9g^CjH(jjuT#`VQNqcQFSz(6+Yv
zFlxh6dNM#P>^GuPjM#!s)Fos6pv7wXh1OwWLPzm-B(-)!
zo8alrWHA|~5=_T%h}9U?>zR{i^u;ce-|dB3*sYw(TTB<`?Op=%
zHG55K;^Xw
zraa7k;c&RZ`;(AT$I*5_FMBt1p{*k?STz=!3zmu1sCS@f#c4z4=C-H6N3nH#!TENo
zBCQEw=HrdG1K{>x%94TGImwopV<;U+OzhIsAm|Zptb`&e6B*E;37!vu)TuyNsM3>V
zx12?_=aNlf#QNmtZx?O2<$E+uvHaT#2a8(C&#zd}?M!#M&7%(6a*zOpeZv!OlQ)(=
z%_NqgsA)&7{~Zu+Q5KqRCIP|B?2jb?8_O4>Wr2dPLzUrP9pt3f#`w4WF;m7wS^|ft
zEiY?96iTJ%kM=&&5)|ug4k`LW%6FA~v6gd)aAu7(SU53adK8zQQ(Av#ysy(UZpY&p
z9kE8W%ic>a_)V`wbK^TQ5g7;A;+wPBY&DLi4(jg~2!K0_m9RJVj2!HfpiNIFtXCO3&`&QCoQ$CJlq+I8mGzisAINxT_h>3rMs
zE!8OfQT{UHK6J{oDal%^7Ar6i6eC?D=VReG6N*Wj{Pkc+6v*p}U+rC7Nse_pXwyDq
zksX#b>F|ChwrPFK+#Yz6)_R3$dm-NzXQlie@qI7xNX2Wv%J?2exq9q=@#q@Qp?*!*
z!|Hm@BP#tm{244c(aZHbpC`g@x*rDZSr@`vrB*KNFm=PK<>^n`I}-I(qw7T>*WbU|
z(mdm-FK(g&Dh>ib)P$UgCkKj^uY)H9c8D6OPC?0)y7Ec1>R&Zi%UbDEL0bRey`Hv
zd}RK7cluo_*_!z~qy5$P;Z%Y1<+k|M{XI^`V1Y!GqiHRd_XdgEXbPy5{+cRM1d&fh
z7`Z@e4ACP}evF7&#jIbpiXt8>a|G1x4r
zwuP4;??axOUbFlUB;FS`k4#x*Jd5;$DlORB(
zqw!2BXqqC>gumj%N{vxboOUiR%L_#XoN0z>DZ|ZEwS%wO;%(MJBs1z{zVJGwdUAPe
zrPM2gqbxK^3he9f$)3p+4J;Yx{)%Oi!fjXE#OLTV$vZW>*-F}U0)^8@O{`t4V3WJH
z@;*j7ubHxW)U9?E@-SjTVK+@F34KIhh+f3HySu~NM)6J>d8!A<3%otW#_2o_+9l+d
z>s>zowwQ{9PL>sCfG<)5C_!_XI=R4I^M6&6%OHR3n53k}#8p*YfY%qSQerh-s<`%0
z*g+P|O4@;Os`bEUUA*z;&*?EjLy2v8?N*v(~1fRR?w4@9M{bc
zgKXu#eIRJQKB(HrGd;Bq387OfLH+pZx982_)}LONvJND?uYuv%479uOqJ{AkvMNFF
zSb^tQPLTfKU0zT<7iOh)XW=Ja^XM+2E$%MIqX4gIPPcJc(c^AU3;zo7VFHW<+pnFng;OsH6CNrHm5+v4+ly7`q}`|q+R4z4ay2{%IebFY30$&g(4KY+URmV7kD>^#AGW~nw}oD#SFlaLjO(Ib+X3PirzomD___>E?Y{BA1;#4iqGzx
z-1L*NI9At$Xo5MJFE7q0Q9@iSUu1rW&x1w=?;0N9tsFlD3^~n-zC69b70v@dA
z-djZcs;FL8Ph+~($D{a
z-8&!ltAW!-5Hhi#366;Xxo)$g36IAXe7lDk5eHezXsB5<{cN!)E3GP9XU!y2s$?Ba
zW-mXa$B%`z!!X9LVWAEkI~YF)$MuUW=gO8=b;r`#`zaUA?$V0lPtpBce2Nva0&52!
zE_W<0HoJRcN7MSH$CwN{%p*QQL!)E9(k@9txNAc6DW5L?wz2&J
za@&|pMx-r0sBW}DiV!h)nw=D)rw@OAr1JUi-<)Xz_TPPYQnl$@W$Xle7C69giDEy0
zRxiV3LGQAF3&&c>j~Wg=Ql7OE|2-Mw>4Z;1_R~b@<8y8SDDUpr;Y1N|hTFs)&*9h{
zBfuSMR1EuYLZJ?06Qz#_#22w2kx5RXNxXq>U8lywuKLp1+?VHNaAS!{?frGLnZC7}+(n;#I(Z+K#mR`q
z03gmUnXY*=*xMVRkku-s(`XCCQ>9eeAQVs|^^zHmCx>1Gyae-2ljm{?D*XP_Szba%
z{P}9vqNrzn_b5eNH>VJASWH%xV@7Lhk=>S=SCbe_cZk5-+dAY9%{@a-`Of)XzzQJo
zRvAtRmLrly`pmH_@J3xiQ+4QTc(|}5zZzuaru42bhP8KEF$(XsN9s)uK|a;4W$0y1#xWN&!eCb~;L;W^|d)9jg
z5BE+up+mTQ{^nHRJ`+c!a%DZea%&`(DisxWZF6b~%NGh#$rW%^uE4@H++o9Dj-PPY
z=AjS^Ob#Vb9SuZ%zPTI4lNpU(AcvWHf4wK}4@C>G?|=1sOx!-q;iKOpTZ^wLefl#)
z<)QQBadmeyGIUK3H*z9tRGN)Bhy~S5ZrJQ`ReL|s?5JF-0)&koAx)U|m|HDQg|t{l
zP2N`^O1=W}NoH%_{1y&?-oF03#LDAxj0s9`%+F|oJaR-Fn8C9z-jYBE62>4P%kYd`
zQ034JQ$!P}FE~?*4K(l}kJE-(HO7@%`d~aBx*YOG_C^wc-X3K=RB6fAZakz6njx>;
znc+BeGN<@+xrswIZc=4T#%3{to7@onhOkg!cR5^&UCWx%cyN5ax501PE!ojIT&0Mj
z`w&5_hGih?0IKuQ4e;TA#TW4Ax;U(?+B!%|jawV^P#@8@Le)ka>(=p0ubnk+%D0~|f
zBJNVR^s~MrgtWY`ZBnjbgaHz>qqLiDv0x87zj1OV_vy^U07Q*YY<0D!PLKT|b`a^s
z*w0*QLjvec_%G24P;CXBUcu~`^@iJYwI%k=zOPOXqF92KwllxbTY7Z1#{r7JV
z3+_DL+w~lbD22}Q!sR&QDKY9{7gatX=J(xXyE{(eaIVcZ)$=}IsguZJ
zEHQUf$RfOy8Xa{TO~@O5P#<5+UR#yHVa0fRIG1+!mL7uMuR7;APz2wSDrr0tG-R8~
z#I8QFCj*EKLm58CHyA8&hBZD!Dce?9Anv5n4CV-a7YUR38ZOf7T}tf)k3%zb=;Cz%
zA#pv6*ozL;{WAtN=`R&{)VK84hw{vVG^qVi11XYPRn@cF^eiQ_&wxMA
zc9Cf>i=^iVbus*f9O7%F-l6tj0Rn)X1n@7f?tyN92B6<>kP8hRfgT7rD}QLLUFc)o
z&eui9cy9|u2NUa5;8)^ziZc1!{52|83QxaaK76|N43E)Gi-G8+dK0fq=a@`p#|-_<
z$(Y2bp(UC2el%m2-f4}s_FAiK|GIC3)(Dlcq4aq?+x71a_wmJw&~mMDF{$KQmL8Gi
zyE-`885cED@;c7H@M*ZE=BMPC?3
zz4xPIRvtvKyunA*!AKZ2ygWg-q%~(Sh>;mOhtJd(j|XEJl5S_KXdh4C7`LZf{ppOH*-KgJXWq|WKyTgy+Fj1hwZy--eF@&O~s^H!?HoA2~5ASNvGynrt`#m
zN2us!4v)zV&v29JjBA-%XS33NZ@N5%I>g5rVkPLI0CX^du6di4r8IAN%Q?45zngFc
z73%bf0R0XR*%g$`8uP_sj?rQWB7C!EZ1o4_Ly!ZiH@YVY|4hZPHFj##MDqvOzPYim0^TWclN2YF4{Afkl(A-vwE#=5URR$d;jEe^)B
zpa?_w&Hp}x%@!K@Shl@jx4)zNaFUX&j7HkzFk0{}4+Cy<7aOVejK$v?4Dhy@lW&ie
zd{q;G_w_=vcAu&~E5Fe^$
zAAhz~oWH83AUc6?1BI#kpCG)JQO&wRA+`)-6Y`h4rISQ
zSr7X@Fze)vng*e~;_3;vDgIRLA!|O+1FScT;oLJLuuf9@r*Cp)zIe)lF}pD6uocH{
z3Tbl{5u38z<4UaS@1|NW2)_QBI#Uu4#tK7Poc_21@m@xdT(sfMIz;L5u8okukb3o8K#1L9au>Lq;9wJ8y^WelaEabS!mB
z0?uWi+iZm3kyNSxa+igyX5(eU-d?cUd!;f$Nq)58REc1Q#~jwHV!!d(fYW21lf(%)
zN`@C0R0LtmCX$IlsB~8_ih5EM>8=IGE}y59D_8zjjfN2Ty|-H}Fm655ehmNOo%Dv_
zFOo?nCM;8_LnKRWP8CS9CB9Z~{eV=E{JY-HE0Xwl_?iB8P(DYJkmHj|b>;n1UimTk
zYCN~$UbB#T7t@dQ!g~CF14b2N$vID!dP}=33yEr(FXC*qcpGIFQ*jbwP?%35~NZ_
zu=j3Ja{M@lM~1HIYhJnrm3+8;Uu)R+qLSDXZQa|9o~>>4O56H~9Y#|m-tsiZxI_$%ollR<;9)%!K~AjzJRn#JwctU&r=Ls|z60jK!DF+F6VPAM}DT9c8n^ZP#xLW)7HC!oK&H+P%N2ft+S(sHZxp
zWdF&XIs^~8`^y81{IgYA*Q7$i*H2hiqryYz6o%%BIn>canUKjk@0UC>o*Mm|pKz0F
zrmv*{;$(j$J&tJSM6)rItE3VRkFHY!JN$A4@MvSc$^bLAO=u^yo{Ns99EI_?@pepX
zTKWuH@{z(o_n;R?6$Jd+B;&-mw*?<>3SB#4E#~I|cs(Q2>u`wqnZs8w53&V00zNc)
zEzbGgV((rTo8a<$?6<>7bTqQb^de~Ry^OejrHYM!&saNsVG_E+QZZ;`!PxOK@UohT
zaXr&m7V)dQkI6+R3&Oezj_X&>f%1DaaZJPofR+7+6&0K|n^HR&#wYJR2iYFSbrr3a
z;b)sroMHdnd%R(3i}$x@rtSkUg-qQR=dhQT7pQ%=jSe3g)grlkU9oq_;qEqWrw7&q
zz|&h0ys4C&#rE~}X6HGkV8gyM_v+!L@y;NtFouL6&{>zo`N)jWWY2@ziWced>S>_I
zQ2=&Cdi!I}%n_iH{$feHp+@AMYKik$beu69QPSe4Cxc2=^iVlP9MR@G&~wG3vZ(Zp
zMp1yVl;BEe+9$BY`F+y^D)Qz16*C|k@bd9VLk!Ps#x+ls`)q!7m?l(Rs05IjH1wA|
z*tfSnU2S%9K3!&S@qO>;%1_vt+A{i#iYk=BX;<`Nvb>MBB4NG?Op_|TcSkyN4-wy}
zwDie_?^s%-lK=*eM;=c?rq=IHS4ySgNKG9Ok#5JdGJ>DXsK-zhMfVEXO56C5mv+d2
zTkt*V>BMne?;UljUk$liT3ZpULh{8UBvOF^voNZfTz5YMXt|ujk
zRU@<-{K)}kT;qQE-EyU&vIaBZqu#M>K*q&)f15f41$IZ$Bg{_{UMlPFQXK}fbOJ?$
zK397@!EE_QjsJ7PCKEVyoL>ymyMUi-7mqYR8|?%Ewd%?YD_i=XqW*TYM5zb#{FCre
zL|<@M2?PPUTz^@mh|V?5fo!Q4`R8fHMML&=kEX1`JxiPs0B4j
z{F&gYDm6?G);&ESW~Nw?XxYn%fqD$fQlK~!uqg`
zO+c688;M?glrwuuu(yYght|00IE)eZQA)zW(Bseyu%^^o)uY6&WR2ZwUa8&l7WRr&
z{iC=uhB=eWg6h(
zfLB;e5DjycL9MQ?_K~7%N&?ZyBqm#2P7eRXk(qM-A#g#o?EH_Xg_4BCWWmC{sn0;=
zjl(#Yfi}0XH9%G9Hm
z#eW-=K~*G!bvRAr2)Z4fXR}dTGlTvjdxJGz34$atlTuA}2xtmaOGk5N9eFhhLti+<
zd%WZot=%q9L0Y2B>
ze31s5-{lJ1I^5td?W`FNAv4t)0qL;}0l*9Ozv1Q5-XMjG%STefji$3lfZ$cK^Y!lY
zRzE3^kmhFS^z;2WnM|K0Zy8h%1-O~gp|Rn|*R7EWd6d$=n!t`%sHyxPfs{cZ%64sR
z7-;PqkLl%!eCiD%pk{a2?)&({Hq3F{GL|~Tyfj$xWD5N^k+l|=x~lKwiNVuO-9JCg
zHanTA$Kk?VX=bSDL-;Lx@~&V-`%>z!4NhG`t6o+zA!L@`CUyAXlA-d-<SNtfdWHlUNFe#FdkWdbmno1bJmB7Cq^4`aBI1-QzXK+^gEp^v+(0}_S
zrfoo4V5FW$jHjZxQZE6)wSGV^i2-zYeS1FNE=0lRw{NM8{|}k|i7*b#(AD1C7nCM8
z>qX{bvue~U4mjN=dw-B@q*MU%s&44%=~)aX(m+h3QyuxAeB7QHXyxc`C_HxR<>JYu16AHP
zy8{%QmL-onJ%3k2sUi404{Pqdi9)0>>iyGgm_eG=`cVzT-l34m6J-HRx8+-Z(CsyX
z1b)CrPJ=oW9@vqSlS3Bq?KI>6)~==;>h1A>xVUF@0VmipubK&s{Y(x$mEWsDD+NNJ
zcd5i?r9mW&%V}TE?4)JcKpjFGPV7^V05qg;g3{j<2A$+8gU%|#9DGm?v=7Vg{q!in
z=xzUA>;#C*&lJg@E@!hcYm&HS8YSnu5exdh$u%ivCGof%i`9PbM^6oaK_5hA!?R<0ME`K_U2LUZ_ha`PE5p&XsG{`#ig*bJ(QU
zx#+EI^he;Rg05JiTp;nWiDA)Il^`H*cF^!wQbo$v@{UlYjcs~nH
zxg54b#ZeNUnw<_fG}Pq(Kgm^=f$&?hSg0c0Don3?F6$YZZ?xNhhiO)}jGyfmvw)-5
zZHoOkA!aC^!fDxIyRpD-qvNB0m+Q=8W}-z;-EWe>IOc0u2-HiY6DSjzjYPIEV2~TP
zcol$$4B!X6?^x{zBNh@Mxb~z~D?wE+*NF1F1=3!N79$JxoNsiZ$SaA0AW;rt9I=@U
z2P795EUtEkphgXE-Sc(7T5A1=HTh*wyQz=kRilFhrCd+Sw+c}$J3Vl>pc_(&A>z*j
z#J)b=VvKWv`sn1wSQ!TOl^j+5i`y=%W`P9Ka+3oYyX|VKJ-XsYP;(-Jo)ewh1uz_Z
z*i2^QnIv|0cC=bGx!X2eyrOd=8%t))$M=F|)Hf`u9|zbj)*7nxn6mj9~U(lq+attoCGh*ZzkZd
zS#lq-%=9b`2LB4`=#^^!_c}MM3uOYmm#x4}Y~6s?)>h2o#D>EVEV@aE{wL|NBXO6$znswKAa9mxz-|TE%@5eIkNlW%WhCmu@RtkGh
zOI3?iuJ^~tN~l(?#mKtwpU}p@IW6Bo%$O+X3%I(t$OYNCrheH-vPH-hGi#Jtv_Fnn
z+9ZCDVPt1#B3#*AR`c{_`t3a3!kidnU9ij$kziz-wy;eF%*1@(2b<0~RM&j@d|xhM
zT#lyuZJ@Q=+#Ns&oyOL=!x6CUF9rv-sw`??L{MP_l4T+G;ih+gmt@!s3UDDR@`(F!
z&qB8VPd8z~D8!7POzpF;$2d2*^6?r=F@CCK{xJLwrHcjbHa)4yY`a)P^0Ac-y7W+V
zAg0p8@V}8@!#0&O6OS4p`81~HE1RNrbabfH+JO4jmVF=8O3LGz+!mmG#mx^j=&%IB
z7^6X)(NDsFfvTklL}n6q&N`tlQFt4j-~4&%o@LbNLv)(-(9u0wdt1mRxRT-#%hf@Vt+;hQ
zK*gN(y{97ekRrt#M)q4Ky_Q1tBwZA;O=c*=ecXxI5+M*AB~aFmXYNFOCB$;H8
zk>n%GFCoEUMeq|aznv)W$EzWnBn&DArEETrigG_E41^mG^d=q%`|!_H3Iv+xMS8F0
z+P`TlwSR}>bK25#Ql6GYBCtaIyMPK}oBLxKYGrCP_^hUpItbU`H0v3VkdRrSjGX>8FpuJI()i~IXxu`u*zl@x+yoYaz)4j*2Az)LDL
z3XvSKfVaK5m##-2Sr^1>=z(zIAvkux$D5WQ*aIS2njf^bc1g`D
zC!);Fv>_yZ7&+bD-5>rxC&{wyqL@^l25(k1x7)+BfMJUCbk
zlKR2;wz*#ttb|6)>R0JJR
zM1$1-K=Ep%|1PBiC^8HQ9!uOs4-hc7c?OaDAh87q7C}(-znz;REW|MCW|r|7>{l1uK|<<%ZVCJF_$bYa1F8(i-&Ncd
zg4|B~iQ}9F!<50je!Jm>!9r@-hmdO4ZDTotIUr9Wfkms9TdbJFgJ@O7wT)`%59l*G
zpJovBxC%U;Efqz;VJz$oLBWWgXR!zYL%8|f8*epDSRqS>Yf2M@VIcnx*fk`&pZ*S|m21bMrcAYy%%D>uzwnI^5%JI6>(~
z>Iv=mypQUabp~1#F+3+;OZxAF(ej`*EvV9Kod${IIcWk@C{hA&?8A^>hKG}RA7X)K
z_JjY;Awo#`Ay~*>Vco&fNx(?-&)3_JHhkZKzHg5q!^sTg-_~^F1V9W?9Rst`j18fz
zo{PeW$*8MElu@Nz0PO=}43KU*)*jtj|F`nR@SyTrn294zkzT1;*U0vS^1vXjSwSct
zo%q_=I1`KG4WgFWMmvM2>j}Zr%IFbRh;))J2p=D6ubV@XoMPr;q)vy&tG$_tF1I*z
z@cGG@MC+N38MBFOAgOq`l8_J>sMVr_9+yQcG|8+EPrAD~8H2k(CaEA&Ro8fOCvbW^
zix+4!+4=S~qJw+Z{okJTcCRClqPQx6;vJ|n9m}VXO_r+O!J^YB1OagC
zL6?VG95T2!*I?9V&gom75#)ohj1&+hi>jBYnS}y$pCeJzP{YM^t(okBd#PCQ^
zPfA?hL3E7dAA!R(eY(;J^s&@D$zXj>pjIY*+KCY?oZzR19*7(wr8L`~u)&iA`N14!
z&a@k}t|KSvltvyN?>ioy20aZypG0^%$YKmY_^^K=h*pC7`SWLfeQ*Z^ph~wH_OnA0o#qGX
ztAgBf5_!kh|4Zf9d!U$#wyF$vG#AvAveS}t4%i@5Yf-1arXn!Hj86q>joXN
zbm~l_>an7BIG*oLjg2~Xa+BGOaq2;ie44T`kD1t_#$uokwh10Wc?kq1D8I}m7=xZf
zgAwyUC5caY&Hta0%Eg129JQOw^`A`#T?JVK1qG$lJ8;HPF4C|33z!qpr0wZsUO>4s
zP0uq6$iSrYxX|jgx-#@U_YfL>G&BIYEBS4{QfF@{9kdYX6QiY!5g#gs-ily`#XQ{Y
zd_C);yT;?%)Pi$^g~6dR7*@!<(_v?T*;UA>I@3)YX=?j9R18GxPlsM!UL#Qx@L04`
zc_JX|a?5%__+ONUR1nstgKU1x8DQX+zkUqKnQdcAbcMl!EP)cB6DXH99ah=ms(LgH
zEC87_70{9Zzsl8{PwBSyuJSfYfG)pPha=MLsBjd;iJiPUw)N0+IxeW
z#jt$u`_d0rrHAEcRE|om-^K+J+O%?M1l*1nFfzDmT%@FEAUM^!VKo^h4|)QTIXOYC
z>G%KItX>36+Y#wAN8ty6oG#&mCykxqf(I3lB7$c&1dly{5J^IdSoil9qh8A-NO+mQ
zy*ySei=Sy?3WI`6X2GX41*s--A4`Q)=I&o2nIk43@a`C?>2V-@yg9;%!e#k+JXcOd
zxY=VJSX^AZgTI?DPdZty$;AJ6r!TBGI#fQt9}X=Dgy~UaD0y%CMykFbBYZXC&KPa-
zUySsUn;?jzSgbQK>aZBy#J%4eNhxBrhbP+`w^Tlc4fuVEV&N_5rYxTnGd!{=9BysDF)5
zH#!Y0_j(zrRQpIK^TiR4Fg1_B6=Y|t4Gfz1=*TwFVQYjo*gkLa5tyQxj-`w5490L~
z#i{-qC9c0cYL~Jb_3CPiO9cD%m|;JJ59_9Vd@q`xp^`kS>8RnUH~SapEs&^^uH$9C
z9_fP>!vqO~M&w5}i)pF6HxLv_uC=(3iFU@!&kZzZ1qrcrmoSZ6My@z8Ldd2t0u2C<
z;%BkF0N#~aC&srT?dPo!V%TU&0g$tYV74>CmBS&
zG^V4eH!qJ)7Oj~!ujzv^s=r0SAdXRp1*FF2KyF$YRJlhW}^JycLu*-H|
zSSd(SPv6wbI?o63ULMcU_1JNMte=W8F?@+!RRpL+^inuE?h#w#vUptbL1-$|?XceF
zVN(@>U>VaDhCx+`ja{G1>wG8#s^#LlDEp{LKP*^!gAQHKrooX$5qI~UUOa`2g-RVs
z?-og6%A4@9CQz#=gTUVWPb|q-@aVzog9*jg=lhh@;Qf8ne@%dzSl7gaB0T1o#9$7A
z(4fbCv!0n^Mc_1oTwsqgbzrGdpfYT6%>C&~@#<9)nmsjylDq#96?Hol42HGP>(~D1b1Qri7B=7P@>dzU7|Y2s1K2br6U=I
zPr*JhcSOJu4YsRxae~35&qt#e39>??ut8fO^)mCzWW-%`hRazY=0kmi=w?7)I!2^K
z3`BiCQ4dYgflc^$e!M%B(AVS5@2*}2Q&a+XkGd}^
z+X<@XV{&9)aj+BmdEG^RA~y3_J{!nE5fJVGrwSymmBq7Z{-uXN6hl7Q+D8P*IL&H(
z6PTv0Q3P{MjY{pLFJOVc{IyeK5(4bohWcQ+5d3p4k%GN8fXWOTp&_Ew+|(p=>oYSIeS|KR-VbK*ZmB;-Lw1c
z{)oMpN{J}^Xp2R;N-l_qtSJyV&VYDi*MWx~0u?boCwz#wEs=~!vu|swYL!ltLcclx
zTQ6H*>CBu_p?~7n7T(wQs69z!)k~>5pn$Q^}Zc
z)#KXg|A7zAMHKW1MeweP!S=~P1ep>u6&GBR?Ba4ql7^h^t`homuEXc~7lt&M9Ik*j
zR}>y=uwRb01?d13g5vAHVbo%>F
zFXTUK63+l9`H^o3fYA6r`kF6xcQ)ni>7EwCSajMzb8eA|s8#{!I-WP{o-~-;~ko`y7iPK)O_k>=6
zErW;(q=jTbNx+4>wnYO|F9rXW$tddYs?VzdGe*>tFuN2U(ujoIYFF$lnS#9-xf^Oa
zV^`pML^tsr(*KN2;Dg4!e2S9nOooEtmQ3{LEH!fZj-CTSi@*F*+L>S%Gy&GyLa^>i
zVrY}XymJ8Yx}_yMF)_ax)m9M-VF<^q^T8B3HKg(;foE^zMp)Np5B}eLFSE_qY*zAe
zi_1<9%|H1_KUjt%CbNSYBoupv7Nd9nf@tnL^s5NG!BieyUthmEsarw|Hpu5Chr^bG<@xMOZE9uAv?zAUj{5xC7Cx4
zQiGiL_v+cSc0A&HKL1lgS;+7g`7=y3g=|Ven3KsDt|PDvXHy^^O(yC?7pg~#(v3J%
zZT@To=TN=rRKdqoj+*#3>senKr2>AAq)`Jd2yjs&t+i)_$#%azHg)#{bz1W{~Twvh+&w`6LL@7<2|q!Xj^kVlOkdSD8JKEZJZp
ztww=CK#17|#NJjELVo1KV7Ez1803K9d$~pj=bBVnUgw#LOa*%fST-R(dJK0#2MaQJ
z58-vd;m&GVuQZ7;*;_~_geOI`84hC|(d;ZIk%F0Kl*I9;pG^)FbjaZrgQPhM{>=fX
zpSs!LTn$mBMp7$qK4l`#@lfG-q|$BDX8a!lv(bkn_B~x|Z%D}9KdL>-g%Sq2A#$bS^WI;pa=P4@
zt8YVo2Nrx-JaX`PAw~K9=z
z6|1tPCo1B6q$H7VGJxf*G3HV()lgnEbWo_anv>q@3xy@33c4aLp_kquov1NYG)*!Y
zO=453F#(ts111;s=wHBkivOSf5FP|mZ)V*~7~&e4MAXFoWOnYAg9!o+-u@}OQ)&O7
zXe%aLpFJA$J@Zh8Ic!1*lzkx`sE@i};x~DFeKC;sCT(%cGIIk1)d9FBD-^(K?h-$z
zktpiuLE?LVfPB0U-Au;HZy^pbszUUBZ@aA>z5~s;Y)ie4z7`1|A79rUBN$8HdtTkE
zn{k=c@M0K4m0EO?E+}rUt^lxcFa?`kyLi0C!7z``4lB2Ae?-%pAabP~8-d`baU7hF
z#>Da!{9cQ-j_3kZh@~bJN2mkoeC;<%@BU+TPM47=40K^(xZ&e?^})+*J@~^KuS>ta9%M@3
z2i8GbPC`oabqhtyvF^l|zay3c6>!iSz~_sAXM%RG7Y)wokP-dOwq^zhFwDnGR?I`5
zPG$!jIkBSMAW&}U7$NY&t>b;lN4$XUWlPvsPVNg9?zRgq(>TGv=9+%LTuw1xb`V?f
zc!*3kMa{=M)ja1=By_eJ)b~Gu68Hp^f1alimbC_BV48N=)()jeQb#65)|kvhjD
zvr?zadjmePg%zCbXeyU9cMU#26U+U-T)g{o->*7WEB3j2iB-G4x7@5bB~pLCjpVnf
z?y+F1?1|ljqwBI8JAc{6jtTrtZsSh__uRoIZJVvIZBa4Ni2LE+6t2Wav4wbxuM**k
zhiP@m%k9Ex|E3#rwhKgV@ThL28l-dr;ee>6+fPqt?K@m1{b2+13b%I28TAAhBeE2*Vw
z!-CPx{ii_Li0Xm>?HkCAA;T4l%U(B!dLDPTyOo55Q98>9q1|}qr0_`i*Z)0qSB4Dh
z*&$~>uV<>AXpTx>&|i68e`{2cit=@0XvSB?bNSN~8P!98HsTVvI9McDSrbj~%s3Wz
zOA_8-dVbs;2teEd*UcegsmQM)FqB_WtP$1eCV&r@Ad+;#$kSwoZWiq>{3WUXE^Czr;q65k}(mYp{e{}W0+0)W$c#Ly%fAigJ=};_*(-uFK3+N3OPbb#?yc|lSLM4{WZk_q#3VX1sHyHEV5G{w*
zU<_$GM%vy$WVZ>Sv8djaearm$e3K&{=14+}9a*qM__KDjvmJ5}s#d*~UZ~P6#Liv$
za&|eG2g6*Xx^{<1M!aq!1EtiX;F7E5Z5|&(#E5>{uIRTN
zwyAd#v9)9V`uYo&_Tb-LA19!{Zar+JUG+zLcw#>XrnNkZ<3#n$JdIMfGa4?^w6iSZ
zSlGs
z^M9B&jq{z^iLFqOZS2R~4zd@ZeK-v@w(k9IWf4SxKD5HwgNg6p{P--C(t04-dRXk!
z$SF&}E?r-bd{Xs-MZb}YLaj-%
z7_QiSTs$N(6bRn%KPIn*9~K>7$U~D^V>N2cb}G`u&(=H7Ok-eAp4Z+IM9k-`b5@pK
zW4TI!BS})jDkT_U_Qrs)wviH;)#ZI79a^GO0yEFKT?ha$Y>(T5pO?peuhywi(@o_Ir5p1Cy!{U
zCfmGijvCn3!s}sp@GHW{
zo07(>gEji4oDc5P6`dqr&=V?WUVlvH(mXf!b$^->*GvT<98*HhV|Rxf?3OOoJhWR@
z9lLZIn^c_fx%{mw(I%*htkwk?;d0qBm;P*$hgo3R*VQ@h8$UxG)ozJtE7XdH+a5Qj
z3m$CnrJhoFsw*{_YM5jm@9Y*U5S-DdFj4bW8&>z
z%ENSRKcmCpfY)Gm6Jqj|fbJWi2h_{66p==S3l2}4d#XPIkqpb;NOlM1T`q-~ygUlI
zWMbIk>(n1nIBuEAh>h%ht#j-oCS9LGI;8M`(b3aXuF2cd|X{unf
zoI|bDEb3?qT`Q7&J6&pQO`}~`b4|kJw246yu|_+bj+wOm=`(UREw5@}AQN=8`{V~3E@u8XDecBuM59;hcGG*#{DF4gS3
z4vZbbJV;GIXUev}oW?oi(y8DcQZ6;yiFodVn9o_D+O-fyJ>JHmk4fKL<`N8O-p*w2
zZB04?GvZUd@&Ojd1bwfHBHBe9%=8r(;Rm-
zxbVX=@+=8sU#Ma3CI6Pybcl6vxi9%LVHm}PqE)aChzSXA4%(OEhsoOOAEjD55mS2t!P_*wD_#CIZNEHc
zbB(ugINhq9L=?#YxH-JpH-C?@^eiJ8+)3b4onp}T;6XmpTXfwF_IR?$LQA#6VvT8m
zu4k%~XR-?Ywvo4m^i6tURfUXSQ~lX8yq=
zXB|4_M~bPgA1t|j_l@w_u_X#7o;r8z)E)jAgzP;kbn5)w0SGx>o?e;1e`7u8Xc;XhP&7zq*RI#LugJ
zEZ1e|kFeVrA(1)hz(IrAwaduVZq$PsN~O);A4Zf+Vh!Y;$&7(Q<{wYv%RkiP6|yIy
zl$VSt7G^OX&t}e*JhY^u+L=~OnRhu~_g?-B=;tCufG)6ERUscW##N!G?+rDyOLP=<
zUT#9v;c76$JyaU_bGbc6&wlWX&;Dz-xH|(;v6AHhH_tZgy{C@?>9PXv-L>)2xXgV-
zHK3X$=cDl@TMhsk+U1qwi;nWmSPI{WD&{#75@&RI9fIGU4}HOcN?+o8WiVp8OvZ|t9t#mg#3=zn
zOXh;zbt<7|%xw?#GJ?J>U*txueNni?eidmJ=Z&y9ZC6BG?n3MidEHs+Cvn_3FE)8n
z&S2rbzQH2Cq^)oUD8*9&zWVUv`XWNfhaM{LunG7;D%`(HCaeMq3@-G#2-R3!lbJ=}
z0&!{%1PWVPwr^&N=Nul7f8eZ=*W4k)+OuUWIiD)n*i>-D+Gwugz3y#n2!!@f1PG2M
zv0@mB?uwr{!d#I|ThG>1M&a9?tQLPC)vi&p;{5>1h2K>G*d7XyaFA(m1MBSYc9SR^
zX<8ec&sK-sua#vA#5%NwhR-E?vRIQjRX_$>L6NS=y#KT6T-Vs^fbs18?7G3RGL6d7ta
zGKhAppsYJG>Tw+IJBB|Trg=Ed7Jvf+-t@)EMQftJTk8~mR8`YvbFBd!4%&S<@8NKJ
zHJHezMzh9N#Mxvph_-k=uAp%vq!N#^+x~Q)IL6~5BS6IK7o5rF+FeQ)W%w79lv9gH
zEv2qho4mE(e8v+dxH21?P`O-1#B864!$Zxa5j)~`fe6O_+k{8J=soUNg2hc@A%V(G
zR87-Q8cT4>e~6tyoJES&pDkA^qm#VWqLNsxGs%)$lW5_SPZz3P(T~TwS&^g%Jt%!w
z;m9!r-G!L3^0&W9OA*)1hN4QtdsiG>Jl^0vKyNr6Vez4`!-YiV8*`f`iAo6Z>BxwJQ&a$eRagcQIt+If5$&0owOuCYn^4Am4}C66e`%M=T?a
zx~tTAejItyU^Qp6fwedlM;7f`!Kpeo{9e^1x9;38-UZ$t%H2_ijm1jAiG8th`A^5)^^rfGlHQPM(d*uiq?)}n
zrd&OZhSvV?zcuLF1^p;ea?Y@{^Hw%j6Ma)t5`CQ8UltW%#%oQv+^+W(&hb_`%#=z6
z@B9*@+f`nLFo)t|edfs|&_I8`ZRK_^Asm3D0vZaS^IS1}V86gTRvS$wY5+E4j|M8y
zsZ>f5SgmIfnvUGnlRCVk)#kO4!@FpoQYXybRHtvkD~yp@=z~PPTy7Sl(|#yYsJFPs
z%PR0AlCb%nOxOfye~=PCI5OK%a{!Wh1#0n_zOWb}e+mD7h~~_HgWmNLyMoa>mp7!u8r*
zg$blI`h@5gYm8c}Day$`?R#r0?M*j|*I!tYSo*IE(WUzG#=1R1?~T()?~8W71>k%X
zu%W|?4_D0wQqo}Va@+or-1jotu-WHZ(L7%&(Fv9&#E!H`*-qh4BQV^U%X#K
z*75mfpSot4N^sFlXk{^rRfmZ+o0nd}SE
z66M(wAJc1>g|PA|X8u$^pu%2!AU28I8;R7K2c%P4Hu3>a<%Imhvp8%wQKY-RWFHCI
ztQE@vsbl;p>M;c2M{P|W++BKNMUaSijFj+}eyU^mCgPgJGip{2<;X9qr5^_tzsXqo
z$UoP63+-PgR{M?yVp(!cU~SNTzsd
zWHNa^tM@0wTar&+EDy#eqR_Z6p?=*9>h{{@@
z9S`*imqLf*;Arf)2=c4hD7UWRKOn4bXylyE*c8eUvGfx4vywr)RTnK+OME42s8C#X
z-_`cH!~i&Jy4|V7^0KQDP*nM@%a~~hzmnMRIi9y0z
zf#=3vc4nyX_1B^-I#X?`qLzbWcPoFKRd30X=I#mS
z(P@%g!~T@|rQsS*%k|kmt?gA@-+49HopBXp`R1&oYI07?n7WEvOBJb)v!
z;*?9J$zWjrtyAyeV2&`o3a|$viS_vGOB0$x%3>@#Gn~7#jKbE-@wFCR;O)w
zSF;bdkC_i^O;&G8+7kNNH=9@)B7&q_)YNk>4!xSo|qFl0s)^*^M>-}
z;=`~;lSRNLBuCGXMxnKDH0K=e%Ws5!gCa%_tq(M+c{PB54(5v^U_rKIZqW7JWv|2D
zBKaY`?8bAwCFM$+d%T;|ncgKb9)v%u{-N*zaidrC85M9+;d=W(QKQ++^>4V}aJ;7p
zX!mV9A&h{JBp8{HJxZg-wES+*{v_#IsKRzsvdQg)6O-dqXb8nf{4Jkjwj|~d=r741gl5dlJIy+a%n^w#jviNz^3d5z_V{>8LpjuVcFRQr
zEdkQT{|1ey+9)981L-;D|L8<&tWZ0uEkD$n_rWyCtI?^3$jpr^Zybhd;M}wz5NLk_$bhMEebOpzU(EzxCRAU
z3Qyk;smNb*>Q$bULoQ0Br6$1exf9T|8vIpsg?q$KQCrB@{al~co<0993$Rg>SvOHl
z5ls7do_OlD!Y?1+cyUY+J5W%KSk_~Vl@ITc21#8n+w7Kqp5a@WJ$V4iXEO!K>sS^|
z?F8>Bo+p3s5mJ}0A0s$WoPamw)8=gfUH(HUdpf7p$p2oFa74b`M^Xr^QV?
z|Fl#aMLrC}k6QTY9K8-TMmv~{BXK$!@U65tMK0D`(nn_=gf$VaHi@oH7v^HXHYU*)
z8~B{KC-_K2Q`Dl7Lk)fR91B>Ervpge!rxF|KW|v#F1MGrw0hh^%6#Eckccv}D*FY~
zONhJE@$q&|paS~sd{&vW${IVB*Ns06*~o$4U;mIY1^#K}HrDj;Z4=(o^*mRonf-9z
zrrQRKE90!(VQc6|BMyqy$3F-aK%mQ+>}9J%+*m3H-6*lHzA(x9@5;Vl>~M-uJo@g2
z@6WWFO)+!|c}fyLpu>okX&)MFM~hJU8*KVo^bQ2w?9M))EIp*J-{34(8h+}GQFJjl
zNI@(ol$!cprX}N8)mKSYJmczOS%b@5l&TDKJy%-SmLUMUOB+ecb5*lh{4x0<4pj$I
z>Af|P_8(hW^mxF@I&js`vxoft(QH_+!@iSe4`SH)28eh{wH8`PxS7aeTwuewm%Qlu
z`1q-m^UBvvNnEUzCNExV=qu!2^xdkevmE2H;`U2z6@xWvB*X^Lh4pAc2)FVjY|>)2
zi+ili3hHLc3eMY0FFZjIgiF5)_gDT*0oOt8QzL_R(qXXz?$s7w8*5`C{l(2K$n?#K#kf#^K46TELrSytTS#aOTx1Y%3#98d
zaZM+-3Sqi9^F@5o`2^kYtqtx@SF%516&jeiF82r_y!}0HuA+qWv~|<#eFoQD?9acP
zxH#tIJl_Sk%cvoE4s*+145Xk(XDHQ;TIInX
z`qkp#2lnXeT0NH3yS8&dPm{~#L4{~(%k%-NyDsP62h
zVu-vz<8LXmVJnW~Od6}*vgdJO4>wbGo|rw_
zZWzM7%w!R!VrMNmC4*ZEbfg?4Z;b)vy{3N{6^c>kww$#AN1KJ$hmNx{bmG@gawp`@
zGZWwaq@s7}b4K3^(fbpJzHWJzuVtX^&`2zu&L;i&rt%GP;@^4T2wwoJhOWK
zM{Jt@D>f_IZN-!m;Z)T$3x2cQzT4Hvd^y`)WU>|{bVScWkS
zqFmhM2e@bK$ufnzFvcfJ>rM4u=p6drcwRhjscUHq?bF=07Bq1yPmpS9W0QuAB!p$)
z@-paEpG~jBxTdqzXf!;YQK4F1#AB=clF*nJ8$OZNt`|0&?I-(q_C`U)a8!;JmKSpW
z4fwS7$Gz_TQIeyE)f|@F^RWMSjVCU&`I<#=p?JWQh>;iFSCf`3D)=g8##?R5N}U;JNdl
zPe3%?Si2sGHC&>Ma3wprlh%HSK&a^L{o*pF_)0a+%WX-!67-#Umy
z?j~m*<+liAFGkWfkbP9Pbyh1MskjXA%<-*2_B5vpaH`0|g0)Lgn>jRrKYSK(C$p0%Cg$w+*z}w@nL2(S~_oLATh>^AwABE@BFtB
z!EJWKw`@>FQevTw#_&>2jdk2E$R|I6b*xIn>Y(9x5PIY0cW1XL_+-T}SZbK9TtXk{
zI2{k@#xPnFCb4hH(WunGIG)ajCDoe<84kzffsDtd!dBd=)TeT$%f}h7Ew}uy_b>F-
z$Iqhs6@MZ6KB7~|Wv6pK%hd&6ywCG!luAWxelO(!K4lBFIX3L$Wu%>%d4#X-W!oWDJ=XK12_=Td40S7
z^$^h6fYsa@noNT%R@HyjhH-nsL-niVOV=+?2h*awNgpYPTA(fU`gi47oQwUnoqlU7
z$FBLVLaSq&WDjb{mpQ*%FDqT%0aT~j?vsu=1eTsY!5Y0aP{E`=9laiFdU!bUOu2S=
zSqhIXD^Hb(;Dyz66Pul*`IF?hn^?W5um@&GX#4?4=gZ0Xk$vy3o#3pV--Nwd3`W1y|Li#RE<9
z)`=o=+?E}z9jOzXjV=!a?3ZC{KOAL7DWxgAv6FEeh|?mp?eDHYTw>?C2mN-R}+N?;7;s
zGa6T>1jx&ycv8w@J3MRS(P-C&{KB<*Z$x&BhD9cSx!5{-R`DIg7$i;kVa*2*#3A9z
z#@*}#V~~yy_IeN
z*TE*}x^LlY7OfR1ZyKWi^l$35x&>SUh`qkrEg!4>@JK41*E6dqme#*Uvs9A28sf#y
zP|_~}e<DPdxZ3U`A43k2x}O>sGjX>6bnfs~MT;9VwrI`lc`xeAZWI2b
zD?Di_*QW6kUEmG?m1i^weN!K5Ew84)){H8y!ZTD0-F)!Q6vY*;`OXwAaW8}CH>;1UV;h#V3QC{hCW3<_>tXJBE
zFo$Aedn+!^H+$i`U_T1^91y`gx!>VYMW3z!`K{O5V%hI>pMag4xiP$2oJD~8OzC!q
zg#d$*EePwkWQYcf@iV?LAuJ@M$K10BY{U^1O&|L}%h@3D${oK9&4#fmNHMC
zyBW6#d|S20ce*-%aMpTz2@?#;JA;JgC+cIHN^W!YaL^_SIMppm1*7fG{C-1
zJ;GnQKX@}!hDvD-=jOV(Iws8qSVK*1x84nk!A!1y&#3C>_7!`{T>{2Xxb0b2&2&sFj&
z&gl6^z-J@^-I`VWh1P?4Zyq<(g_deKbY9zoXTM;Qizj5lC|!Ld=XT4fzi_+bL{WF!
z2P_U}ubzdr>FhEZhoE9Haa@S{)8)*bK(xVY4_Eu6+;+E@AX%9H#vfJ2EKRN%ssOSk
z+uat&>2D%z%eN(=hi(Yncc0c2PkvyVI3_PE(I|fHHppMqc1IH)mgC8!k57jOl2_(*
z+m?jd#8@GkIvT`2m?0kKT_w5iCNf~Mi_o_DmnQ1}b5Suk^PylZ%fzgp74q9X57frH
z*dJ!5%)n@D)-C5q-)`9?fycm`P5Tm5oJ|jB-m4mn3nE4zCqm@MP)=#N3+g*=kW%e^
zvhpXhEl?@TnJ#U)X@jp#0(%I{dpN&o*ygjzG4qet;sai|+fuTr!+E*Vf}af#O?y?%
z7nLqAf67Np>>kpLdr>0?euzFa!#Ii5Z3yyE#X+_>Q%!UY^D2yC5UvZZId5}JW^9yZ
zT=P^Yid4NXEsY}${Pol2d*61m8-%C$6Y0>ltV?RR^uZ_gwJgWIbU?Krbq?|CS+!h2
z^}NMz8$#UGHOAJj-Q8w}S*mAzlVIfZ1FaB0R+~B{x%Y(oTQ&<3Nag3D;h;~o$EPVK
zx0gxkYLd{NTn?`!|6Po-2l(%d#VW#;QNrWhO$3FbQzG5mD;uYY4l~3vsHgZN+>e^X
z4n~n9qMsccNQB}|XRF&|%kKnXQrYK)2tIj59Mk%~WW4u_5cGufA}Ze2@^-@A9G_qN
z=}fVKBLkVtZ5D}9ZU@lFL?mY=>Qt<)4&t2nm#;sJ+e$#GDVzQX6R;mw77u*=mm6SP
z#OpL&V-+a=%5Y@u-Ced`mu?baNPj9et})x@5YumPw^WAGW?|8`6p(2j`BbTEaGIE(
zzVw``;=#5Cm(L6VzyrIVnvJ&$F2>w*tBnh>>*Gs7V)oMsN)3x6?Lb$g9VqHmHG0rls9TQkUPem%?aS>wutnQylHT#P$m
z_R6GP>a${CG~iXIjhKNX2+LB;alx%jWp%3o)Y3?2QXdfa0sj&O#|8ixn*0W`5Xv5o
z`$zPupuM3PS3zxfj3!&Xdk!xPnm!Qqf*YPkZCID9FW(^w^HJ9GCBJA=izY^^*KY*)
z+^6Caj!&WQO9i$gGNvqfJe&GW6mgR{rB*I2*Gqmb+AjrBwLkA8j-m=4{`vliI?KXf+b(aF#N^16B36Pn0NRV3-a&zFwdq$~XfrzG{(qLg1453YwxHja+2Vn{-*<
z!4ik7P4*&5F;{}3()83!6hv0SbiX7$ob>pF+4|U+=`7txD_miOJdr`F4r^{_xpZ-(
zz$_Lpxqj5}Jf+@n8Cu6p-1dB+3p6S2z$Sv6g}d&Q(|!Imt3M93^LCC7{g_awUTN^L
z_GI3#?IjbFTs(pwZVH=TOP0`*e(~_Lv7SlHXU$i4DvIVxG$1Kq;qO4M&a2etOfQccBX#6W
zgTEbzT0ck{_bT}^8`*gFQb4uFb9Sd;wJ9XD7SZ)I+iS!C!)$I|iu9$?}6ICqCO
zMAuUL?9NX_|z~I2n{bq<_;P#k6f_2ik2if-Wev1DE!_w
z+^~$z?N$u}g}y3AX-a7|>+__pAs>-z&*1WNgu1v#?^m1+z339)MIM2I9iXCyc#PZd
zc`q>wiv7dr#I{%H4L&C+&|fW-gp$slVy-)gR*)?p-;4QRVX388KXrhV5
zE4lW#+4cb_+26*;Md{Jo3EyT#AA|@JSJCjWGt{~mKZGeCAF0U0T>;hrBtN<6UWZ^i
zdG`G}ShD9a090WvFg@r-e?8Iee0D`R8-js=U~scpK$DhMy`N
z#qka7+U-p(w6wqXWK%h0`$)KTj|U=74Mrlod7)Y}aA%6dqj_NBsdLK_iJ}&U_ysrU
z?#@?Jx7`ouQn_psP9J~erTnN0n6`Ub6j`jfJ&SO6%y5sFy?$&==agjh(79vR;JA}(
z<=E*Prp?7yEXAFPp{j3adTgmRKIxHVniVL~jIAcm%s@^rIMUE=+>ITH`F)umD93`t
zo$*O90glsY4-*k1ImZRkkv`sk_3!fn;#BRRHWVt@QI*`()z;YdqS?eb3{k%-MCWJ
z-J?0&AEz!1zB`Iaz}x0*5A2_1v9yUE`I?XVWlYk@XPS3;YNVSzA)=t$wR|GKT&#gM
z#>ZAz#qd&8H{3^sY-K`-nifuz?DnU7^@Ml{iC({39t;1_6<1Z5AYKK9aYh9W#LyQA
z!6SS~L;y_It17djI4KWpUiL3ckd$}h-N4p%EHM^6I__~sv7r6SC28s{=V>NYTZSu=
zt#Id`#n@*-l;1kh8Lp^|WKvyLR3D0?xeDj$XJAc>;RM_o*T$!?CfS|j!r{FE#hZl&
zKVn8{4uegFxrn;xEpWe(azy&t{F1TRg+KA|V+aiOO*<-s|LElq2$*uIgaM)~?-2n7
zO&FrMOD9fg2^(9vP1{krI+ukHeQizht$%YDB@YPY=ax>J4=S5cHULFHy1xe+wUTeu
zk|DERgBeZKROIQ0+EvPs;yaE=%Ayi=u8S#n#9We1!-mdhDKu&B=a^c6sMH6Xi5oIHq=)+s7S`UqZjBu=P9hEP0#UF&E
zNiw__QdW9hV~Wc+w~91)pl(GEabl7EDTGX~LYj-YUf;%7bZYoW`sn
z{hmGzd49(g#gP5=)yRA?NPyYeQ9QdAB*QIudoB>3VRSwuv64BtEZgykhCHX5NzFML
zX1a$)TsP!jT4@MFaf{Ya@^jF+_~*s2LsrX87u|F(V&zh0xuj}p^_DO6T1=Eh-CbMt
zpTKlxss(jhmI~YZ_(_+SDsi3+lG>bdmLmb}kB)Yx!lpL~5yaQ8o>2{YGJ!%v2PEWj
zF6NXL?TMOav2HfCPA4tIPDKIM?J|C#-h7oVR0ei)TZ9Mf%HUR&lJwaKwTJD`m28F~
zJBDqCX_5%vToIb(@EHGsw*pJO}^vwj7*FEpLGR*T~TXe1{;gs(Bgq0VJ+EuBFddDpT
z+VBW{^4c7);AkP+VWrra8b8sp+N5aN9h#c@^_k%WpgxqP@$BlepU#<5c^pbzz=oAVa5+
z^){41!R1DdDA0-+aBQf}8;vav5KMFp4^RzUzMd@e_ivtWpOO}tXxZo!6b2c}Q4**H
zSy~}kzME8^1QlPh3Vyd;E($)kL(1gde-=vRkx5Z2V3cK9NA{^d=;HQ20}&5VvST{E
zD)FDL3I(^Gjs!ev)IQL_@J>Amx24Mzg)SgK{o^O0g
zZMTs~01b*>Yz3?c+ZkyuH`0%&?}#j5_M4t9nvQ(deCEj8tJ+w603--^OXVkZX>yI;
z{3x7|1Vq=G?>SYPhoVb?a9tI#D%}y`u#gcfLIfMaN0$fqcbR;`NZbO+36!GGaQO
zxxDUQ5&GZ{NN4QBuym>WYV(*JF#fohw`v}Gr2lan_D~dVeObQ7Cy&+ZY`m;CSs>Ag
zp^$-WsA3NJ_{i4$N-z#Jw--szQxjiq0$SN_-nBtE)5NB5Jn-u!eSn~ta?9tULe34rpu#4zD4`$z{r}M!iK)}Zf-1yw0P8_BxRs<)T`|Z
zUG)SfW%2_vCtitY!&GszOEv{sMCSWh$>vpjt$&XaT%v}}6~cxwv;TCF&G({5Lo4Nm
zAs$I0Wqss>5zthvME&5#AdACqf5TOIga`0VZ5{7g$ZL)a2zYI=oxML@ZWSXRQqEdg
zdd-E~8R9Hf&ptU`0si(6#}&h7(8^l=VXq*gEB2OuYe{XpIibib3NPv-KB7LNrAlsG
zp30G$gAx|7DKx!U3;LPf94nWv#8^ICmg=oXiZ$$}7kfZhUQXEx;p0OZ%D;v$8scZT_c}6~
zFB6rYcU5$VL-v~s!~Qswvu#YKmRjIRWVtDuyL{#=qgPkEe+puyd
zC;H32zN!mNir|>gbUX`LZqTg@tV2^rjHNLX3VgzV+yI!7B&gdyR@yk-OLACT$BgNh
znu-lRU9M2#Hl487Vzck5!ao^IiMBF*%FM&q%q>;$9=d@reHn){>wuY(NC0n
z>J2T-frS529VU}l*vB>P>p{1f)KL*#1#BccEI1JljCKz}7MXeAXXLUQ97wE^Y4FZ<6A@HKh|g8CvHV~`L{wIJcF
zKojt?nvd~iA9gu;eco_zzYc4*TTwj2%1#E?c*a#P=EhY#B+z0ldfxmlk*Dw{-^7~q
zp2!;2uAeAj6y3URPPB4Vx--#7>ALwPEyit!vNeUe)p6w;JXh0>0ob!a*X%)xFKxYYl|)5yDct^~g3pi|1I!m%`S1TIJo
zU}3?4+Mj$!bPR`=RjMV6AG$qi)61284~#f^QprZBdQo3Hcwu`%e8@ZzcBwyqW#avg
zV^aU8Jm?bs<8_3jPa5aU820aj7U?LBhpB#QftB4Z&f{o|`#!=BK9SV}SFrN=FuI<#
zLA2mEUkiJT!?_~Nj~GqY8T9=nJ^jL}3TJIV0_dRf)c^Uogs%&j*gY;NPTNx6IL+dNp2%q%%u8)>H8xyETmi&jA
zG6nImi@_@qW5u3^xG7Ue<*|u&LA<|pKo-ko%~MJ!r@Hlx`Uk8qhp5u)>*ESO>^$tu
zI$kh9vq$YgCDsxniVYabMvO%v8{{E6CyTuAw{h&TI#?!eq5Nh|`&F#pBPjqDrsiv)
z4!X#uUy8bgsmEQoLwS!U1NWhCR+$gMaTA%OV$@Q*(yV#|#6^J8AdAaqzhKs#J3f=S
zRnHh^5jYy{I@VA^Uf<2WP$_$Cx-3gBCy7uuU;X|tnfB{_Q6fIC90|pIDc$?K(WD8g
zNFt?jTNZ$|GN;Ws$_Y9!k;O1aVoOx95Utm+X#@CvjX6|s{1hmETsTdS)oTQ1D1x_j
z#HeiC|3=Ne-F%$)@%jMWa-kxLo5zt6-+ay*WAQP*w9x{DMbVw+n+8
zdq}4SkOX%ikE}Pa^ZeCeV7~AV&6DoH&DY(siF(C+xTAKFlm(2V9IsMM%Qr=FdKD{~7
z%d#hLV&!Wv(5cydQ@-Cw3&H-F)n#5Bm(A*Hu54Do1;R=cfYocTky^Y-2T
zx6-AGz&h(+a>_KyIsXq6Zyne4`-Kln2qGfV(kR{12vQ>5A|S1Fcf**9(jn5_-5p~a
z(%oG{IyZX27<=^d`~IHi`TM>1UiZCo&ULP9*L@$BGQmeLFlztqXOVBne*_+0(Vv0+
zLe0Rz8n4f(2n5=M?3cU_rd^)_bW4&~UZn~FUx0%yRX2zl1d4AnIdm4NWM{`;KUq15
z5oD1D!nN+Puo6clE69Tx#Jy6v=1bK~fd{g>>`P*{UBV1lB!6qEn1b8jJtXLcZY^K~
zH`}Y#G5aTTkXoBWeNcri!EzwW6qYp?M})9laScOu*l
z4r8@f1yiO8Dd#$NyDUrQtZ_Rs2G|EmAFTZRQ2
zHwE-lzu(sFqART)xOl@^G^z@7n^+smvs$N1J@6dIann
zcqQF7QQs2Al=?yVcyty=61Dc0$tu!qt16QSv{8J}z{WuPXh)RfS6>wCcl6aCD-B@&
z+cjrNUp1xHD&u=uYqga89hhh_mVBt%=b*@~K~Bfyk@uERyZE=Hu_13e!NHr*_qjFe
z4VLG4LrFsTNt~v?XoQ_LO(mKOCYeL0xL)h)9p@9VF-;aB4}9+bbj*LuA$oW_)F-RF
zT`oIDx41u*wpso>-)ivJ5cJCzs;#6K6*F^6fvcdq!jw2yx~MNg^O`~8TMeHQUi}xc
zCRw#QxiSS)61}k3(T%}uig=JvTtwJ@
zy%Ev+-X2mq_+MSW=hD$!={e389O!ZDjvma~t-As^^tkv5y6S(N+nt_4AJLr62i1@X
z&kd*Wl3_o?-FvnZQj7JVTnAjcm;K&d3K7b31~sPvY00^I}+}Od}6CFK#bsS(SMqnAo$OQD!j$
z^~=}(9_ZV;#at)x?;ot#+zu`b?H%yhNz+B_Suh&}L}&prwrPr-3jJB@VrtE)2UxHY
zxh-nRP?H>|GPqk@S2H=GE}qTMj3X=ETseZi%}4r45dX--{UdbuKmKWf`+q|KRSt
zg;^9UZ51ZEIM$Fk>HdDfOGhDW?(t)akQOugxTr9@>-X{`G1Kxsp@9
z0qHs3xY2dEyAp(^`^)A(3o}gC$_PoG{U>m}?Zo*qu)s<6L#sh>_7B2%dBKlo*%ssH
z>Z|)`F*EJqZRy5b|ED)^wszBK=nuQ%?M-WK?>m*&!kvBh)<^^XTXB$%Eb
zV1WUc5(kQpz*_h{w;3ERPs(3xwYaH2)$N}9S?B*5ZcOE!1H4YZMF(=L3so5JHXB1K
zY(ub?)*Rx%pN+g4!uhPHr~0n)o(j!34l@6^VUz(B2b}Z$8c%4MqD6NeP!sTn%S35B
znCP+$7FJOYy|aH~Z-K6~n2ggjox7ge86LNm5akVZ0n1Z>$);K2p9;x2-?yvULb9=c
zrj6W*dBbqhxo{0&6q=m4erfx$m_UbDn3rI(wRl_MF-1#goq8$z4X58wp*&N&+xuw{
zOxCDA=pmRg)*i72;{UPh@3CDd+(k~j*EecBbw|`VLrni#)%Dw&wp%Uy-S=wBVT^gR
z1%>lMV;Y}H$Gg?QOUft(uWyTtBL4nGg{l-ogt{`^cC!afUN@1%O~(^yeOlRyuM#Y)
z#}5LY@tM|(6H%XJOhWmxB%a%pC=wk`c83KE`)rVZ?UlL>9F<5vLor$~zrfiZaJf-(
zOpPq~^xfq+S^~#lPSv_+NGfyj-gR5A?X^Tzp4e}5)iAwzVM7#CCkzJuDXvFxpB_qQ
zlxb!GImnKnx?9r*CCziR+47`wT}Q?o&jD`+My;)%rh=M{y^~@iblqhrPxE2rHn{~B
zp2eqKY8Q%`*Wr&K>4)6^*ME8oeJE<;rzqcJ!;9!bSsTOUcR#WXKh=p|m*OTl==909Tz7BEeG=w_RJ{HEaxT`o+{*
z`Y!~8D%o&}fVwvgRr5p8m(4-tKqMUfpn6sltP=G*J(w*otF_q)=a1?a{{psjF9X$>qeGJ1QqB6eHj(IV_@~l*^fM)htvDZ=~M&xZot?%z0JdiSE-dBx1MA(H?OR4|EF)O`hS^;r799iYgL5V)B^M=<#u{{C=a~M{lmIP={SruDCsZDr|cBcFYEt+@x%i{~DakcpV<17zD^m
zf+NpD#Fugu9opp2J^N=Lc*p;&HTmAN9&IWwX!u;2Blgo1GFr;lx&hx4TUi;pnC!yd
zItIT5WF^1BMGFy2C)(HA69>FtctoA>{qgw5BP@fCj{*b8va$Py5vvUur)XTjHILZa
zM4q&S-Xu9Er+2t=Cv@4bBwyZIJ7>5FgWN~C{#``zaYdzMuR&FYDMu45LmmW*2q3mVxx=!_bv3
z*C>&maiMdVAIZ&sv>^=#7>~AeH18hdWsroUCW@2D{Cwjq{?4vV7!a{
zt6%Q=Ui9J5B^JvM}!GM`bh#(9W{@$=lwfjO7Lcy_4jAdMe=e)J*Q)%Z0A
zY!XFz!GH1E_p5tMjk0lA_46ONcwM7+=>x6=bt^uR9@Zlad
zpJk`EB+IbyGmh`OyRpPpf@IONSF@Sk*_4E%(+;mMj<34P%<2Y9H1xqGYFHJG4;qTX
z*>Q>T_z}ZBp}oJz;!2%%z6V?~-M0W;hcteXQj;mu&hNc*;n(`AmAeq_r&UJQmzTBi
z=S2OH0I7c}>_*N|T~u3eTzjtbrl4fucNi1P%SQ{e65%o7!FIE3xO>+^ilL_yl`Ua!
zz9$&=vO(Jdn1aUAm_dU)7+b`f^#ZPT=^&<`nkru=0udV8k&nK8DSZ?fNmR$pbp@};
zw@`X`qpf2u`IxJtlb?gF(<|@8Iry`
zuM^rpwl-t#`Zi%=ZT@+pAlhGAwGoYm_8^;QhMKGujy9)f=myg03?igwGeZew$5NRT
ziUb2dxzmhP#7Tqmet1xLDrERk6j)-_3{C1EwCWNM2EOb^{zR3mO}t+f{DK;eVt3AM
zTOp2ud4LDv4h%30B{F)A{8%n2Z?^6uiQ7&|&Dy-M`tN=+g3d_dDbzRAU`eTU2v5Bo
z?MX+E08!|)Wu=?&AAUKY`14Qm?Bt3|JV7_u%-k#Q%r+N9mbh2}0Y@PZ9;If;mRhG%1bqy5BX8iL+8&_#fbb4OXH4g+|4%uZNMR=GHhb;aDZ2~*
z1$Y#YJw9mk#pVqzh4{yi)>mim66<+lvSH)j2|c60dPc-zoCL;QJ5c*!OSO31^%u|<}M
z3GzzJaQg-3G*GcBfJ?$XxC9tgP8U8L41FJAz~#?1g+EjtlF`mq%Ct@@t{z+&)Qgk9
zQ8AYGdvOITSrn0UDQ=c=kvEdk@($FR^ADOPzd)BzIl2t;Y@=iBoa7EKt|4ZHb(6AFZmv
zT|zxUe^UUiHRhbv}Q
z=$&UHKe!Wco)W^$gF1`kV9{tWq^cDZ*SyZh*}p%sK*yCVb-J
zi7R@QgGKrUBoz1L#{&DOeaU0<>9L_WXuDq5>ukl)+bIryN@vcJC()6_8vD#F@BsiF
z&!HiMx|cyC_Zw~LD9MN8WWd~pdZV;o$l&a#FfA#cQO;s;T8>K~E#x>Y
zkb(!0M@a{_lT--h8blR4o_spP%GeGkwJlD{6183ywJKVA`t%hn3Kp{JR}>n==ggoK
z;@7+Sjeehj`eGaPl_a2vmpNAM0d6EgE{xkoXGyAKT=kdG@PAnZJ@?H_9BH6?cF#g&
zYYAIC$64L&Z*?l!-+0UZT
zk@hRZnR+%jcm6zJdr(8r|Gg(aXF}x`sJsWylhx)3VVbnK=H2)(@WUHP+rH*clNC0f
zo6K{xYZ0ojtTCu*uk_1iOOhL2ZXmM$V1?Fww*@W_;7Jf~+BI3u!yDWhHa)UgBO?JU
zja~ED)B7^_|1tTZrC&H@3}n!T`-C>tRKtILi9P$_%)n1SVY6yKeCXUj;kUtUMwYgn
zE925By%(1`R)8GC&rb3Znq$-nj%^HVtdc#F^e^!FXd$@^znU2#vTXO+6bOcUzv(tR
zV3G^m{!tMTz4hj1*d>-&JEqTcO*tz$p7iu9d(1m4E+zG3et%*!Cjq>u3OZ`0=k}ZM
zx2(-KS~Q8cNh^KvKSgfBRw*MI&n9wfRuRYgF-MEhxc(U>iwLOyh3QO1{|~O8&W6d}
zN>h9^L9=cSGqo>b7`%l{rB4JaHZgBdM7OT+&IY5to_2nDgxj06?&j(t
zEk{`aISTX8+pot
z1i**nA~BUK*h`(K-_q-&-Yz4g6(Zi+v6J6EbY0T&F~dUJ!p&%ws%fNr4drJAS(plx
zJzeB&lC&or0rxS9J?>AeLYsPo$l_sqvHAjs-V~*O!Rsj;9`@GQE#(K;>+1^Tw^c}m
zhSSQBsD7Fi=S$JK;#p=DlN~?i>fl4R;;cUqnKbw=f>=`b$Rs)Q
zdu7wMm3aKfJmWiV;FT5^>I+#4QSvvtYUdt34QJ7fAShqV{Wc?-{-7uEV>Q3fO7Ar@
z78ai^%*vIA$ES*U{P`H1(~7za)WLInJ0`M6oINX%jl(0|qju%2U~??9HXkq}pTlOh
z8)pVwo4k(%r+Gy}3omyb#0iFI^KUR4Mc)XC(X+Z6OnwQ$!nALvdI-6v@M(UTDUlRQ
z_G2IlxIuy=CntQ8zY-f-)Py|f=)x;*e-I=y=coT_SpX{20%ze}V1
z>IETCakxmiZi}dt=#^ZG^)hh%0lMyl|Mqm(Fma&a)YF4UfNo0>^eOt!|D6ROus+OQ
z?(k`B1ZfLuy2#>lpLS!o%N#kCvuPNPConL!sT(%k=u@}&IzT9=7~6je>lm<)ReRZ|
z3l=|=EaFd2o7lKgH#0dSVRVMNF|v*~Q>0OEl_H`g8B2$+7;O7VwGXJS@R9pN^1`_a
zbWjehNkJb0;e()ez63|;zHewRD*VHwdky~3D?n!50p^03qgVRE;j3uAE)hk@AMap%
z*iuz;q;*1kmJ)j5I5S)b?6c!1(R4E$c}*7_gEKT72o!8lA#pd}jTpd|XTCTX6pE+B
zeH-)ORthN$uY`4@MowKj>hH5RMp2by(DAK;2zAKy0Zf=?kyQ?`J2?hwAZkQX-#(1e
z2iwc&P;_%dK=EZ{!Zr!dzEivZzF`20(act{MMyGe^^{)WfQ?FFT~OVvmPOAFiQA{m
zyF9xTV);3Nfkt>I*i}$uA+NptB&9IvSj*SB|I>@xa6tfHrhf9uF8KG}?L&}(SRP=%
zZD=#v2I^gGu)7s*f}Tsl>b`t)jhW#O+zVS?njpD{pYS2kf-2{auWZ&M!lj*$LHe%5
zzR6a0#N@z<4IE9A*LH%XdYK+n?uaK*8L^^H>H(=lC#36FNiqwiqkYZ66JQZHQs|YFemFlfNWuW&298sP`ZsJb7g18r}<|ePQ#$2EPH@c=oSI)wQcnLE!Hl
zV~WHCV}b!j`R#0dKN%I9bcC4ISbRLX)_y%<9}M}+ABbq}N4n!C3CG_k7??;a&+>9}
zN8y>iHW~GT9$qc)ff#KjTBd_@FKNYeN}e9?haOfJEi0|Q|JI$TC0+0kc(2f4qt4y#
zb39;dchZUFg`_W-eXVj1tAXSdD8=c-5rQU{T|3I3oG6H0+u35#Ze|F(yd&Hh;|u6A
zk4m0k&$*xH4Y^V$
zItU6Pg$~Q?%=*_yI<2htDTET)1|;z_jGNDWF=#%{Q(F+p3hznu_ji-u4(>n0Mc$`0
zTYooO4WxPhRo0GF7P_0gmUS8U3+FH01a2m6|Hl-XdF3HB5q8GI@qrlp|3fru)|zwr
z7eo-;=g9ZN^9B`9dPjGFc%oi4lY8&~aRLLvne3Q~iD~p9H}f*_E)E6pXHY^Al7gV|
ze(gQ?_`Q;`Tbw~|)59pUqJB5t-%hZY4V7v0)s;vWv}in@`eNayC+mZYS{^w(8>%&2
z@gbO%n<~_#H}3*y^gbWfWdY3|5ct&8+3~cV&)w)^e`Y(DTv4bsrFD7d`}LQEI5Vr}
zn{lU;Wa?YN4Q*scRG1`R<>+wAP}!vIC{Wv)AtKvTEAUDoRc-QaQ+kZ8HJj%Ucv#MM
zfs3C?v6<3ZJaINHGfjosz?n>9eCHnXLebZ~fw^I?X_nwULs}B%xT;F8XG{LbQRf@p
z;r1dHkzVa5>arGgK~)Slh=4S9(f@}Jk^2?AL5lsB$6yV_%JQqkfeLTS(pcJ@p~Z1I
z)aQ6j9tMyB59G@D-eViJIf#HJ{loO@%x{3It9*q2>UJ!QF|x59P4RrqS6>
z=?hyB1M^M*95lbiU0pX;WJ$DER`N>s2CvH9(=U?I?m
zm{;neJf2B}gSn2rXWl#`U0{f##?-sWz_9Oit@WE0m8hIfw%Pnhyw;{SD~m(0RFVMi
zT|;FY>^#*6Xfu`3aYzLqBBAi=+yVFYKsa+;O)5;6_)V!$CJ(Y+V;`*Pmo_`j5W
zTc`LJkx9*KT^lqViXlNcHF~PPgo;g!drco}EEoHmT7M-X+Rm2MT$lgmD5O;%j>zLp
zw4!)|2(~76pN5c+T$xT*_qZ>gd5JZ*Se4Cgrl(N1?Hst8Oxan~n6}*~nxF)?vTj_rNiQY{I@vm{p7@v25R
z8Ze3xKXw%Pm99n*&7YjNcAj{}~Yw)4W3X
z0f(dd{`P&qiiBNaqb4Ug<9TBnrv;D;Vr7Cmj65s(LAxM_`}(?Ey5K%K|2(gKbKJ-A
z9hB3&&jv;&rxW5H3ds)0=_+4mlbIKW_)w~gFX6eWF4Lr+aooBFow)q@aHaGQO
zTMi~{okkGob&N+YCkNM5LQDM4rKjiRp^PI4p$oz}^tYDdk{?;Ee%6zJQOWYsA^F}6
z#^3f=<9=GDq;n^1^o7oJKU!UCK9+4l=s)7_hp(#zu)7)5#4i2F^t~`P0LDoz$F^LV
za)&^kcW#+;C>zAS0;Zy=&=9Gt_FBe$LRx(bvDfSz$tji#Lhq;tw>W)9na!=#n1Kyi
zEm=zA9P^5w`4pEE1z=yG>t}By|5nImQCY?$a;2wXTju77@YeD4efxLY`tdX*KDblP
zO2JCM{kjmipLFV=h)VZ9ob~TjKW;<9vbVsAX=y4-v2;aYj4fNdzM((ra(Y;zhRFb$
z@_0mr=>%7RDT6wp40B*j19}lUmRA=irLa#;eK8rvI3eX5>y!M3rTV5})#!F?y`PJp
z3t+yOVp_BxXL^}sF_5xeTC}hAYr0_(x-CGF(WCd13&^kV)N3>>-3-UnoI~c(^{)qO
zws16MCu$_6s`kQv6QCd9GX{SL{Sy4!5cLWrazR)v@*R5m>7F(WFrOm&cphjvrt9Yj
zYZBZGtY_)=uj>G%j|`s?yIHxDoHp^ZGqyMJ%kfwnCWl}5j&Y0I
zdLP+Y%h|`=oMp~0C^dmBTvb9r$b;#@qoDmQ9F2g>R%PmATg)x3wXJQ%xQ1A?_UyI)
zoKyS#&T^@DMe8#TKo`KrgQm;U5^l~_!n&+!r^y`$vW
zo15pH2#s`gE$Q=b9=sR7L9U!Rw~9NArT%|^s3Dlwe&}9w5h&fZcDELk;pPKmgCJul
z2}1b*+lIlR??O|gweQaRJs+6v6IyG3xP7Yx$f&!bxU+H+tx@
zd~`Bz`eZqjih>^0a887ONnv@PvDrVF5GQ>!#`_|n4=Qvq`t>p|4YW71i0v@qG9LU#
zobC+CFE^;R>^rteq5fA)6;Z6rn+^o;R_}kz3c1lQ;0f*{^!&w{gc?~}u{9ecSsI;3
z`qz*Nbr2U(HV0_zu6JQ>0Q?clt&bQw_WQY6UDjoXn)x#5NIo_Sg+NRIsiO2lFWq0W
z%cb!9gdpzeDEK$207OgO_8Zw
zc-jrhAR~cBFDF~;Aj~I~$AHez^LSHIyE}@Q*B_nuh)_XMk2%)MJNzI15WZ;OP4r{Q
zU}$gU19!%6`JSs!g)N@2p7oZ0NUyzSLcNd7^R00i$_n-mSl`WBk&Uf5%&Ar{O4&W$
zd?2c4k8NHtBYKP7EMI#fX_G+1o4G
z-D(N#Cug`!XzFGyHF5Cs=R7SgoM=9d@Y)3smpZa95`%G*XVo09KsoX5-^}QcG*!mX
zuW&%_f}2@`4fZ9dkQi{eiaenHHBXO>&TD~XC|{ZEU38R6f-mlbd}N4MeB%xJhyj<6xa;fxW&1oLMhs+MP6-Xh(DF|#xoQW(XJVg>uYK~LS-5-
z?d!75-|c%}8**^odC>D`R5@sSRY*$Z5&TE{xiYHES*~!Y>zl8MMZnY;W
zr1-?~v6()-_%s(ho`Foeqf7GS0sT|k{gwXmL~fJ22cXDW_n@C
zx5{R3IBGIl(7FvbzN+YDoZ1an1LN)=%&ID3g5b=EwykwCbvX51FB+xeTnadTp7h}*
zp-g{+T1)L`dF+v^?ZqpyP(zyqUq|hT;E%vWmRB(&@6ZuA+e440qC#lO0#U+cZ_mYF
z9aFsl)iq?Te;FHkPdOhwQ6@~Y9I8v{%{eNx+++=IILycb`6S>%X16Bmh^L5JqKnz8
zBx)UF&$#@K*T3TI{YWv>;q$w=1fQMB%^ZK@6TYwV-LE2GoY)TGE7%#0?0%t@B^aT&
z;>W!zeDE|0psxNT;*i$AxyoyQC1T4m@vh<4Tx8+Z!PCRv52)>a=0cU?=r?7d=X*4!
zt65UU=B~E;*niSxma9}inrFp`yEGI;`Kb%)=u0RImKDpV)(*d2b{kw}*e-xuq`*vT
zdutks(waowjnPb&g_c|4Vf_4eTC7w*Qz>6zSM93|F!|5{v8LAiQ%+JCsXOm)VZu#s
zHFD;nQ8qIX)UWf@nY4|tZP2vmJAmX2V63#6yYh>ON;{F1vc}ZAfE*GDftSSQdPcz&
zZ!XUHD7*lZ)nqO*cnF~u9sjI}=wn;gfTvZj-BdN^K%4nX<*_<%1hm+4!sCIT)P@Vm
zl#-NC0{tVVfw!l{&=f>N+R^0aQ=F_1KUD{1^4^OZ%&jGi_*Pfm&kY^4RSU0c7(Tmf_S
zm+RXA8jzQ`XS4Rhud~^+twx1QtL|ITY9Ybf1FvZM;oe59A6E_sVg~OX%D%Ii8;-`5
zo}0saCuFSCwh1x4POVh0fAfDQ6t}BpBXL@ftbX42)t(n}-gIuW5~+pxMyF;9=FIWJ
zwK;yWQZ^qtJFLb6szp6)WOpZXUg}V2TPyo8n7_FN4orffo48XbS9Ob2DQ<1|&At9o
z;>d{DyzA@Me8>9fLU(4P1`7h#;}r#VO{@1AbG&8sR9G%mg1+vxz+KT2&fBZDRF|Fg
z&e3#H_w;JbtfdcbeP~7wOP}4)mgRfU*0^mzys?5XCm2viJ>3bVgMv%KT$C
zkeLMs$2-(&RHBWsHB5Zzq&EPXM;71BaPqP_m>(rlU{RAMe;;-wZ>0zhScb7Ta9g>9FW5mTCL=iCEq2?A<%xV2O`OKv7dX`2euv*f
zgl8Z1HGlDyiKjjWO-VN`_=Ra@7gdwZ8HpGgK)qR$VRG4jpplz5=w7l8*8A=+lzRK^
zngsec;u4RxAh2w}{7|lWn+Ju2W#mL`?{$$bWf`Te$p?bg=iSUF76WdqqmJ(y;+fky
z7F*tsUR*-M77?D-6=~5P_i3;z4n|!n(aLv%y5@x%8b0Xmg1a(y&oZCwM@9quVU~9*
zTsof>{L{U|j&D8`l{cNX30q(8s_#xE)-)mqM(#A}d1X!--_VFT6dorpINT%EOxDSM
z8B`?m9+?Vmt<1yFvH;XY@xY_uPsu(jAY0;+n;Ii<=oJg1$i);Su3?MVNUmDmLu_*|
zio_p6g#wW&m#^(x@~MEzOqQ
z-{Ko_TJzqGjPZ>6be$TCMO%m{Ax7V6*(|ox#3x633n%!fDrTgYcq{y~_
zt1;G~8q<6!j#5;=!oRf)E!D8@>nQo~cq*$K73Da^r0sA4~Q_7R4zNu_J-6E6C*rgpnaa
z(ig6N_|(R~Yhl-3qyT*%ZGj|s=(Lhz?ZZnJX}c!L9gZvAdNw`EA{1oxdbiLqJt2tT
zXOI==+>0wolM+?j)gU>I+wVjMg2p<8k;@QG9^MWpGpz>A1k=+8H>12&rKY@Tmx!gN
zuML7Xn@hKeP6ta(74vatyXv+zgonFNlB+(d5H2i0ybM)Zq!r1FtfO`s{eGL=T8?2B
zlC-#&fa^@CU0?zPy`k&PGoG9ajb-8C8L2V^UBMZHFZB)<&+dkAkqs$5GrLwk5J{or
zn9_IiE>s@13)6~L)*l%FH$ce0)~>Rq@SC?G@T@$I{qpA7J*lh3wQne}VMZ*qzM{Ex
zW30EkTnpUH)>RQa<+~f%TpohdI@tLYKtK5n7|^VP&ZfS+l?09}555_eThy958E=`8q9Zr3
zx?@DmFg0=e_~o>{v|8~NYGy23?RJ{L1b=XB{{ie)JJEISG9%2YFjXR-2oC>?@vbNo
zO&x`biHZZ9^{_r>cWTGI5!Eq|uXOzT$!^R6LQOa2lM`XlpHInw(GG+*jfi%Oqk|R~
zbgOazygqa$95be1Ww8_a&
zt=jjEH&l*+BvM7uN(YO+5A;+@l{Acf3)!=Sv+Elg_AmB%Xiy6YUdIE*d)xDCxxGMs
z-Tsgo8co7`qb)9D6Qsb`y!q(d*PyDHLh0bb;Un;|r1ehbapiG3-3P9$7^qE38_097
z{10;nUorJ%`Va)q_1;F(S(%d|Xg;XV0l{lak+OJq7Mp_QOtM(LQxpmB(8}6AjC1@O
zRYvPxAz3Ev|3#IuwEd?C?JOcqr>4rWcj9@r^?b-`u30A^T4JuQHN??KYN$%Z*e~a%
z#yX=8ZM91=LDt~Q{m5b^p5Q5JOzVNunDanVF8bNu_t&nFmlcW=IzDlgqZcVYO!v*b
z6Rti2LlbavaSNwPCBC>OsIC^sL?|dr+hqA4StX707DEw
z9IYp)EU&;w8ChM7x74s!$nK%DFiMxM%f$_1=$UHEXu~1j!+nm_nLdaweArKsbKRn)
z)<`?y+qr1}Qn
zLk~7U``NR<>b4j$Pp{JtNKcpKsJE++f_x@HW9
zw^-?kxzn?6sMjz{%xVk0G&ir+0-u|0rv0Q|^q)c8fw^$V?rjUrG(Jj^6X2@LcSfVL
zTM$CiotL#tidTVuzmPpRM`}Fu<1uqXm4c#=N-pgU^EYH#m-Y@Otq%Q8?@0kmcJ9et
zrj^7f^s+bSEbwmC?n6tu1FQ5>+@X!-fBb9MfykbC+})|CDy(5y4#7
z1gSlBzg!dh*IVO-ctDHz)X~k7wD{F$hu5coS*+&mG%}l?^EGyrPlyW^?f73Z^UFj#
z6eJrA)NyEF-$o2$1-BSS@EDNzqiTKyPutYFHZA-6uaS!Gz^*T^@=)Jy9|KSgHok1v;~a{cbfu|QhdBpXoC7wwwU
zD3Oqmm{Cnf2PBE2@2rXZycZeTx4MbSV@c|E|w
znAhz$@zijE_J0-cINU~;g~9HmTDhs>X3#CJk?A69;I01Bt8vofm|o@-
zcV2k>P=(szee1F3miRLup=1Lh`PH65+4~=B!LBc*PL7y`kykq`1%=kU8;JXP&~gH9
zb5i(tGbA{NenR_UeNudYvA@3?v#6JFy_B=}9Xn_ve(`2QHDAsIH^P`eW5XYJZ`s)R
zyy*=CmPAEmVGYtnX@r43`Q7z~iJu@PxN?We1swjXrc*mf5ZsLN=+#vQ!7?YXDT
zHS0RJRG@U()=G5=pDozS{`L)f3Cx~mVL2*LtN)?)!Q!Aq!gfn;>$j~jNkl$>aq!!O
zWcuPduw&Y{Ulow^)ryQ?BSv-Xg16{NZ{C?-voV|{oL^B2qXt>v3mb9d;z5CzM83qT
z=Quc&qCYx-ek~7=$cb7j&~N*@a)N`7e#pq!1=qf{_723%d*%s7RK!vl4|jwIGEtFs
z7*+pO){yL#C3JTbr9SR?WqN!WR=kQ-fZ8g84=9+pb9MVDfW*m;emVODNyiZCWi6PE
zUfF*pRj_q4)l9v6%a^Oo_R-Y;@OvZJ-V14T5LkUZF~96%LUsy#Y<__(DpPpX{4hqI
z1Bx28gUw#N1T-H9npd~oR&jHqvbSLk^^cXDC`pO+$f##P#qyUgUHO*$(V{fjOwOhtKo?
zTOW6138C@=?4M((cXlVoaT#bcS0S}J@VaB1F8&nbFVvDl`E=tT58dj*{`ZL(S_-n;
zJ_mM={Nzm7e3`hZ9e`8=j5@%Q>R*%ELcO=M&z!oOkj8$izjYgBn8P;D7BE?rQsZWf
zpB<-brRLz5^Kux9yX~8-+cR0{E~C^f2o_=%qFZxcMnSR^2UIRFKOwdS%Jjklcs0-`VrDcB
z*|1GS^8-}%t^`+<_KO2ung3xBT;_Vm{~)6mqAFFaWO9xjhp
z1ubQyB=4nisLzX<4T$Jfg)jzmm9uoX-zvyq>L@wta~PdWZ19bL0D)z&+w5FR)6G9K
z?e?GofO^K>Cg!}9cy9`g_c%o6W@B2CAnOB%C7y(tD__;JpAA?fla4}=iOyc_n!YyR
z!QH+fPR5Iah~zg~#`*!59fhqpxNpNA+)BX?+ADAWGrkU%jVN|wuXIPcr~>z>g5;Wx
z&Uv`O$;CU@wERRONCaoazx|>9?k}u7s=r9g#^f34
z>~f%-81z1ycf54Mu%{vT8J*3S4{9WrTfjCwbz7ggZ#@>x5-gVw6&ViC>sxBF5~|EJ
z>qxlsEgq^4^C{83ujfFTJYV(1;9PXdTbBI=oZG@s#9yfx6>za9HyDw@@&B;*9#Bzy
z&Auo}0R>SINg{}nCFh)T&UpZloHGmzK|~~rB*~I9BN+w;FaS!HzS*%>nn92$;w?T@GrtoLNjR>U{-dw2prhGsVYl_FAc@5&sbOcA
z$R1XaUq9y)z0#YeMoIBqt8{~S%>`i_GxepCvF|EwYLlxrK)h?Sx25`gyK+FR`xtVW
z8ZbA(1z$Jbwuc=V{WLm#PT5;9xnG%(kf}y-oIl*weS3h5ty$Wn$*Y#7z%(XXFA{)N
zSHLEBKIw_n9ZJ;zYqH|t1OG5E0u~oDW5Sf_l9|sa&S?~IEwQNl_T+SE$qQt28wo38&
zEt7Rv3Rgs*tq^`{eRe{E)TH@fomDTEV1k1yC1yy#gGXUPY%zY($N70fd~^Pr>?#Jv
zKf1bdmb;UO>l~-nvH~wPLhuRw<)snw7DSO1~
zM8UAByR~UxYaMu|kg{7>`V^-=Q^ZfRRJXiH4-9xnu_-(d?KlDQ#Df~vOR0E&HJT!j
zh9XL!)TZ}wZvTUvFE`x(ac1}KUo*S47d$qtMzyW@+b5-7TU*y%Y}zf=I=V|?K6f@iHSoN57BuFZW<-n)#8ifcv9FIvhbkSn26J_K)x(ZlvV~
zO|wCdL!L!0Ghb$dTQXL3Edl`*Q=(6A1no$^JAM7o2GR1;(h!^~u;9l#dAqqf{O##lBg&JVi4eOg
ziURfI&_I@g3K7IWjqPxe_wH%xR>HbB
z3xezXRcY*%1e;d-O_M1iht~Q9#$hD|Gr5$T^@eY3kpdIm9sm;|Q`IRyyfIe=+Mvg;ZU)`#Kuu7CO-Z|Q!<+Tv78wn%b6dU
z$v`c0H0ad)+A*-K?~CCCq5bO}O8}h|9%B6EvTSyLvD0><%Cg3NRduq)mWaoGEUK&8
zwJr!BQANC1_KtHXlW(@`Op;^IzQZ`J+{}@HLH8
zu|U!2#{xsYn6Ug#Y)_I`QgaU!GW!mQX2mY{F!4#*gY3r2iiss0DQ8x<42~<-4+nSc
zZEMS0$_xd!Q@3>kO?Kiphj_n!{W{`xvNJQ%0`eQJ03L0~@QgS7jmfCfT=$7Bg(nM7
zuqX&I;8;H+>pe?Gx{uLviLghk<;L19cO~Xma5-V!L<9XkJ4x_i3rmjA!r9`|^m1u$HMzIEox7}~ULk4LO-!XiI)t4!vm)W)RTiT<
zydRKjN6HFWqnT_03BCStJMXrJ&5eW@t~H-#9&{;jwn!9ZGE6$WAngqwmGFE24#)b2!w=1M^tu(koVB*TX+4D3`+zzb`SOZ
zGZdL~oEKI2${3-cq0dEE=K{_h?+_3W^eB!qdbURpuV;AAYIw#AO5P|FG%So*G(1Fb
zAy>k97*-2ngE=BRJo~{j+0#8n%lMkwkQv-8nf>rNub`k|xvUR-P;Q&^i^AV@!Qq1-i8+5kiSP?~v*WD^kF9YI$YGxl
z@>JVhc%yF0DYvZb1xj7BdNimLZt};dy^RsQVcInUnIA|gg%;c0V0!dL=`Oypjp3nh@hm)S3e~$@A<}P
zDT(7;OGb$vc<0xQ{Jp=S#dd8nk=&UUOc;cCynK9mjZ+*tU%HCe3B)egi61^3So#^?
zU4-TH;%1l)Vn)$uqIt6^A!i1wF!30aAt@$X=dl~XfglnKB-hv1yO0A1%iW`0p+if-
zet%e?R-UI;L)faBRzds5(@V>VPJY!4$Mlrz4;J=5}mt&
zmuH1;a&W{ZnAwYwk#P*y73*fXsncQ2yA>*bUd+CvVRxM5hHO?e6&W|WJ*1>madmZt
zcThMF)WQIpU>d9u_kgoy65z#t=Q_$mc&zl033~6OM{u#DvZkOC>^nF2^cQ)Q(S~T0
z7HK^Q_S_sRPC0v-F6y;KN-5~^4gg%IHEeXFU0z)!$-jG)(cCPIKt+~oC}1ia8PRT6
z{Ag)y9nb#eZ!jT21ZC6R0BnX}0^E0ThE^%|9}*K~jOv~81CWTpBbX^T>U&`GJq{m>5=_^PJO(1S+X0L?s22&_@yUe(<&I
zX~Kz8zn3p_pu^WMX5FZn*4LUz#{Xnh3?n5a#p7Vjz;&3cQMmPWR7?_M5fv7s?KCXjL)va}
z1y4COMF*U3*Aw>SFd=2T*3A1A5LTcVd5H2dlN1h`>qy{#|-wFc-VlX(shL=0thO
z4MT&gRc7cta&Gxsk5m;EbCLl-1s$!$+);zY=bd%18tzHmEM1{O$D@#Eo<+1@FTf-
zN}tG>FSESumE>5ej?=1pxtWlZO?mezvc=1!uc>->E1=Pe8DTwvVy+*>++<7CV>@*o
zN&M$){hKF8C-x{AZ0pz2e6{S1b0{%lNg|A09exX1=iaMAhxgoN(?H>-W8`|WMc}%Q
zK5tZ{w&&2P4e<#7&&q8cmgu_C!qGlkBP6tL9S)nGfX#uo?N4kwtB5&r0X%fq?@8YP
z%$X=11(HiN1TK8XB^O!|MOe%}1{PUkn7}Fp+hyO}GFL&G=9f)lAZ!;o5H8Ca5O~!W
z_%FFEbz|dn(>oYY#65R|jZyDf$-WyQN*tBlnGNKG2~
zI)3rx(e?CZNub|PbHmSUSt~5Y(kNp&VG`5hlLA-Zkh#yy0eK2We%!h#8UBdd_LU-U
zQbkp@02Ov=TnUD9?^Hq9SI*V@In@sqf&@KGR~hW43h?s_3J_bi1Eu`~Nk7K3fGqni
z={LJ!A*}lkVfPZm#b~)2&f(4PXDP-KoYvCsuZ`k3o{
z;55b?8clYp1C|a?KaQN|Ee@RDvIw+g^c@2&D
zQKq}kW=;l~$CUR}@u85t&63{K8k;DuVF$cSp|Y{lW@AJRubcdOsEd3|{QZNQ5-a)Z
z#EM$e9Tg4nl<1|kG!(GbmQpU(r&_KO4~%`YF7lM*ZkOiI%HCBl)Fb}U%a&)J<(yr8
zwz{^CezfkRD+)dJnPbbrUSzcSE@>0-D4t
z&O>@gPO*3O7s_8gPO6-s<9_pv?zem7mA~M7WhON?Fx)Qwcr^nDVbjKw=NUae=~IGR
zMVGl}_L2AOQtr!y7YY2|q5cKu@Jot0V%))Bx7_P{Lce0H_n`kDL4>
z^%9OYP9r(CYbflpDWYOi%VCK~PQ1(PwqAKKQ-BCAJV(4=a<0)S9}-+Jc5{0W8O4W@%|YqiN0Sf%(IO%
zS~AkJNB*$pLRgs<2z!8@x!CxaaB$IY6FYuak}#st$(!)nzVQU8A8ir7)DbWt&m
z7Ie7!nE>gZhSp%d8hyT?*s1F_0SgES_G~j&aAqsthUN=Y=rE?=qY^MoH7ef;286%X
z0P8Atpx+pgHAtW$RzB)miXRqJ3K^<$_5lg1D=W-cl@Ho|^&(woJ#svk8T5pMXBoRz
zcybg(cNC*PMhL6FSF;{pR`Y0RpJ!tYV)`5hu$rBU$RiQE0>;71?I2gmNNB}z?$>TD
z;EJu@Ph&tBK%vp=4lfjsC&y$xSyIF(~ToywCY2OPkw!&T_ws%sdjLhs28XC3S^;!TY%4#{|@7N(ZD#F-ZB6iCQohfm{
zW1BJ1n(fW+AN;2i8N8|wM|;cH#+-?4+Y(eZ>d?2su6_n2nKh>foYhb>hEzBhshCiV
zFl9d3(MqO6W=K+j<@0~sS+<>??-(hrIPdqUcQ6$E{hgqpFWhKNN9|iR$aMXCwI(&s
zO(0tIfWbS4c7%zt>?j&csMOtAlg*^j5i=}LJ%Q@Is=#xG>sP=pH@cbicwiUa%-cQ{
zO`Lnlv|5aFbu}+aOa#(HHr_V9=u+T??=DQ7JC1|iv%S1$bL)rI+$H7Qlzd_Tl1-sv
zZRj-?=G);l;NFBNdAOwZXZW@WmZoQt$GE$ZgJdZ`ksV)=U88-&$j$hkia=K<&RQ^~
z)=GzeHtX`{{Fk1~t~lZ%C6UN}kg1)D$5X+5-A{TaIJZ*)Go{3u!8eW?as+>flZphJ
zv$f$QSSU4>+f(G0>}|a75Y~5zskpMEh`z8B3g3r!rx}tZg49=@LGEYxFb|a0V?{&f
z7@*`vOn;}l>{4RyY8-;L$DiWwbBP_scqL@$i4%7x>s2pM(y_LOR-c8|h-~9s$D15m
zxnla?aWY8b!k%Rgls@C^>p$0|@5_oRy(t>YR=!Rc<Gjm&;~dQ?`7a
zntR$&L#2iNGQb;CmTM1dhuPxMODO%(Sl9BXCB9~*jS=zk$XqZfvCtU5vH7PeMy_$S
zIM{0*=)(m5^!{x?UFgfN-`27tH#PHru%|XGQHuz>lOVm+{9@g>6v0nK9)?TY$Ti%4EoJ!q0J#uAxnML%-X)rp(%;*Tt-P|mmX<198*mVp$R
zTLQhrF(bZK4nHT4cFTb7roT?g|u2zb}_%mi%l*ZDxUTypUr-&$d>YR
zYptbs!G@TkM{h@@IP+*|Ue-ebbdGM<`YHr*vKgZ;NHN;vmU@NfAn#c!TjVCdT&+0Y
z9(Xn7A2n8(PS>%X_+vYV$oUqqwnoO8-$^h*hRph4{anLQo89`Sk}g1)k3OX?&q_n+
z-riOPX}+F!R+z_)>B*-DLEV(m((jzOst;{@)PyGVO@r~zju<2Gk8%rXhkmB}cd_xy
zd8?`Acsi`L#^N7I8Ft2#PPAGqLgnu?{X7Ht390~+B$Ih#bGfYVu6;=nTKDDs02nDE
z^eiG$ioa6LZuMmse>K{o1^+tzDsx{IKHu)i5IG#XMBOB(nxO+bx&k#{yzeShU3P&q
z$vJ)q4rLabQ7&vS?&~V+mzvVM?Mk-w`EYP(tsZPT@g|_P9&=`@y6)0Ld|SxkVFa+e
zhwGqoe_o^XO%aU*gUh@#_&!>Yl2;67C9vM~8TRDaHO$t5_R-
z)w^M}U{z<<>3-Q`;fPoPIsE?q@!cO1t}?H@prre|%|+wmBFH-4k!D%$In(g#ujORD
z!7}|OWA9a09}T>wykfGrXjdsII;X|i?1b^qGj7i7?vGe308?8oaZCp5QX#BY3PIE1
zN(4j1jtjOM_$;Gm=q!tC3knoAtYSh{lD=V<6_0Ia@+{5fN5zhbm&@nk(3+K=Y~+bq
zuC-`X7<_^WALxRIM3-}a);B|zHUkN?b~_S_WhD1~w;|Q#8bHruJY&oolM_Fzvhu8B
z&3WUt`wY?cNx=>bV1r^BlYXvcHe16`#J=*m9BZ*0ZsJ!RD1EGrs+beM;@oF+?bi`q
zFZ^7aa>IN4$MEt;bC^qpk*P+;dEaH<6q8-O9Eas$H?2Ovtd&PL$lJBCdeo#08Kxr1
z8k;3QiJyU!$&jg^<*10i!!=m5syXe~EvH&#bV?rBGbUd)Z$(mhnnql6rl({yvRP6C
zoawAS$U1W*V42BKfO(avE{1WBS#OTDpm^Ifm%n-3Ob~;8ZPV<{g4086T;yEFwhj6D
zK=K4)fpwK9bFc+5$of#XKAvQ7+|hwLS(%>kfuJh4I*o+d*eTkAzE<{G+
zS+?lWbKf}sN~Xlwi$n3j%78-KTKUxOmV*$DbdEbR&?kSr#a+9|HvZAsnVN}o3Vq`y
zLEbn19Rq=qI$%IwufW5u3(!`sa8$^w?xJ|T_437K0BrKfvEJ3!W%En0Tki82g2sa$!i75n{c<V1}Sn
z{isIdsmRTvs}>g7wV3YgxcV46%$9FLs>Wirr@vAO4gDL4VZVaYzAo
z?>I?+2PsI0x;@q&t6r@Ef@<$H_!*s~)@T}PsN)7qQ;d?zO
zvzk+=YdCAQ{VFibn{OZ`-5C7jE)*Erzn2z0RQl_d)Z3eEP(rPZE7DSFZ?%_ftwj;$GNNbOCa}IlYl}cG-X$SIYfCQ=wW18~4phQjI~IhvY3u__
z5Ph?FJQzTcFgJp=Fkubd-U`$rAq{npM1yZUqI-Vx^dA`4>}?ASM~XfqL#|(4*@twr
z_@yg!(Y9Alc&L5YXsbZ1_lpXQQaHat!NAV74S;m5-iWb2n(7@0i4Ws+pD=mfKG%x1
znTXaB%sRlV!M?#WFh$#S(=ev`2Q)kcZOVKUCI*2^(4-SSlPmQU-963IRk#yx(cI~g
zS4qDp(s?aioN!z(@WN&eSYu;-8`OwkcSLz{7NVT>%*`ElON)>PmSPw(1f3{Se{gZnTH1#mSlW&Tx#pt4X*diLpBgop2@
z53nQTd5S?OgnOj>9*JypgV`CfX|f@ui*TbhPeW~CU=kfvO|Z{&mE
zOhNjOd|UhBwIt{I(uvc?vrw7!_Mlb|2kSco?ka|w?^K(K*G6+z+nZDeLd14+Z0=18
zUyo{3Dq*OYq$rwRrL8<#AV|s^z!oOu`mjq8C}T01oT>CpbWT#hH`LCoV-Dz9I`E&m
z0xU#DkGSY+9^uQi7Pk#6!;c9X2^&2V7!X8tZ&wfty|t6DA_~Hz1)Gs<{lGa{u`(ge
z1l`tW)pO3)&l&+c7XjFy`hxKWCfnPf=5UCL{`J?}G#m-%QZ4$4imqE=)2NnY?$VN|
zdd{>dlW)5PBg1ff*rJbard~U-$sZ;qU2rMj6df`H@IlAlUmX2X+R*5iOEXNuSA1*ZQb-_oaZ~$E2*`lZw7|Ec1@5In{`)=xwUxm<7eXqo}SI;`u9x6
zKUe8gb?nC_{BSayj_Wqk%bnZZ7N46Qn#@?T%%szX10PWAnm!&Wdgm?iOJFiBPKtP1
zpC!&~<3)ymkvz%zq|BmS921-C9OljMd5BY^Q6Yjq(S~&CbCe02{Nzfrb`+Jho`_cF
zr`TC_5K(2Zlh!z&
zY*Xw+F~E8KIbraMlbT+@?vZc!{uL^|T2~|9RtlP;CGcbt^s=TG5J?jpi>_3hd=p(C
zxWxcSG4teTBaoBa18L}`?fk~lhHR$jh_mI$@be$5_ZpU*JnKV)u323?s_@LGVwUdkj+$s~Ru;
z8b`ph)GUJoe+>)*Z@bUIy8gMFw2R83wN-}>AoARk(9pH-M!|qc+<>Zu`%un6s`ECm
zBG`ELBY71wt+zDSWVe}CtZb6dcS|)o0^wTay|QgGRLav8&3$vqLys2Hk|~R`)&1h~
z7p2_*{75cA?EHyCL$bjc&GniXvJTWpcV)Q`WZMhg7we|mQgJE03-QK8xIU$nO7og=
zJNQh)T$toV{}zSmjk>912w!DkRbG!)A+%)9RVH-$jE}8%!+(7G(jir-fLEtUFi|lZ
z*!8SlB2%~@&sItTEau2x5_NXH7+eRuV)KEQYvaS8tOa$HtjEb${VnlKvAR#l;UYkG
z!kaKUC@2`!8+`Yz46K)9x+xhK_l`$x59S!T(~yg54G>$6N_(9<)>%u92|!JZcaH}6
zGjUnm1o=$K+9L@MVoqnozHlAPUAl`kZG;!{J=$S|ylDC9$5Wv!R)ww`wRw=zz#WR*
zr_$Livy`<5)qN8(6h;wKN{-q=)=0`Zg_xtBwN{Ua3Me-QJ^D;4h@QnhvzG|VeR1q9q5r#^8(Wy8V-@5ij~
z^`?GPT*dbpP_RPql|NB1beYb(ZM8zykZfO
zPOhpnwvwLXK<
znj+%lVxa@@?QlKbW(y8zZ`+JXs4i&~^41?BT~0-%=zGSf_>No}XtTn~Oe=qJi({>6
zXmeAKy8yZ_zG1M9^r}sk5b#N?L
zroH}5*ekI*3>%+wqOhfB4gCG&R4H>~PP`p$yj{X)Y9uOKo;8D~@Qc4S*mfNoFT)XA0d%-J#bH~k(BcDRZC=NihVkmDd;=3%4
zbuLRX7}Ml#jP`)ckAr((yN_Ut8OCw5aGAO#4xn155O@i@Vq^k9f|
zh^3s?^)SRcyK1VY49%j#VSvncsJa1pJD>1(pa(L?H2_`dWE-u50W3hT4#SFDsQ{Bv
zY}k#XPHjoz0=$nPe52#Xs*lrR#QhWX4Mhv^b;6mZLP<(}Z~LhRJPi#?q~C|`G7EI~
zpN3`8685tOE(iowG>kupBSj}nuYWCOKPG|Ck6c=F)9FL-!qM8b6I&R%acXn&F=qe!^?$Zk7t}*f4uIb*+$smI-Z+x@k
z4d{Mj^I;x(J=*ivs!c~D1{=Kr`(Ov7XgXE_`KAfyt33PE!CFfDhMp-HZqz{d17A`N
zH6v)3{(RVDAD^8%kFFnmg%w8@>+QkO@-GP*p8Z4gs)5Ow>~1zyh=3J7K6a+M1N01g
zFm&@YLMEE=&YCUOX@J1)BA8ruK6utMGU|+x-pzJ$k{hx-wQYAB#M)oxt;btvHE1IC
zT;u3t1pLdM)kqESVf0_?_)yR0TL=x$>+UlNam;tS`h4^iC(R~$cHoj+U+Mvb27G=V
zDmssz@Ndg5jsID8iStKOm-T)H5q!SCNn3mw_gm^y-Cwl8|JeglDAs(V=^sMw8V72MVJfzLEBq0l_^Y12zl*Qw
zccb=h6*6{f*=W}razp%B5qQU`lnUN1Q5Obg4`UB;qcmYv7Xd
z^Sajn@U75Q8)w;j?Qg;antHimQ3Iv6Od+pte5nNQI`JUYX=JsjWlEfO^Ij07Kg^R?
z?Juz77+X|T8F%p)X6_db^-%xdaD0mm-4RO6wZW6J`0(AD=QD|S7GJxY^3x=?=xo`~
zdq@N_jKm3grxn6D`-d_}46V*f@6sXr8K9{(K!!&*DOEVGzLBDevHHVap5RFyqBZvZ
zbk(9IfVw73APE#Pec?`PeDo%ft?+eG5@W|`OA4*v(-rHBkZSH)?=Ib7@Gy;
zokA{pEA$1WXbptgC*`H711uyX9w%d*O4bR0?lKG?_M2u1+OB7ogLF)OzV_bsdRi{#
z16c}16V(cd{)Ww8Q~}`ZThb~`9YkmqOq+H>8a^3$gS)L{p?0$8NV0awgL;7!fmp~g?1Zsxb}7P10EUYn-9b8
zwn}zMFyAw-kjgDFB(fcQThvXukUZgQ%jk9Baa^!EU>~@Z8?`080c$Rqfm;bQim*g}
zh^dmeDk*Jd{XlkvxZ0j^v@z0o+_`URXSb*vmH?r1)MM}+)ZDA&)E9cSmXOT`Ph|7w
zc(Y!|H%#)HAtKkEN!gs;h5x}@92(5Tr#7AoE)}_(hQ~o;jwY0)4=FgSOUg|g(|pxd
z91jQb%zYsWtt69vPj;ZI>HT70Ri{1!b4S3rzEQO34%P6lL%m#6$VXg+c)8s1O2hgB
zV*Bj#PD3wCR;};|iM7;W;7seA>f^3eOgA!^|8czavmi7K9P0mgNGL?JUrW|7W@ctS
zLiCwTHWn&jXE}Lo{V;ZGKid8=(9l6)wb->4H6eTyicL{-(y0Lryf|4i`ej0q=IJbT
zoYp((xU_iD5%(T2GnijeQc^~>JJQf{y
zJ`kKsrWTcqCb1Z8KYa!AdTCrL;siRIHi0-G-Co+!eW(FTKpKu)$1`gcvZ>z3Har62
z)r9@e5~Eh1#86&K_PY20-1TbmTq|`t^A{cUF_xLKsc~Ln8!dC$EI!wzJ>saG&)5wd
z&ZNf-Z+AUQ99!B)MvuB$YH_&>`Y`30UNscxjHS)2zP?&}QyX*MbG5nMX=41T$J$zQ
zGVq`Tp+=SHTP>;=7QdKkhdftafPCzCi`aa*Scr#?bzdCqpANy=s5TXTVElCefe79T
zktla^Q2%Z?Yc-6iq*k)(*03E7h(;p8Ep=m3{jVfj{WfRZ<_KHy)aZ~Yna~O;F%yl%
zDPoY&g{BWvs_C#cOSOkmb}S!3tj%1d20lLi&)t*NX5R(FOxNBs>`-i)f2By}L7NcI
zgzdGYv$nTkVYRN`qzva6pjETGr}=h@JjSrHWh<{M_JUp66%Y6+?kEhl>dEk(2i$x}
zz~*7RNTKQ~^#Z09Leb?5pQ#M`Rw&*LL4A6VI%nTa>e
z&F1gyID4pRMH|}Bkl$5fP~X%D-RyAVTd60fYaE|opYY_1SM`o9RuXwJ@bbHE88yKxe^0itJBv2+tHSsCgbSnG!c){U7(b)NW`}fWmy^Er*oHK@b$JM7t644NxvKYuD5!QnP8F6vMy)Xvs6Zo*~sah$v~F){IAb$99!
zcPYn&uNGBW-KL5!-MYs~8sz)4y7KRk@nb@8WO?^R;Uab1;(#
z;(?RAkbn9wnE7*<*GH#=Q&Nx%zP+hLP}}Raz@FX>(xuv;JZPv*K}J=&LelIt+W$K?
zX>$*ASz8B7cQ#g2ODA_r$CnP4&R&*oUbdFrtS(O0|10WXXJ_Z-=BBTpt?H
z&}jd?oTH_OIjW4gpuLK;f`_G}o{goOrL3E?qx*GvR9Q4(G&JIWFYoGU?q=@fVe4dx
zRwMS!wtW
zY@9qioNOFS{My{|GIl_D4`)+deswn)S!MfI0utAzVtQZhe?0#Wk2<(nUKLV{nKg;Rn@nuAwT
zl2cYv@;?yzcl%d2_mQ_We`V>$_P-hj{xkb$=i>b*`{&@{;XtGI`M*=?e|7u+x&3po
zQ#*RtI$DZw@bGc5^YO8B3b69>adL8SaF0zzYG%^TF~!*iRCl7
zHRG*stn;;VACs#MO==fP|3(}L{{Fn~qy9o_kp>J>XX3rRxoHUV^zG^3kG*A=1#
z{rV9{_q$>DfxmqpHT?H=@wHpOZx7u4eKVWn_eqlN#*1ig0KZQ-(0`=%FE9SfD*g-g
z{u8yotM@-r`)^_GFKYh<*8ZaQznQgvqxQd8x8W;YcSu+>e>R$kHT?U^;2K}77$)2{3|qlJSsau
zqak@({q$3kI(#Q2!A&X-j1?9oiOFA>Z$S`7I0Zx+jCsJ8mTf
zA^$Y9jr|vm=b?vdE~Qya=5?2twG9pn6@3L8ANT(}sC%L10dFTE*yhB`5IjsI<_K`y
zH9WpNOj|9rq5rKGK@W(0k#rqwcwU}FB@y2Q#2(uFrL(@+`sxZ0l`QI341d7iU2NY^j$MVbo1eD3gQM+RaWpA%
z#2S9NpF5aF??=~#jp=^ymG~_KS>=`5Pou}WQlW-8$;Z=0T3M7oHJr;we>VT=)R3gF
za4%C4D2p#Ls_b3JHb;HE;zl_1&&_ANNd(i{)&)-14Ilnod86o8@jif5MiE*o@y7g*
zx^w!rYE>9{oC+|a54P+C?DT17k8<;otUTPF03?}v+L81da7jJOm((#l7>;}U=&7{F
zspT+@=7mYVq|Jt1Y~gZ%HK|jElH11G2!9qeN%;NAj;|j?(yIA*aG*=tFk!*d90d)9
z%bh2V67o}qx}K-vm(FLsBD!tupo5-7Uo
zq!zD0nAF?ssjHo!1QYns8QWskc(9aex-&60;5psHSkH1d4)Yo$!QKp1@eeqvpLYxNgjeHSac)^4p(_9;@5IzE7oueM=T`8=SdJs(t~0zpVd0
z^`$qk4~wtqJYr%SD5-W0E9?U)uomY0!LeiGZ+6i#UVQueyvrt@Ak|;9&z5~z$XEA1
zPYl0YjJEj7`k0Rv+19cXY8U#RAJVjrm*Es~5s!UGr{d=z5h#k0`}z7tICdQsd;UN}
z7~8#*jdUpEmDv}LU!ruhOq1}S&@X$h!fD{dVz|>=h1Ya4XH@*Q{FdX$=B4lqvZL>Q
zvAd@gi{JW+`B)m+CZz17?7i<9i#{9mz;{`>lrt4ea`MU2tM*>p@1}ynnW&_ORP`TK
zdm91H{Rjw57a#YwmNjjMA&%>L!iLoX78d;aE~r;cdS3y4BfI!otr9D106HtuBrZY?
z1~@AoRbZMOH?EyGzB&%jNC^uY;L}elVc8#Z0A7S8_jEi7@a{jh*0lD{o1K5?>LxWM
zG{>wZ92L(%&;nVU>J#Ahn?)RF>L_JpEzs8Tz;90r_tsmyv^+}&Qwce6=p2$z<@H0D
zX(1wqo&B)~{ajei!_F9xu4!Q8K_U-DK(YBBg8C;=p?DX;WHvJ|ZxQnCpoK5gpMF!SR?B8%~Fl?L?M$C;5KVeTS7;ipb
z+({I3T_VxCmn?MdftO6G*6bD4;n69-+e@kXku0Hq9zW
zqTlE>UVrd97M8KL9xQ~O9lnwc`|xU-b{L%T_`&x2^KI`7@0z2NVD_w_K7z^1ixM->
zPZMsKIgi*rS5D$nwR-z3Y%eZEXF&-&0g3#_C6fo+Vmkqz-cnpf5B`95Oibo7NY&VX=IjaTWW8K434gApJ+{Q+Ui7Z7ggQS*j
z@!o98eh0gUeIn(Mx);@uHh*!Aox8CsE$>NB!rVLiB2PF)$o1dEA{V!h48SXoTo_q?
zGe~`uIm*KgAQJ*s#=5FY{UT9wLJZJE%L0-}mo24!0*xNQa%>2{zGpWPo|%@J>SQ%^
zWEkUDwFadm=miS!?7+Cm$+~QcOWAii(trSqP@?Fd6X
zd|%RFW?;V}ztTPW6*{_wW)6J{J;1S-jV+(@RT|Bo7$yvRayaFDco!9Au&>Gg`Uj3^-j`N5rgm&t2{B
z8r?ZeoGBa|qZB$oV9vGXeQpJRoT;F{2Q8L
zwBBd&jJQ*3@eHF-y1^Ic=?II=Ix#jrvX#^JQ=u|9w~j~P^Y_Y+))=KjawG?*7D_hC
zgj0Vs@Dqq->lEwAtkYtvg!**nH1&_dswVHkmF^wmxY(&R^|B=9^I$Ur9q#*nGIe{G
zR@n4i2FKFZQP&C7GUMb4Hg1YBiTx9EtE@ra%eK7_o29bBDs6}(zIF(0@aYAd&lJYL
z5gU!P?ynH~u?Y09=(Hj(x##XU{+|FH4bKSvCozbr0JH%H{^_Zq6^psUA2e+Ne>s4U
zRgkYdO>;Sve^MUyX&aB!dOp$IAqJZ}Z2>MouIr2JEi)$ob5GSEYa1D*r{(D5f2kI_
zOY_nII1I)wsOTgQ%^E}7!WQ{$4(oQhye&;<0#D&ebO1`CkMbe#c)9m_d4<{x#G}N$
zmg$uDephE^Dmr8}dKFZqsMv
zX+QR1-OuzHRBdyE*+_&$rv9t=d3&}1%R1*|d78+(F-G&C?tIq=gijNqZJoiOy?0hY
zB+X{W6d{J9e_|MRjS2OGvkF+~eZ5}DaR0L(?C)3MlV5ob9-%N>n;VzDF;ZHd7Jc#M
zOXKI;uI+vvHOg6o+@FVAzQOPKUs&($RjS_o{ZDBC^WUF+yW_Jf0Ot2<@2J2w#0^mQfGgn
zdv-&=`EM6pi!BGAQ0bj5$LpBcmRqsMao){3Tc)5Mmj})kUi*SiUw1NcOIp2SQt7oj
zJjP2G^^%qJ14PHYI7Q@*4ot0b_tuIe5&RZyke7fZnOHt(XKmg`Y6Dr(xAST?z%AbU
zyXv4ZD6i+G_Jp5l##OoSgw~{{hIh{B!75g0WxQ^&&e1?0NOf50uXf$f(HXtyD+pXa
zg50G85K{@)`R(L{BZhJY$>;2JW{k?}7@=d6#yPM?hs=V)za&Kcoqo!+!j>ua9w}(>
zdue&pldYajvwg<#-5i5Tt*^}&uLz&MksZ`dG^z5_n?(p#+cqOvQy)BDc&_bPdtRUP
zN}o-a*VEkNL)!Po{QZHquc0QgbZ-CxEOY|uI}?wd)|AsGVxKHkbUWFtwTyA@$`t?N
zP?gj%h93nZZa+1xd3n(o=cbqK$iOgp=C0xUQd~Wr^ZDes5UXcw#si4Fyrm>fDpUO{
zZr&flD}DFWJmN|1OG!9mi-|B^YwNDs8Lb~k(fWlp@7g{@)Q!deY)W#5u1G$Y2Ji_1
zC4)+}aa0_82Qdw@$liLFftN|*?)xs+47@UJOw%saE!v1=iqT=xv0B_DAa>B%^ZW7M
zePGkp6D5^TR9?#tKF-6f#_7gXE$~FN{(&8$>5Jp9*~Q`Bdz66&4KK8;RB^?
zTc*~hE|(v&lCiyw+aW8}k%3iStqa{rm;MrF=iD1Ltss~B^@PRwBC#vi%OP=k=4$PS
z{zqU_)8gvKK9^^kNny3i%9dY@0{h94YdMWpN5grEwjJ2rK;wK!)>XsGgti5)q#t~D
z7rCEjkRqb)e=}~o$0Be!mOG{EjclYoU1`m1
z;vD(ExcKU*wx1_jYF{c8YC!QCZY}P#P)hOQZp8@@+=8@)LUAi@#X^zb2^y?OaVJ0^
zxD$dzu)r(r@11wf`#b0CPIhN!?##XODI%1YR9_8R9H0}f02-{me|Q;WU?omm7w+Ah
zGT$iqfz@f=p`V(Gl%i|Xf!y__p&dx7j(CT
zf(9?;is<{5mNI)|d+qMq5G1%(%vJAA#s
z|9LF^GME@{VjH}KBuhR(mBUlT0ps=Xj^a1#7wav9X|u=b#va&?L5m4!7Ami+d+??J
z-plu0nV;=hPGCM-GPW5mYFDsWGd~=B=#K}YUF87I{q3)pPrvK{#kx-By$Dzru$_H)
z?YW=3ta-$@2sk$P`#w*|+eJx&8b7sUjT^k#kR(y8r@$a8_Wjpk0f9+{*|Rz3gS90%
z^Wfrbh`?{&cH>d?<1bt97TJQ|8YC06Q}1O`D=1+@@LXKIY;iaAvrD)qzGr{9Y5q&V
zxP{i@*ciYQ*9iD(%lX#j84&on*U@Ol($u2#u1bACR~D{OOp5u!Zp4QN%))^@A?A2k
zrL|5!EYDaHg-?RyeT1=q_^U^=YQO-bE>mR5JR
zc_SFU5@Y)E>(r3vw?pcH03T)frErEo$MGCVRyt|P@j6$S{OJ?INlo>(
z^+3k#yb7+?cB4v7a&;#JU8v4mcxP+ZuXg9P1bTKto6$_y9%rpIdd$q}p7R$i+b=HIRkk&{BE^S@yzlwWJBDKkYeYIRs$|?U!LAzB
zBFUahOjtVl+~TLxN$zg_FX5RbG==vwk^3(#EO`5G`8)9(@at1v(AoIXjs5(1F=icx
zSSY;f#b7HC;H+ryfcn|nU%>9_sGt@8E-qNVPai8CC{%EwQRp$_^24*_)Bpx0^(23L
z#U=*YdC<0*NjFHLJr^dYr;Tq*4(UK9-R`4xi;TH>dne=V(#OLzBp;$E(P5hLK~B**
zx{Ws~taczv*ULfB|0X$l!==8{cs&WPK}<`n-B7hWmH$se@L
zFP7=)(t9xSba}|Z>fKk?^!C}X8j9XhWsQ@G#XK;@BKj8EvWr&`Z1mW}sbYG(?baQ?
zC`>gU!kr8oDZnFFX^y0uOO8v#2W_GTTx|kNv?ceD^Dws3Xi?~(Gm%7PjM@pCQ6F-(``ud_(q6D~^L&xI9;#S@TLSEhZ=T$$DOX*|3Q(P(aJ>rCISk?8{H3%t
z{!HVeXmZCChiYXHYN@W%EjPC*#JS`jXcCb-Rdbk5groF18lU2>6y<4UY>t;qJ_g-`
zCU!0eWayt><}|t82utB+cF*p^+xc?LV@*bN2cnphb=ElKG(gF9JF_<^fEQbsu4qD?
zA(gSi4hQ1<_)3EBtxIh3NwV0n?93J!7<{iK-9@RwE7+0$u6DCJ;t`>ybTUH>vge3N
z8n@RD;Yeh-D6m9X{&GvYnn(Q~(AbdWt=7l&`T?5F`D}w#!5OL%n`VhlsntX=Ubuv>QSsq?C2E0*%*Qlp9f>01bfM=}c^fgwp|$g0m4w75O#KX)*O(s<
zIZj$br1d`vj_%H%t(^R~IBT)O-Se}TrU?1F=Dl#W+JlSy4mU1VHbWvY0kWg?w`15!
zPV_rE+!_6qRo09cBq~Yg9y~$xDtr`+B#_76@z0zX=1flID^j#79__*gC5H|PD!=mS
zuP>?fRTaoW8U-F8me-hVRWNA>w_)_U8Cw*jg^~j_bzsDr!9WJcw6vqN^97akf8*bs
zd@c%E#{_p2{%w2yNXx2x?jq+Nc}sR+jL{nV4hxQIa{RU|5E85MSQ67`Y}la2FKPwb
zBp+HW04ZzRbDv2mNqT9Wrq(v!kFa48r2)e7MN{wvJ_G!>)x!0FvNc)XQx1N$VKwzn
zcU2b69DkrbUWIc<6+v?{K`DRm5srBB0Lgk&;2ieB<>)i17dnhQg?fo{f$OcN+hSO;
zB-GwBk3UA9>M}*?unC6O3Q6Tn*io^V#y^spF561V0=>8MV-mxIQ&s1{&hV|?w`2Ta
zy`I*o*`OvHw*c*KLjt?-kdV}$o83StM(`=p-Bwl#%XHq}_*@KcF8@$e%dXz|+h5Hl
zZLFkG!LLrO`(gSu`8jX%_kKRBNtvPmq4m;fQ^h4H^}ixQ$FKPi7k-laay~_=5cQH|
z(bi16OlrKsy&(sq|G?!i@xtwV7d7C)P^)-wtNA?ZP8ojz{Ny1(xP*JFn%{pNdP~We
zE28u_v7kl?s+2vi;otW0SI>~D0-^K_anao3)zX<~E6W2}Ge{b~YyID(a?i`eu0>;p8oo>Vk`SnYgObATRIY(iC
z_9Zv_)SE}@$XJ=jM3!f8_=*dFE^-@kmYe!8#qa#FIo3ocICx
z$w0aiY93~O@b!*(^qN&(S>sY;;(*oafPKBqr*h><(HtR6rm6I+B>mvl|e?{8^I;7U6X{1b1b}KdCz+wccD=Z;kH(n
z8H_ss=d;PK7ZbiF`wKSa^uQss;XrB8i}ZKMLeb7~Bz&;N>1Zf&n=w35FFN6Ufec6D
zO={Nq#>PO)qkcI`x~!9v9t}^(|HUwOJh&&Aocz^qx=h#5$@=G_tK%VG)uf<~*~fR|
zjb?+MUwh8=pj-?y5K~?|`75P*(L1N}Q^+-GLw^)-^)LoFl`BBEuuJob!`@Hh=;sH`
zB}y5#;w{k6LZzV+vt{5dW5pwL!jLsC&euarHfL*Hz<+YMvr9}#av)IhLop>SS{UFx%-&C!80pY)Nx9xi@78B3xz~
zN-gG_n~8HXjB6@tr$ryn20nMi$WsJPkMVmbx-r&6{Wdf2lrb_rn3a@ZqtvcHPH#b;
zR~>)RGO&1zC{Bd0cxv*Ed#dEY6F;0DEYmeHy*J*lLe&?S{j^1Q3rHr;+YG39O#USE
zASeD}_6qUjlB-0oi_Z^Yn*3%aJZ8Tr)S9GaSlSAge=8bUn)Aa~K3)HT*tWX_K&=Z1
ztRTR%g+19vvg*u3O*dPUbLHe>fN4}wNml-}@Os}}#1ThdaSTKG0L)y7mnjtl`oviT
z-AolT8M*RMrs*m&d$PpAo`bh&#(3seWLeb*;M1)~%n*OACm+788OSMIDA|@rHtHIx
zY^miNn$DxZn?+|*$w-s+a05Beyl1QFeAAAfESOtl?{^+Z*eiRn-Aej`r(>ALh|tMO
zeK9PeXV}&lIMsg+@W1$*UrB;;(;m+6g4Axh%r!NtiSCp42*Q)bnNtgxQ;AuSW-zqb
zbZlMQVX&maZZv5!K54AvwJCFrZdcN6LiL}m0VeCQt?w;<6eB&It8o~3wQOT<#hc+(
z5o{g!GFjTHMbl0o>9R*`Ck_*mz;9Ix-wX>!Z5F*nMXVXm`THp^CG@pNAa`zg+SAoXOc)j}&>YN%}n(LZ
z$4zhL*dA|j;s2CB{lDk`$kNxc05qxLId^cK3iPSLtz>&+%F+K
zYHaQCWo;&}#VPQ7Tjl}pKS&@eHFY3k0Wki!(ILm4JVLus-n6rRoK-uFRWwX9p>B#Ppu_)=<(IyH4_$r(_FULMk<%C(eX9Mi6{b%_I^ZY2~~V`
zb9@={IGz6Qf7%$&xAg`dy2)%Lt2VRrN}IR#dAYb2#J&1Hwc^{hc}WpJc9EJrk?~hj
zk#X+zSuT4Vd2mfqR!wJhcW7*+&_w5eFt|GjNX*Y`OE1(
z|Eg2y^9QPOM16eMp37c&ND6%8^2S2r8Mu&o2R!iVy2i7@?i*1#4F6-z-h~4RaMKTo
zXw3a8*GYyTcxlPn>?W5-&cLgwi~=6rf%*i`zby;1{5}rJGr$S0JZU!_)hsXFnc-Z>
zY1eGk#v5_2DM9mZO4|pJm;jZD5aX
zES<5Q*PgQ{zH8I{bY-0@k!U`00d=(gw@>itAE>sKR`M|@o+rb5^#hNW@0qB1^Q&wJ
z(D9x^8K(0!Ewv!ASTry{20=yc3(wN|+kb*R-%h50of)pJ;Pyg--uZUND;)-K@bl_`WN0%6yP4_B`O}-K+m-@WtiRAf|O~%ah(~
zh9c
zf7m?!zc!i66e8e_o6|blhaDT!)$-Z$LI3m`W=BW+!-b+&G^pM4lC|^H$$$Ed*1b60
zxcc3NZW)oYfOGgR52L`%f7XWFDrmrEstj$p&vY&)v4*Jrcti5|z1>w3d{J&a&dd%M
zHWSjDr1SM%31aL;MPg7;tOD~Dr?N_^tcS8^SLclF5=^AeZ1TQHnUL{Nq#jo5_G9VJ4l@E(PtUgD
z81IIYbH`%Fl@*|VtGPW)*5K!*2Ib$=mEJ!xNE
zG<6W*W5s;$O5J_bP2O^~`FQR2u1`;20<|`ns0+xpp4~2X%{ty4GRVZotV@du`Gxz#
zWPSYiDMF?CI0noTT1jXr?(hT&6@C$>(^`yX%8bCsCfdq^wr1+#9X
zg{)yN31(_~B>=NbG!hfF-rHR-9c7zG9jQT7tIc+pJ=49%?hpbhx4EH7Zk#V$s}#o$
zIifF|x>VTjYuZJj6c|B~oJ=GcKFHg>2U}^jq-0lgi(a4*fR3{BnVJN7CBQVAT>i9g0Lcg^w|atz-msVCA+7
z
zs4>p^Rn6oj{^P09e*4yJ>8QI930YL>D7Jm;;r)P0fg3s0pNLC-N@R0gLRGri-DVb#
z&wMVpm(%m)PFMNL=Z9JlZ$zfn!wac*S39i93L@L#!Iwt6-Xy^bX5hjB2hFI{=zB{u
zK3<@PlPO2Z>OSw+_d~RF^iOVlyn()S%^}hy`tWPARpl-QHwLZkAzP5Bi}62TrHhM(
zT68X}#9%_Nt3`+m&7!ON{BPLlUJz-7OcQnN&^>0DZ+(UHAJ?_D?H~D>K33!q61@y3
zPq0g+N$Pc8VPyvIE!%+sWBp8`yV$
zB(uUa3p#~u#m!QsN0+Ja&MvPlxe(7zckncF|Kw+~mdbjqdR}2weIFpO;!5UPSiOwD
zXX)3%>bymjik9v7=R!)`wDdBe14z1yqDeNJC^4W$KfR~v?rjC5(QU;Y(pE`R#X|RD28F&xqF1At2XtQn
z5QC@D?u--N!E%zGJt<}u*i0$@Hvwn1%hMH#&-Y-+f!+{;kTh#>z-GSo$d$aI1MZRY
zqj^`t(pPvS@aMo(4FNi!tu3qbe5_pCpiNEc&$&I`B`qSELj1S$T7IdNIx~mxcE1N8
zJsxmyY8m*;W~lYfY;?e6JGD*B9{{D{ujJw9d~m$1TUCA6Zw*Jj<$qNzd&O$KnQwQa
zgqLYndHeHaiVUgmQO_ybZ+^X{?SiU`
zx`XX&tE~NC3i8`^AC*7k*GAO9&sz#8ous46Paiq-o$Y@(_tdo^Y-?h?oSj9)`OE#+
zg{rHQ6S*huZs7xT!urQ1`Z{+wK5gV9%OO2S3zidhYac=bx{P-H@BYaE=HCLHKKbtg
zDqUP+cvWwD!XBqQdgzLZbg}a3#u%pSP7Ym_Y=3r9Ex_l#yNItg}hUM${thi+E;cQuCWE2d!D+nPRP^|JvwXL
zzE~P>W&1l{pbbMenoU=6o!OX^?0ZYa(}CCo(kdi5?5rmA!YlXi&G3~goK6OH60MK@
zyVdko{SM?-W6xI?e~?`3k7FS*(SQCgymJe^)Jd#J`3zfHwnU-gH#U*wCR<_*6
zic{|t80+d6pzCPWsZx({0}r_$l&c?2xnKryg6WlzSU}R?cD%Rv`&phw?=G^TT~i~Z
zkwtgZh4b;oZW!XdnA6q@Z_>N1fo9r6+A<jc(`z3~`T0sYN^I!aJ}t9y#@Xl#
zVYo;lroc4)=br+==5GPud)8j$QgmiPm!EuG`z-D#C8oIa47&*}0jT7yS`^li5$CPW
z)%P)uPc8v)>&?^8Jzbk$-bsrV9u~J0fB6C&RdS`38~vHW$>Nd!tVnHTxyY(SNCycs
zBiQgn3U(G1FMS!pQ&JUMt*g?xsphBKH8eFH=#2PkNq#64-fF<&$|M5)zHIIc%*1iAoohnCPVh}oSRjvJzFS@KsQ
zj$%PfIbl3_&&(nX7p>syXpeRT(KaAHSpUq?aS^F5MGXL~BV+cRKLm{R_JBP(IoNva
zzeg&O?F$R(j@FFyoKjel9Z;Rjol~gkSRMOZHx)f4$0PXhUO7HWZ~79Y7)}euwK?8W
z7{f=}s8zxh54;GrE%+R0>DJ!ov-yK`5c~s*S=7VbK0_^53F$edr&nd;h6n$bldP9M
zIp#K*Km5MvgP;~>T2p`A|I6?~M10e#4D!3?o_}+q;z>-Dfq;;}ByvLA)-XM-Pn8Ni
zZ8C|}@#6qo|C9g7R$N3r<;);XmDT*{Xo|@vlItt;dP)wnmpQnE}{KuE>E|qCL
zI3npmPv;C;07
zfsf}zO;1+zq>iHNg^vrE(w1Ye8P`W|p>n&k3%i}v-y+ofyy%58V`7tyX~P7q4(06W
z{{3HULv^oAjUGAryq&^s7K)>|fGrzt{)^kSDoQU`Jl8*2smEgzf{xhI|6n9j^1#J^
z3_TxfZMC22SoQ!EQffPLQ+UkNd=+b7CSd6lqQ
zS=mK5P*m!NC9fV^8^iF(J3^f*6KaOs(Z&d%xVaxsxB$>(oI0h?hSQDPmWIAiugAP!IA%2Lghw?2y)VAaakmM@kR?}I#IKO7G
zs~y?5S_
z*{D39=+@fnWU3;6H!7?a5>VgW%sH0AimFA;&ct~mwJyA3idg4gO&JP#(-}t;Z_fQG
z5-;@(Ml{sl-SHRg`L*l1hzODf)!2*$fMQb`l?vk4#|KOxHYh-9M>%G(kFkEk>B8xb
zO6Em2PZ$DgW4Yf>%KCdCVENTx?kFt?C`elXrBr_`Yyv^$4%iI&t#&zuA&xBchBL-n
zc=7a~1gjdUFQHItS6};5lVsH5yK1N9p52E#)9I!*Tw|JRmSY9hv=-r-6HN9M^d*z_
zC#F4`aPWY7O_CITZHQ1HNGO>a1rLhB@HHPC(eu)xIv+)gpVo>DJ$|~%$UrUMb)0UX
zx56M6HwJB;l$V~$EJV++Xy%CxwG5)NGc5|TioM9)QBE9U?R`nK=ilvaA2-Z5&+aQk
zbWUM+M!{pJo(|X^STZ1;)8cA3SE~2K?`g!~Y(e(gd}MCl)O58|git~73Hs#$$_)Ks6#;>PGKs?r_twin0%;ayo%SA
ziYXUdZVaTvjc4d*C~pk9NlaEhyWZs@xdjXyEC&jayQ$q)5g8C%3P)AnEEK+0ksNf5
z1aDzqvpJxK1zyagaQqOzvDv1z%dtVUO
zR_wW|CkYtsHBvAKJ2_7@?^u^;|L)-gD&CFtvUEu07939xh`Cps+vQ}#Tk_R_DCu-djktc->x=0!JQ1K0UnDKNlVR)jtm&&8flddi;AB
zHRue~fJlQIMq>4D`UTTKiR#nBXbQV_{=H%(l@a68i!EA!%8jg1+P@%kaO;=voS8w8Qkt`txEua`A|W
z%T?gNN+vH=b;z6Qx>vx^q9-kK&;&AHn$=?h+H?n$SLI_SUq(jZgvBEaEw(G0Gyuwl
zGizrdjB6SIN*VVIwqmr93X`DwI8x{0v-X4@emYwB>pf2Exn0bCiBUX^pGRzUkX56B
zZ0vE_m7%G_z-$97_Rf9hLxJVgq22>tQ)$nsj}HG6OcxNJT1M|5=A9j?!m@oAKvvJ=h3;mBo+>j7kFfya%pg#yc2|h$N#ui)|0itM9lJwP?jk;N{r|tr|
zw(gj4(*8buA*zrW+{}#?8rQwREe}wbHk##>KkfCMYhqhSHf>~V3V9pjyBsfayeW{^
zTyBnL(gGiGavm?^4*Y3x3n+QADipQduLB*
zfH^$cUvtC*{DlG`Md!CFyBII^tehKZ{mpc}Ix~jhcy@;C#%95d!uofA>pye1CuK-z
z00H5V+kHx1M48@438P%40c5c-J1uv=GgKYMuBBPvuEC?NnY7gc5=2i!Ui@k}ic!%!
z0~E;Fl-q0s_CG8yW=6l2Q?=5J(kL1KR)5%?zAH7?*H$|qSpasj)K&ppxcw~Y_GXQP
zz=Vp@bw9lleV@tfOJ_gDY)s*zhH1FQl>Jj=ZJE&-mmwX6g`>$viqjLtp3)o7X7uNN
zR&uB+545&Jd`f;yW#`A4E^_W!22M&ex`_
zaZlExLlU-o-*Hn2(O-BQOHC7#iJ7`xgW31cKtI!DFYF9Z&A7LSr{@yz9(ZKxZ9+n+N$^t)>qih>HciHUi>Q;EDgoo(`wo^l)5;HX6`
zna?W4s8zy*tIT&Qv!#_s1odgtLIiF?S$lD4uo~JYVhuJC|`y<<&GtF^eZcJ%k
zlj}5qU!14>uBnv&;-c}kIM~hBls7csd+7*_cIHh1|J>byPVHRG_OzI<=C(dpS_3(h8F&Cbn;
zO3Zk-EG@hW6JB9Fud3U1K4wq$+@d`=95Z+>uBw$3YC`Ra-CT-}E*wO@Ez&A5UO$v#
zH4^$H`m(X2LE$=$2C(BueXMq;(suF3PYT|!?dGO|Zq7#uO_mI9%-qQOK;0N%CUijd;5$oS3-PRmFCp-p69fPeyn~)Hi4hpP*~ba-YVKCD)(4FiNWe
z_%|F&WV+dtqrI&)^EUQH=Vq$eCGz~-uq%`#fp>Dz^{
zzb}_WTyhk;|GJ57J-N5e>I%A$C`x*^e6l}Orf}MmyFIpfw(_%&Ro;7Yef~@(%dWq3
zUA4u^p*u=*opf{XxM^7AU7Sf#8aq%r#U0nU>vgAWTjblgU9f82gq>DeZ(5qM9ji&-
z=8pDmeN6!UfS2Lwnqc=BLXXQJ_5+btvl1N2abs3;2-7gYU_$}A5qfTOeDv5R
zf)6KITY*X{|Iz#c&Ta>Onyq-2Rp;C9&&0rQw3#BtyqZ&Q{Lr)8$cmLP&f3pEZoK3%
zba%|CdFsN)L*AC+PK-BxX=RYVrrJ$fcUnu4(VRw?CEtMj>4MDBW>Li75KJy5box(;Xlv5fIZ*+~>n!*m}txv9Jfvma{g7ZysDCX{+6rP-Xq5W#`42N^2*pK(1
zm-RlDi;x)ex09F^qu@^C;$Jkb6>Mo329^fx^U!
zM17+txn%3PAeC=S@Tbdl+;`Z1mOi=_5~|+Q^)`ms0N$>DUKLAutIEW1s0RwFg>dav
zO*y$~3#DI$`8o6-W%G^tmLrQG_Xl~R1+I`_e{-JYY=&{NL>f%23bOB0jA=3&3J~i>
z!>CWhvL&~l(@`%iE^V;6xVtBV*T+KG8RHf`~-c&-wrYcUDfvWN8=M-Lh|@~Yls?Q=J`
zjV+Co1zM)fKU3bqAXB{s6PXReDd*OUhotfAQOvZ542C>0xA0Z(n~G=F=V$6i>abD(
zTd7I?dC0j29IGYtbQ^6^_wq1PWe8TB#`<<3Ey4Q=2ub6M(=n9W6ix+0h8#zIKJS9n
zDWh(H#}e`qfZw2wE|gZ?%1Gb$OW|d1$7LtT+)kd!ZHp~fl<8yY36HWv9`|h_#A3qc
z%pk4F1UzJMB+$?Dm@b)CcWZodgX`Gf&Bl$^oHwTBuj>Jr2~V#>12L;5*?q<}(ir^d
z=yt#A(r8KNG%+SK=*t@va>mlinkxV_Ke+TR3i{;{gxiiLl>kpxchK2so;4V|4uvXK
zU(L)m5Pg-qyr*4lj_&F#KdV}hmgYVhqHr@kTFkw0^x=h!);z-eT-g8Lfggw!6#b-%
z!ksUw;i*zpEQ(w3rX|8n$?9`x?7NAPnxNz+mx|s!g%U@ogF-=99l#@UtJfW99i?G|
zYG*pH+pL#lR+-t2P%_*b$&u{({==^1$o0(=s0AYNRBN}W~>Swhs%uO}#
zjS+O1Oi{HN(HDJckoQXPTLVN=%`5wq9X*y&MaC=K0$3xa1_s*fQM@VV6JDW84Tw(7
za8+MV2}jf-9upV~Uj;U-?7Qq5uid4m_S&B7lk$6k>l&m-o*lZ339*N@`1eRIcBzF~
zFa|jFP-aPvFYB2nZTc^j$`29Hb(WGY<3W;x+WY8NV=RE5IYo>-4|!Dl4?ovOT$}xJ
zZSG7J5aq1nGQ54PK+>A(wLR2`uF!dtb1SCsh!(|3$xX>2>lGT**xxzUM5<^jyd$wE
z7evhF+D4P_26IrUGK(dHnL+v_S!m8UZq)Bz#|$1fg!k+>^tNF{?D&i~B~
z{Jpgq&{tOWaP|c^H$)#|ZO&th1(z(5F@*SlJWZG2Cn6$CM%gdO^!6-HQ|)ox4b;iK4mcN6q=R$c6AyCws$z_
zymScO)#WIFtsXSXQ2D{`CfB$mB3~DI?wYgdA7C|E4Yq^hMDN7=Ys&c
zDv2X1p`}n0=;mRV^Z}+Ncqg^oaNVM^%4_~uL+5<6G>+i}BRKE&-i}D}jfI57Fyprr
zK7N|!D<4G*2;3>}es({xKa6m?^txcKJa{b#4Qff~k$BP2YfV|Z^*Bd$zt#zzQ#B)N
z1yeze`u;RzTKRcTZTVH`5@HTISz_L+IkPrPX0S4=Dr%e1Qry2|^tF1VRh01Q4tlJxY&HOqc?g)rS0jBzjJ;qxCNOm(J<7_C@ky@r|
zBim}#Vy5@d)p)5Eh#En=OoFwwV);L{h$xg@_&33F$?DGS&pvL->>~ruCEA`h8Qz*$
zQ)?*u?Z0xkKq&hrIA^c2wmsn^Sq58cz2dPm{siNMo5Jlz&&Bj_ofV7@49q)}&4C`}
z*@SpzmOLZG#auIK5zDDoec7kgaKM)u!Qwb@{$9xJ+6RS*H0t+SO;Jm`II
z!K1?-VOP1hQO!`U{g-Pu;lqX8--}TJ4qXnFjg>3kb}X5&6SCalOYi!qC3ikX@WYL^
z9q^SHU;Mv;FyFqbwGNHD@L!|`4TEFfGmi(2=P%IT_Pr>ha(h1Ni>gUft|rq$C+f3A
z%+D&{nTsL@GXuTNySo(SY9si^J*HEcB=`3jP9I&Xc;`uwp%8)7d@?=?f@h)5g7>j&
zJJ0QgC_Ogm6XzPe{k|-A%c=}$(p5nun}%LJZMFGQ+gM%_5jy^60~-~-RvO+ACFt*N
z%zXVw1$T&zQ^<60(1~iOE31@o3^=L5J6k2>UX1lw9;BY{?nXB>fV&(8GjIH!m0UZn
z7rkTb;(=?kHU@pZ!XMOhEt4U57$GD?&sHQr#uPh~<#IvSRwz7lPN1-{ZgaXXJ)bt}
zW1UmH`aA5vf>K3=Rv|Xq=A*k=r0%-pBA(c=Uxw8=e@RY<#58}n>GA}Kq@z3Eec|D=
z5uYsE^gNVkFu&jS092kU@bUwTR(QCW3TCXXNI=^|hmp_nNzC5mWSgz@oZj@ThQ5JY
zh0NILm(}~9HJ?eG1?v>(P!=T93lM|$lH~4;ELUA6T6ci1*R1M9)UTlUE=<-AlGK#<
zD7+rXKpac$eh)#0FeS#S=21JHoQSezifG*3R2Gs?SAIM+W`gm?5uF@%X79x~3B&g?
zA5m6nOdM@y<r>wju1`QI@;I~
z5_)Crkehhx4nKLUcW8FQ+Exq<{vudU3Js2n^+rzncgKU9ayPMsc3h<>eZ9Hz8^TmN
z4qT-buuQfC-&9Dhug%xQ8-kPUv>+rtjeN&ZY`A3$2fF++#-~MRTQf-fE-J3e>}M&k
zqG?9mcdm558st+`Sl!(#+5_F@C-)SM=@mbl-ZU4QrPnB&gdz&F$4WeAqVLoT_!dml
zriL_Le@g^?PH2QalXp91Y0vm6DBE
zpl|v)v#q<#7Mp+yW%*98_{aCf@=E6tfkqngksj%^FXoKPHUk~stf93Kqi2Sm{PUeR
z7deW~Jm-@yhESQUf-Y+p*8*Kx@5MXcxOz+Y+s&V!NS}S=)Pc;5!M5zq)Wl78UNpFL76qMcA~^H*z9(8o_@%8K9#
z_h#<(iR(#fdQEsjRZQbdyqNs*C
zy(^foz+8Lm=x2Q+6eUymibp&9_|S`jpN5Mn$nh9J(_qR!P`=4UHKx(ZH-S%^kOUL7
z{$uE_B)PPxOSbE@urz}`;2EKXthAs~sxY(4&#p?3J}j2$*}_b-tOdA3dNTFU1ioze
zpq>lXpRoPQyz17-rb4)MI=1%~Q}91mDZhlcxen+
zdfE)a6V{tG(eciwtL{;SBkvB<8B}pspKRVBd19&N0u75b^BZgvvh@i9jDaBaxA#$8f|2d@Loxf+PiPA#(f}^nB<)fBv^4}G5YR!
ze}&51;;->rO&WfSf}+LxwB2O6rQc6TZd{ARNG|^dw5%P`Pp-^vYW+w5?bp=i%WrpC
zcoHV-9lFxKe4qMea$Dwfx_PFy0x$egX2$_
z3@>`}^8(V_MfwlQi0lh9E%AsE;6enZ7(CHwd2FjO&CoJz99d0
zqZ2F3s-e&903JuxoPp<2zTu;r5IdOQ{4USDV3(8q#p;Lu;BD_|A|KJ92JUg4wO+q;
z8x}a6dE&o3JEw}Np`h)usF95IGk(&kjA(hDWx>Yr7
zcUF#O9$~M)bejFpE3Wr6VVM!EtZLaV
z>&Ptz{4b?rH>GjWb)bGUqz}a7SMAWXweS8z*@kaZdAGtT+!^)-mK~M9q+qY#9Tq?-dqK3ZVAh*@?8}iIhpA$K>WrJb|>K(oeB_r0dPzF
zq4Y$|^3U6_@e+{wNBya)kR{pM+V!rnlXsy_)e1k|LhYVI7|Isoj=UXk%KEMDyeB^x
zsO&t~YGI3q9=a}V3U`)u7<=cNMYp5o=g+((9B-0b-okaS(kj)WiO}1}ubSI)3JehU
zG8Hy_f$s>LDpwHC1(jiX92F|M!joAWo74#3O5QP$apXc`?yjs$R`P|LTb^tRxtp@1
zdb-rKDB{HYYH&^8=?RO|
zNz2lISq=`TEABz)xzyq$!7Epi5KrUAq#o5~0?}Bi+ZMr1Zyxj$*CW=rvZzK16R5>9
z6>cUBF1*ywGIILXB2o>-m$UY6hnO@2GctY9&fo1sAUC|GuXLNITM&;`zw8<{D6iJjFZPQfr~6^HwcgR0y|zI`+T0j*G9ig65p(n$%Em
z8w4p~$9Fpk0e*Gw-WiBaQ)KK8iG?FjLGY=i{`W}LJe-u6v$(5|h2s}s?5}LnsCV22
zl#XY%+$$&l;oHR&Y+}L!x
zaN4Q=9M`>MOidY}R9CxdXvl`Y`UwyR
z*2@-RoV7fE?8D@0Hfnx3K_H%=UQu!0(v#Ar-$l1sq}9q<$vepr2hYI80nBm%0bX*x
zm0^`N`6ufv*X|z^r37Cnwm$m~t1~(unf8^#9N66-?Pm{oa3$1ADy00o@Xk;&lO~yW
z81(&$zwXBmU!=3UU-&{`MFr@$0RuxSV%u76xlgj*E#J3-2c~W^H%ZFKrmUGp{~Wwl
zpxF8L_=C(@Zt{>(BXNh?unINnjg+US0)BMzJDS{=n{}
zlkla_lIvz1R$L?ndqMQ|4Op{OJw$rtR1RTZ+N~(Mc+HOlk8+aLJWT|;I|heN&Ajt#
z7J0y=o}dCZ6^GvD(;qji?dbLP$Edwm_@>wLMJ(`OM>cdD-8~Proje&XTYeTb{vm#w
zxXEn}3(B{PYb#!2!00o5dql^-Izdz{>Ig3n^*vWaynqr~$mrsmjjo*Dy!xAqsbkc2-)Cc0n&vp1Uj+T2hC6FdIaevRb46rN+%!1LQ?@S`OrWUVB+H>->7(PGk+qc|sv`nVi)0KQ-I0FN9%d_>vG>qir_8Gg}&4dAK+Q{HH
ztQ|jXw=EAOkIBfht;%%ya6nHQpoQGA*Lrcp_ZLz;t}5XjquHWktUD8((Ar(ctup3k
z=en>+#BlTW{~_xwqvGhIwZR4wLV~-yh2ZY)!9CEpy9T!qEI0vzOK^90cXxO9;NDC(
z@BQYxYldIFSas^`v-R0hRS8|$wU%X3=TkfXG|?26@h}r+H7@vg5yMF;qsI?yT`c_L
z6ZNWp@w6Ldc`L!X2d|}8*)A9Q`uOaeLOwIuxo4#3ME`p|w9yX@Ym!E5oTzZrN&6SC
zmD0D=5B{r@J|Ll8i{_6%52AOap-=_91O-y05#qk8DXF@j2%TC4(+*&oFvJB_|0>IS
zd`jh@+%#Kbe+fbnh{B2cGfVA5w~$F1Fsz^u+vmYla#}ee@1ztFAQrrmfW^@bHm(3m
zo569K$8r;|@$ed5J#^_!S~Qc%H9jKd?{&qN!zag%is@n>N|OkQ828|tZp}o^Y_=Ou
zVlv7L;<)03r{Dn-MheoklVW?rJ!{%y@A%P(3%4KkcS=c*w||PdMYgBTp3*;EhWeV%
zA(cPZ+R*<9i_~ta5<*RxG^-IYYfg&4my=(x2%
zVVq?=v*?TVlg&*Jh>3OEj#tm8B?mM$XLSC$*R)^h+&s6!9kohizNl|^^?W|sOn4LT
z1>AcDQgI|qDzsid95$7ClYq+JE93+R-#uxhC1TV&*5+9skKYlWxcWYHH7cwTjhrqG
zQ8q1Q(J-3iE#Jl^M#n`K5iYA48!TPLv-CtOH||z`S{8goolkTUdwK$Lh`BAjuzr$H
zjjd>^`!}5sX?yDX_cz=2jUF}iT^WPB{L0YmW0R;RetXI1Llpt0>yUmH^-#@L(_bH@
zbrZdOMR#58PM(DPN)e?Q0RYk*oqRzh9R+dn+dysc
z%=cI~EG}@n$vGKq6xd5$JioO#$&T+`B4V*t%&Dt?%{(3ovYk?2f>jnY;e78W?Zk(8$K}ywO#WS?t#uz;`7NPR
z^x&1-qf-#0oo8d&d58fR;savG{ig8JI=NUUI0uGdCL4<6b@qE7$HKDJK}&U5(YKL9
zP!Uon(z0eBlZZ+8v?Q_8zBHWZd)H>9mSrU1>bs*89mz4w-q1LB^@lLddYh#l(bq5}
zEw$y>_~J(rF)EGRpeE_TH`0lWBycm8;&OHG-?fL*oN%9>jO4Rk_#$0nzbj3?>d@$4
zd%8SmR)U&taU(miD2s{plqIZRia~@GXS@;c%o%^F@!JfuMn*1ax%780DzCSBoum!@
zBEL5mUx1j3qA6jiT5~f~a1x($oq9t4q7h9NFu28vh+)9=F*EM;Q&4aQwg34d^(ND6
zx>1^U4?D|!I;>$Es*sA7OOx{cZ=ByJM*Ot+aTPwcx1iX$%Qe*UWwRD7anNR07=|Pt
zrL{}IfTJJ3pf28PY?lXn!+N#CqDc!qyX5T-alsMEZ`ArSsl>5E(5195ANy4x`qRZR
zKO`8;)WC|W2d1J_af$FWB6UOAQwXSk4y@{FvAj@TrXfB(g#Ocg)2dCUVFLRt(!RBY
z%X|1>Cu4{EK%0ZTD?W;OKfk|nS?U~=f+&VVcZ7X3jEVSVN(WmH@9FehWITi6dB?a{
z^F~dOW}4!`R5ruINd95+XR?zXk>`2l2sZUZn@g!+`)&DnefgO{$PsP^1Ix<{@A2rb
zG(+z~E*F~RmR9!KBK95?=acKFlq=*r4El#7!S@P26^6N#qwt@AJKjh3$3pJ!{iWG7
z1Xk-djpuDC;I7f?Z+t6s6znCm$W2IOVKXtP*IzFJ`DB8fBuxct_i7u??0$AEqWV$G
z(9Nq#&5fKLvXs{S^N*0Mb8hDrZGj>7avVG&!BYC1Vm;dI>61sVIV5=V3~XGCpBIk^
zcHJ%ad@1uBO>jZEh*M+kzr);og@-V61W9A2@J2N}ifKr4qwT{7027fm&-a6QYPGDj
zLn;H*z@{e;kJ8R1#42GOo{2OUkL)0nHg*l?%XvWTs&`LJ;z3H9wS&=NF*Sv+^i6f!
z{%9imcDE*La?x=6QtX`xxh)I{n@F==r;6{=?Zm^S5p~&#Ld2MuthShY#lNpwUu$-$
z6B<>pU%z{0=%CqqR^1fA6+P_uovpj?q_np>RCfo?wtD|HxQK|?Y=DL$g}Lk2zI&+0
zmv_?0-)!dV_B`+uHr-sk_Saas${&U7lQ-SqV+zh}
zOEdL0yH)9v@wVBA*c{=9*|}yU=nUDL)3HA_&U(u%y2o~(fCBsppp_AOr*z^pTpF4$
zow1)*un9>4Ka@<^l91wNQAFa!q)`z8I6=DYAxbL391-#pPJ9TRvj346(C|jq8zVj`TaHQ+0_un#Y
zC4{td$?WXHZfd!%VV`(4VY17wD%%*oEbNU;;my_o9R^8sF!l(%E-L!iPhE>7L>Hf0
zS=JbmLtp;P=}g3~zr@DQl$*H@xBm;wHo@Fiy{q9@L@{h_({kE4pnj3T)>*KSIe4%Z
zw6;4#Fax9i?`cIA=7%$1?};u9iN#<#2$csM*1UqMi#yOMs%77MNXGJs8y;H#sZ!3&
z{roiy_}Y#S>eRaknRaJS5?mLexI9T+9|BiK4lAwp0~D?f57{{k=T>~#PFL+SjSPLh
zi86CqD#Sc
z<-Q`7^to$Zw1Z5noU%G6tGN6xsLM5%GH>K-+6av{IwF7Cm?D=rz%68|Xq5@%`
zSqb^rKrKB-m6!~p+Ss)h&f}1Iy3U;yo!{5cfa}WooMe1YY2ccDM~Jc
zWi@NQkmzWtV97Z;VB1%Vw6>0gIb_yJEL{ukCFY&H^LE9GR}nv1wC83_Rx?e;l+;Ql
zhTEZ&^7LJ2@%)
zg&EAK!zVA`(6AKz5wKy2eVk&o#M{z%5=Y_-7LpE9PyZ@YoZ4sGnj7KXRNmEogCxvq
zle(PxysANA`p7+=NvAL-MjI;Mg<=&o;MD=O=N;$r89$M}3F7uux8oi`d;@anVA}~it&iH
z6V-lykXw?~BO;BLEamFB*^=*yVx-~ocA2?Z?EO%0Y~$hafX{{pat&Wu+n$XojY;OU
z-ImupkZiLhgU>3R=fDrW)tI@B+b#DtSe!~XTRTd*9A=E;)Tp;`=7C9KKlkHGu=``A
zM@||*`H$TAvtY5JSv9CdZ!y2y<>|22U9NXw29nb>#kLS)BTyCsRJ4)Z69jN*q)cA+CGd~3eOgLLV)&mk@K$ZA^U22@#p
z2iR5I=1RH?>War)rb$!?9%U!~$kk#PxElvmz{_B$#Tvo1(%x)#EnW3GyFWSBuSPd~
z@?08tiSe}kGc)yvnbZ9U5f}w$s#R`dTfCrQuEAokq`Zu~Q5Rauc=D1nZzp2l+H=F#
zU{!(n9hAOSSvWb<>ZFNI9OU~hW2dNWt|uKxtZMQn7@qz9&JUa_|IJ9#`0|l7tjoVq
zHVFOoIr;{gWY>A{i}`C)KE>(Sz?mpP`s?;xxtC&Yi4CVm2;Yml3;!gJ*4lm96kD7Y
zsH)h0&|8JxEIm==Fbr8H!&*g^&G`sdR?n_YY~Z*|bq8BGbnKR*G8MOQFe8_0XqSHW
zYwN@^r^HA#eKlKsPE~$Tq*Lt@?(g{YGB?+idM}5R`1RUvFA2n__sxdN)>-KU<(^X+
zUcTdD%i})jI4r!17(df&0(gpDFb=1UMwYXptQ{-~WES`Xd0aPvPcB1VEv1abpz8ft
z`TY5@ot{u$I@TigTN=f3WldtWK80V&FW@Yv(`=}?h-8(f{xc1S!G5Fm^9S5^dj&>7
zc&LY(nXM>%m^=`NN|+6>axCPvf;h59YhQa=+KWxRtRc5J+Pw=;MvJVI<54d%d&aErSbFfscBZ2o>F>rP_h
zac4F9;2qIAbcj$_mcI1F&uuAe89lH|kKc;-WanxoGrru0MP|CfnXBhlF|gixTU=^I
zp&MWCXp&o(|17FfK}dS|9&ljfD*b0tY)S<<=-W*5$@qkTOG
zZ)4GD&kkqpD4a?I8>{E(7MmO
zdI_|c>ew9xY2!4CFr^zgE%d?I_}PEz4(`x%~%Zj=!WK2>ZevG
ziPkAJ3B6zpHEs9u+^nP6w&x|)i2)ZNoHFgU$~KXesMMV6NL#a4ZF*I=pt^R{%qorEtFi9E
z@$@sjmtp8{3Wm`38=*DV6y(;6F8Xt3T)LAFcO$xippD)iGG{N*E)kp(*0PaA52aTUb{YcHKV3e55>!>9
z-0YKxzwWjB<6o{gytN76`zjp2l5P(7i`^BY&vC*0=Ut8U^@sMKSw8b3#i5sLzRy>8z$R8v
z?+j@E`n}tf@3k!1YDW1tL{MMnk3UqGlz8g$e*Y^xeGdD+HS6oAO!~133Cl4dqI;J}b*3B3ZfPJA(ui83xT>nE@dSdLxv$m_RD@vw6
z^Hs)jSw_vVz!T%T0n88P_?O`2AbvYd(N6s33>MVOkx;qcP!8MXrSq@k-7k2Q*L%qs
zAu_Xd${IAdSX`WOj7;lFy6$afag^vg%>H70S?1-Z8BA9V!ek;C$xFf>`b0J1!&Vp`
zPQ3waHI@V!fj76w==dmIRAQ1ez@@M7z;bPN4+9V@g-GRt>BrKqYI@7xzsfhxe`s+R
z#cH!B=z>_!*SN{=6(#W~RTRn`Ik*>B_yE8)HbbNikI~JTlb+4s()E@Nab;5K#gt<`
zU-``3)ZvbEgN?~0Kcx74V8weKX?Ac>(t1?7?;r2r9-*PRh_zd|J)|NUs;eogt_kBZ
z{ZTPbbl%q=&teCGQFwGu9_xfW(8H0gdUcO>Zo>?68N#S632#!d6S
z!UVaKL_)#HUq>p{;!SlW
zkb<@s#mJBdNLk@ORB*FzaV)Ni!a8vr$K0=QyT67?A%JAaa7gj9=7@bg0OG<9CyL=D
z;xEjv0wJ+vI1;XZlQ`=JTgDk=;BpN#x|!5rMq36W`r_%X|7oV&ekR=kYoN6{d&6cM
z71p~=NeKia?qBbPxfsO3qAbhj#6kWu{E<|KESFq6PIZG@kPU@2aFRoi%|1YIBHR@f
z^wZ=c=|CS-_-F+kDMgna(^q__uVNC)ZcP&F7Z019tN&n^+!pDq0XTh*QqiM*zV2YP
z-SChsI$CmmwEwM96J@_61B^vE?KRz8#|oOnZ&@f||2VneMBFN%w9_d}pkEWh#}%cU
z!BwPIBIh|0wy*Qeb5nl}`K<1-dZp3KBJr0p88#UlST{O)!>Oi0|Cvvc4A#MRr}q~d
zSbup@mmmW+SlF~%)g=UCWCj=A${%-+gje4X2w*xs8!{i1uQMyq;jE@0JZN{WUir$v
zJ#3D|Ki1*ZhYA^Y%)DD9@Om)BNg);Hpk~qk(&i3^-9N(8%51EUCf>PS=6$$tvP@EW
zX?<;A>J97eYG*_!BW&R+5oC}}Iy#TwpC_Z4dG&8*731R&fyx%{G4i!{l~6wTcAd<@
z|NX2b5&t^QMP`mFd+;D&kJx`K>i89gaBl~sI#~!
z-s%z!@WN)8TaQh^B2%)wQL{O9>mv;(h>(~P^-oXb8
z-iv0$>C3llRzTY6*x6WdA1va{O06-ZxeeQ$rI{#Ebi4KR+
z%_z$&BJ1E`;J;#nTo2_P1^l{{y`yr#PHU8IR9;vYEwxaY+RJ!XvjS$FE18{
zxi9lfl&XM6-6)6vfPF@Z<8-8%c>tg5^s)g+1qB6AA#^?pU}B3POullKljS*Wsvr>q
zpI}AOSZDvdr#EvfQImYd>+>=h6b)2g&H2o)|LS^0I;E}PPIyYfmCW&3<8a`
zqawb!UD3I+*^C0)zo=DQWH^%m0Km1|xMI$XPbQxgY47D5#kjZ-008zj2&14Xyw*uB
zl`ZUnJ0UgzpaU`RUYtJTfmTE|gK#ZJ*jMRo$SO5I7eN!#QDgv-YZM*;u(e*iZWejY
zN?AUBb-g48;}FPXlbMOy0Ppn-hMqw{Jpi)l9e+nI>l`>#4eArpuB2%=<3T||sIzy5
zjXf07L?kt~h(mx}$VrVFxDr#9j&j-i+m)y03z+v{;yTn)OUmqK4}P}rZ9TW5LfGdU
zwq+ObDctVZe9hz88_XxjW%{xKC$y!c0d!$8lPE|401mP_+mbhYx+!a8z6{cC93l9y
zU_%8U9Ej*R*?0ycN5!J;M45vR>$b@ly~2DaEHa?czh-r=3XYCo9s>xLfm$|aHJU?z
zH^msi%0vy$UHYVB)spL830`=!pxy@I{I;-69Kxnzdu3zali&dWOlK>xAd09{_sp%H
zkSuR7Pr$DDN1IE6Dym2sGV&G^E(ETH^RA`)xQ==Tpj?g)MgeE#;Nae_BXU%ewDpEh
z5!S=O3V@Kt#m;Ft89w2j#&feDye`7_c7p2#6>S?b3OTHUUQ<%Yo<1*@plz?Re~xnU
z>L4UB7=2{}vUdO`lt76Don@i?-~R&=!>x4rCYNgZ1-Spc)|vR?QOn(8OZ7EC&5oM5
z&zhCZ=vXSTo8RDhkue#w;RO#~Y=&jS=$pZq6P|a&ng`ct4KXyTA2czC)M3a2n3x1uj4UL!2$mzIJgt<+
zrxLjYogd~c)!9?&d5@^!#MV63An=8NS5>z-6Jh$f73=~b_Cf`Vj0Bk>LKE@fc}mh&
z;)caN*XS3R!LJEOWYV{P_{(YD&U0&EJ8N;W1hJ!mYl-d7OD&Ossn0avY!QGy$Y$}?
zIYvza##(ZvSRnP%%RKKi5OT@WqU
ztMMkVch#0-FKxPhf8;b=(${%BYM;Gqi;?!-n8QJTSM@_#!IH*-jtv28D@VrO
zpEn6!*newz!s3&Iv)O+r1+SIJ4lUZ7B6DY8VL>or#TaxdPL`>(B8Qk)+8<7i0(KD5
zN|Ik2M5?k<%rOBY0l)}}?!6MZIN|fl-F_#3k7P0v}#e`+Fl2|~Dmy(egL
z7(PoC>i~{u^tle3F{3iwqj@-e6}+?wwjB@=bQP%=UEyPDQ<0Oa{)Dj5xWa2stI*Ck
zyUJy`PyrPYr(IU3kddoT3=Ys?s^Nf1YAh`E_XE5h*v6S?z~SKrcN}yR+>fLOayMOQ
zDeROk8x)cq5CW{cAG?MYs^3Ez)pRHKmj_}Hg+p$?gL;Bq$MQe~c5TEy%_{jWlgbXA
z^g)%xZ=TRn!2;3~eizp2PVlBsfAG;X0>kp_itt}Ox5)qVHL^lq(m-x%{1MUkWI1m#
zXfiUy^VaEOrK8&4`SXJPMLyii??4u(djWajH<&RU`d}<}gaF2386=FzZ}-gv82tMW
z{*F(&s^L*WLo*rtEyn|p@c3USiqn(s1WBdty$AdlwA!pm8PgzOENT5WTX{#MeI5Lp
zeq>akS$cluT8pKf7Q0}~cR%fln);49%3RhSGOe^Dm8+ebo#pjzQb6QwGkUzt)#eQg
zhGp{n6DgNLMsivAV3zx_UrkOHT}`3JNO)KO4sZmP|)k{#BKQav+}*J
z!39nUjfuAb$va&1EJ$#ssya#T+mp8;$w)xi2m^}&kkd=Yt>;2l|K%x+{G6|DXFNOO
z>ei4+J1iAmbFDTisQ;u#Qu(Y6XO
z*?G~V>ilTlS+=<+*M0JRUcB`l>C|u`WG%lEDA~OPINw>&{2j9(CH;I;dG*MK+{y{I
zqaT~YHs*!uMdI4GC;H3~)7K~c-{{x2oY?i0c99&kS4SxsbtFB4tZB{%7QS5Ibv|kX0Tvu|OPSRes&|Zdpxbpdi;d}f|ZJE<~
z6K*dTHjpVqUbp!vrpg}$jjGkAj;=GRpCUlH221zz!ywBVv;1=x4Lo@(3>Sh3!l_#A
zqNWBCO-P7CtxLfBGO14yS6A0ZUvUu9oaL~chLGcn2~~}4GK;E?^nW-0$aqC8%$%l3
z7m$UWEGl-A@g1`*ak0=jBStg>4qSvsU0qFEz?cxN8sQKbu)^%4QZ*kV&c^8_EtC9!2iIY;>76$`?0TTXKAE
z)v=qsw%Yo2jdT0FmY=&4!tTqFttn@bsBSg
zi@e*ll#W;Fjg9Bw8Q04PmMWuh2b1?d&
z`2S`)=$VLr87!OsG+MY@r}7bb_4cdvUF%;c=}aGP%!s}DbeS!)TVtzSpe{E^FLLLW!TwQqu5!tVnCP|A%tG}xUr@k1FTzo=@b#tI
zyJ=d`=B1ukuWg|i_vqC-9cVjBn#CgWLQr3Ex?m*;qS2UU-7Zw|Jw5yt!tU(X9z6V}
z-iuvjo&5~rM%j0!5MV-vH)P)+A`hoY5Hj5Yrcu7p=GLyKYa=Hip
zEW&Qa$+IW$yfFEw#PXOwGd*2kRpTnucuqG->SmW1?@?Xjq4##a7||Peyl;{s6YG35
z42qy@t(l**K+iZmtq4r1w>#2X{cQ+DR+kO=yrAaX=*yOBoypqnMY2oyQF{%6>ChOQ
z5YUL0WPH{JC!&zIbNi3(lnlG=MVu;}xBhECBrV+B`bIsqW`M7+ThzMPnZxs5JSSZD
z>JiT7b!!&;3nQ0Synmyizi5eFBWu23m-3A?=5;^@9ZpM{s%Khu-+Wyyis5=l+#u!u
zY@WUNyPFK+aAQ3|wn5~iUBx%&)h2?%$(78Jbh#ek0
z@(7fm#&qj@@Fa&Qf@l@X!y3tXMNr+*x%}&sWym;6y5@}8T-$r|R;t@=eO;aLh<*a1
zZevH=GF?m^M3aN>O@UPCqQ<;5mJ=3|)eLu0675X{k>|YAOdJ#keh_W@6%yT=JG(dC
z!UFM0`UAmZ-}fE^$J=hV?R*=rx5F!Dz?j+pWtwnE@St-o1y-+Uu#-3T^is1{`B*G1
zPhh*?s{hrcAPvEdVOddQLi_ML{Kv_6Z)k;UW!5tH%NF7-+g{@BN+UTgfSRrI?PcQ;{m`brAkF67YaH#+-Lr2A568L_!zL#qL8avQa`CNI*wh
z1S*TeX`<2OPMbrZRL4KM6d#QOU)iuCkT3Nsv(y*X6?s=Vf7ij|k*mlr`1qyd4JzGv
zHyMHCjkL+ju4O;*mJh#fh8NIuTL^NpNFV$aQvSwvDDAe?L^l-<2fr+B_`hN?@gdKQvy41<_rbf(uFKUX>#Ukx0D
zoBfam^KtjjMTC=k$jj2vK7VTDHX$V45tLk0vRamGV4ME+l?N|>O~rip3P&l5&Ufi{
zBxs?~cctc8-mg2C)61)*cgfBL_68gQ0McT@s{80xN89x+F$atJ0bI$CXK$
zl`*&?CwAmPo*qzCfypk;x}vm@x|XRh@4PF
zE%>^z78P_u9Ck0aSKO@ol+I@90`R`B<16tTmxcByaCoX1BqwiIO5U^tZH(m{PKnxP#0)BRy~oVpiRhoyCMO<7=b
z2#Ter2#CXGD>5{Rw-1S@dnK#Y*g@ww|F
zA_Jh*7Z+#d_p55=*9(h~{<~VOi>1OaJ}#SKU+lm{uOi@
ziOs0#Nteg=<*5;GNYDGRxjmz}oRv$9|@m9Wg_h*#rj
z_uLmWYo=!6ttAi+k$<->sm0U&I?`*_iAro+rPB6w-8+~UF%yyR0IH@}-7nmV>b#ua
zo{y7MAXO^Byoc)@@n!KiRZyNl)jj3%!0VIO!$f=+Mp&a*zJ;P95|v}4Ar$FXxo*Jt
z=w?`hPXHSY0bh8&Rgr8e+0`W_Y?`h{8qjmB4h5CddZTfK^|9}R$3qGu+H?^w-Y)$<
z9sxV!b`#@yT*wISQtx~nsBLIP?oyRFi8DnM3}Z{pid2h=OWUgHjB%groM#U
z4K2M&uCX26fiZ~hZs^f;S1}5v>WJke??8T|$%eISZ!|>fd(nNzUWr_+WiN7u+bRZ2
zptw}ZmMSHh)*FSUkjGoVDJM-tkUn@(cwqA$zxL;LMXd}4Be|>{g>0TW##bBxZ+l=v
zKDNH$kezj@xynbb64k=Dx*`rfku|9GP=O2yVEf(->`U_a!4WG7wNsYIXBkX94zI)1
z&h^5s;BqH
z#()nY4)8-72|`nxODRaq8Jx8DqAEjH0bsjKJRFk5GA)S5X*V%R5Kx&y)TdYemQ8Pn
zz6TQ#+v#`t-!QC3*IQK-oDn=60?$LPk3D7`9Je_c=c3x#e>H4BVq*tMoyOpf_MT`6
z3<&xoV@F>{{%Weusy%H{R{J~2jvH|RMnJj0PyaE2{r^C|@L^7JM
zK-L|B5Z)qxz0u(W6{Cxzny;=)`Knvk6t36-_X>prhbPrt*W
zN+6QPT;YvsO%%x#@W(eW1GJ}8U01#v=Th1(9IAU6x^V>?$ZLqHbT#Dhsg1Mg4*1{V
zmaAuz(Fp_#CnCHG??F?vG9}K{IhnBQ;Y9ERKx@ks4z+756A*bt)!Xk4B$zV&&!nNm
zAPNwo72`Z1ekk#~?78Rs^ps?sw>+;sSLYb1q&>7e=1_V%2@cz%Fr)9KJPazr>vN9u
ze!+yJWh-eY-;_&Tru^P-U1Ue3Pg^2kjB=|)h6dv|`-q4PYOKcH?msi7c%k<6g`r!ZvzFzkHpWJVWD6X-cF>hua9lA=3$p!Olu{*W(iDcN&6VPIMsvJ}{NgpqVis~&vL
zArl#hk)lC_dJg~yjq$btf&crd6AQC)X`<(|n)reg##wG&Y?8x3DAmFG%_37A7>Vq%
zuWj^HCzA$@r1fpNb*
zZTG91z!tzrv~WnnHQw{GD~R(cJ?4u{{wM;TJmlqIvQ=VL^Mbz8%J-pSnl+qJk_c(1
zMN3sfMjtE{^m5hm1cp(?g6R}K{rb20bg?LF_*b2ai;Sk;kOYG^LBW}PH
z4RCpgG5%!0BKEHQ^b4h&PT20{^mI-o5kgwl&B^a2K0`9?a!uA}7*#Fl#C-PHi||DJ
zIO#;~E|Q|YZuir#Wg6>^kG;YF_qho)(*O0`#ET$r*=MIrwp8)&?6d3pH8dXuugVTN
z1OB=9{4FUFxY*t$#7myGdMD+vs$#Xw{SK6rbVQti@m-niO};ZQV5nnoSLX)2M#3s-tKGJd|)>H5f?oQ6Ne>?T0p
zw%5N|zOymQSaw=iQ~O-TwoBo&M?w7MHomK8WXAhnlLxwL_U(}OTv|DDRBhc$entmc
zFuF!nZr$s?rsU4Mwk!4U*PY&goEFP%^bgRZOiD1SJ!T2fV_KLVKtV(2YYTUNS+3Ps
zW8PC$Y^7ovK(^i7k?*GC+BAe}ymiGvHD~st6R1D^5{VU!yM7pq1Q`gTTh%|W!5)xS94UYD#8G#!X&ID$NU~kK6;$5nJuGWl
z&bBBvFtnR-JN($>9?EKZ;fds$N@Jl}GWPc{{cnCrIjr|lS-#kTTiIDr_Vzl+$#gw@
zUV^}JxwyWaVB#sK#KwYP(DEsKKE}AXkhBC<8^|@yOA=8AHBZ)&Xy+)Zo`2M!I6sm#
zU5*SdfdE#~6>3yMnKjh%@{l9x)|$}g?m%psNVl&GCvd-W%Z+?m*g(|c>D}>Kz~f^Z
z-q<1y47JpeQhFtPF4C&-nF4bwD!OJ$8h+}hUpDXEJAE5yCU{aZ&n60O_VJ4se&Sx`
zn^1+moLD(l8Lv+;HvzvXJCw!K&f+9OwXrA+pFcn54ehyzPzQ9X#(8T<4%GNKIbV6+
z7IKHwx&D8K=WR9F^FF^^%#2{mpPWwvLQ$x~#@Q4DIw=B_>u7O^e9{rxsGS<#6fJnN
zKp3Utg|m^La+XSDN(8!vi>?O+?5l06)+$H7Q&^c&FR#sD`BYpC)SYpf{c_##S76X2
z>>>MP61%N_6K-v(X7wxoXyNdVbD#K(M)7ji_fz__e(zTqM;wg0fu|V*pqY}0U-?b0
z3*D7*A=kxTNn3~EuN)xFg@LnuZ58Sso1~!RY8j6Sl!y^C&AOzqY+8_
zyJuP8QmpUML%m})eR&uj^7)Y7-J{N(R{GuN^r|m=_ibds4`~1IK)sm;K6Go3%_&u;
z^?#?^Z~jCI;LM*mke_9nurPKAz<@-w#8k1&C`!#Im#k#hJL4A9Qs)fo=M()Hw$JTa
zKDCsCoW7phrok#lYXTD$_Y##4B~%(DBsuxP#>?XUL493oFI27^r}u?}XRTM1AB^t)
zeuC{U+~aMce9iXs{|uC0L1(j)Q;Jg}!KnQ=zMhe(I9E&CJmofDf*Q`VG7ZlZ
zdCFLv8J=ex(%$|7`#<_}e4y&)AF$i(RdcxVAOFEI8d|H~(Fs`z`Wdv&g`AD_b{BbYc3x$M#v2)K6a6L-DLW^U&EIr!pRlg80Q0#AnA(0&dh@O>;wTi(cq~m-PO&g`;7aMU~<)obA_3rxKT`r+6m*i)2b1ucSx#
z^*`~~#DX#G(2$uL=is5;v^ayJjL)hAk`N?7WiwZtU!t*iM9Wx@-01Foeq}PMmFDt#
zQejVO)ghZXPgk1pe5HMf&YE3%F76O=*GIXwOeR?MvXIwZdfH9H?FN5JotXNKvDpgP
zyD5XsOPpFyZi|A#^N2Qwv!?anGZNOKDJRt@Yo2fZmQOeU;?oCYRY
z*MN^J&BvfRE*_8Q@cP%K-E{YV3c<*A2cCca{mxmhKk84|E5Xqn^ee{a8$+T-gfcm>
z&qC}5l5_IJq*~kOVe4!2DCw-{%?88q=Ra!2GhLdnPWOdKVo*sa
zZS?WVhtdFb`rSeQ+(EZ#u<`zhQVz+7Sxet
zw%bk5?Z%hZn4f@SP5$5HZ}w#pl5>PU-R7m~kBfb69hKIm>2?1Uii@{e#tPdpVR5Fe
ziHy_3N2Thtq~4u^LKb4XmyY+{?c}dJU-U!5Q8VkIfCO<0F
zey3eEwQrIkF^1@=$R!s4)Tqd{T(wujb6hXIH4HkK;?9e}PAFbJ<&X`}MHLd~&a4mJ
zm#+-T|G!hK{*01l70$nA*H&D&e$M=bIx8@AzaR+zlThrld@LBMw6-E+H!!ESsai#!|QF#X^f8g%5?F%)Sec#gIT#|
zmGgF1uHm>)W$?o0`AJ^-Z`oy$xltW!+L0UJ)8(UvVFQv`$n$A(hJc&>6Dy4s%xQAZ
zji-I?sn)^SS#2sKw-kK2$JFAa_9Z7Wr%LNq_c?0(NwD^%>W!
zXT)au+U9fUJA1@anlh>YZDOJx!wpGV$g}}}B*8Sy;MQw84=X=;?$f>Q|DM)doGwR;
zVihVbnZHP<835w*%(F01Tm(jmsHc?|-`!^5Z69+^(CzVtek6Tv$oT+D{;nV@TBtp{
zNf{pc@06QfTk~Et%)*@VL!m94LtV;oafuGREpV{1*X;i{^%?r#)TcuoVz(YSB;SE4
zR4=;119BrgcKtVJ0ifG?KW_c;-cZL{0Xcy{K`_$$qo}5O;kIpnO|RDQYg74VO7GU0
zN)v(72!T>~DNZi6tophp+@bkA>JY1dY97Vpu#eh^WQgfwnW+*SGZJ
zm!f>gR1x9QD;ug|J~_K)Q?hyEoB
zlMyM^@TQAz36c+}4465>P0q3F5f$j}%OI_YWO8cA{%TTF;qAGpDocAs2vM~5(JEKZ
zEh@L+q0Aqc$Yu+7Ce}&eG&YzQoeImr?sJ`ba@HN}yj)T(bbAS>`GpC#IbPtM1ME+5
zGC3?N21((hJbvFcG4hQ1
z%Tvc?8X@fA>A9qw?~TI!u}?yM@wnMxJtDo(h@_Qax
zu|wD9BKEYmW&E42ymZjRi1FzMZ*^*bo(bEbyR23xBcC=M^PVYp}v7%gQfE2x}Ew0ka2;vQilxO>aE
z2P@{tLDJ$8^NOEqo6u)(zT-8LDu?QTW1Id;5~lm=_u>!
zqbT<{w{VgKUqas*p$WlwC6}UlF=uiNv^NZ&gD0d&e2-AV5;>@aUca8{t_3lFb
zMA6#4{mA$kPfkh-o$UB1@h&+mE41)}0D8!rM^RlAPV;9oB>eU@ax2aM&h>rl2elu3
zM$Cd@=r?ubGzMi_vBFTaZlk90&-b$i$bU~eYCUa268ZJv6-!T;FJX`(V9Fv?3xk(#
zn{Z@U_7Zgz)4mshhZ%RNcIivtrBJhN2kQtC5vK=y-sOTWo5(Mv=}Gs>+Ec|3vzubY
zeD__zmzFL~1S;kl*8)b4y0hx@u&}{FlrUvJE?qXKloX&#K|!%#MdWDnq~5|RQEORC
zsk2^jUEg8@b8O4QYewr+q}6fL(+8iL_gW-ShEgXs|GGTSG5eDutAkm6MlCZa&O7*m3i7>28zDfxb$3c+QosFUSrFDfThx)Nvm4p`*m%
zDy`_utmrMIU9WJ7@WT=?QGHr7H>X8*8Fdoi!{YeJF@H6;aq}yCwa!+QS6sf>x(1s%
zo=g8J`IP1C&ML;l&2u@1H9DTKld=3*@F&KYYIQKvlYxt$=dmw2^i>V#_@i`upCINS
z>i6R6I@@WJ8{u;sml{QjYP4-)j7m+S^St0Nbxp)z2DLVuOVXUKgZ0|^XfEn~Yym4g
zYrc={2YJ^G_*^x%c8mv)Q`*iC^0o9#y!~Y?+)Zz4gBHTHS?K(#k@tm9t*KoozSufG
zrIa2H{ysv)fc0n3&qTVzYuv%G*q$V~k1=aX>Gr&G0X0+>Y#_wg_XI)p?p#^Ctb`GP
zN84wvWLqGGArp7{=*}c;5#M!IW_YBR;edmZwJ#_X4IXGRaeB8wxd~cwnAZdEfFHBa
z)0zMHV8{PMSzqpd{uiuV5*kAlWmel+ovI
z2P_4&>yDkC+!_#P!YJxrJt>CI_?9;kN+?zy-?<{e9`r{gg>-5N-TRHuzhv*aQ-(Mm
z*nO>*BsQ6N(7sgHJ$#evc8^^pSAcKtT}}9k=)*OZcEO{@9*P|HHKSeiMaRE$xx;i0
z2A2r8T-vpy`t9l%d<2!IWI`m(&mq35RJU7i&+bYKZC}i
z{o2lN^q^mM26xl8)$%Do6<1lSi-u)VPhrOeRY5~ZT(B^t<-<=$+~gjOaxusC=+<*)
zoI~>EffFVCMhFudd$+c0yCG;_0)K8pzRe2I|9{we%c!`bWowiO4#6ElaCi3v4esvl
z?w&vs9D+LpcXw$>aCdii4bnKTo1Am+{l<9zk})=W*OHnwYptr;=i^gL>S6fj=yz!T
za)|Y=LuN%PIf@|bR7d5c{maNO3i5+tVTU4M
z&OF<9mUkv`mb9g&MKPb`jEah$$TIh7&7SbWiM0oLl0DF8V-@Bu3@xA8^{dOtL}y@A
zdn)sTdt6LL5BIF!UXi~|plKkV_fh5U><)cf~dUcp$AcwKwl7cFw>zqqR{
z>?*a3I2!pqvdhRstd$k?oQI!CljCLtdYosES_^)BxP0R}bvj!8=LV$S7Q
zcp&^E0|hciC1>-ItUNpl!pGx}XU_a7Usx)-x)e$mcQr%gfL${V*i_U!3ax#fBPzoE
zIky(rkv{^V|3&>XT%)szlZBK&YPyX6txE#2g!dx6b4V^`MRZ?WucehXW4BHZaxt<7
z;+tY=on=qB1k9N^T~+}pUl1><)gooFPTMp!pMzn%ljLPvl+`NhFQ{vqj;}wu2hyr?
zHruZZBbZ6J>z6c6!Vt?BS~IrFtl7RPq*(8QN~IT0Vab%tWekn(ee*hvsMNTt%fY`<
z>^a$e3&%Moiihpa8u$HA1&C)g+onZrtG(@jvv~Bq;cBs~f>KF;bOlU1(+gTo{{A?V
zy*Wj^Y*KQ^$HRVxnpr$~t3hWdV|qh`P+9L}9@;%_bRFBv_dqvw%&<%8j6uTk5s~L<
z%mo&GY?K(y;=-q3YJb6VG@kh#uymZ7lchBgBfi&^)#3QfKPp9(K>vdbKV!MWlv=RR
zIk8e&C62Y{bfO_({9IeO%bw*Rdcb>wjXHh(JF{>)g4!(@)k2acP0?7t%9X@W>FOk`3lkd4%mP~W-C+R`E6lanksIc9
zwSQhkHJhZejz2Cf@i!#uno^k4)4W7k5l1btXeIBvu%eS%j2(-T(70Yd=~D1jY2Brv
zP8vhID9vqQj|KdE4_fVkbr4xo-pxRt`IERBL6^vO1;L~|EN2~@RTzle9z
z3G50CY7B;!W+<4#KdGQRVa%A&?j`s|L@7r76$=4LZRvZ77Gc2aw*!MSR|RW+LFrB%
zd6Lm-0pb>gL90(1dnr;pt+RP&h=EHbw6ap(0ugo=Dt_9kpty$%UX>e$$tXEp;dC``
zZ^d2|>&PLBKepOgiGn28V@Y(+6(^HD_@g*F7(^_=hJqf|_Z=VBtY)t3rw+6|T8G6$
zB?lg(a}GJ0IXmkA4yztWaMsDoSAD{6)&h
z7aO_Yl?~<;twDTzKZS$vcdc9HaxvIh)5i?z^s*rZ$``em|JgkgBB$?pLlPLIpIj=K{qjlF
zOq_=^W~K0N-gGEZ)ALIAM}7@W|4ETTiAM2>{HvFQ)Ts3=;T&IW(~}vaTgnMKmoyOW29
z{>55{p(q>|pr_R>JnSqNp+d6Qg2gZ)wF6$_718{BfY>q-apvkrE=qrda36x-FS>C(
zhB3>Z^4oXvgax!od;h?J)0&cE_OO$^4w?#v&(sDEccD`z95d{ybL^_b^oue03H4@w
zVX8mDd_Q>Eds{daO6}oFdUj8z{f3qyT`puqZ*C~gfX5{@Mm@r2UKS_!v}8)fUsY&Sr|X4Wf?W^9x2Yk5(;18*evu`BQj#d+Ta&C#urEdyJc
z0vQ{WR;R+K-I>P%Kam!(7r_~*4<4QG*@cM^JDIDdw_K@myN6=NlJoa-08^aDiMz*a
zJl3kRIpS~eERKtZ8=
zw0ViL_X)hbyxsHtksam_?;qXR=iAH!VWg!ivh+(~l%*?*DWK||xR2<3&Rl6l++j$J8+pRRAblT@z%}p8C6xcSzBji$0nn
z%_{LB3ZD+3&qbU^$o7yxR6X+5VvyN;s?VBICf-Px)nZh
zf>Z9_7BNgiuA54FxZ=cM$%Di#5!N5dhUS5TN@(k_l`pFp;A7xU!5#Tf368PSha)6%_(>`!Q{we7{U0
zGFt`JXgZT`ip&fW<=3=Q=97Mctm~(Df1Dv5Q&B(8XaP{-VOWx}WUM=qLHDV3x2T*L
zkPrPXCWpEW2}8ZQ5DS8zVwMMnuD_R!{?N#i&L?fk&G`zaKHxysKBf09#>h1GDlg|W
zM5dXOBXP*&6XYXg5B%unu~1~tUS1Ewo-Bs=B)UMr*W2`g7}D^S>O!6GxfuUF5!0xp
zl@x~K(NMLnS$HjmjzR9&F~*mCqS_+(!36?a@B|~C*sb6vuJKjT_wso#b!I>b(o|$X
zUf2fdURzUz^3%?sXDG2>RnOK)MD6-zmJxV5iUEipf1%b;^4xSeC%GjJ>`9SM%prnDz$OU}HsMUPxXJ}$!A~A3<~Pf0;&VW3K^s$7QC9Rye(Zzs0seN=n%`4vHK*>Xyb5Z
zIn=GMs^2och2+ukIFK{SQiR?p0~lXLo#|25JMi04JtI3}Am=oA_#sV)F+pG*(1)P)BTq!}22M6E-P4IFO*%=t08yWVNQEmk7@$
zR9Ev5vdtGoW^Xf@ZkA>%_2T2%V-rzy@Rv22z+*beQeYM{rxqbJ;C63*!_>+u>7-M^
zORHQi8@W@P3<_9~79$HSbD#se>2j=wFE4gY!3~e9GHuql}3p$U}+&3
zsPLN^HE8?1Ij{wJ=FntdVewu>wca{5{|^%3BBVe{r^&%H%EhSeGo~m7XRu;nWx`x0
zkEZu(#5htyh<2ryZ9~7FFO25b^lwR6AGH+b>MDgDU9AXnz$oVcjeM|@#-^0jWO?{_p4s6j-{dC1
zGC&LR^?uA?-PZlyo4CA{a*-x-n-2Jqh!C<(3iwCiKYUJY;T3~?t)s?FEEXV^OVR)tnOwIl-Kx%~-J6(H$*oE+
zdO`C`5QvENlg@W`3X-8G_lbeRHAX>sI4&Bi4}HHK!w8ES{azU*&
zW0v)Uw7ObtII3gAf;ewLCD$VV!;mN&t64$aPFH9hw9pVOssuTZjO%RYeNUo_-}pG-
z$eZM+Xf_Yc1;{z@9S`PR3FbjsA>%2SBS%GNE7QQSg12_UfM@Wp;e%TJn_-Xok`>~Q
zKDonNtAEhYxO9ZJ3TzCm?aXS8^Nb9j9W{ymaB*UnwD{!R`y6pBHR17VNoN$)ERCeg
zx=pUa+caobiGjDJ#qxR+O^W~AK2zyYG&_5iEekjur;PA~@mGWk5;ume&J`t5ChOdV
zhjHJdt*jvhC<#gwa1h4P;B_S$B{KRz4!&fT#LDhIc!k=gHiHFSUy7A2Jfs^cV
zfscw}#)`lJ_TU@qB4LHlNn&QT?JN$bw|3Z7GbR`l9p5jrzu{sCS`JtE{k4IT;s(ej
z);hR-X+&0x>2d=pcZ-SV`h75RMRRm}ij(@}FVvFXh=qwQ2rP|);&LJX79aECFLVCVJmL1qP4tNvqQ{F?J)yn`u2LR=
zKZJNV%Pc~Xam)->XxX)Jk+_FZ*!V6sqv@I~#ec3w7i)NwYrDwG^pZ#c^V4_r<`>`A
z3`HDyFTCIVp1)^0A|E5g0
z%U7fG4^p^>JzeyC_=k=P3kFA8Kl1-Nc)s~-Dx`xmLJ3GVeepa_-We!8lZQK9HB{4n
z%<9OvodtyfMETG6b4qjP_4rn0#5~V{u72iyi|Wrl8BJTX3&ic(9eqo(&9mjto~j;8
z{IhKxe%G*|pT^Z!8^OEf$mSlJd2%;h)(>m;r@;IddfA7|P0x?lwL3Jn7-S1TPRZJb
zi}K15_k_>KVoZJ)O~bjvddUsG?khK+*DP(OGWZ6
z!TTL+6KR_~8T&+TL&nqF+Vz>9vHj^#J?SR#0$q6if=l=E54jfzy8I?8^Q+p`4}V*a
zD)U}xxKJq|h>0iMr!18HI>^7Qr6MV!vHl|-hCgVm$Gvf^%ucCgb#GKE}{b0I-
zy5cN$+|g5BL{g^ivA`Au03coI{hTT!aWy;HkF3k4C;YWG=;{tCXZ>rt1bnv@hdZV<
z!K^@@F{JEct5-%6z#Ni4j>kkz@OlYH_wz6BJ039rMYFP6Icwu$A7PQNBPo(Ah+8X7
zdeH5~;VNSc&PQHwlAn^e!%ks*DUVa+TS^!$o
z6Hf1x$V}&`cV@BnRlH~*mh)+-+&9VSW7cnX>59`-&H$u$cucRL;3u>xjpw?eIZAix
zvh_K$QME{0m+|}^sT}KQIk_W7_`4jLkGLzq@9UV<+oCMo(&2K*V
z-m{xGb60O8zOW;hn#OLI2GnHa*MO$F?gzGOA~DM
z8GgE%2xHJs;QxDHQh##bqOu(OldUfmXo~2@mpu`p>+CeuV2-$?+_|Of-}+39N2FdV
zE*FlIRibdM341znsoso%66mzF=KizlhU{b)T_LC5fy5G$f2a2EPl6g|J
z+soJMSwD9!y0oz)(=r?D-le9O1L7>2x@IBX%jZzCQ@rjN<@@5m{Zq}i>KHQS&-F*;
zyA^E5Jh$CEAff_Xyi)XT!2XB$_@>p(Q5o%JSDQ-?{eQA#x
zb6PSF@;=kfMETHsb*Jp9G*?UGrE{bpAseE
zyw)ydeFfKvqcYk%%Xb6oO*Q+oZ*7G#OP8AJfD&9Doi^VuCFbDvpo?iLcD7-ek{WT6
z@i30;cBTiS*{)WUi0C%twm?MU)~mCFPJ0GiUnD{-gJr1ciN+=XlRqFHj4`hP{s?B09?w*
z)IuwjeWceuCFXgb%@tl-S%V>Z6&Gw=BDRsW9+8|(kLw1UB+7gw@-1X~Yhzn`;*>%5
z3Y|H-GhC~3+Lf(LAyT*ErjzLT>Cx&D9K=SHGE&a8u?`|`Sb=vsmXHlSdfvp?GSaduY
zmgc^Su6nAuI>rqe$0m!z53I{5>`ph&(*#V!Jgq*5Tko9bSq9h(Cy&T%7Vt>b)fpoH
zxL`G%HEqTm88_Lo<&EO-HnRV0maKT-T$1)Huefe7T3R4udMqIu)>B?h4M*M)Y^(#8
z3pO?mnZFA|hscX`lzojSEGiLD)^;yv5#&#fgk&s8Xh6w_gX}ptk~!J6rn{>GsKjL@
z=j<5zgQhqPGG|xeZEq|o;-rk9?Sm^!B?-Py!|g9F2b&SCEM$Ut5`ME!`ExoWp!ZQ<
zT&ajiHVB2;SIvwZSGZbhv>8qtQFQs;eTN@!>O8C#CmH%~Ppe0?7BbPt?&tZq#PT1%
zhK>pGar;`wJAbD5G(rRq+0PM($VMYEAw+UDwJNV=AV?WZQEl0_mNMa)b>v5jpOb)3
zA;M>VEUvq8)xM)dosvk7`6;mqpz@M!eYYWx;PAEq-!t=@ym(K&y%bi%D82*LnO!gP
z$0g}9B7a1RA@jTu=aq;-H^vv(F?58T)916FG{t0Rc9b6oNndp>%D%=b&}=v9%-Gbt
zHdlyjSX@*%CrAV)r?=ufX9`olVe+3_c$X0vuz>O6aKZ-M24dLhKF>%Wtv$$E_2qY9
zgJko{7JrB6wD>WhVG)=JIEIb&tz28zJ`{nck3tQGhp1oB8p$chsWgpjEbCIf
z)>FQ%_)@s&v?(Cm&)`f9zXNbL`SLyEaBco{VnG@uAo|$Y8FqUEDCyMj*XS|EWKe`_mg(h#=%mitfBkJGC$SigM{GdxaOTb@TV-W7$7{rV}qI}e>4A%{nUswKV+nSZQe0#3~$^}^(
zvN@fm#W-hyv=)X>T)DNC9Gs8KZn*L_{e6_3x}fV>+3Cr|aCh@*eCuk?K3z?%E;2$@Jg!x1C@SpS`qrYa%Z(rd4Fl?~Ib)V?cX*S2$h9
z5M9pct_aGcjthcyBqhXzN8N(DhybOpyWrV=2_wsV7lDECfga)u(T6JGP99yNCf|
zjU(PJA?m80i~jEIh|nZ#fxlSr>dc8f^QIgcb_IbaPHISdK1~|er3%U75Zk`tktvo;I
z+Eer4OxQl(C7v-RUEz|HF){i)TvV>w9c#>5owihrhJW|=nHEgZaW|{crf*kiZ`ntQ
z_J1+d69Fc1FQ$bf3i%$^KFBwC=-(JU%hmTC2SKl;cxi^$JXS{A`JI$aJ~+n=7uDM+?<}Almd2T{e{j;8g86zF`_Ftpit~h1aalZvDW2Z&4L1c
zs@iXnAQwl5wc7g2_uAlrZNkslBtQMylK}bG+3e4ag=5H~S3}F-SI=`Gr%uuWYqIWo
z*o=9~Q<`T!wf^E=@uS-NPYlzn433I
zy4M%hHsyXRJnUub`Cb0xdI^>Qv{@uNiO9T)HN(~6Mw&Hd)Yjwz6BzdWBV-gcJKf$7
z2!5!Dl%5j*oc|a+>-}i~fiFhj!r*=acxQ
ztxhFW)YsNCC7CcxF~>jxtzoKdtJiK)gLeM=GjSj<390eSHu677C)_X8fHKJ#NIo+#
z4mr(ZhBBx7u$OpZKDj`EBOsaYA(Y`})X9E~aRme6(0o(5U5$DbDu>#mv`5Z4ov%l*
zYnBm_S;U@ks5O4GFNk{}zd)qNIwWSJUxhdZk3=_^gM)buqpM1IuU(``@+ePP6J$MB`0SW
z=w89rCwsErzyB#2)#;&a?IKY4Aro&d$*7Y9NCsO0<^+Yied4xEF<4wyQurJ+!fzn+
zIu0A$^*r%CiGgz}eL5j-YwxN9N(uEF+B*maV5Z4w$D+!@Ehq-id4o;8uRdAEQFZMh
z&AcsG=(wk56fwvGDd
z&d27uKsjR@oh!_Yuv5}k0}#q(AQ99gvZcBQq8kL&bRfGZj{+2jYqq=Z-5>W5O6oPx
zmcd=;S6TOmA
zCXnZ&I;j*)rj)WB;cA&N4tdW
z-oJhOy23!!0VU!7KP8LUX~U}BMb(m&J>?8c)Fl0
z5>PvbkWgsKte@+1c1z+lAcVHJT-Q@uD*i6OV8ZtDBa!HC#OJpWm|&9l?L%ptw)2WF
z7DD(6MB?T5ZJtHh?cK0B;3wGBQNxS>ESXzg==;!YeqfJVQ3*E02*-Z2Mn9Kz)7GvQ
zYG>U7a@c>mZ^k33;9V`Sl*^|JNdh{*PCC;4H~#z
zh-&*0eLE$0)iIC*_Xy&D-u>&q%1<8Ak@7@n#R7M{C~1;2E(cNSlE9IOOXqvIIdr1(J=IhyjvA3)3rmw;ay@9n@!SV^ugL$=Fj9#%)e-2oFYXm~sQO;VEAS#s*YDT>FEjJ0P>JfnU+jsI4C
z(h;o3k?OiIiFY}r4qK%`Rig)Xtt~F7-{E)FiaM0QGD-L1*M#$1wbe^cdC*@rord)&
z=xFE$6NKzfZOrLaLy$MUFk4QW!m~gw=bCD8z(!Lhh(C**_5pq_?L9>iW8%+uYDRn#
z#r=Wd)+}h?!3O2gNG<54p3+v--dv-JL)<9>rvC=?j@-w@L>>!9Y$KAoEIfX4zlWmN
zCDdnOq||CV$b#T)RJMCdvSUr+e3i4q3t`8!6cD_^tNo>5nEGU$+kwS#e%v6e)PWWFqf`ozQ8*#_kc$mBb
zmMoh{U6xvtr0?H=UEnsd&uBPKoOfaYjIzcd{|5)SC?(90C$wU_@fIdi_P=$r^D8&F
zn8k9xli-3NbO@FtI(~vvOF_k|g)l_I3~3(BFzdxh&?xjWocDfK?9X~eGat~uC;WjQ
zH`?O5N(Q-vgTnn}h-$4S16Vm&*8DN52hQ=Y5lotcnQRNEaKXqfR`}tBL6@Qv0GlWv
zJB}P`Nlhi(|86KV5e~z_1(Gd(*9N<`Qom!0lySqWBB0m*0EgS6z(n>%S5S+dg1tdt1E{)g`p*l#`ybua%ajxj8>K5|bQm@}t@+{iIY`8sYR
z+e#a2UFGF3%posHnNu=_qEY&IZkM-}#Ld4){ro@0ruPAxF1%>aQ2W~T;xC}Cy?K|#
zaA|V~+}mm}MG((2?r9YgxbpuZGpb+VFVP=~`-~7@j3A$I%Jg%mgdHKSK4_hs@KJ7s
z+D81R25KBdh9H@zV=OlY$pxC6-kV?k)o<P?swmC|7^i-3s`r
zz02<-#=6?@z++A-{icc3K^DA*CazBOBqsi1kqc&!5M+Sq$eRK7%^njLIav*6GEew_
zIQBc-ghWV{npwjZw$uu8%ZDrGwr5N8Upk5TFY1k#&d`9-F4HM!WaCwvPDO!T?a^hI
zO}Jkd@#9{V$A8om^dZ===5$JP--~s~Y4z7X>SQEP;wWdxlEUt7!va)|WR}kDv1=IT@ed1&)gT)0e!8
z7CtyW7E^~vMrtr%cTvIWlXQE|#4m0BmtsJ*r{ZCdAQ=6-m4AgGng2r2SABsyvwtdc
z--Ooi!4!a;hQvgo7@{xm4Q=11XG1x^gZ1Fp8yn0q)?IL3AI#eA4ZEWkH?3{2s(^oe
zQ0#y4=VQov9~k-|)BD)_ODTy;3XWvy)tShK{ZW8c>_;_uY`cY(5XldSUdec8Q;`7-
zu<48PY~HK9BY<1sRyq3-DT$%U`H%5GmCtYAji)T7H*?`hmZVEwX9HdZ{|;0rND5dM
zJad2irSPAhCH+M&8Dch3bazPEJo^n2JQ8*s=7o6AYkJI5<64ZI9tr>Rpn+pb1hk`~
zt@=$)Ef~5gs%-b#i%RIQFS@x#{7}Jzl6L_>LBZH~=9u`Kjw~#C)n`)#U;w^BMTnbP
zjSM)cW|n;^XKB!|g@d~d7pAs>VGu^;SENl%p_U6nMe--2olbJgd`T6~UadWsJynhbs&lfW$;^YjE+a8Je>|}je`bj
z0t#1mJMw-kO1J$}sRZ?>ZXF~jWv(##GEe9D&
zmur}F*ao0)P*RbDT5O(y8sGMuz^MI=M9rN2$<&`QNcgTV_8{8^`|Vq1t#oQ-yZidz
zHuYqfs1EcnC)pI5wXvkZ=C1mLo=4~T`s6v?LVA~!8N8c6Saye;8q;sbtr`vbzI`(u
zPd+kVNn*%|Qn-q|S#};HK>)L2LKt0Bix^8}xk5HN;SlhoA@mK!SDv*S)NTdah&6j9GseVX>BHe_qj;{AFpwl4iS~{Pv}4pWP6ZPK>-i3{~om
z@V)44k-xY88*#5cKb?37mL3quvqjQChyO1~k}<&1Mj8!;qNKlvB|tivXxQn7MSHZ^
z2Ks)M2X9jBte&#jeQOtS{GeDgZ1i)I?~@36aLIXp}Y`RsD2)jtg*5P=Yhf
z6EnCB1yZaq%pr6mT=v6H(#m*RoH&VFCU1U>?soo#v-|1sD4>5IQ1(&%
zveSWAU)R6y7)oer^#4-I;cW`C+6@XR{$FrERI6G2ITpTgLvVp4{v&oQYGxjsqC{~HT3*Q-3p8aB+*;I)Z=#6h=0WYKY9FnV%u^{Rx;`AO7A62w0=rTe&I
z_z?te#1vZL+e=+xiO1|8#>lU9UTi7{
zk_IZ_X^l?L%7%X+)*!hS#~8&l@R%m(TQOiJ)M56*n%g}YdHF7tzf!wT1CtS&i~I_Z
zSfBG1)F`9{hv-%mX$c%Ats6G5)tF_5x6b@5xsihvOL~lMx7XCsDQq_EkDrNX~n{)(#sl
zFEd$Jz;fC_Z$6v7EXFuIi=}F%uk%KeEB|TTi@;S0IJ=I8WY=ctgw`w@h+BclD~$~+
zlQ!{m^`?s(e)A1+L;GLxIz`C(e!IhU%@B6N8Qr(#?~ZyLlyu|rl9=iKG3R^HjF8BL
z2=!_Z5`$Agzv_x+soZD`2Q*YD+#3w-)vh5*df5+#pj#8tP)(M|$5A^PnsN_!cd?27
zqRNr$WG?MJ?{v$7?BfQV3<6pyWYR6>1A+e=fOE@F-^rqMef03Tn-mnBG(u5=*r_gNskgB$X0rz;77c;Hh
zs&q;ysHn;hDJd!by7)_DXSVESQd*0pmMkxKXEFBmWX^00l6y}*M2U0_cNRxx=8WiU
z@QylzbY2}kw;!+aC;K~uuF*)g8$3I9V1uHP(z!&JK}e6|Wv+CqnHo(8<<77d6s+^u
zw~K^veSa)!H%*|yZt%HvCzPc_Z8}|R)EtKRLvnIdwJ#j&JOPgt9zp8AQ
zS#3fuYLG*op%3H86P~*lzqfRzrKl~w`{dI2=BMKu!!c~&wUn)3Z%!x+o^5oGY&J)`
zx*i9=wd?=nFLJtq6n>}F{If7E&HTM1sgcN9*4jdJks~PRRlFc=^sKw_j7aD#wHcJZ
zO$E0ZINPAIJZ3_ckO@ewBOxbAaVRo=8e)dsYrx|+~sb;s=T|UlX
zB3z#ab(pgS^x3h_j2W?OQIllMozBkqK5dRAcIbFZ_bJ$)0xaqG6fgHJ=Dx0c^<$T?
zrk>dBfP?s^o#>zA6c)6;?GjznzuL8NXsF*5-SIOqaDdcexIMN6GviPlm4O=;hyYSB~b|O{eGeH%~k=>!D^zQn&@H
z6#_M~p9@Ml4q-1(lQHA)df{_oZj8A0W@z9IJ(gtn{rXcq-Xz)CTUgvz4>I@LP6J)b
z(?n|cdH7jFq(XADvoAEV%btAa7CkR>=gpKI(ED6B`?o5i*wDPR)pJK&nxqN|0)W#t
zY`#9{`X367d%CY_VivefTA2w;I9JgZFZiGYLgcRKJp;#U2*PUfS?uEPwYENdY_Yk?x9^ZUh
zqHi&?z3v$V;?#SDphIGxts13t)R3x#X-R!?wGXy4wcOwv18Q)1Kkoyy06gDtxKF+=Og|s3=
z=sIeCww6peA~%P5driG(l5go_dFAqo78lH3<4|oPMeq%7J832d%g$X4aDAt1E#f!77&nHeb{dR$i?^?M(L(u1BTdmp6OGmS5
z3gv9;VhliT7n9c?KRTa1IjaY=wX~(sr?A@p)ph-w_3V4DY)=qb>v1DAOVR71YjVGt
zo(3Yk{cS=4m;nkZ7YZ~nmYjU5l8^C!s>|{IK531!lCYe}n3g{*^n9POXRe9y?DMcq
zQ&xc}BdzTlsOq}CFVvyj|8|0uSVKFxjE5jeD4F7%KXrtL(8>sY}uDaNE=3mIFstsF6;
zOLv~1qX;Ns&KjCQO@Hu;)zLTyeVIrJ=5EP!HcIt@-4)ZQ!Chy4C`uAE?6H)5Rqks1
zc+qfY4K!(E{E`h?5%h)!YCCA@ulijF*x^+k}tixBivOt0FY~JB(A05&7PK4inVlCj?
zsx;3GOxD(4!D^FoTgK6)99
zW==E8f;^|DJYZFQOFT?O7FSs4Ni*dw*l_6kcy^{(ucbE|5U
z)d}&l@o9KGG>V25kD9C$mn9;>{v>&AS_~+=b@N4v8v#)dy4n#~uDcD;@PW9Dc<+4Z
z^#A8}UZ$0gm)$Rd*h-MW>QHU8jR&Q-)xT_fJCoB@H|G3Luyf6FWvW%w37udObL{PH
zeJN}05(ZnU!RWgw`L>_lb!_%BUibKtd(`l)?SFM#(Kbv>`g&^H^8N}aWIWY3I@5ev
zp5mf(J8dx=$v4i2<}~w=I%7u;(fL3{
zQerjudce}aWD^aNC@bvZZ`pvK&lwb;nw^n5nIBGfl1LVbCLaQ7_YZ6cog
z`IM*_&*qYX7rmc;&hly+OGEwel2bQa+xInUj>F`0zgKoA+mJizq94Y0pMME9dyHmL
zHe}f@A;@L=>T;YHn1=D3b`V_6eq_kM0A6ESmfz^fM_Z^&PM$CB6*k~$>yuxkxN2>Z
z&w2$$q7GH|{ncw)kD1j|YD>?09EzklF|Uc;-tX^{t^a7<;VZsJ)NshOXFI>}5unrw
z=X-a9$Kzwwkv4E}@odn5mVugcvCOaAhpeivMBdv?m;f3n()!N+^mggBYJ#Hjrt
zYe2}Sp#CqDP!(yRE#e_oKIn-Z(B;<7{r5*hJ`0D{X3gsRksM7MaD92XN|ZSCA%^;w)+sW$*@8kr!K>vA#NA+}U|WKEvo_ZRcYn?8m_+L=9)+i^
z#i%t3<2;h>n#JpwK2)nkR6RWRHjlC9!Q3x0qbrF1@)tW7)y5nL*Qi-f{g*3e+oskp
z8#(LKSsfK8_k^YA59Uqcf@!L2?b`8kJIY;&@|Sy$jIHG_>8WkZjAc0`3~3mvu_MkS
z#)6Fp^J>wm8TSrH7>)W*hwN`wUh?J^E-Cb#&P?X!KX`AgCdr-d*X)&acpRj4X)12U
zYF+=*+E(u^<+-2B(!bybL=k%P(Cy}5-Pj}~)O_N9A35B0J
zm)(l2?VzCnoE?6xd8+9s*+VLHqD#}wzgR#jL@`BAqa&)f7`1Nov#MWuEY=gEbIMvB
zZ}r3MUcbRdcWVKhl%qNs?H!y&JgfdvTU$uQ9xM9MoGV2;nv~?GU$wI!FRwNa?91NU
zB^c&($~fS9dK$_}t&p}H!aU`5@GNUTV*Yh9nFNz@n1*qNDr}c|S@n~E~x6x*zXJj?8WVCXG1Gz%X4=|^9cQICJ(K4pc)-nB5i^)ezZ}wZ_OJMe8
z!Hn5hv_*v>)%0y^;@TMTm3o$b-{LJG6HEgstz+};Gh3Ppb*i}qQ{{S??4Zb|uRK8c
zzE{n)N-0{Ct_oFw=efd<0uP0)A3V+WT}XI#7NSYEgOI+^wV;?ZXvqSuQbCJtKug7^
zJtyf7x=rZnsUSj@q(7wHHl&wE3z>yIc8bF8=EXar#d8EBsrU@!Bt==|v<6H4MvAwJ
z!6ZUSZvT6QkSxUJIe>c~YFTxx7P#;QE@5UT3b(AA-f%vD@f4VPz;cQhGOtW2&^_+X
zH_?c0_n|`qIq0psQ00d3HMD)b0_Rc)VAj3W=kOpBJE(O|Rju;0@$Q{<46zMwexsNb
zdnBOS!$*kQce#mP%?3$`kvzH2Mxehx-s1b5*rCducQGnlkVfrUgwAt
z+cWel9x&1gQ|~yNXmzoqSbfo%9?ztN`u#1MDf$S7G%l{=w^ta|FXj_M_WUM=fdbNST!T~f!6
z^8I>2m**|%sXeM6KbVW#-&3(fY^50U!Q!i$|7t)hf3e*o?5*4vxLZr4zJ&hMxwQL)
z4O?6oyZqu=*j91!z2eed_x9Yx@L_{Ax#MbKC`%}I&DJhDgN5Jj&S*^y4Fmr)#^j-P
z=4maO57te?G|9I)-VltZTi#VkUFxNSbFs%aF1)0yf$=)0g}}PdxzRQ@3H&k8g_ctzx`OzIWVp}rb%SxfOp5e-{&z$w21QyFQ4}Lb{3)^y@
z@3uU=Z`QTG?hmkiCsY_6Chz;QNM0e7ebcm>Jeu##w7_#q*i~t}El-p#l5#yrFw~g&
z9FOWW7eK{LzlzH{{T7vmmtp!qrBQ(CXk4O0r^R%Iua=LoGGUq7*>W`$NOW2MB;k60
z=|5v;Z&{h$vmuD9@rB*oAIFz_<>auY#V_$lqr@4lb!HMo8fc;Tl2)*mIwibJ)bAEP
zy3oMV$|QC$KO9gb@my!)>F{EGq60v6I_Sh;IwEfMzt2&KcEr!#Y;LSbYhB;eX3APQ
z`tTNVo~meOWs;nXk;huMJv%)T5&HQvl(^gI5qC9c8)YPK4t7#Y7FLyVEo_s@Yd55B
zT@s^3AX5D0kGojvAC8lBVUfkNRnf1V(^t?E@`P7BAD&hj7IJAO?7hs|u(yJoTrD%&
zCv3Qu`X7VhmO;g*eks+~-+L3^++2B;j{RZ87M&Ky~x~QT5gVQ9WVXFp45dD$=FW4bm;$-7UFvcdC?<0@5rX
zT?^6;(h|!83rlw_EFDWN3!na;_xqmjoqy+?GiUB|pXG->f
zH3WlAY+U@h?XcM!e<#phi?D)Z|)5u@AT$PMb^I&z+uKP-g{A7eatUp5(YrVHlrVdej+
zOxI8<4*0RUjenF7D2FtWB%yA=_vKg^S$Gi?EFHAoFwqW=pb~>ooov~xZ(VcGu7Zgp
z{air^&q!C#g(4d+@-mpc$1-Ln=ho!c$rXjAnm+<<+M!}-4>g@)!#Ay=gCW*OhMrhm
zRDFPF`!z=`Z@H7$A&R41ZCf7PKH~=ELPkf68rt$HD?UArn{P*y16gIq5XTj+$y&*qX$Ay1Mcqjh@F6V3x5qq#E+6d2-FqUqJ9<
zxkK#0$#9PWwud?YHthOj+T(@@`XLm^J#+6r+<6mTS{=96BB{M!vAi62YD7sCVgfWa
z+M;12OzrxK((YY+sB;z`4GLLWZW>le*=XZx^i*?CEIz9^I|TA~Gy?D#P7@`(`cSj2
z6#TzPZis9q8)u!fz=+$VwM*j3BT>jOLAqYxBHt09y0xpfs%`M7GDBE~{iKU}kE$u1
zwVX8zNy9>MU-~c1O*(h+B}1%q`SxQF*)Q=vtxMn?U(FAkO)dLWcO;PCCk8Qsw*0cA
zYdxjUp80X!?_@pc8ba--L4rTZ-+g&rt5RxrZw;EZ4%G<-78m}0GikO@`fw*%NpKHA
zgeO)nUG1Kk5pbZlyCU;U^U&G!3DDB1+gys=MV-w_)18XO*ya^62WxFbzD{eJP(r|R
z(X=33ArMccyA~nlXB8#c^*su?R403N`LG|LlW+38*;ds%q=B0$XqF@M
zZKpHiF_L^&;OwW*3c!8UF`9-G4a$JsG6Flf6G+9smRg2|IO`Oe0|}G50?#;vxY0K*
z8&h5HTZ6*4Y-$egWYI^H@&*QP={B2Z?|t5tF4nyyN(nu`ZiD0+dAk8U1oyq+&TqM($wemG<^WwHER1
z{YGARkk?(XP_99vY-c+eZNu64@i8v&BTMARbsWJ}rO5q}GtBfhBL(jYe>Nh3mbFm<
z0~h;)$&*Eg#UXG~#6zjli=msbDmGk}?;#C(5m
zycVeYhadk+D{L-=XZv(mtHBqZRqef@T>7%t=FC6m*INDnKLf;hoMo4@xtiE?-2bzh
zi6Hb+M|85KvoEDfq16?h_!iY6^kc7HM=+?b&@Q+)8Lw||67Mhj9D^3z
z)D6_KX+PdaA5!1u0m2c*6iuJ{3h`D%5NtC86yR&_^TWW)DdO{;5x$Nlqg??vp%b~h
zTUr%1iCkmut1}N$V*x?)20Y7_vHOtd+9mlDi>a3y?TR&b2itszqW!oo?;MkZ8
z^t9q}X`C%(Kya*&WjPwR?VJA~=;!^}4kddn;r&eONYm}ue{SV100#T8m=L=OKfLF|FHLDEm+H5?~
zG#^X<(X-EttacMDv{Bw8V3K$n!Jd8gi$Ic4)HG<>^%#){nbo5Fxw_@v-ZfEd(cyOx
zfeu`6c_tULY5n-!UHSDdj%X|t{+we?$8C2(VEaqwq~CwrDt3U=tDD#XZ!UE!`s4>{
z
z#fTCY017a%|GuzsBX`97$c%)qSFUX>uRt?}hX3c;G4Gypjv`yOq|Iyw33BUc%KUI=MhN
zOl=q}W2-a?%sWJ#_)bGvI`Aa{{?G$?Fej~|7Bj|y`pdMaK9r%
zSx+G%Tb6YYcd-a7ofW0`V_v)pN(}aIQm#JE;IQ*mBtFo96*qP
z1JL+bvlvgV@`^;Tk*!r;tDNt0eQL^YNJ6*5WTQU?5N{?_;3r_c8nr`0t_Niesf0S<
zn@Ly;#`hUCG92IeEDr4@Qc`e%eZzJC9UO
zYhum|`B~9gTuA&9QOhe3P2vnvA
z@*C~wIUcrWzz|^CbTg#B_~R^XVz7bsLB@#jN2p2^Xw#(Wu+uF_N-7%X!`YW5u36U`
zZ3O+dt@chgYBnw{(Y^jo7N-5t#&!vPq|as#Pc<*6{4_p>DGmq4p-S|)BQnYWz=YK%
zDpq*jf3G`DfSUzz%XnXr=C`8|6qxx%U?rUv(IddmH7z)*Czrf)Hy)U`(Gu@h*x1yx
zvzRA}lBYEn9t=LNy*}8}W16~l+`{QlfOC>MTVYTOx&IP<>dNFhi
z;OQTqsRX6@QWI?Vr2+zL6WT`ON&Z;+vdgq*(aUl<>C!mZAFjU;x<7b~I8QnNO6J|h
z4Uu%TU8rbteYe9t}F>|>VOX9~p-uZE{8sdb62htcOYKks@i*}dc4=dJB)sMikTe5)xlS}y
zRiWpj??ZLfq^zlL12bm*tai0tuT;Mj@QM~96o#LtV*Ropjc?~FAFJ(jzG7LZif;Gm
zr=Q#DRcz5Qm5k2_a=yvf)uzTmL4N-bT#MR|rwS^s$Zwbq*sjhGjEt$nSqQmZ&w%Eyn?C0^Wo@IYA59wp0
zEu|Se-u&Zf70F`Kj9Jdale<;Nu6q!(kv6j_p
zsX8$^g*eyqX_k$=c3+OtUwsRPN&m<9J
zy!mT}b#OR1i(<>GK(+2zd7$%ol1zoOM2G7acTR<0{!{(W<>nvQoyi%Ihaz-hHZk9o
zzOE}9?IqkMtharhb3LSJ6f&W#W!ebmI93J)ch$6pkv(LbGWF8c$IuHmdbTF)><=eC
zM?NQ-3-lKbd&fn<<`A&}yw`Cs>nd$OKHEzy6u`ic`kS?r2*$t6V@=Pg{?4B8;$>at
zETU-Xd3#!?V``ngH6Plv&3&G|%dDnu<7C%1xq$;*Qf?OI|3fD46@tjYIL>``XB9z@
z4{af|;^T9~jc_g#vD(H^sYAEkI=Z-LduwyfwZ<>sW)X9CE`GbMEnHINtu9?L8&3s7g_%Ck=M
zlZjMr$){Glq};}qBf+D
z>(&QU^@!NRkJ>kr3=L@K$;!u4@|Lqnfj`UO5g01w`36t8J{SxqQxfrRsxLk?0~gb=TXtTsqST`n`I>sC%0p
zo6ML)^oI+2&u>$m!~q-JfDw4eP_AtY24s<@?>pKjH|Km`IBKnJJ*pQIR<$Z~wpTQ0
zh#XH^S%C&eN~n@b$HjAi9rS#5PSxcCgBXL(#OJ)DdL}BM`T~v9j)zPzi;9#@tX0Lt
z0FJC5F_Jj14RH!%L<_eJ$6ZC@Pd|(*U7ShEk(D3JXU|>@0haN
z8%IE4q;xlcF$R-8@y`PNS9JQT5MRrlAp;w9QcUP#M6DpIDeQx*ylh!P`B~@mi|23a
zt*p}nMWDCrwpk4}OhaYdpP_+jtQUlFBoNnL;gZzvuIW
zQNm{i0zNv+j-ieiAZ9;IZ7bx4iCjs9yH$S^->O*|9B5|-Z5JSI9yMZh+A0IJT$-n+
z{nPyZbnkGtjC*1eDs+_0beMmfO>gA3U;gnypMaCne$qkTt)_u29z@Y8U8v{i{i0Fd
z@^
zV+rzxd>U=cVO0z}24~4IE0^ePKc82T2<)Hj?_02}l*_oPtGS?)8|Y417@g7w|J@&7
z#&k!+4w#r0-kF!{)}dmG27l+!(1`Rc>NmarrYkoscRSg>UD3Q>KV6q;<~BeZl(=?mnqiH7k^TH<+TA
z)H3It9@&92b!AtUW=qTZRR=W)CdO>P_9@QOZ`}VpgglD)fN0ix%MrF|nu;`9YJ1ow
zUd`YRE$?PP;NXxxBh5B>dBu?WNrqUq{YrPo_dF$;1GM+*`Jx#&qz@TU4tCc!^3Y!t
zGtJ9M$@%yqml%)O&CGeX?~x4PjkiHT%511_iO?vOn5_ZY)>pc!*gnU`BCw+Wbz*DK
zfH5X=2hmH)l?j+N{+{N+61cSab0@tCabENK&P_Il1c4{dDz9fuH*@*9a^Xn5%BN^$
zB0Wf|nv>HRD}L0*-O!seW!#phMO7qvc&!P&c$wbU5WH4LRswOM^=HQl-w`?NTfXl3
zECnsV)2Dp$%4neDnqjm=>~FG!u03}BjrAWC{q6z3?6Js!toRjMDe__0vqk(_1T-~6
zzwve3=`X6;2Wr){uj=MD?$wvvFMFjPx;}aEU}01J{=+^LJN>Bkclv##>};v7vGj61
zZ`m0!zT1$!i^IbF)D1ybWo0cIWSWj2-x5EFUqj;>>z`d|Az|idAhw-8pRDmJgo6}Y
ztX2M79ARPJCu?BdrDAAj2+Sc;Z)$>BYf-`Y*C!WehsK!|Lc>E=9Xd<+XQy_?{xpY$
zd%gWzf6P?5(#&w}1Qz$k!CKhw8L=r&cm8M1R3yvAl1(n3KHI&yc)yJ|d1paj=c<97
zOR=5D9hkQK)sN)SUwp!_2Ykr`d=f3KkW(TXv(&cPsn{=72e;paegg?EC~Ca*K{f6w
z%f+V{syBDG@Ob1z-{if~;7s~+n-()smD1WK`Ge}Z?Qq9kqs3cr&`8(T{^r2jsBqH&
z%0Lsk=}X1|X6{*+ALaOU`5jpJks(4f~+87$&@eaHXj
z%aK@f&8=I;(2$961=q`)pKPl=a>-z|DN0K-qDgTC7K)D
zZ$PTQa?J9k&)CcO{^HNxoqs~@@2(;JgzWtGWZa`Dxc1V7Dtdn~UzCr)F6)b)etXcb
zIA7>5a=wFJ3}9=(omu+G_z&2&n*?$`oP{f2`hQXqa~^YPulfb2@h)usnQcjf#{JQl
ztNw=CZ(=GrDTJgHrYRPV2@kk1gI+RSVi2(3D=<8Za1@L$vv_vNL98_$Y_@xvj?Efj
zGY}O$6cw#P>F8^lPOamO^vrT)o45DY$DLIl^!w9l%xFMF06lbCxI2Bzq>*ZzH)AIu
zrsdoTaw&71G(Df8X5TA}-Fm>$AG%^7qpDEg<
zaB-$*K~JzVRpZSIb@fw~AZaWUK_vg;#V3h7&9*5uGt$MOyfcW#R)-iD=h7XWvL7vk
zvG~w+0x_)z+%D4+ApiPjhH4m5!#K0un-j(?qd#;Tdk{J-JFPWzc8
zbFOXTj)fj6@V&Tf{5&dh*JQ+HF3DK6DY`z*@w6SkW_gwSmGr7GFKkspUe(E@Ebieb
zm`5f0{N;xU91I~HvV-@3G^We@(BM+4IL${XWmFn#nK{f9QRsqw!B&5iW$lud=@=al~jtZREBPD`|SO?9+Uu4cr_kpi@EBGMjNP8{*!rF$CS5=$a>Tt@zOKU%NN+PyK6)Mq{Fzuz
z)!k`oRC{!N16SWMOJ01pGnesEGWzi%mrlFwTXjZvb8_{{0yLkFbzyJ0tz($r>>HP|
z+pGIH84wVTff*nS&mx#y>2Q54`+0U*1%59yG%CdRVk!J{lhw|(+fmLP7?=W+Z57V`
z;I{7y+~x1HYrO^JS7kLujqU&pOofusqP~M&M&9U@%a=6;Hk#@)tCtlR7*R_~HxuoX`wg;>O=i`i?Ba`|t1))`rviD26O2#JLcxBX
z4k9em*)J@dD6--QC5k=uiw5oXhK8UtIhlNYcaQ0zq91{(4!hh_bFsveDS*JvWp)Cw
zS@U6k77Owb_>ETvuJmhGJ+0U}t<_Og%Qb-`r{3z!0_7F$8AomxboG&#;Lq&CG5X-C
zqIA#8Zk<9}{Kk=tY-Sq3t=iX;nXn9tC6CN~6TE(KFpDCzfV$YE;}cz5(4fvfL1*XT
zK~(MMZ(U#B95{CNndiPzG~;y?X*nK#&@&cMCB%|AUx<(8Kufr9l)4Xht%o@2PL@r*
zRH;ewj3kME{?ARhf5BPRb48ISI3~MFX3hr0O@<}IRY?^?Cu5Ti9}MRj*fL8OfwjCH
zY0tc>9+)+a5T8~|D
zW28dX76W*$itt<{Dd=q4n!G8kOZp)na-2ox+FBsn?V&#*+WiARHqfQF`rQaP|BFzjPrl8+N<*cL*$uNsD9YJotM7C7&7!M>&w
zP8r^uj@Jo~hJ)Yo0C^6=`?KN~T^D*{H=(JGYIF##XaDt-9x`IK*RQuPRiBf2I-le4
zbK^#;sXFnsk6vl{?392CCu?|V#-@(JiVRC_9Ivph!U&CY9{=!5-3Jb^Uh2@;b5wE+
zQdL>|u}-#`=9#r6wa>=p&QY__Cj|9;SyQDdf;5lw#0+7Mbc+4lQnw+tn~E1R7>W}F
zjwheFdGC@e++g=2O-;qvjwMk_AucdnsuU{Nz*m~Ce{UYmn>m>v1$bqJ6PR7Wrhori
z;%v?EoNnD&145>qC5oHYSBfJt1oEXl#C;-jUYDx
ziM3BV>tcNC-GhA=6kB)}m?6&n_?Z7PUHJ3xH>|#QVb-D#3vW!qs$oZAhb3bbMJ$if
z@(fDWx{P$vD}N;GVH>t?1}h6L4!>oU0vXPFwEWKZ8Ue
zb!QdwCbkvz#S4a#=Xi#&FZ+xTO~+O$65y^Ml{&h!Q3iaqRS(rFdB8Ux%X=SY>ffF~
zXKJ<=F=6JwE)qU!_bdoC=2UeH-%F-&J}-2o6%=`J*0%_ell09As06fmia;0#SN+xv
z^E3(YOI-jHY}qDU>B^TjvwqZumjbDpYaX-LpM>0KqSYH5eL8i0McO_6ItOn#!d&yS
zl?T54Y7#DP$4QH-#9nT|XoJbffz@EauC#PplSn)yo+L4g^G#5|GV#$k{cSOEIkheLI(id#y&c2^pUfGisqw?aFlWX4rkw|RFQOjZHquj98Z;m%r
z9HXgMZHm%0aysZca#`rK|GNK-Eu5$}1s*l%X|HNKIv`^UKj>Ohz~WN9z#~2F@K+U?
z+}}Ezx~ofxt{4{vikb*
zu@d5Q3W#e8+>I7~^@==rHE!bhimSJ@cl^{r!=1(0rpt3`^N;X6vLgyJzbgRVvjwXOZ8YRcR!zv=H;znP5mjS^0kkvb=dIcmllV-xIZMXMBF~=%R$is1S*mzO>FTkZem&0)Cu2RcQ@^=*S>WSNRNkf}U<+%y^
zQhFb=GvbFWx+ech^mdv>g9FLyHm1_jy>uI0|0`g)quKS6N3D=LyK&D{ZSm2YqiUnD
zOmf&MTIu<>+Sv5U7{TMvT?qcHnRbapJKwKE=AW0dy4wof0&a|rq;gL+w9}3kSeRIN
z&G|HBE>cDWB=nSL`t%??hb_0oi!^xfg~L{TIJFK2hGe^vtd#EO=QKNsp;j%lx}
zo-dx~%N&FMqAbDTY+zY+btU2euMR8RxRA3l)mcp~=*}>-WJT#Lx$X{-umkv$)40rE
zb{>oK(8hJ%u*0h3&@JM#=dM=2ux7Jyz}5=MVRmsb;11ZPbV#l9h3*?Oz{DD{FdO4o
z+arMSrBV9(S0bQ22YI^9$vvJ&X>@V-{jKMT$D-$dR-lOo;x}tYDU}38E*!2!f$v{k
z`$yp?CD>YiJvpPac3H}=aM_(}66WD?7>$`rsAcncY1cc=62^|#eXFYh%9(w;(YSUK
z4JI|bjG^I7i$Z+TFEkLLgy2*AoeV4CzvLnY3~4NRcBXHZ)pmpIX-`GaJyWBV1
zn3(xRWvZU5a6OGlNFqVp(8Hf5v-^zqsBOiKP_)nL9bo+xB8k%HaQupk>bS24C~K_d
zvIZ>xhzx)3;vOENV0JXtbD={YTb!MWF`Y8`ykvGBn2IObJA_wHnz(A!-=!}qbTF#D
zFk8Hs=98_|nMbTwz{y(fR{d&xzxsn77tjV;pNSP(11%&`LqmVB)yM7r5%6Hl?wuj=
z{-`i)%!&3?JNQd_NBxD51ywK>_0h%S-ZQl}4+_X8eU7?RXZ9H58N4vCxlg0J<5rjyI^RwtsZs+?&@n
z2ED_QSFfUE8LaOtbb3}liQvPRvx>P70M-}y`E3-Rhu6=cpnEaX27r8UI)zOmQvYK55>gmk2;AQA(5@Ag|
z9+l9ornl~Ovv@7JVk*|=z5|8sbOztP*)ab%SWOpco<7SwN;#(+h1f|$m|w&@GdsBM
zRdH?W!!2D(H_OF}DNIfdxQfONru1TbxhxsG7um*dnKW
zm#=QWT|BR0+ps8PU&9eoG;X#<96)fF0q8u`Q@$*f3`8wGD`3~MnBJ=W#smg^MJ10w
z`r<_fu)x$9rc(`~m|X}K=XI3z8cjU@V(T0T<>>e$~c^B{oKL@ww#@Avyv
zEVv*y-7;*0(6gn|RXf^xwZQFCTF*SR^Q6_N$#ZFVU=fYe0XzQB$!`^F*%q|o(5Cit
z^iqfo$KW{Ixg*?uellx3xNUQLbjO!)h^fZtw0c|1VM9=O>3EFTEB>Y?BS*FvG=e&q
z*@W$@!@otZ(m_^Q#6C%7W$l-5EPE}BAxarVkpp42x%^?{HKI*|sCyqj(EBT;J)%x_
z#PsK2Dh#ZbS%R3qzl&-8=B|2H!R6GOB{*KW`Q+#4_2z3s%3p{EpV_8{LYFIw$Km2N
z-`BEJ`rvBLl*xzh%%zP557?daU}0laNyzEq#Pbr?EzVcKYv8}07$*OUoiGYR_D@B8^pf1j8H
zI2E9jJ?J
zHX#+eSO(&smKuQ-q8&k?HM-r$r^lX&>ggOE@{51>DXDac6Gp~FOpXWOgIXC`cBKD!YqOf%=amztm|e^oB$@$l#1
z>_;D&bCdf{N|Npjw;mLQkS|D^nLR%KsDECm>?Kj5S=cLoj<0W_?OCVp)TJ@<(d$b{
zG~T^0ga(v-e{_*Sz1o7cA
z8G2#GxsXVE-X>x*p36oF6yMNJAd`h2zk?Dt_^svl#Ac*PK&XpPb0ifL{!hm+?J1l%
zl8w3S)U+0+bC0I*e{S~d%gaq&-#{Ph6U^GrJ+9*fd@&?@pGPCR`QL0%AF_IS-3Lt8
z?(>)I7NSbjnwws!cDSVYmTk9MXRdj
zMm*cbz8#MnmY1^@?82Gmo}QbN1Tc0}_{PQ;86$<}rRxZe4HuF7g|t~mvcI-`$if&E
zS_h4WPThBS)|Asdm3*j$#fn#)MImxYcNAS=!Flv5TO`f9vzzUz21RL%T6lhy8m9K?
zlym_%anX#?QP0UPIIExf=nD@XUB%-=O~F(S%i^r1IrJfEe=pHOZ^hDR3|w+wu`6LMAaU%+Q(2Ph&e(sw7{khT
z?>Fzh06xmR86#Fwmr~ogS21y=3S+{zJ8WH*9LjSnwde{Rt>kybbr2g$=L(ej8oFqdn(Tf
zB^|yUKG`a_m^lvPlbM$+o$gd;ChM!GRM=EkB-<1+8comU&5oM!9kx9+00&+|Lu&3q
z*HJO$_yK@iSiW{0xx<@^-lO1by$?Csrb
zRwr;pY|E2Q?D9P|#h~?FNqm76b&DcUWt*wRR}jTtL#+Lv!jQkAadTo(=l)&q!jsBrt8ax2oe$%lQ6^GQ~r}?{`>p&Q^zW&lYn9
zM?;rcySV3__
z#r$DbqN{~vd=Pp^qqr@93aGLQTTG@JSG6GFgI**7`7WctLpMaA?9Qr-Pw4%NjiF`P
zNP#k*%jg}nH}M)MytWcdjt+KjZb6&$9^n49{tYeeSx
z{mTl?s9gvzALFL*o<3qrEpv_p_Jsy~Ijeioqg3+~o2fAHaU&J4Ig(s#2)2eVAms+<6bRt|$OU4w<0V
z84%xYDw{gnHT>#xJ2QxIR^u4e#CZNySE4DJjc6I1fhwoNaA{rwk15y-Yn^ruWYK
z4u7nA!@IN-|MDx4(+AMFUeqR1EwRn;4LQ94-v68`coekPG@b6M60)9_!o@CjJDW`)
z&L4Ij(_dN`AY*Ap8FYcgH1wfYG;qB?Rld;jZ>TWlk6G_*My^g6*DBW-ma
z!O)NhI(mr08y|YU0PdA!2?VbmFWn!-Yt+?jn@lLs@ws7r`stVTX$Opf8Av1XBQBd4
zevmWX4V&#e*Qs58K!hE!<#+%hwl4<2sJ&>{0NEAO87p^E&(7ronnyD#s}GF_|Hf;r
zjfhGZLIgQ}wiz2wRuTLA%iqi$)ezf~q|LMHWSSS}at{V$!VVKqB;fS&fX9L&llRvH
z)rpir9VBsPiFN&nludc<_w)M&3(JlrtyACFfb!TW1DhUwGU09@PE)jt-8lD>q8m;3
zowkvqgV{Z#R3hxn_Kjr_`~K=3r;AwXDP$g&U5$JC3C0-b8<%7Gs+256zdq0
z4+s=)U4ac0x8#~uQ}zoRlh&W4PoHlBH0os4?vyGGsI}M-|MbZVZOLBN^Dv}_!qvEO
zU51=m_u1Ei){Z)@io(|CHxd%TKW+J26OsJMuc#8SULO5B@cy@ET!$VWaTsd1|ItuV
zd}O8ub=kyxyyf}z`n`1YoF1Lo)!rngL8>xWeIt?GH44S3@N8b8qix}Ch+y;p=W-w~
z(svow^5W@^f$l!v_=cf6)69lt!H>HCwfbR?&+^S`rJxs^
z*bjpv?gC*Ki)ga;1HJFQ_Yudi|Qs^eOU4T{oPpa*E$Ae
z)b-^80B+xBD0M*~IfmlZd(V$B0|0@24t=-#(ZQD*NM5B6>_l%hD>bd_Ae_#G*dS9G
zeK%MbRIR8N$V6lz7UbI_m`=G!T*Y}jxr&p$pjMM)z|28Z=7Z%(p6dHtlv0CVU)T7Y
z&xg&0KH}y+BxN9Nv11{8@je6DGro6>7C?8pq&&19MkNviXEXln4YD-+Fhyv%88W9D
z^7Qj)iI^((&FNN{jWWxnN&Pv1CctX3l)Mos!jxTG(R6#4wn`i5xv!FGJY{%xWO0w`
z{pM%_*cJiS8Jv%dT^GWRRu5E6fJlm&JLsio^VW`^Jq`z?`)&`KEzwvd+}X7Jq)EX`
zGELp6gZ)W73!QrD%!Pt*JmbG3r)0I6T<6B)#`zOIYlWdO9iM0GxjYAmGl6^cu
zrp!=B+ii-Hf%!!5;=ltJ+m$0@6Ss5mlGJ><-63dP#1F-S=PSa2w>p1Q3O6IF}e}Cf8PseP-F;0mj
zS06=ey%0A$T@bki)_sS7iYp)XL~WDegRRub5m<$5jh2tFS~sLNJ2F+EG}lr65GhhI
zz=iz&50o)53ctPVqpQGs=tfac?eqVHdl(oW{)i^K
z%mwU4qjXOpINOI!A|}0s!twdP
z=0A`EUX)>CeB#W3jCHZ`E{5;tvXgvWCO`6KOi|~nfQ`t@={v|Y`4(}w{Rr#5yDmm?
zx!hdtqQbh-UW~v<7Yq!-i^sp+69TlJV$AjAKcm>@KzSL<5BIJlH?UT21Q$2>k7gqM
zEUsR|Brq^$9
z2HBVze(D7I(tn38%9TW*3SNEWR-@*9lf5L)o44Nn?SLLG$E`niQ<76%1J;&5K8QEtZiBwD&hIamh|J&nP#w@x5p
zD(&ZlQ8xUg9MHW+85ax6O;G`Jv8N`FqI$?T26P6m69$Co|A|Hgf?HgLf-T8%dKzlu
zMIbBOpTYU-rC3;ZFBrow5AcYvyh&6TCAa9XtA3C?_=av0R((#L6(26D|Hbh8gY}dV
zTTRd_l6Z%Y3$
zQ7K`Jyv05ER>dYh#4dut5b_RfuT?9XTedYP&JHbn4(-U_$sTRQpS*gbHPxT(VVa`~GAO;=yK+H2s
zbLnc*xAZlvbTuN@@)z;m6xND*a^IQRHD3qTZv9xv?~8hbPEZ0!dh|KzysK8w`=Gpeqn&%ilnIji9X*5Tal+T?nGjV+%>4(`u4->W+Umk
z=#l5ZO;$;rye+zaf?Nh9{YBPNX}#XjHZ{3K*_mb{@s)ImSjhQe?dNZ0F|Ms-@h$y0
z7&7XSs`qC-y7B6X6`vmU7D4ZY-ITw7imv!m@q=cJQ-h;|C4I!CQKW;3GB8X;Nk~=x
z`!he~memf!)89zbnWx{X&JGVN`9)^vydKj2LX$lVO00A_-`kfy3Yp>4EAwIflH$IL
z?AqzIYq@#xc(Z>vg>EPAG+r+CKY{H8C$ri6$M?sR!p3~{(UsYd6~
z(BTL7`8$L3V3xJ7nr|^aNS)hZkQID+((&ezU}uH88(+RaK5eaiLutu?1~KX{FoakBgFV82kDcy3_nl}05d3s%U0{|UHI=1272YwIg5bPhm=l`Ijonl
zC$ICBG30+WY;5imbb^cOAAC`Ot}h>SvHvN!G5+n3Z2Om4KweW}pC%Z$U(*>FP_RD1
zFxC23hrw;do_F8!6|Q5P-Ytq|n#TJ7Zi@U5VgpD?XFJIy)2Fw8{*U1AZ;4~px<5VN
zOsBxT#Z!Pecsu@W>Oa)}d)Ibak>2NmU&cn;uxVr&-hYb({vCb$;=fs^r&(;9hT})0
z+wi|u{D+DE&OQF|f1{y%;JE6iT9kmE4wV8@%+P%
z{xsHa`^_(-X8%=fm~l#8?R6Rah4XYb6ustaf$OJb_RL+T9$op2Dy%hnPI`uP6Q~Jr
zq#WRzzeG~#=EfJsrvXtvmH+7w{+2ms3FQ3b0r0O?(p}LJ>ejdm!3=dH;-q22HC=If
z7-;4ZN!Ol;_TZK`%
zpY|?hQ!hV`7&8Ka#_r6Z>wXMyamTuHs1M`2_qgG_TLyqM!Pl0@N4U9r|9ReIIl-r%oidKlKaIUej>jT@jdi=3^j@UsC?mJ({h_(
zI^Lu-9mi9v{URS!U42&TU?3II4S(Ums%Iry_y0y*3He)jHN)y1XUbC1)&#VcUjG3v@9JZQKl%RG84V$K9$v5IIh2w6nQhL@G1v`CVNT_1F{RfN
z7L)0`iNb)nE*9yv0RQS&$;!Jb@z1hB_T#dI
zKMj``=3tX(l4Wa>-8lM8%joGR7&EL0?MZxAG?=~+Bz@ji{W){P|C5W?#PJi~zh*wt
z&`V=2AZ)yDAewsJ1D++j>Aq25478}nexuUb2)()kSJ
zI4`WWB#&GUXcgA6_10C*9uu6;oO@3!mRMQ6tZAeAI9S4|`BHHdmSpcYWaqQ~FiG&Nk$`
z&z}zJeG=969i24IFUV>&<3h0kgB_d<+@F&0?&}NZVCD_G0cWP55cEFHgaS|9!C$t1
z@~{}mg^Z=u9W3hZR(({8n!3|qk9+l5Z+l>hb9|A#-mb@+Ji9e}`MRmHpK)`^K7nY7`hry>fom;$*lCDj;eb-WrawvGIeieRCg~dani+rVS0&+vGaz=20rM=A8k*H}i&|+|
zwuwJpDC0s>jY+>`h{g0LiJgdm8?*T=U~d==A9RL7#_f4IX0D#r<5&yAepb&JGsx!$W;eWL4*qU{!SvkK2IBvl@m#w
zDK@4P?YlPdF?$)7U!I=@0PI1PS0d!cg03qmDR_^+F!s&Up4iiLC(d&RadB}=exDsY
z>TVNb;^VA%lg;@3o|$HXiklb*irmVE6UTWR(R}OMiLr?uPsN
zK9nij7m;C40`#pL6=cE2AAGEQSYb6wpW~``Dv@&B+eAI{Ygt+kZ0=?r4$_c&!~Z;F
z@{v`MJRh*uXA|g=Eq3E1F>z}&M(iawYU7A4$UO@>JU?Hz3sxKdChgot=9s;u!+}0*
zjAoPuL;KmXLf=wa_9KrZ@C1=6w55eJ_=a8}F
zSA6g7a$GPmi)`w#e0JPE{~Y7@$GY2%xE!_0)hnT!A|5S%dk&KrP1dwkegw8TRw151
zhX;M6&;tl|pfVJF9j|xaKLpJ(8NdAvy1g-6FtnVOGG?ao4T#-~M5ZPzC-XW~L4UvU
ztB3Z=Y`6d*%h6)4^|7#wJK&@JHKlPnj7mx8zyL4J%l(Q^c26c2Qn*qN=hM3n7suH1
za7R{%j+pn|Zf)M+i^|m8xmFt1dR0|3H&KicpKHlMBwQn)ONqcY#gm~_=$-}lx()Hr
zGe@aZD#UHPe3N7})Hld1X`vGbPpjc+H;3qHlH_S;vMyHD?B^*Xr>cwOrew_=rj(aH
zLF?yW()f6ZY)u`8Eaji`p%B{OeTa_eLt7Bbpw*<*k28jT)`wwrFQlXn6l3?K2+*|;
z&iCo)0DCPGAB8lXkISO)l3qt`nX}a$$3fP8ukR4E2z+Q9-d_7S3lf@JzeK<@4cG?G
z@!WZV^`N4CP1|E(GLQ*)ENt26Qw6%ruvZvxOW)csQ<~2#?K+w~jEbRwZ2R|jA>zLL
z_8$a|Lvuv`3aS9%Q)g^=&M-3sH&QWV9^@XhC62UtK*TWrutEs4WI=rY;!
za#HX6*#%16kU76pn)Pooo1YGyYLY%w&k-_#EA>}uy-|_T2H=daW-D_t&9e*G{kCTd
zR5vZjDIpr)n_Y=vlcdxR(kGpFK8lGiad$+f*3W8-sm|gPs3TE#-ypm
z+CIiUL1YALztXz@rZi!gwjB>LrJC8h-}-KU7n1Zh!cCSR5g7#5&HGN3b8%*}c|g)Rsl%jR$<
z?vjETyezaZ&xC9v`HAi%X1&uX?Yi3}Sm@PZfcpw_N;dSi=D4O}%e0V
z?fI7Te4ok{X#Q(K;eroJ^;(Bt(*l337L1%k1JDO-%?<+Eo
z*BgA+QZL21@>*W1^sVZrJ=o_`U%csbO62D4viW$eRP~66FQ-QltVIBw!7(+hKhe)P
z={ufY!yx&oj9J}S;Oj>VZo!04Om}8wKTN&usdNr6J$o+^ZJE3ZSnm`|?3xp}T$Fqw
zYTCBu9IQZfo(>q@uMgLTm9Ta|TsO8&7UHs;V071)f5GNWUDqvAlY_Sha^FB4a7jA<
zws&jh61#KPB!*GrYMV)=Z_^K3xx1yervbHbJXBEC2uV7_rH%Dj-iinF3ZCSeW=apx
z9s+hV82MdU?!II=RxP$;&Nh`gcvy$TI$2f<3+gDG-`xv)#V1COa@Hbo6{`K
zzb+teEhR
zlpZe9vR5)j7&iX$4NS}NK~}WSwqO<}m?F)Y*qLa-BGqQMeupiXwr`Zd6&F%<@1(F;
zIn*wXPR4h0lmr(XC9&~1dgh>%F#1$mN{LuQtYWkqR=EYmLkGtJFT<`>hJf@Fv~pZ(
z>T;0MP#Y7PTWx2ryvfUfB_?OCXpyX2qMUV&F>1p<(<}U4cfalOl)tc60!egs;P)(3
z3O709?+ISK$Og#w+--iXitvIQVeV5?C*ZVFErleOTiAOj8Jt5(h-}5(4>sJ$98l52
zKu^^YHBO}y$2MbpkJJP&G;=ORWL!oqVNl$=y2L;yO{?U@1c?}!`@FJLiW3DZYgMXb
zBv@u6@yC-R>$~qhgQk-l7{>XADw96e-E(UmsGa2Q*EdJZ>=+Yo611&}3`;ks5CI7w?F{-3V3eeEGBTCZ(8-4z#R{A^aNd{so?#Vu#YJ`oq-akJtRs
z?-Y4?!;m(^Q^2Rke4LqZ~5$ez$&(8(ocfQ!c;
zVqpzz%wBzZ+!T6wBb<6C1e|Jegp{MwJ%wXiym9Ah_}Sh>!$lR$u30C)n;IlU?=3NM
z_R^fsf1Tp~T*6v@Y-_E>n}5kPq(SI!A9LacSy%Stf6cQ(_)jPJImbmT4Y=)KS{2+sewd_GLSpNXMoGaA9FZk7YjVryQD^99rv-y2O)6V4AZP`C-Ey7BMA
zI(vOGu{HWhn~bwV`$`@QLnh$p8_1)ZNb?OYj>gBXjo~Z(Xfc^i$tYE_i`=;(*9FYX
z024LSUF8?&7Ni7D)cCZHR7H^w5NlMu@+x<-i|;vA!Q<>tY=d%*?=>^cidjT%MfYem
zE(Ka%IE!!Z1-G`5oE}Y>V$Q&lk(o2ixIoQKpC2+aI;-T|C028+TTVE!h*S1H(;V2(~d?&`@Lz0qx>AZu)qb3oLM7#m+3`pBc
z0ZKQ>`K2DbNUeORr!g)5)GJ(XUE9<3jhmvi+aP11bnQ@KR=4ruz
z)wZmG3$2=0xa#IS;Sz_0>K1<0zODU&m{l%7o>)n3f9q!&=#o-+V;k1#om4yCck^V*
z(|>Vm_LQ06J-hI}On~vwrVRh$Cr)D?B*-DxcD;PnFn>dj>$w~6(BAU#J@}*ZhGXlM
z#bj_v6(J*Dm{uhbg603ZIw1|~F6|qh^ra-cY(p4u7RtW`?xcJX=X;VMDEL5r)W9$t
z4uYxsfSp=he!4)?mdQD`cLncLG;2@ie|=%YcV`01|2U_O(WB^SeLJhRxV2>KHQC;|)6Bqd)<9SzPJJpUU3q|%(
zQ&{WWz%t*$51NX69m+I0I(b09PlfRP{lfvOr$sa=QpI}d?J49*w;ztZv1byuv?9;{
zwJD1;O6HUwO@uk5N@BWm>{6o`n|DE!wJEai&stp%?{P3aI?THU$*ivw$KG3c=!o$Su)AnXCsGH(QVQfty^P_?XY24P&P~!WH7Au
zT@5FMLs074$4Rhk)CiGep0SDdZ_6~q6kMLWJH}bwX|G6i-h8sXdl_zOC4HPE5gf4*
z`BT>Cu-WT$Rifp7*x4;iUDH4!YPu3nWqoIbx39BY%CMNT_th9chEkF+(q|;pV>Zi;
z0brH@RT=Jzp78@Ctn>
zLqz)pQ?Af1-PdXRkcXSAlE}L4!7G4Gja-h6(c1TBRe-;%h*xGsjEz7X$XnmY&;xas
zGABs@Akh*X;NqM5#i=OwMLy#l+Mge0dtS{p2_rX35}ptQyoykaif
zVA9o1Dc1Q>iJps|mdq-E#oiopGx(84!V*hg#8^;1-gYZ@h-aoy1U;Lj9wEXW30;`t
z|M(*l`8iFd{*Bc;)t-Eg>Qy!G;mreGmbnj{)!m%0-mw&P^AMq{SmOktn)YP0>q8-b
z6XW;CB?WzoEHe1(O$YH(B(7OM{UgM#J2_C`LsRcBer@1Q)ojsp;Z&*eS^xf2JuXhH
zkGsbnZg;^gu=#V2g>~rln~^E$D{tEjXE(_f7?;NS-X0mvt10}bL-#eZ8%XgkYkMrW
z-prCC-pdF1CiKxlRFC{+PnyAQY)eDQMjZQbcTbx4V+JPdL@AjN6|YvCm;+pV1|vNr
z_69BI)plc^^Ntm*jaAKbRz}Lm=T#78^e%0@k-vA2tTc&iKCiQ9C)I|5Pd)EAus@mT
zx#_th^t3H*_=blG;N%*v9`|8D+6A9zB8#gX>(NmSl+-5X!0abG+twP}FKOi!PXA;_
z_*dcp*M#yH3%N%N=`>(D0d6@=<)`Kx+mbUY8^()HlZl^t(Wc7wFP#r{awv4Fa)5tg
zF>y7#yPlq$NxbM>)b#2~8^N?H$}6|=AhzCEzUD*?CtJCAZ|4A?R@=pF=x}!-l42xy
zW1#?UUo*JHCoZ)KUI}-yYs=LG1{y2kPsQy!WviWjmv-e73d4
zC!9)*T4KFnHsnitB!!|~1mO5SVk
zh~o+P{APAek%!rFm$oV@)?-$1_&9et$!?ws9OrB)v!3crI$VA=>2#Fa&q#4{EGFo>
zI#O#>@rvA8HGQ<=!ndM~?C(?rZA4^MK4y-)c)Tw#!nqEyZHD}@`Z0SPhhMMqJz@8sRW?r|Q?8-<
zV7&I-Iiq7+H}=r^=<-QzklX!B>E~fGcd)4;V#$S!Po+*eSeqFnygC!>Q#TJqgnd&K
z_pXVUC@n=S2I=wKJfZ(03Q!a(QEpR_k(n`JOb|>jYhL`RJ-ph|
z2duugLi@-y=c*JvIT;V@dHO}E&cSipSyOIGkRW^0v9Z8$Hcp0GN_5zQSDoXqQK6tR
zY*1O}s70|&v3l@l|Jq_^Q)mciQvWDcJ1~~h)pKWRp7FGJB(fj{qpV%b%fYJKmCzGd
zsXs2owOkf_*$r*HxssRVTf7wsM;+%L>4bj~UwG=^yXj`QP@sJSohW9G9Uv__`(E)M
zfPMx$=uR++NE67zVLkUUktdCg<#~+{D9t)jah!oy$oe`!2`W1?InIbEdgR99HlMFC
zNbH@qfsOl*dvM7#514YN(JAVSua1eYjzgUotFVMXou&Twnzcu^FB8IV?ROJQ%QYF4rEE^(KqdeGL+_(ct*A^7owYY)zA~E^wzkus8|nw^mH?2
zXhc87+i;;WsJ|0eR9wzD*4O
zvic6GUJf02$-*K)$8w_(-m}qPpcayPr&ybmWlVKwaZkbcwcI6h;p=8rj$i;}FrY90
z1(aKl*l8lg!H99>B&NtS_ZI0#%BI0lgVpKp3>owyERx#ehZW%RK4(~V`ZEXeR8$PL
zxo?Rp65QY`eV)TxpM$|trNoG$w~%w$@%>+&q#+}y_FN&7X$1r{H2XvUKKe=j+YO%|
zNkBkGa)I#?DiFfIXFIR}Yw~TC{KlLShlaLG@&vI2K6AupEpfvvhd6(L5`pYHG7`1(
zM-5)jVUc+@xu#6NQOpu-AI?`kXcwQD%Maq}x_r!LoY3;%d(-ln^!GkWRV}goK|>lF
zpxgA0k=Z%wZAj%?sq%p{y@EFxgYBT6jd>
zd(_o-Dz0~U0pKC<`PFj^(1?oB7kguJ@Z3c+N8kfZMK9OML1?KN$-nJB&?;qhkeB0j
z>levp)$%MT$%n;0o=4;~cW5Oc$@cGbce*lv3TFrWHMV=dBi7(TV>Ov)g|e
z&)-D)&y#=a(mx+wob5kO{^uqB7wcY}?dv%H+8>sKH?(`lAUlda-=w6VCSN3H>PPV3
Y@!#>^@!#>^@qcFg4}^t9{Q$HC0I&TR?f?J)

literal 0
HcmV?d00001

diff --git a/doc/src/DecisionTrees/reveal.js/.gitignore b/doc/src/DecisionTrees/reveal.js/.gitignore
new file mode 100644
index 000000000..a5df3133d
--- /dev/null
+++ b/doc/src/DecisionTrees/reveal.js/.gitignore
@@ -0,0 +1,8 @@
+.DS_Store
+.svn
+log/*.log
+tmp/**
+node_modules/
+.sass-cache
+css/reveal.min.css
+js/reveal.min.js
diff --git a/doc/src/DecisionTrees/reveal.js/.travis.yml b/doc/src/DecisionTrees/reveal.js/.travis.yml
new file mode 100644
index 000000000..165d9ae9f
--- /dev/null
+++ b/doc/src/DecisionTrees/reveal.js/.travis.yml
@@ -0,0 +1,5 @@
+language: node_js
+node_js:
+  - 0.10
+before_script:
+  - npm install -g grunt-cli
\ No newline at end of file
diff --git a/doc/src/DecisionTrees/reveal.js/CONTRIBUTING.md b/doc/src/DecisionTrees/reveal.js/CONTRIBUTING.md
new file mode 100644
index 000000000..c2091e88f
--- /dev/null
+++ b/doc/src/DecisionTrees/reveal.js/CONTRIBUTING.md
@@ -0,0 +1,23 @@
+## Contributing
+
+Please keep the [issue tracker](http://github.com/hakimel/reveal.js/issues) limited to **bug reports**, **feature requests** and **pull requests**.
+
+
+### Personal Support
+If you have personal support or setup questions the best place to ask those are [StackOverflow](http://stackoverflow.com/questions/tagged/reveal.js).
+
+
+### Bug Reports
+When reporting a bug make sure to include information about which browser and operating system you are on as well as the necessary steps to reproduce the issue. If possible please include a link to a sample presentation where the bug can be tested.
+
+
+### Pull Requests
+- Should follow the coding style of the file you work in, most importantly:
+  - Tabs to indent
+  - Single-quoted strings
+- Should be made towards the **dev branch**
+- Should be submitted from a feature/topic branch (not your master)
+
+
+### Plugins
+Please do not submit plugins as pull requests. They should be maintained in their own separate repository. More information here: https://github.com/hakimel/reveal.js/wiki/Plugin-Guidelines
diff --git a/doc/src/DecisionTrees/reveal.js/Gruntfile.js b/doc/src/DecisionTrees/reveal.js/Gruntfile.js
new file mode 100644
index 000000000..b257e8f32
--- /dev/null
+++ b/doc/src/DecisionTrees/reveal.js/Gruntfile.js
@@ -0,0 +1,140 @@
+/* global module:false */
+module.exports = function(grunt) {
+	var port = grunt.option('port') || 8000;
+	// Project configuration
+	grunt.initConfig({
+		pkg: grunt.file.readJSON('package.json'),
+		meta: {
+			banner:
+				'/*!\n' +
+				' * reveal.js <%= pkg.version %> (<%= grunt.template.today("yyyy-mm-dd, HH:MM") %>)\n' +
+				' * http://lab.hakim.se/reveal-js\n' +
+				' * MIT licensed\n' +
+				' *\n' +
+				' * Copyright (C) 2014 Hakim El Hattab, http://hakim.se\n' +
+				' */'
+		},
+
+		qunit: {
+			files: [ 'test/*.html' ]
+		},
+
+		uglify: {
+			options: {
+				banner: '<%= meta.banner %>\n'
+			},
+			build: {
+				src: 'js/reveal.js',
+				dest: 'js/reveal.min.js'
+			}
+		},
+
+		cssmin: {
+			compress: {
+				files: {
+					'css/reveal.min.css': [ 'css/reveal.css' ]
+				}
+			}
+		},
+
+		sass: {
+			main: {
+				files: {
+					'css/theme/darkgray.css': 'css/theme/source/darkgray.scss',
+					'css/theme/beigesmall.css': 'css/theme/source/beigesmall.scss',
+					'css/theme/cbc.css': 'css/theme/source/cbc.scss',
+					'css/theme/default.css': 'css/theme/source/default.scss',
+					'css/theme/beige.css': 'css/theme/source/beige.scss',
+					'css/theme/night.css': 'css/theme/source/night.scss',
+					'css/theme/serif.css': 'css/theme/source/serif.scss',
+					'css/theme/simple.css': 'css/theme/source/simple.scss',
+					'css/theme/sky.css': 'css/theme/source/sky.scss',
+					'css/theme/moon.css': 'css/theme/source/moon.scss',
+					'css/theme/solarized.css': 'css/theme/source/solarized.scss',
+					'css/theme/blood.css': 'css/theme/source/blood.scss'
+				}
+			}
+		},
+
+		jshint: {
+			options: {
+				curly: false,
+				eqeqeq: true,
+				immed: true,
+				latedef: true,
+				newcap: true,
+				noarg: true,
+				sub: true,
+				undef: true,
+				eqnull: true,
+				browser: true,
+				expr: true,
+				globals: {
+					head: false,
+					module: false,
+					console: false,
+					unescape: false
+				}
+			},
+			files: [ 'Gruntfile.js', 'js/reveal.js' ]
+		},
+
+		connect: {
+			server: {
+				options: {
+					port: port,
+					base: '.'
+				}
+			}
+		},
+
+		zip: {
+			'reveal-js-presentation.zip': [
+				'index.html',
+				'css/**',
+				'js/**',
+				'lib/**',
+				'images/**',
+				'plugin/**'
+			]
+		},
+
+		watch: {
+			main: {
+				files: [ 'Gruntfile.js', 'js/reveal.js', 'css/reveal.css' ],
+				tasks: 'default'
+			},
+			theme: {
+				files: [ 'css/theme/source/*.scss', 'css/theme/template/*.scss' ],
+				tasks: 'themes'
+			}
+		}
+
+	});
+
+	// Dependencies
+	grunt.loadNpmTasks( 'grunt-contrib-qunit' );
+	grunt.loadNpmTasks( 'grunt-contrib-jshint' );
+	grunt.loadNpmTasks( 'grunt-contrib-cssmin' );
+	grunt.loadNpmTasks( 'grunt-contrib-uglify' );
+	grunt.loadNpmTasks( 'grunt-contrib-watch' );
+	grunt.loadNpmTasks( 'grunt-contrib-sass' );
+	grunt.loadNpmTasks( 'grunt-contrib-connect' );
+	grunt.loadNpmTasks( 'grunt-zip' );
+
+	// Default task
+	grunt.registerTask( 'default', [ 'jshint', 'cssmin', 'uglify', 'qunit' ] );
+
+	// Theme task
+	grunt.registerTask( 'themes', [ 'sass' ] );
+
+	// Package presentation to archive
+	grunt.registerTask( 'package', [ 'default', 'zip' ] );
+
+	// Serve presentation locally
+	grunt.registerTask( 'serve', [ 'connect', 'watch' ] );
+
+	// Run tests
+	grunt.registerTask( 'test', [ 'jshint', 'qunit' ] );
+
+};
diff --git a/doc/src/DecisionTrees/reveal.js/LICENSE b/doc/src/DecisionTrees/reveal.js/LICENSE
new file mode 100644
index 000000000..09623076f
--- /dev/null
+++ b/doc/src/DecisionTrees/reveal.js/LICENSE
@@ -0,0 +1,19 @@
+Copyright (C) 2015 Hakim El Hattab, http://hakim.se
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in
+all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+THE SOFTWARE.
\ No newline at end of file
diff --git a/doc/src/DecisionTrees/reveal.js/README.md b/doc/src/DecisionTrees/reveal.js/README.md
new file mode 100644
index 000000000..573b19597
--- /dev/null
+++ b/doc/src/DecisionTrees/reveal.js/README.md
@@ -0,0 +1,1052 @@
+# reveal.js [![Build Status](https://travis-ci.org/hakimel/reveal.js.svg?branch=master)](https://travis-ci.org/hakimel/reveal.js)
+
+A framework for easily creating beautiful presentations using HTML. [Check out the live demo](http://lab.hakim.se/reveal-js/).
+
+reveal.js comes with a broad range of features including [nested slides](https://github.com/hakimel/reveal.js#markup), [Markdown contents](https://github.com/hakimel/reveal.js#markdown), [PDF export](https://github.com/hakimel/reveal.js#pdf-export), [speaker notes](https://github.com/hakimel/reveal.js#speaker-notes) and a [JavaScript API](https://github.com/hakimel/reveal.js#api). It's best viewed in a modern browser but [fallbacks](https://github.com/hakimel/reveal.js/wiki/Browser-Support) are available to make sure your presentation can still be viewed elsewhere.
+
+
+#### More reading:
+- [Installation](#installation): Step-by-step instructions for getting reveal.js running on your computer.
+- [Changelog](https://github.com/hakimel/reveal.js/releases): Up-to-date version history.
+- [Examples](https://github.com/hakimel/reveal.js/wiki/Example-Presentations): Presentations created with reveal.js, add your own!
+- [Browser Support](https://github.com/hakimel/reveal.js/wiki/Browser-Support): Explanation of browser support and fallbacks.
+- [Plugins](https://github.com/hakimel/reveal.js/wiki/Plugins,-Tools-and-Hardware): A list of plugins that can be used to extend reveal.js.
+
+## Online Editor
+
+Presentations are written using HTML or Markdown but there's also an online editor for those of you who prefer a graphical interface. Give it a try at [http://slides.com](http://slides.com).
+
+
+## Instructions
+
+### Markup
+
+Markup hierarchy needs to be ``
     
     
    `` where the ``
    `` represents one slide and can be repeated indefinitely. If you place multiple ``
    ``'s inside of another ``
    `` they will be shown as vertical slides. The first of the vertical slides is the "root" of the others (at the top), and it will be included in the horizontal sequence. For example:
+
+```html
+
    
+	
    
+		
    Single Horizontal Slide
    
+		
    
+			
    Vertical Slide 1
    
+			
    Vertical Slide 2
    
+		
    
+	
    
+
    
+```
+
+### Markdown
+
+It's possible to write your slides using Markdown. To enable Markdown, add the ```data-markdown``` attribute to your ```
    ``` elements and wrap the contents in a ```
+
    
+```
+
+#### External Markdown
+
+You can write your content as a separate file and have reveal.js load it at runtime. Note the separator arguments which determine how slides are delimited in the external file. The ```data-charset``` attribute is optional and specifies which charset to use when loading the external file.
+
+When used locally, this feature requires that reveal.js [runs from a local web server](#full-setup).
+
+```html
+
    
+
    
+```
+
+#### Element Attributes
+
+Special syntax (in html comment) is available for adding attributes to Markdown elements. This is useful for fragments, amongst other things.
+
+```html
+
    
+	
+
    
+```
+
+#### Slide Attributes
+
+Special syntax (in html comment) is available for adding attributes to the slide `
    ` elements generated by your Markdown.
+
+```html
+
    
+	
+
    
+```
+
+
+### Configuration
+
+At the end of your page you need to initialize reveal by running the following code. Note that all config values are optional and will default as specified below.
+
+```javascript
+Reveal.initialize({
+
+	// Display controls in the bottom right corner
+	controls: true,
+
+	// Display a presentation progress bar
+	progress: true,
+
+	// Display the page number of the current slide
+	slideNumber: false,
+
+	// Push each slide change to the browser history
+	history: false,
+
+	// Enable keyboard shortcuts for navigation
+	keyboard: true,
+
+	// Enable the slide overview mode
+	overview: true,
+
+	// Vertical centering of slides
+	center: true,
+
+	// Enables touch navigation on devices with touch input
+	touch: true,
+
+	// Loop the presentation
+	loop: false,
+
+	// Change the presentation direction to be RTL
+	rtl: false,
+
+	// Turns fragments on and off globally
+	fragments: true,
+
+	// Flags if the presentation is running in an embedded mode,
+	// i.e. contained within a limited portion of the screen
+	embedded: false,
+
+	// Flags if we should show a help overlay when the questionmark
+	// key is pressed
+	help: true,
+
+	// Number of milliseconds between automatically proceeding to the
+	// next slide, disabled when set to 0, this value can be overwritten
+	// by using a data-autoslide attribute on your slides
+	autoSlide: 0,
+
+	// Stop auto-sliding after user input
+	autoSlideStoppable: true,
+
+	// Enable slide navigation via mouse wheel
+	mouseWheel: false,
+
+	// Hides the address bar on mobile devices
+	hideAddressBar: true,
+
+	// Opens links in an iframe preview overlay
+	previewLinks: false,
+
+	// Transition style
+	transition: 'default', // none/fade/slide/convex/concave/zoom
+
+	// Transition speed
+	transitionSpeed: 'default', // default/fast/slow
+
+	// Transition style for full page slide backgrounds
+	backgroundTransition: 'default', // none/fade/slide/convex/concave/zoom
+
+	// Number of slides away from the current that are visible
+	viewDistance: 3,
+
+	// Parallax background image
+	parallaxBackgroundImage: '', // e.g. "'https://s3.amazonaws.com/hakim-static/reveal-js/reveal-parallax-1.jpg'"
+
+	// Parallax background size
+	parallaxBackgroundSize: '', // CSS syntax, e.g. "2100px 900px"
+
+	// Amount to move parallax background (horizontal and vertical) on slide change
+	// Number, e.g. 100
+	parallaxBackgroundHorizontal: '',
+	parallaxBackgroundVertical: ''
+
+});
+```
+
+
+The configuration can be updated after initialization using the ```configure``` method:
+
+```javascript
+// Turn autoSlide off
+Reveal.configure({ autoSlide: 0 });
+
+// Start auto-sliding every 5s
+Reveal.configure({ autoSlide: 5000 });
+```
+
+
+### Dependencies
+
+Reveal.js doesn't _rely_ on any third party scripts to work but a few optional libraries are included by default. These libraries are loaded as dependencies in the order they appear, for example:
+
+```javascript
+Reveal.initialize({
+	dependencies: [
+		// Cross-browser shim that fully implements classList - https://github.com/eligrey/classList.js/
+		{ src: 'lib/js/classList.js', condition: function() { return !document.body.classList; } },
+
+		// Interpret Markdown in 
     elements
+		{ src: 'plugin/markdown/marked.js', condition: function() { return !!document.querySelector( '[data-markdown]' ); } },
+		{ src: 'plugin/markdown/markdown.js', condition: function() { return !!document.querySelector( '[data-markdown]' ); } },
+
+		// Syntax highlight for  elements
+		{ src: 'plugin/highlight/highlight.js', async: true, callback: function() { hljs.initHighlightingOnLoad(); } },
+
+		// Zoom in and out with Alt+click
+		{ src: 'plugin/zoom-js/zoom.js', async: true },
+
+		// Speaker notes
+		{ src: 'plugin/notes/notes.js', async: true },
+
+		// Remote control your reveal.js presentation using a touch device
+		{ src: 'plugin/remotes/remotes.js', async: true },
+
+		// MathJax
+		{ src: 'plugin/math/math.js', async: true }
+	]
+});
+```
+
+You can add your own extensions using the same syntax. The following properties are available for each dependency object:
+- **src**: Path to the script to load
+- **async**: [optional] Flags if the script should load after reveal.js has started, defaults to false
+- **callback**: [optional] Function to execute when the script has loaded
+- **condition**: [optional] Function which must return true for the script to be loaded
+
+
+### Ready Event
+
+A 'ready' event is fired when reveal.js has loaded all non-async dependencies and is ready to start navigating. To check if reveal.js is already 'ready' you can call `Reveal.isReady()`.
+
+```javascript
+Reveal.addEventListener( 'ready', function( event ) {
+	// event.currentSlide, event.indexh, event.indexv
+} );
+```
+
+
+### Presentation Size
+
+All presentations have a normal size, that is the resolution at which they are authored. The framework will automatically scale presentations uniformly based on this size to ensure that everything fits on any given display or viewport.
+
+See below for a list of configuration options related to sizing, including default values:
+
+```javascript
+Reveal.initialize({
+
+	...
+
+	// The "normal" size of the presentation, aspect ratio will be preserved
+	// when the presentation is scaled to fit different resolutions. Can be
+	// specified using percentage units.
+	width: 960,
+	height: 700,
+
+	// Factor of the display size that should remain empty around the content
+	margin: 0.1,
+
+	// Bounds for smallest/largest possible scale to apply to content
+	minScale: 0.2,
+	maxScale: 1.5
+
+});
+```
+
+
+### Auto-sliding
+
+Presentations can be configured to progress through slides automatically, without any user input. To enable this you will need to tell the framework how many milliseconds it should wait between slides:
+
+```javascript
+// Slide every five seconds
+Reveal.configure({
+  autoSlide: 5000
+});
+```
+When this is turned on a control element will appear that enables users to pause and resume auto-sliding. Alternatively, sliding can be paused or resumed by pressing »a« on the keyboard. Sliding is paused automatically as soon as the user starts navigating. You can disable these controls by specifying ```autoSlideStoppable: false``` in your reveal.js config.
+
+You can also override the slide duration for individual slides and fragments by using the ```data-autoslide``` attribute:
+
+```html
+
    
+	
    After 2 seconds the first fragment will be shown.
    
+	
    After 10 seconds the next fragment will be shown.
    
+	
    Now, the fragment is displayed for 2 seconds before the next slide is shown.
    
+
    
+```
+
+Whenever the auto-slide mode is resumed or paused the ```autoslideresumed``` and ```autoslidepaused``` events are fired.
+
+
+### Keyboard Bindings
+
+If you're unhappy with any of the default keyboard bindings you can override them using the ```keyboard``` config option:
+
+```javascript
+Reveal.configure({
+  keyboard: {
+    13: 'next', // go to the next slide when the ENTER key is pressed
+    27: function() {}, // do something custom when ESC is pressed
+    32: null // don't do anything when SPACE is pressed (i.e. disable a reveal.js default binding)
+  }
+});
+```
+
+### Lazy Loading
+
+When working on presentation with a lot of media or iframe content it's important to load lazily. Lazy loading means that reveal.js will only load content for the few slides nearest to the current slide. The number of slides that are preloaded is determined by the `viewDistance` configuration option.
+
+To enable lazy loading all you need to do is change your "src" attributes to "data-src" as shown below. This is supported for image, video, audio and iframe elements. Lazy loaded iframes will also unload when the containing slide is no longer visible.
+
+```html
+
    
+  
+  
+  
+
    
+```
+
+
+### API
+
+The ``Reveal`` object exposes a JavaScript API for controlling navigation and reading state:
+
+```javascript
+// Navigation
+Reveal.slide( indexh, indexv, indexf );
+Reveal.left();
+Reveal.right();
+Reveal.up();
+Reveal.down();
+Reveal.prev();
+Reveal.next();
+Reveal.prevFragment();
+Reveal.nextFragment();
+
+// Toggle presentation states, optionally pass true/false to force on/off
+Reveal.toggleOverview();
+Reveal.togglePause();
+Reveal.toggleAutoSlide();
+
+// Change a config value at runtime
+Reveal.configure({ controls: true });
+
+// Returns the present configuration options
+Reveal.getConfig();
+
+// Fetch the current scale of the presentation
+Reveal.getScale();
+
+// Retrieves the previous and current slide elements
+Reveal.getPreviousSlide();
+Reveal.getCurrentSlide();
+
+Reveal.getIndices(); // { h: 0, v: 0 } }
+Reveal.getProgress(); // 0-1
+Reveal.getTotalSlides();
+
+// State checks
+Reveal.isFirstSlide();
+Reveal.isLastSlide();
+Reveal.isOverview();
+Reveal.isPaused();
+Reveal.isAutoSliding();
+```
+
+### Slide Changed Event
+
+A 'slidechanged' event is fired each time the slide is changed (regardless of state). The event object holds the index values of the current slide as well as a reference to the previous and current slide HTML nodes.
+
+Some libraries, like MathJax (see [#226](https://github.com/hakimel/reveal.js/issues/226#issuecomment-10261609)), get confused by the transforms and display states of slides. Often times, this can be fixed by calling their update or render function from this callback.
+
+```javascript
+Reveal.addEventListener( 'slidechanged', function( event ) {
+	// event.previousSlide, event.currentSlide, event.indexh, event.indexv
+} );
+```
+
+### Presentation State
+
+The presentation's current state can be fetched by using the `getState` method. A state object contains all of the information required to put the presentation back as it was when `getState` was first called. Sort of like a snapshot. It's a simple object that can easily be stringified and persisted or sent over the wire.
+
+```javascript
+Reveal.slide( 1 );
+// we're on slide 1
+
+var state = Reveal.getState();
+
+Reveal.slide( 3 );
+// we're on slide 3
+
+Reveal.setState( state );
+// we're back on slide 1
+```
+
+### Slide States
+
+If you set ``data-state="somestate"`` on a slide ``
    ``, "somestate" will be applied as a class on the document element when that slide is opened. This allows you to apply broad style changes to the page based on the active slide.
+
+Furthermore you can also listen to these changes in state via JavaScript:
+
+```javascript
+Reveal.addEventListener( 'somestate', function() {
+	// TODO: Sprinkle magic
+}, false );
+```
+
+### Slide Backgrounds
+
+Slides are contained within a limited portion of the screen by default to allow them to fit any display and scale uniformly. You can apply full page backgrounds outside of the slide area by adding a ```data-background``` attribute to your ```
    ``` elements. Four different types of backgrounds are supported: color, image, video and iframe. Below are a few examples.
+
+```html
+
    
+	
    All CSS color formats are supported, like rgba() or hsl().
    
+
    
+
    
+	
    This slide will have a full-size background image.
    
+
    
+
    
+	
    This background image will be sized to 100px and repeated.
    
+
    
+
    
+	
    Video. Multiple sources can be defined using a comma separated list. Video will loop when the data-background-video-loop attribute is provided.
    
+
    
+
    
+	
    Embeds a web page as a background. Note that the page won't be interactive.
    
+
    
+```
+
+Backgrounds transition using a fade animation by default. This can be changed to a linear sliding transition by passing ```backgroundTransition: 'slide'``` to the ```Reveal.initialize()``` call. Alternatively you can set ```data-background-transition``` on any section with a background to override that specific transition.
+
+
+### Parallax Background
+
+If you want to use a parallax scrolling background, set the first two config properties below when initializing reveal.js (the other two are optional).
+
+```javascript
+Reveal.initialize({
+
+	// Parallax background image
+	parallaxBackgroundImage: '', // e.g. "https://s3.amazonaws.com/hakim-static/reveal-js/reveal-parallax-1.jpg"
+
+	// Parallax background size
+	parallaxBackgroundSize: '', // CSS syntax, e.g. "2100px 900px" - currently only pixels are supported (don't use % or auto)
+
+	// Amount of pixels to move the parallax background per slide step,
+	// a value of 0 disables movement along the given axis
+	// These are optional, if they aren't specified they'll be calculated automatically
+	parallaxBackgroundHorizontal: 200,
+	parallaxBackgroundVertical: 50
+
+});
+```
+
+Make sure that the background size is much bigger than screen size to allow for some scrolling. [View example](http://lab.hakim.se/reveal-js/?parallaxBackgroundImage=https%3A%2F%2Fs3.amazonaws.com%2Fhakim-static%2Freveal-js%2Freveal-parallax-1.jpg¶llaxBackgroundSize=2100px%20900px).
+
+
+
+### Slide Transitions
+The global presentation transition is set using the ```transition``` config value. You can override the global transition for a specific slide by using the ```data-transition``` attribute:
+
+```html
+
    
+	
    This slide will override the presentation transition and zoom!
    
+
    
+
+
    
+	
    Choose from three transition speeds: default, fast or slow!
    
+
    
+```
+
+You can also use different in and out transitions for the same slide:
+
+```html
+
    
+    The train goes on … 
+
    
+
     
+    and on … 
+
    
+
     
+    and stops.
+
    
+
     
+    (Passengers entering and leaving)
+
    
+
    
+    And it starts again.
+
    
+```
+
+
+Note that this does not work with the page and cube transitions.
+
+
+### Internal links
+
+It's easy to link between slides. The first example below targets the index of another slide whereas the second targets a slide with an ID attribute (```
    ```):
+
+```html
+Link
+Link
+```
+
+You can also add relative navigation links, similar to the built in reveal.js controls, by appending one of the following classes on any element. Note that each element is automatically given an ```enabled``` class when it's a valid navigation route based on the current slide.
+
+```html
+
+
+
+
+ 
+ 
+```
+
+
+### Fragments
+Fragments are used to highlight individual elements on a slide. Every element with the class ```fragment``` will be stepped through before moving on to the next slide. Here's an example: http://lab.hakim.se/reveal-js/#/fragments
+
+The default fragment style is to start out invisible and fade in. This style can be changed by appending a different class to the fragment:
+
+```html
+
    
+	
    grow
    
+	
    shrink
    
+	
    fade-out
    
+	
    visible only once
    
+	
    blue only once
    
+	
    highlight-red
    
+	
    highlight-green
    
+	
    highlight-blue
    
+
    
+```
+
+Multiple fragments can be applied to the same element sequentially by wrapping it, this will fade in the text on the first step and fade it back out on the second.
+
+```html
+
    
+	
+		I'll fade in, then out
+	
+
    
+```
+
+The display order of fragments can be controlled using the ```data-fragment-index``` attribute.
+
+```html
+
    
+	
    Appears last
    
+	
    Appears first
    
+	
    Appears second
    
+
    
+```
+
+### Fragment events
+
+When a slide fragment is either shown or hidden reveal.js will dispatch an event.
+
+Some libraries, like MathJax (see #505), get confused by the initially hidden fragment elements. Often times this can be fixed by calling their update or render function from this callback.
+
+```javascript
+Reveal.addEventListener( 'fragmentshown', function( event ) {
+	// event.fragment = the fragment DOM element
+} );
+Reveal.addEventListener( 'fragmenthidden', function( event ) {
+	// event.fragment = the fragment DOM element
+} );
+```
+
+### Code syntax highlighting
+
+By default, Reveal is configured with [highlight.js](http://softwaremaniacs.org/soft/highlight/en/) for code syntax highlighting. Below is an example with clojure code that will be syntax highlighted. When the `data-trim` attribute is present surrounding whitespace is automatically removed.
+
+```html
+
    
+	
    
+(def lazy-fib
+  (concat
+   [0 1]
+   ((fn rfib [a b]
+        (lazy-cons (+ a b) (rfib b (+ a b)))) 0 1)))
+	
    
+
    
+```
+
+### Slide number
+If you would like to display the page number of the current slide you can do so using the ```slideNumber``` configuration value.
+
+```javascript
+// Shows the slide number using default formatting
+Reveal.configure({ slideNumber: true });
+
+// Slide number formatting can be configured using these variables:
+//  h: current slide's horizontal index
+//  v: current slide's vertical index
+//  c: current slide index (flattened)
+//  t: total number of slides (flattened)
+Reveal.configure({ slideNumber: 'c / t' });
+
+```
+
+
+### Overview mode
+
+Press "Esc" or "o" keys to toggle the overview mode on and off. While you're in this mode, you can still navigate between slides,
+as if you were at 1,000 feet above your presentation. The overview mode comes with a few API hooks:
+
+```javascript
+Reveal.addEventListener( 'overviewshown', function( event ) { /* ... */ } );
+Reveal.addEventListener( 'overviewhidden', function( event ) { /* ... */ } );
+
+// Toggle the overview mode programmatically
+Reveal.toggleOverview();
+```
+
+### Fullscreen mode
+Just press »F« on your keyboard to show your presentation in fullscreen mode. Press the »ESC« key to exit fullscreen mode.
+
+
+### Embedded media
+Embedded HTML5 `