diff --git a/doc/pub/week38/html/._week38-bs000.html b/doc/pub/week38/html/._week38-bs000.html index d43a9b66b..3c1fdc334 100644 --- a/doc/pub/week38/html/._week38-bs000.html +++ b/doc/pub/week38/html/._week38-bs000.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -325,7 +339,7 @@ MathJax.Hub.Config({
    -

    September 18-22

    +

    September 16-20, 2024


    @@ -350,7 +364,7 @@ MathJax.Hub.Config({
  • 9
  • 10
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • @@ -364,7 +378,7 @@ MathJax.Hub.Config({ -->
    - © 1999-2023, Morten Hjorth-Jensen. Released under CC Attribution-NonCommercial 4.0 license + © 1999-2024, Morten Hjorth-Jensen. Released under CC Attribution-NonCommercial 4.0 license
    diff --git a/doc/pub/week38/html/._week38-bs001.html b/doc/pub/week38/html/._week38-bs001.html index 4ca649286..9864dbaf6 100644 --- a/doc/pub/week38/html/._week38-bs001.html +++ b/doc/pub/week38/html/._week38-bs001.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,37 +321,16 @@ MathJax.Hub.Config({

     

     

     

    -

    Plans for week 38

    +

    Plans for week 38, lecture Monday September 16

    -
    -
    - - -
    -
    -
    @@ -359,7 +352,7 @@ MathJax.Hub.Config({
  • 10
  • 11
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs002.html b/doc/pub/week38/html/._week38-bs002.html index 5f67d06ee..487d5e7ea 100644 --- a/doc/pub/week38/html/._week38-bs002.html +++ b/doc/pub/week38/html/._week38-bs002.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,7 +321,24 @@ MathJax.Hub.Config({

     

     

     

    -

    Material from last week and relevant for the first project

    +

    Suggested reading and videos

    +
    +
    + + +
    +
    +

    @@ -326,7 +357,7 @@ MathJax.Hub.Config({

  • 11
  • 12
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs003.html b/doc/pub/week38/html/._week38-bs003.html index c1d331831..57141c7bc 100644 --- a/doc/pub/week38/html/._week38-bs003.html +++ b/doc/pub/week38/html/._week38-bs003.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -306,24 +320,21 @@ MathJax.Hub.Config({

     

     

     

    - -

    Various steps in cross-validation

    + +

    Plans for the lab sessions

    -

    When the repetitive splitting of the data set is done randomly, -samples may accidently end up in a fast majority of the splits in -either training or test set. Such samples may have an unbalanced -influence on either model building or prediction evaluation. To avoid -this \( k \)-fold cross-validation structures the data splitting. The -samples are divided into \( k \) more or less equally sized exhaustive and -mutually exclusive subsets. In turn (at each split) one of these -subsets plays the role of the test set while the union of the -remaining subsets constitutes the training set. Such a splitting -warrants a balanced representation of each sample in both training and -test set over the splits. Still the division into the \( k \) subsets -involves a degree of randomness. This may be fully excluded when -choosing \( k=n \). This particular case is referred to as leave-one-out -cross-validation (LOOCV). -

    +
    +
    + +
      +
    • Repetition from last week on the bias-variance tradeoff
      • +
    • Resampling techniques, cross-validation examples included here, see also the lectures from last week on the bootstrap method
      • +
    • Exercise for week 38 on the bias-variance tradeoff, see also the video from the lab session from week 37 at https://youtu.be/omLmp_kkie0
      • +
    • Work on project 1, in particular resampling methods like cross-validation and bootstrap.
      • +
    +
    +
    +

    @@ -343,7 +354,7 @@ cross-validation (LOOCV).

  • 12
  • 13
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs004.html b/doc/pub/week38/html/._week38-bs004.html index 448fb6c4d..81863a2ad 100644 --- a/doc/pub/week38/html/._week38-bs004.html +++ b/doc/pub/week38/html/._week38-bs004.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -306,28 +320,9 @@ MathJax.Hub.Config({

     

     

     

    - -

    How to set up the cross-validation for Ridge and/or Lasso

    + +

    Material for lecture Thursday September 21

    - -$$ -\begin{align*} -\boldsymbol{\beta}_{-i}(\lambda) & = ( \boldsymbol{X}_{-i, \ast}^{T} -\boldsymbol{X}_{-i, \ast} + \lambda \boldsymbol{I}_{pp})^{-1} -\boldsymbol{X}_{-i, \ast}^{T} \boldsymbol{y}_{-i} -\end{align*} -$$ - - -

    diff --git a/doc/pub/week38/html/._week38-bs005.html b/doc/pub/week38/html/._week38-bs005.html index ee6df67da..551bdd8e1 100644 --- a/doc/pub/week38/html/._week38-bs005.html +++ b/doc/pub/week38/html/._week38-bs005.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -306,23 +320,22 @@ MathJax.Hub.Config({

     

     

     

    - -

    Cross-validation in brief

    + +

    Logistic Regression

    -

    For the various values of \( k \)

    +

    In linear regression our main interest was centered on learning the +coefficients of a functional fit (say a polynomial) in order to be +able to predict the response of a continuous variable on some unseen +data. The fit to the continuous variable \( y_i \) is based on some +independent variables \( \boldsymbol{x}_i \). Linear regression resulted in +analytical expressions for standard ordinary Least Squares or Ridge +regression (in terms of matrices to invert) for several quantities, +ranging from the variance and thereby the confidence intervals of the +parameters \( \boldsymbol{\beta} \) to the mean squared error. If we can invert +the product of the design matrices, linear regression gives then a +simple recipe for fitting our data. +

    -
      -
    1. shuffle the dataset randomly.
      2. -
    2. Split the dataset into \( k \) groups.
      3. -
    3. For each unique group: -
        -
      1. Decide which group to use as set for test data
        2. -
      2. Take the remaining groups as a training data set
        3. -
      3. Fit a model on the training set and evaluate it on the test set
        4. -
      4. Retain the evaluation score and discard the model
        5. -
      -
    4. Summarize the model using the sample of model evaluation scores
      5. -

      @@ -343,7 +356,7 @@ MathJax.Hub.Config({
    • 14
    • 15
    • ...
      • -
    • 63
      • +
    • 65
    • »
    diff --git a/doc/pub/week38/html/._week38-bs006.html b/doc/pub/week38/html/._week38-bs006.html index 039379297..90e9dab96 100644 --- a/doc/pub/week38/html/._week38-bs006.html +++ b/doc/pub/week38/html/._week38-bs006.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -306,121 +320,26 @@ MathJax.Hub.Config({

     

     

     

    - -

    Code Example for Cross-validation and \( k \)-fold Cross-validation

    + +

    Classification problems

    -

    The code here uses Ridge regression with cross-validation (CV) resampling and \( k \)-fold CV in order to fit a specific polynomial.

    - - -
    -
    -
    -
    -
    - 
    import numpy as np
-import matplotlib.pyplot as plt
-from sklearn.model_selection import KFold
-from sklearn.linear_model import Ridge
-from sklearn.model_selection import cross_val_score
-from sklearn.preprocessing import PolynomialFeatures
-
-# A seed just to ensure that the random numbers are the same for every run.
-# Useful for eventual debugging.
-np.random.seed(3155)
-
-# Generate the data.
-nsamples = 100
-x = np.random.randn(nsamples)
-y = 3*x**2 + np.random.randn(nsamples)
-
-## Cross-validation on Ridge regression using KFold only
-
-# Decide degree on polynomial to fit
-poly = PolynomialFeatures(degree = 6)
-
-# Decide which values of lambda to use
-nlambdas = 500
-lambdas = np.logspace(-3, 5, nlambdas)
-
-# Initialize a KFold instance
-k = 5
-kfold = KFold(n_splits = k)
-
-# Perform the cross-validation to estimate MSE
-scores_KFold = np.zeros((nlambdas, k))
-
-i = 0
-for lmb in lambdas:
-    ridge = Ridge(alpha = lmb)
-    j = 0
-    for train_inds, test_inds in kfold.split(x):
-        xtrain = x[train_inds]
-        ytrain = y[train_inds]
-
-        xtest = x[test_inds]
-        ytest = y[test_inds]
-
-        Xtrain = poly.fit_transform(xtrain[:, np.newaxis])
-        ridge.fit(Xtrain, ytrain[:, np.newaxis])
-
-        Xtest = poly.fit_transform(xtest[:, np.newaxis])
-        ypred = ridge.predict(Xtest)
-
-        scores_KFold[i,j] = np.sum((ypred - ytest[:, np.newaxis])**2)/np.size(ypred)
-
-        j += 1
-    i += 1
-
-
-estimated_mse_KFold = np.mean(scores_KFold, axis = 1)
-
-## Cross-validation using cross_val_score from sklearn along with KFold
-
-# kfold is an instance initialized above as:
-# kfold = KFold(n_splits = k)
-
-estimated_mse_sklearn = np.zeros(nlambdas)
-i = 0
-for lmb in lambdas:
-    ridge = Ridge(alpha = lmb)
-
-    X = poly.fit_transform(x[:, np.newaxis])
-    estimated_mse_folds = cross_val_score(ridge, X, y[:, np.newaxis], scoring='neg_mean_squared_error', cv=kfold)
-
-    # cross_val_score return an array containing the estimated negative mse for every fold.
-    # we have to the the mean of every array in order to get an estimate of the mse of the model
-    estimated_mse_sklearn[i] = np.mean(-estimated_mse_folds)
-
-    i += 1
-
-## Plot and compare the slightly different ways to perform cross-validation
-
-plt.figure()
-
-plt.plot(np.log10(lambdas), estimated_mse_sklearn, label = 'cross_val_score')
-plt.plot(np.log10(lambdas), estimated_mse_KFold, 'r--', label = 'KFold')
-
-plt.xlabel('log10(lambda)')
-plt.ylabel('mse')
-
-plt.legend()
-
-plt.show()
-
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    +

    Classification problems, however, are concerned with outcomes taking +the form of discrete variables (i.e. categories). We may for example, +on the basis of DNA sequencing for a number of patients, like to find +out which mutations are important for a certain disease; or based on +scans of various patients' brains, figure out if there is a tumor or +not; or given a specific physical system, we'd like to identify its +state, say whether it is an ordered or disordered system (typical +situation in solid state physics); or classify the status of a +patient, whether she/he has a stroke or not and many other similar +situations. +

    +

    The most common situation we encounter when we apply logistic +regression is that of two possible outcomes, normally denoted as a +binary outcome, true or false, positive or negative, success or +failure etc. +

    @@ -443,7 +362,7 @@ plt.show()

  • 15
  • 16
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs007.html b/doc/pub/week38/html/._week38-bs007.html index 266d8a8e9..cf2dca7e9 100644 --- a/doc/pub/week38/html/._week38-bs007.html +++ b/doc/pub/week38/html/._week38-bs007.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,7 +321,23 @@ MathJax.Hub.Config({

     

     

     

    -

    Material for lecture Thursday September 21

    +

    Optimization and Deep learning

    + +

    Logistic regression will also serve as our stepping stone towards +neural network algorithms and supervised deep learning. For logistic +learning, the minimization of the cost function leads to a non-linear +equation in the parameters \( \boldsymbol{\beta} \). The optimization of the +problem calls therefore for minimization algorithms. This forms the +bottle neck of all machine learning algorithms, namely how to find +reliable minima of a multi-variable function. This leads us to the +family of gradient descent methods. The latter are the working horses +of basically all modern machine learning algorithms. +

    + +

    We note also that many of the topics discussed here on logistic +regression are also commonly used in modern supervised Deep Learning +models, as we will see later. +

    @@ -331,7 +361,7 @@ MathJax.Hub.Config({

  • 16
  • 17
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs008.html b/doc/pub/week38/html/._week38-bs008.html index df8419c98..1686433ff 100644 --- a/doc/pub/week38/html/._week38-bs008.html +++ b/doc/pub/week38/html/._week38-bs008.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,21 +321,31 @@ MathJax.Hub.Config({

     

     

     

    -

    Logistic Regression

    +

    Basics

    -

    In linear regression our main interest was centered on learning the -coefficients of a functional fit (say a polynomial) in order to be -able to predict the response of a continuous variable on some unseen -data. The fit to the continuous variable \( y_i \) is based on some -independent variables \( \boldsymbol{x}_i \). Linear regression resulted in -analytical expressions for standard ordinary Least Squares or Ridge -regression (in terms of matrices to invert) for several quantities, -ranging from the variance and thereby the confidence intervals of the -parameters \( \boldsymbol{\beta} \) to the mean squared error. If we can invert -the product of the design matrices, linear regression gives then a -simple recipe for fitting our data. +

    We consider the case where the dependent variables, also called the +responses or the outcomes, \( y_i \) are discrete and only take values +from \( k=0,\dots,K-1 \) (i.e. \( K \) classes).

    +

    The goal is to predict the +output classes from the design matrix \( \boldsymbol{X}\in\mathbb{R}^{n\times p} \) +made of \( n \) samples, each of which carries \( p \) features or predictors. The +primary goal is to identify the classes to which new unseen samples +belong. +

    + +

    Let us specialize to the case of two classes only, with outputs +\( y_i=0 \) and \( y_i=1 \). Our outcomes could represent the status of a +credit card user that could default or not on her/his credit card +debt. That is +

    + +$$ +y_i = \begin{bmatrix} 0 & \mathrm{no}\\ 1 & \mathrm{yes} \end{bmatrix}. +$$ + +

      @@ -345,7 +369,7 @@ simple recipe for fitting our data.
    • 17
    • 18
    • ...
      • -
    • 63
      • +
    • 65
    • »
    diff --git a/doc/pub/week38/html/._week38-bs009.html b/doc/pub/week38/html/._week38-bs009.html index 1e4937f49..df145a26e 100644 --- a/doc/pub/week38/html/._week38-bs009.html +++ b/doc/pub/week38/html/._week38-bs009.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -306,25 +320,27 @@ MathJax.Hub.Config({

     

     

     

    - -

    Classification problems

    + +

    Linear classifier

    -

    Classification problems, however, are concerned with outcomes taking -the form of discrete variables (i.e. categories). We may for example, -on the basis of DNA sequencing for a number of patients, like to find -out which mutations are important for a certain disease; or based on -scans of various patients' brains, figure out if there is a tumor or -not; or given a specific physical system, we'd like to identify its -state, say whether it is an ordered or disordered system (typical -situation in solid state physics); or classify the status of a -patient, whether she/he has a stroke or not and many other similar -situations. +

    Before moving to the logistic model, let us try to use our linear +regression model to classify these two outcomes. We could for example +fit a linear model to the default case if \( y_i > 0.5 \) and the no +default case \( y_i \leq 0.5 \).

    -

    The most common situation we encounter when we apply logistic -regression is that of two possible outcomes, normally denoted as a -binary outcome, true or false, positive or negative, success or -failure etc. +

    We would then have our +weighted linear combination, namely +

    +$$ +\begin{equation} +\boldsymbol{y} = \boldsymbol{X}^T\boldsymbol{\beta} + \boldsymbol{\epsilon}, +\tag{1} +\end{equation} +$$ + +

    where \( \boldsymbol{y} \) is a vector representing the possible outcomes, \( \boldsymbol{X} \) is our +\( n\times p \) design matrix and \( \boldsymbol{\beta} \) represents our estimators/predictors.

    @@ -351,7 +367,7 @@ failure etc.

  • 18
  • 19
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs010.html b/doc/pub/week38/html/._week38-bs010.html index dcc95f7a8..498cca27a 100644 --- a/doc/pub/week38/html/._week38-bs010.html +++ b/doc/pub/week38/html/._week38-bs010.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,22 +321,26 @@ MathJax.Hub.Config({

     

     

     

    -

    Optimization and Deep learning

    +

    Some selected properties

    -

    Logistic regression will also serve as our stepping stone towards -neural network algorithms and supervised deep learning. For logistic -learning, the minimization of the cost function leads to a non-linear -equation in the parameters \( \boldsymbol{\beta} \). The optimization of the -problem calls therefore for minimization algorithms. This forms the -bottle neck of all machine learning algorithms, namely how to find -reliable minima of a multi-variable function. This leads us to the -family of gradient descent methods. The latter are the working horses -of basically all modern machine learning algorithms. +

    The main problem with our function is that it takes values on the +entire real axis. In the case of logistic regression, however, the +labels \( y_i \) are discrete variables. A typical example is the credit +card data discussed below here, where we can set the state of +defaulting the debt to \( y_i=1 \) and not to \( y_i=0 \) for one the persons +in the data set (see the full example below).

    -

    We note also that many of the topics discussed here on logistic -regression are also commonly used in modern supervised Deep Learning -models, as we will see later. +

    One simple way to get a discrete output is to have sign +functions that map the output of a linear regressor to values \( \{0,1\} \), +\( f(s_i)=sign(s_i)=1 \) if \( s_i\ge 0 \) and 0 if otherwise. +We will encounter this model in our first demonstration of neural networks. +

    + +

    Historically it is called the perceptron model in the machine learning +literature. This model is extremely simple. However, in many cases it is more +favorable to use a ``soft" classifier that outputs +the probability of a given category. This leads us to the logistic function.

    @@ -350,7 +368,7 @@ models, as we will see later.

  • 19
  • 20
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs011.html b/doc/pub/week38/html/._week38-bs011.html index de4d3f35e..5076b002f 100644 --- a/doc/pub/week38/html/._week38-bs011.html +++ b/doc/pub/week38/html/._week38-bs011.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -306,30 +320,86 @@ MathJax.Hub.Config({

     

     

     

    - -

    Basics

    + +

    Simple example

    -

    We consider the case where the dependent variables, also called the -responses or the outcomes, \( y_i \) are discrete and only take values -from \( k=0,\dots,K-1 \) (i.e. \( K \) classes). -

    +

    The following example on data for coronary heart disease (CHD) as function of age may serve as an illustration. In the code here we read and plot whether a person has had CHD (output = 1) or not (output = 0). This ouput is plotted the person's against age. Clearly, the figure shows that attempting to make a standard linear regression fit may not be very meaningful.

    -

    The goal is to predict the -output classes from the design matrix \( \boldsymbol{X}\in\mathbb{R}^{n\times p} \) -made of \( n \) samples, each of which carries \( p \) features or predictors. The -primary goal is to identify the classes to which new unseen samples -belong. -

    -

    Let us specialize to the case of two classes only, with outputs -\( y_i=0 \) and \( y_i=1 \). Our outcomes could represent the status of a -credit card user that could default or not on her/his credit card -debt. That is -

    + +
    +
    +
    +
    +
    + 
    # Common imports
+import os
+import numpy as np
+import pandas as pd
+import matplotlib.pyplot as plt
+from sklearn.linear_model import LinearRegression, Ridge, Lasso
+from sklearn.model_selection import train_test_split
+from sklearn.utils import resample
+from sklearn.metrics import mean_squared_error
+from IPython.display import display
+from pylab import plt, mpl
+plt.style.use('seaborn')
+mpl.rcParams['font.family'] = 'serif'
 
-$$
-y_i = \begin{bmatrix} 0 & \mathrm{no}\\  1 & \mathrm{yes} \end{bmatrix}.
-$$
+# 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("chddata.csv"),'r')
+
+# Read the chd data as  csv file and organize the data into arrays with age group, age, and chd
+chd = pd.read_csv(infile, names=('ID', 'Age', 'Agegroup', 'CHD'))
+chd.columns = ['ID', 'Age', 'Agegroup', 'CHD']
+output = chd['CHD']
+age = chd['Age']
+agegroup = chd['Agegroup']
+numberID  = chd['ID'] 
+display(chd)
+
+plt.scatter(age, output, marker='o')
+plt.axis([18,70.0,-0.1, 1.2])
+plt.xlabel(r'Age')
+plt.ylabel(r'CHD')
+plt.title(r'Age distribution and Coronary heart disease')
+plt.show()
+
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +

    @@ -357,7 +427,7 @@ $$

  • 20
  • 21
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs012.html b/doc/pub/week38/html/._week38-bs012.html index 9463f2bc4..2b9f5b825 100644 --- a/doc/pub/week38/html/._week38-bs012.html +++ b/doc/pub/week38/html/._week38-bs012.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,26 +321,57 @@ MathJax.Hub.Config({

     

     

     

    -

    Linear classifier

    +

    Plotting the mean value for each group

    -

    Before moving to the logistic model, let us try to use our linear -regression model to classify these two outcomes. We could for example -fit a linear model to the default case if \( y_i > 0.5 \) and the no -default case \( y_i \leq 0.5 \). -

    +

    What we could attempt however is to plot the mean value for each group.

    -

    We would then have our -weighted linear combination, namely + + +

    +
    +
    +
    +
    + 
    agegroupmean = np.array([0.1, 0.133, 0.250, 0.333, 0.462, 0.625, 0.765, 0.800])
+group = np.array([1, 2, 3, 4, 5, 6, 7, 8])
+plt.plot(group, agegroupmean, "r-")
+plt.axis([0,9,0, 1.0])
+plt.xlabel(r'Age group')
+plt.ylabel(r'CHD mean values')
+plt.title(r'Mean values for each age group')
+plt.show()
+
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    + +

    We are now trying to find a function \( f(y\vert x) \), that is a function which gives us an expected value for the output \( y \) with a given input \( x \). +In standard linear regression with a linear dependence on \( x \), we would write this in terms of our model

    $$ -\begin{equation} -\boldsymbol{y} = \boldsymbol{X}^T\boldsymbol{\beta} + \boldsymbol{\epsilon}, -\tag{1} -\end{equation} +f(y_i\vert x_i)=\beta_0+\beta_1 x_i. $$ -

    where \( \boldsymbol{y} \) is a vector representing the possible outcomes, \( \boldsymbol{X} \) is our -\( n\times p \) design matrix and \( \boldsymbol{\beta} \) represents our estimators/predictors. +

    This expression implies however that \( f(y_i\vert x_i) \) could take any +value from minus infinity to plus infinity. If we however let +\( f(y\vert y) \) be represented by the mean value, the above example +shows us that we can constrain the function to take values between +zero and one, that is we have \( 0 \le f(y_i\vert x_i) \le 1 \). Looking +at our last curve we see also that it has an S-shaped form. This leads +us to a very popular model for the function \( f \), namely the so-called +Sigmoid function or logistic model. We will consider this function as +representing the probability for finding a value of \( y_i \) with a given +\( x_i \).

    @@ -354,7 +399,7 @@ $$

  • 21
  • 22
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs013.html b/doc/pub/week38/html/._week38-bs013.html index d22733103..51defc67f 100644 --- a/doc/pub/week38/html/._week38-bs013.html +++ b/doc/pub/week38/html/._week38-bs013.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,27 +321,25 @@ MathJax.Hub.Config({

     

     

     

    -

    Some selected properties

    +

    The logistic function

    -

    The main problem with our function is that it takes values on the -entire real axis. In the case of logistic regression, however, the -labels \( y_i \) are discrete variables. A typical example is the credit -card data discussed below here, where we can set the state of -defaulting the debt to \( y_i=1 \) and not to \( y_i=0 \) for one the persons -in the data set (see the full example below). +

    Another widely studied model, is the so-called +perceptron model, which is an example of a "hard classification" model. We +will encounter this model when we discuss neural networks as +well. Each datapoint is deterministically assigned to a category (i.e +\( y_i=0 \) or \( y_i=1 \)). In many cases, and the coronary heart disease data forms one of many such examples, it is favorable to have a "soft" +classifier that outputs the probability of a given category rather +than a single value. For example, given \( x_i \), the classifier +outputs the probability of being in a category \( k \). Logistic regression +is the most common example of a so-called soft classifier. In logistic +regression, the probability that a data point \( x_i \) +belongs to a category \( y_i=\{0,1\} \) is given by the so-called logit function (or Sigmoid) which is meant to represent the likelihood for a given event,

    +$$ +p(t) = \frac{1}{1+\mathrm \exp{-t}}=\frac{\exp{t}}{1+\mathrm \exp{t}}. +$$ -

    One simple way to get a discrete output is to have sign -functions that map the output of a linear regressor to values \( \{0,1\} \), -\( f(s_i)=sign(s_i)=1 \) if \( s_i\ge 0 \) and 0 if otherwise. -We will encounter this model in our first demonstration of neural networks. -

    - -

    Historically it is called the perceptron model in the machine learning -literature. This model is extremely simple. However, in many cases it is more -favorable to use a ``soft" classifier that outputs -the probability of a given category. This leads us to the logistic function. -

    +

    Note that \( 1-p(t)= p(-t) \).

    @@ -354,7 +366,7 @@ the probability of a given category. This leads us to the logistic function.

  • 22
  • 23
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs014.html b/doc/pub/week38/html/._week38-bs014.html index 624b7521e..f28b7d3f9 100644 --- a/doc/pub/week38/html/._week38-bs014.html +++ b/doc/pub/week38/html/._week38-bs014.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,9 +321,9 @@ MathJax.Hub.Config({

     

     

     

    -

    Simple example

    +

    Examples of likelihood functions used in logistic regression and nueral networks

    -

    The following example on data for coronary heart disease (CHD) as function of age may serve as an illustration. In the code here we read and plot whether a person has had CHD (output = 1) or not (output = 0). This ouput is plotted the person's against age. Clearly, the figure shows that attempting to make a standard linear regression fit may not be very meaningful.

    +

    The following code plots the logistic function, the step function and other functions we will encounter from here and on.

    @@ -318,59 +332,59 @@ MathJax.Hub.Config({
    - 
    # Common imports
-import os
-import numpy as np
-import pandas as pd
+  
    """The sigmoid function (or the logistic curve) is a
+function that takes any real number, z, and outputs a number (0,1).
+It is useful in neural networks for assigning weights on a relative scale.
+The value z is the weighted sum of parameters involved in the learning algorithm."""
+
+import numpy
 import matplotlib.pyplot as plt
-from sklearn.linear_model import LinearRegression, Ridge, Lasso
-from sklearn.model_selection import train_test_split
-from sklearn.utils import resample
-from sklearn.metrics import mean_squared_error
-from IPython.display import display
-from pylab import plt, mpl
-plt.style.use('seaborn')
-mpl.rcParams['font.family'] = 'serif'
+import math as mt
 
-# Where to save the figures and data files
-PROJECT_ROOT_DIR = "Results"
-FIGURE_ID = "Results/FigureFiles"
-DATA_ID = "DataFiles/"
+z = numpy.arange(-5, 5, .1)
+sigma_fn = numpy.vectorize(lambda z: 1/(1+numpy.exp(-z)))
+sigma = sigma_fn(z)
 
-if not os.path.exists(PROJECT_ROOT_DIR):
-    os.mkdir(PROJECT_ROOT_DIR)
+fig = plt.figure()
+ax = fig.add_subplot(111)
+ax.plot(z, sigma)
+ax.set_ylim([-0.1, 1.1])
+ax.set_xlim([-5,5])
+ax.grid(True)
+ax.set_xlabel('z')
+ax.set_title('sigmoid function')
 
-if not os.path.exists(FIGURE_ID):
-    os.makedirs(FIGURE_ID)
+plt.show()
 
-if not os.path.exists(DATA_ID):
-    os.makedirs(DATA_ID)
+"""Step Function"""
+z = numpy.arange(-5, 5, .02)
+step_fn = numpy.vectorize(lambda z: 1.0 if z >= 0.0 else 0.0)
+step = step_fn(z)
 
-def image_path(fig_id):
-    return os.path.join(FIGURE_ID, fig_id)
+fig = plt.figure()
+ax = fig.add_subplot(111)
+ax.plot(z, step)
+ax.set_ylim([-0.5, 1.5])
+ax.set_xlim([-5,5])
+ax.grid(True)
+ax.set_xlabel('z')
+ax.set_title('step function')
 
-def data_path(dat_id):
-    return os.path.join(DATA_ID, dat_id)
+plt.show()
 
-def save_fig(fig_id):
-    plt.savefig(image_path(fig_id) + ".png", format='png')
+"""tanh Function"""
+z = numpy.arange(-2*mt.pi, 2*mt.pi, 0.1)
+t = numpy.tanh(z)
 
-infile = open(data_path("chddata.csv"),'r')
+fig = plt.figure()
+ax = fig.add_subplot(111)
+ax.plot(z, t)
+ax.set_ylim([-1.0, 1.0])
+ax.set_xlim([-2*mt.pi,2*mt.pi])
+ax.grid(True)
+ax.set_xlabel('z')
+ax.set_title('tanh function')
 
-# Read the chd data as  csv file and organize the data into arrays with age group, age, and chd
-chd = pd.read_csv(infile, names=('ID', 'Age', 'Agegroup', 'CHD'))
-chd.columns = ['ID', 'Age', 'Agegroup', 'CHD']
-output = chd['CHD']
-age = chd['Age']
-agegroup = chd['Agegroup']
-numberID  = chd['ID'] 
-display(chd)
-
-plt.scatter(age, output, marker='o')
-plt.axis([18,70.0,-0.1, 1.2])
-plt.xlabel(r'Age')
-plt.ylabel(r'CHD')
-plt.title(r'Age distribution and Coronary heart disease')
 plt.show()
    @@ -413,7 +427,7 @@ plt.show()
  • 23
  • 24
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs015.html b/doc/pub/week38/html/._week38-bs015.html index 59ca0123b..05bd553ca 100644 --- a/doc/pub/week38/html/._week38-bs015.html +++ b/doc/pub/week38/html/._week38-bs015.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,58 +321,23 @@ MathJax.Hub.Config({

     

     

     

    -

    Plotting the mean value for each group

    +

    Two parameters

    -

    What we could attempt however is to plot the mean value for each group.

    - - - -
    -
    -
    -
    -
    - 
    agegroupmean = np.array([0.1, 0.133, 0.250, 0.333, 0.462, 0.625, 0.765, 0.800])
-group = np.array([1, 2, 3, 4, 5, 6, 7, 8])
-plt.plot(group, agegroupmean, "r-")
-plt.axis([0,9,0, 1.0])
-plt.xlabel(r'Age group')
-plt.ylabel(r'CHD mean values')
-plt.title(r'Mean values for each age group')
-plt.show()
-
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    - -

    We are now trying to find a function \( f(y\vert x) \), that is a function which gives us an expected value for the output \( y \) with a given input \( x \). -In standard linear regression with a linear dependence on \( x \), we would write this in terms of our model -

    +

    We assume now that we have two classes with \( y_i \) either \( 0 \) or \( 1 \). Furthermore we assume also that we have only two parameters \( \beta \) in our fitting of the Sigmoid function, that is we define probabilities

    $$ -f(y_i\vert x_i)=\beta_0+\beta_1 x_i. +\begin{align*} +p(y_i=1|x_i,\boldsymbol{\beta}) &= \frac{\exp{(\beta_0+\beta_1x_i)}}{1+\exp{(\beta_0+\beta_1x_i)}},\nonumber\\ +p(y_i=0|x_i,\boldsymbol{\beta}) &= 1 - p(y_i=1|x_i,\boldsymbol{\beta}), +\end{align*} +$$ + +

    where \( \boldsymbol{\beta} \) are the weights we wish to extract from data, in our case \( \beta_0 \) and \( \beta_1 \).

    + +

    Note that we used

    +$$ +p(y_i=0\vert x_i, \boldsymbol{\beta}) = 1-p(y_i=1\vert x_i, \boldsymbol{\beta}). $$ -

    This expression implies however that \( f(y_i\vert x_i) \) could take any -value from minus infinity to plus infinity. If we however let -\( f(y\vert y) \) be represented by the mean value, the above example -shows us that we can constrain the function to take values between -zero and one, that is we have \( 0 \le f(y_i\vert x_i) \le 1 \). Looking -at our last curve we see also that it has an S-shaped form. This leads -us to a very popular model for the function \( f \), namely the so-called -Sigmoid function or logistic model. We will consider this function as -representing the probability for finding a value of \( y_i \) with a given -\( x_i \). -

    @@ -385,7 +364,7 @@ representing the probability for finding a value of \( y_i \) with a given

  • 24
  • 25
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs016.html b/doc/pub/week38/html/._week38-bs016.html index fe4204f7c..a1ee25d8b 100644 --- a/doc/pub/week38/html/._week38-bs016.html +++ b/doc/pub/week38/html/._week38-bs016.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -306,26 +320,27 @@ MathJax.Hub.Config({

     

     

     

    - -

    The logistic function

    + +

    Maximum likelihood

    -

    Another widely studied model, is the so-called -perceptron model, which is an example of a "hard classification" model. We -will encounter this model when we discuss neural networks as -well. Each datapoint is deterministically assigned to a category (i.e -\( y_i=0 \) or \( y_i=1 \)). In many cases, and the coronary heart disease data forms one of many such examples, it is favorable to have a "soft" -classifier that outputs the probability of a given category rather -than a single value. For example, given \( x_i \), the classifier -outputs the probability of being in a category \( k \). Logistic regression -is the most common example of a so-called soft classifier. In logistic -regression, the probability that a data point \( x_i \) -belongs to a category \( y_i=\{0,1\} \) is given by the so-called logit function (or Sigmoid) which is meant to represent the likelihood for a given event, +

    In order to define the total likelihood for all possible outcomes from a +dataset \( \mathcal{D}=\{(y_i,x_i)\} \), with the binary labels +\( y_i\in\{0,1\} \) and where the data points are drawn independently, we use the so-called Maximum Likelihood Estimation (MLE) principle. +We aim thus at maximizing +the probability of seeing the observed data. We can then approximate the +likelihood in terms of the product of the individual probabilities of a specific outcome \( y_i \), that is

    $$ -p(t) = \frac{1}{1+\mathrm \exp{-t}}=\frac{\exp{t}}{1+\mathrm \exp{t}}. +\begin{align*} +P(\mathcal{D}|\boldsymbol{\beta})& = \prod_{i=1}^n \left[p(y_i=1|x_i,\boldsymbol{\beta})\right]^{y_i}\left[1-p(y_i=1|x_i,\boldsymbol{\beta}))\right]^{1-y_i}\nonumber \\ +\end{align*} +$$ + +

    from which we obtain the log-likelihood and our cost/loss function

    +$$ +\mathcal{C}(\boldsymbol{\beta}) = \sum_{i=1}^n \left( y_i\log{p(y_i=1|x_i,\boldsymbol{\beta})} + (1-y_i)\log\left[1-p(y_i=1|x_i,\boldsymbol{\beta}))\right]\right). $$ -

    Note that \( 1-p(t)= p(-t) \).

    @@ -352,7 +367,7 @@ $$

  • 25
  • 26
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs017.html b/doc/pub/week38/html/._week38-bs017.html index 3d83c1a4b..3bb860b88 100644 --- a/doc/pub/week38/html/._week38-bs017.html +++ b/doc/pub/week38/html/._week38-bs017.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,86 +321,23 @@ MathJax.Hub.Config({

     

     

     

    -

    Examples of likelihood functions used in logistic regression and nueral networks

    +

    The cost function rewritten

    -

    The following code plots the logistic function, the step function and other functions we will encounter from here and on.

    +

    Reordering the logarithms, we can rewrite the cost/loss function as

    +$$ +\mathcal{C}(\boldsymbol{\beta}) = \sum_{i=1}^n \left(y_i(\beta_0+\beta_1x_i) -\log{(1+\exp{(\beta_0+\beta_1x_i)})}\right). +$$ +

    The maximum likelihood estimator is defined as the set of parameters that maximize the log-likelihood where we maximize with respect to \( \beta \). +Since the cost (error) function is just the negative log-likelihood, for logistic regression we have that +

    +$$ +\mathcal{C}(\boldsymbol{\beta})=-\sum_{i=1}^n \left(y_i(\beta_0+\beta_1x_i) -\log{(1+\exp{(\beta_0+\beta_1x_i)})}\right). +$$ - -
    -
    -
    -
    -
    - 
    """The sigmoid function (or the logistic curve) is a
-function that takes any real number, z, and outputs a number (0,1).
-It is useful in neural networks for assigning weights on a relative scale.
-The value z is the weighted sum of parameters involved in the learning algorithm."""
-
-import numpy
-import matplotlib.pyplot as plt
-import math as mt
-
-z = numpy.arange(-5, 5, .1)
-sigma_fn = numpy.vectorize(lambda z: 1/(1+numpy.exp(-z)))
-sigma = sigma_fn(z)
-
-fig = plt.figure()
-ax = fig.add_subplot(111)
-ax.plot(z, sigma)
-ax.set_ylim([-0.1, 1.1])
-ax.set_xlim([-5,5])
-ax.grid(True)
-ax.set_xlabel('z')
-ax.set_title('sigmoid function')
-
-plt.show()
-
-"""Step Function"""
-z = numpy.arange(-5, 5, .02)
-step_fn = numpy.vectorize(lambda z: 1.0 if z >= 0.0 else 0.0)
-step = step_fn(z)
-
-fig = plt.figure()
-ax = fig.add_subplot(111)
-ax.plot(z, step)
-ax.set_ylim([-0.5, 1.5])
-ax.set_xlim([-5,5])
-ax.grid(True)
-ax.set_xlabel('z')
-ax.set_title('step function')
-
-plt.show()
-
-"""tanh Function"""
-z = numpy.arange(-2*mt.pi, 2*mt.pi, 0.1)
-t = numpy.tanh(z)
-
-fig = plt.figure()
-ax = fig.add_subplot(111)
-ax.plot(z, t)
-ax.set_ylim([-1.0, 1.0])
-ax.set_xlim([-2*mt.pi,2*mt.pi])
-ax.grid(True)
-ax.set_xlabel('z')
-ax.set_title('tanh function')
-
-plt.show()
-
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    - +

    This equation is known in statistics as the cross entropy. Finally, we note that just as in linear regression, +in practice we often supplement the cross-entropy with additional regularization terms, usually \( L_1 \) and \( L_2 \) regularization as we did for Ridge and Lasso regression. +

    @@ -413,7 +364,7 @@ plt.show()

  • 26
  • 27
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs018.html b/doc/pub/week38/html/._week38-bs018.html index e478791bf..f8bb5c812 100644 --- a/doc/pub/week38/html/._week38-bs018.html +++ b/doc/pub/week38/html/._week38-bs018.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,21 +321,23 @@ MathJax.Hub.Config({

     

     

     

    -

    Two parameters

    +

    Minimizing the cross entropy

    + +

    The cross entropy is a convex function of the weights \( \boldsymbol{\beta} \) and, +therefore, any local minimizer is a global minimizer. +

    + +

    Minimizing this +cost function with respect to the two parameters \( \beta_0 \) and \( \beta_1 \) we obtain +

    -

    We assume now that we have two classes with \( y_i \) either \( 0 \) or \( 1 \). Furthermore we assume also that we have only two parameters \( \beta \) in our fitting of the Sigmoid function, that is we define probabilities

    $$ -\begin{align*} -p(y_i=1|x_i,\boldsymbol{\beta}) &= \frac{\exp{(\beta_0+\beta_1x_i)}}{1+\exp{(\beta_0+\beta_1x_i)}},\nonumber\\ -p(y_i=0|x_i,\boldsymbol{\beta}) &= 1 - p(y_i=1|x_i,\boldsymbol{\beta}), -\end{align*} +\frac{\partial \mathcal{C}(\boldsymbol{\beta})}{\partial \beta_0} = -\sum_{i=1}^n \left(y_i -\frac{\exp{(\beta_0+\beta_1x_i)}}{1+\exp{(\beta_0+\beta_1x_i)}}\right), $$ -

    where \( \boldsymbol{\beta} \) are the weights we wish to extract from data, in our case \( \beta_0 \) and \( \beta_1 \).

    - -

    Note that we used

    +

    and

    $$ -p(y_i=0\vert x_i, \boldsymbol{\beta}) = 1-p(y_i=1\vert x_i, \boldsymbol{\beta}). +\frac{\partial \mathcal{C}(\boldsymbol{\beta})}{\partial \beta_1} = -\sum_{i=1}^n \left(y_ix_i -x_i\frac{\exp{(\beta_0+\beta_1x_i)}}{1+\exp{(\beta_0+\beta_1x_i)}}\right). $$ @@ -350,7 +366,7 @@ $$
  • 27
  • 28
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs019.html b/doc/pub/week38/html/._week38-bs019.html index 19f5497d5..b1e5b7367 100644 --- a/doc/pub/week38/html/._week38-bs019.html +++ b/doc/pub/week38/html/._week38-bs019.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -306,25 +320,25 @@ MathJax.Hub.Config({

     

     

     

    - -

    Maximum likelihood

    + +

    A more compact expression

    -

    In order to define the total likelihood for all possible outcomes from a -dataset \( \mathcal{D}=\{(y_i,x_i)\} \), with the binary labels -\( y_i\in\{0,1\} \) and where the data points are drawn independently, we use the so-called Maximum Likelihood Estimation (MLE) principle. -We aim thus at maximizing -the probability of seeing the observed data. We can then approximate the -likelihood in terms of the product of the individual probabilities of a specific outcome \( y_i \), that is +

    Let us now define a vector \( \boldsymbol{y} \) with \( n \) elements \( y_i \), an +\( n\times p \) matrix \( \boldsymbol{X} \) which contains the \( x_i \) values and a +vector \( \boldsymbol{p} \) of fitted probabilities \( p(y_i\vert x_i,\boldsymbol{\beta}) \). We can rewrite in a more compact form the first +derivative of cost function as

    + $$ -\begin{align*} -P(\mathcal{D}|\boldsymbol{\beta})& = \prod_{i=1}^n \left[p(y_i=1|x_i,\boldsymbol{\beta})\right]^{y_i}\left[1-p(y_i=1|x_i,\boldsymbol{\beta}))\right]^{1-y_i}\nonumber \\ -\end{align*} +\frac{\partial \mathcal{C}(\boldsymbol{\beta})}{\partial \boldsymbol{\beta}} = -\boldsymbol{X}^T\left(\boldsymbol{y}-\boldsymbol{p}\right). $$ -

    from which we obtain the log-likelihood and our cost/loss function

    +

    If we in addition define a diagonal matrix \( \boldsymbol{W} \) with elements +\( p(y_i\vert x_i,\boldsymbol{\beta})(1-p(y_i\vert x_i,\boldsymbol{\beta}) \), we can obtain a compact expression of the second derivative as +

    + $$ -\mathcal{C}(\boldsymbol{\beta}) = \sum_{i=1}^n \left( y_i\log{p(y_i=1|x_i,\boldsymbol{\beta})} + (1-y_i)\log\left[1-p(y_i=1|x_i,\boldsymbol{\beta}))\right]\right). +\frac{\partial^2 \mathcal{C}(\boldsymbol{\beta})}{\partial \boldsymbol{\beta}\partial \boldsymbol{\beta}^T} = \boldsymbol{X}^T\boldsymbol{W}\boldsymbol{X}. $$ @@ -353,7 +367,7 @@ $$
  • 28
  • 29
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs020.html b/doc/pub/week38/html/._week38-bs020.html index 5fcde5396..8b364addb 100644 --- a/doc/pub/week38/html/._week38-bs020.html +++ b/doc/pub/week38/html/._week38-bs020.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,23 +321,18 @@ MathJax.Hub.Config({

     

     

     

    -

    The cost function rewritten

    +

    Extending to more predictors

    -

    Reordering the logarithms, we can rewrite the cost/loss function as

    +

    Within a binary classification problem, we can easily expand our model to include multiple predictors. Our ratio between likelihoods is then with \( p \) predictors

    $$ -\mathcal{C}(\boldsymbol{\beta}) = \sum_{i=1}^n \left(y_i(\beta_0+\beta_1x_i) -\log{(1+\exp{(\beta_0+\beta_1x_i)})}\right). +\log{ \frac{p(\boldsymbol{\beta}\boldsymbol{x})}{1-p(\boldsymbol{\beta}\boldsymbol{x})}} = \beta_0+\beta_1x_1+\beta_2x_2+\dots+\beta_px_p. $$ -

    The maximum likelihood estimator is defined as the set of parameters that maximize the log-likelihood where we maximize with respect to \( \beta \). -Since the cost (error) function is just the negative log-likelihood, for logistic regression we have that -

    +

    Here we defined \( \boldsymbol{x}=[1,x_1,x_2,\dots,x_p] \) and \( \boldsymbol{\beta}=[\beta_0, \beta_1, \dots, \beta_p] \) leading to

    $$ -\mathcal{C}(\boldsymbol{\beta})=-\sum_{i=1}^n \left(y_i(\beta_0+\beta_1x_i) -\log{(1+\exp{(\beta_0+\beta_1x_i)})}\right). +p(\boldsymbol{\beta}\boldsymbol{x})=\frac{ \exp{(\beta_0+\beta_1x_1+\beta_2x_2+\dots+\beta_px_p)}}{1+\exp{(\beta_0+\beta_1x_1+\beta_2x_2+\dots+\beta_px_p)}}. $$ -

    This equation is known in statistics as the cross entropy. Finally, we note that just as in linear regression, -in practice we often supplement the cross-entropy with additional regularization terms, usually \( L_1 \) and \( L_2 \) regularization as we did for Ridge and Lasso regression. -

    @@ -350,7 +359,7 @@ in practice we often supplement the cross-entropy with additional regularization

  • 29
  • 30
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs021.html b/doc/pub/week38/html/._week38-bs021.html index 1757278b3..bcaec474a 100644 --- a/doc/pub/week38/html/._week38-bs021.html +++ b/doc/pub/week38/html/._week38-bs021.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,25 +321,30 @@ MathJax.Hub.Config({

     

     

     

    -

    Minimizing the cross entropy

    +

    Including more classes

    -

    The cross entropy is a convex function of the weights \( \boldsymbol{\beta} \) and, -therefore, any local minimizer is a global minimizer. -

    - -

    Minimizing this -cost function with respect to the two parameters \( \beta_0 \) and \( \beta_1 \) we obtain +

    Till now we have mainly focused on two classes, the so-called binary +system. Suppose we wish to extend to \( K \) classes. Let us for the sake +of simplicity assume we have only two predictors. We have then following model

    $$ -\frac{\partial \mathcal{C}(\boldsymbol{\beta})}{\partial \beta_0} = -\sum_{i=1}^n \left(y_i -\frac{\exp{(\beta_0+\beta_1x_i)}}{1+\exp{(\beta_0+\beta_1x_i)}}\right), +\log{\frac{p(C=1\vert x)}{p(K\vert x)}} = \beta_{10}+\beta_{11}x_1, $$

    and

    $$ -\frac{\partial \mathcal{C}(\boldsymbol{\beta})}{\partial \beta_1} = -\sum_{i=1}^n \left(y_ix_i -x_i\frac{\exp{(\beta_0+\beta_1x_i)}}{1+\exp{(\beta_0+\beta_1x_i)}}\right). +\log{\frac{p(C=2\vert x)}{p(K\vert x)}} = \beta_{20}+\beta_{21}x_1, $$ +

    and so on till the class \( C=K-1 \) class

    +$$ +\log{\frac{p(C=K-1\vert x)}{p(K\vert x)}} = \beta_{(K-1)0}+\beta_{(K-1)1}x_1, +$$ + +

    and the model is specified in term of \( K-1 \) so-called log-odds or +logit transformations. +

    @@ -352,7 +371,7 @@ $$

  • 30
  • 31
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs022.html b/doc/pub/week38/html/._week38-bs022.html index 3171e83b9..f2df6ddaa 100644 --- a/doc/pub/week38/html/._week38-bs022.html +++ b/doc/pub/week38/html/._week38-bs022.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,26 +321,42 @@ MathJax.Hub.Config({

     

     

     

    -

    A more compact expression

    +

    More classes

    -

    Let us now define a vector \( \boldsymbol{y} \) with \( n \) elements \( y_i \), an -\( n\times p \) matrix \( \boldsymbol{X} \) which contains the \( x_i \) values and a -vector \( \boldsymbol{p} \) of fitted probabilities \( p(y_i\vert x_i,\boldsymbol{\beta}) \). We can rewrite in a more compact form the first -derivative of cost function as +

    In our discussion of neural networks we will encounter the above again +in terms of a slightly modified function, the so-called Softmax function. +

    + +

    The softmax function is used in various multiclass classification +methods, such as multinomial logistic regression (also known as +softmax regression), multiclass linear discriminant analysis, naive +Bayes classifiers, and artificial neural networks. Specifically, in +multinomial logistic regression and linear discriminant analysis, the +input to the function is the result of \( K \) distinct linear functions, +and the predicted probability for the \( k \)-th class given a sample +vector \( \boldsymbol{x} \) and a weighting vector \( \boldsymbol{\beta} \) is (with two +predictors):

    $$ -\frac{\partial \mathcal{C}(\boldsymbol{\beta})}{\partial \boldsymbol{\beta}} = -\boldsymbol{X}^T\left(\boldsymbol{y}-\boldsymbol{p}\right). +p(C=k\vert \mathbf {x} )=\frac{\exp{(\beta_{k0}+\beta_{k1}x_1)}}{1+\sum_{l=1}^{K-1}\exp{(\beta_{l0}+\beta_{l1}x_1)}}. $$ -

    If we in addition define a diagonal matrix \( \boldsymbol{W} \) with elements -\( p(y_i\vert x_i,\boldsymbol{\beta})(1-p(y_i\vert x_i,\boldsymbol{\beta}) \), we can obtain a compact expression of the second derivative as +

    It is easy to extend to more predictors. The final class is

    +$$ +p(C=K\vert \mathbf {x} )=\frac{1}{1+\sum_{l=1}^{K-1}\exp{(\beta_{l0}+\beta_{l1}x_1)}}, +$$ + +

    and they sum to one. Our earlier discussions were all specialized to +the case with two classes only. It is easy to see from the above that +what we derived earlier is compatible with these equations.

    -$$ -\frac{\partial^2 \mathcal{C}(\boldsymbol{\beta})}{\partial \boldsymbol{\beta}\partial \boldsymbol{\beta}^T} = \boldsymbol{X}^T\boldsymbol{W}\boldsymbol{X}. -$$ - +

    To find the optimal parameters we would typically use a gradient +descent method. Newton's method and gradient descent methods are +discussed in the material on optimization +methods. +

    @@ -353,7 +383,7 @@ $$

  • 31
  • 32
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs023.html b/doc/pub/week38/html/._week38-bs023.html index 725f7d336..9fb3de602 100644 --- a/doc/pub/week38/html/._week38-bs023.html +++ b/doc/pub/week38/html/._week38-bs023.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,18 +321,7 @@ MathJax.Hub.Config({

     

     

     

    -

    Extending to more predictors

    - -

    Within a binary classification problem, we can easily expand our model to include multiple predictors. Our ratio between likelihoods is then with \( p \) predictors

    -$$ -\log{ \frac{p(\boldsymbol{\beta}\boldsymbol{x})}{1-p(\boldsymbol{\beta}\boldsymbol{x})}} = \beta_0+\beta_1x_1+\beta_2x_2+\dots+\beta_px_p. -$$ - -

    Here we defined \( \boldsymbol{x}=[1,x_1,x_2,\dots,x_p] \) and \( \boldsymbol{\beta}=[\beta_0, \beta_1, \dots, \beta_p] \) leading to

    -$$ -p(\boldsymbol{\beta}\boldsymbol{x})=\frac{ \exp{(\beta_0+\beta_1x_1+\beta_2x_2+\dots+\beta_px_p)}}{1+\exp{(\beta_0+\beta_1x_1+\beta_2x_2+\dots+\beta_px_p)}}. -$$ - +

    Friday September 23

    @@ -345,7 +348,7 @@ $$

  • 32
  • 33
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs024.html b/doc/pub/week38/html/._week38-bs024.html index 0066cebcc..5d6b6a56e 100644 --- a/doc/pub/week38/html/._week38-bs024.html +++ b/doc/pub/week38/html/._week38-bs024.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,29 +321,85 @@ MathJax.Hub.Config({

     

     

     

    -

    Including more classes

    +

    Searching for Optimal Regularization Parameters \( \lambda \)

    -

    Till now we have mainly focused on two classes, the so-called binary -system. Suppose we wish to extend to \( K \) classes. Let us for the sake -of simplicity assume we have only two predictors. We have then following model +

    In project 1, when using Ridge and Lasso regression, we end up +searching for the optimal parameter \( \lambda \) which minimizes our +selected scores (MSE or \( R2 \) values for example). The brute force +approach, as discussed in the code here for Ridge regression, consists +in evaluating the MSE as function of different \( \lambda \) values. +Based on these calculations, one tries then to determine the value of the hyperparameter \( \lambda \) +which results in optimal scores (for example the smallest MSE or an \( R2=1 \)).

    -$$ -\log{\frac{p(C=1\vert x)}{p(K\vert x)}} = \beta_{10}+\beta_{11}x_1, -$$ + +
    +
    +
    +
    +
    + 
    import numpy as np
+import pandas as pd
+import matplotlib.pyplot as plt
+from sklearn.model_selection import train_test_split
+from sklearn import linear_model
 
-
    and 
    
-$$
-\log{\frac{p(C=2\vert x)}{p(K\vert x)}} = \beta_{20}+\beta_{21}x_1,
-$$
+def MSE(y_data,y_model):
+    n = np.size(y_model)
+    return np.sum((y_data-y_model)**2)/n
+# A seed just to ensure that the random numbers are the same for every run.
+# Useful for eventual debugging.
+np.random.seed(2021)
 
-
    and so on till the class \( C=K-1 \) class
    
-$$
-\log{\frac{p(C=K-1\vert x)}{p(K\vert x)}} = \beta_{(K-1)0}+\beta_{(K-1)1}x_1,
-$$
+n = 100
+x = np.random.rand(n)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.randn(n)
 
-
    and the model is specified in term of \( K-1 \) so-called log-odds or
-logit transformations.
+Maxpolydegree = 5
+X = np.zeros((n,Maxpolydegree-1))
+
+for degree in range(1,Maxpolydegree): #No intercept column
+    X[:,degree-1] = x**(degree)
+
+# We split the data in test and training data
+X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)
+
+# Decide which values of lambda to use
+nlambdas = 500
+MSERidgePredict = np.zeros(nlambdas)
+lambdas = np.logspace(-4, 2, nlambdas)
+for i in range(nlambdas):
+    lmb = lambdas[i]
+    RegRidge = linear_model.Ridge(lmb)
+    RegRidge.fit(X_train,y_train)
+    ypredictRidge = RegRidge.predict(X_test)
+    MSERidgePredict[i] = MSE(y_test,ypredictRidge)
+
+# Now plot the results
+plt.figure()
+plt.plot(np.log10(lambdas), MSERidgePredict, 'g--', label = 'MSE SL Ridge Test')
+plt.xlabel('log10(lambda)')
+plt.ylabel('MSE')
+plt.legend()
+plt.show()
+
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    + +

    Here we have performed a rather data greedy calculation as function of the regularization parameter \( \lambda \). There is no resampling here. The latter can easily be added by employing the function RidgeCV instead of just calling the Ridge function. For RidgeCV we need to pass the array of \( \lambda \) values. +By inspecting the figure we can in turn determine which is the optimal regularization parameter. +This becomes however less functional in the long run.

    @@ -357,7 +427,7 @@ $$

  • 33
  • 34
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs025.html b/doc/pub/week38/html/._week38-bs025.html index 7bcb642bb..b965cdc49 100644 --- a/doc/pub/week38/html/._week38-bs025.html +++ b/doc/pub/week38/html/._week38-bs025.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,42 +321,84 @@ MathJax.Hub.Config({

     

     

     

    -

    More classes

    + -

    In our discussion of neural networks we will encounter the above again -in terms of a slightly modified function, the so-called Softmax function. +

    An alternative is to use the so-called grid search functionality +included with the library Scikit-Learn, as demonstrated for the same +example here.

    -

    The softmax function is used in various multiclass classification -methods, such as multinomial logistic regression (also known as -softmax regression), multiclass linear discriminant analysis, naive -Bayes classifiers, and artificial neural networks. Specifically, in -multinomial logistic regression and linear discriminant analysis, the -input to the function is the result of \( K \) distinct linear functions, -and the predicted probability for the \( k \)-th class given a sample -vector \( \boldsymbol{x} \) and a weighting vector \( \boldsymbol{\beta} \) is (with two -predictors): + + +

    +
    +
    +
    +
    + 
    import numpy as np
+from sklearn.model_selection import train_test_split
+from sklearn.linear_model import Ridge
+from sklearn.model_selection import GridSearchCV
+
+def R2(y_data, y_model):
+    return 1 - np.sum((y_data - y_model) ** 2) / np.sum((y_data - np.mean(y_data)) ** 2)
+
+def MSE(y_data,y_model):
+    n = np.size(y_model)
+    return np.sum((y_data-y_model)**2)/n
+
+# A seed just to ensure that the random numbers are the same for every run.
+# Useful for eventual debugging.
+np.random.seed(2021)
+
+n = 100
+x = np.random.rand(n)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.randn(n)
+
+Maxpolydegree = 5
+X = np.zeros((n,Maxpolydegree-1))
+
+for degree in range(1,Maxpolydegree): #No intercept column
+    X[:,degree-1] = x**(degree)
+
+# We split the data in test and training data
+X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)
+
+# Decide which values of lambda to use
+nlambdas = 10
+lambdas = np.logspace(-4, 2, nlambdas)
+# create and fit a ridge regression model, testing each alpha
+model = Ridge()
+gridsearch = GridSearchCV(estimator=model, param_grid=dict(alpha=lambdas))
+gridsearch.fit(X_train, y_train)
+print(gridsearch)
+ypredictRidge = gridsearch.predict(X_test)
+# summarize the results of the grid search
+print(f"Best estimated lambda-value: {gridsearch.best_estimator_.alpha}")
+print(f"MSE score: {MSE(y_test,ypredictRidge)}")
+print(f"R2 score: {R2(y_test,ypredictRidge)}")
+
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    + +

    By default the grid search function includes cross validation with +five folds. The Scikit-Learn +documentation +contains more information on how to set the different parameters.

    -$$ -p(C=k\vert \mathbf {x} )=\frac{\exp{(\beta_{k0}+\beta_{k1}x_1)}}{1+\sum_{l=1}^{K-1}\exp{(\beta_{l0}+\beta_{l1}x_1)}}. -$$ - -

    It is easy to extend to more predictors. The final class is

    -$$ -p(C=K\vert \mathbf {x} )=\frac{1}{1+\sum_{l=1}^{K-1}\exp{(\beta_{l0}+\beta_{l1}x_1)}}, -$$ - -

    and they sum to one. Our earlier discussions were all specialized to -the case with two classes only. It is easy to see from the above that -what we derived earlier is compatible with these equations. -

    - -

    To find the optimal parameters we would typically use a gradient -descent method. Newton's method and gradient descent methods are -discussed in the material on optimization -methods. -

    +

    If we take out the random noise, running the above codes results in \( \lambda=0 \) yielding the best fit.

    @@ -369,7 +425,7 @@ methods.

  • 34
  • 35
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs026.html b/doc/pub/week38/html/._week38-bs026.html index b56b2bc16..f16714904 100644 --- a/doc/pub/week38/html/._week38-bs026.html +++ b/doc/pub/week38/html/._week38-bs026.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,7 +321,83 @@ MathJax.Hub.Config({

     

     

     

    -

    Friday September 23

    + + +

    An alternative to the above manual grid set up, is to use a random +search where the parameters are tuned from a random distribution +(uniform below) for a fixed number of iterations. A model is +constructed and evaluated for each combination of chosen parameters. +We repeat the previous example but now with a random search. Note +that values of \( \lambda \) are now limited to be within \( x\in +[0,1] \). This domain may not be the most relevant one for the specific +case under study. +

    + + + +
    +
    +
    +
    +
    + 
    import numpy as np
+from sklearn.model_selection import train_test_split
+from sklearn.linear_model import Ridge
+from sklearn.model_selection import GridSearchCV
+from scipy.stats import uniform as randuniform
+from sklearn.model_selection import RandomizedSearchCV
+
+
+def R2(y_data, y_model):
+    return 1 - np.sum((y_data - y_model) ** 2) / np.sum((y_data - np.mean(y_data)) ** 2)
+
+def MSE(y_data,y_model):
+    n = np.size(y_model)
+    return np.sum((y_data-y_model)**2)/n
+
+# A seed just to ensure that the random numbers are the same for every run.
+# Useful for eventual debugging.
+np.random.seed(2021)
+
+n = 100
+x = np.random.rand(n)
+y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.randn(n)
+
+Maxpolydegree = 5
+X = np.zeros((n,Maxpolydegree-1))
+
+for degree in range(1,Maxpolydegree): #No intercept column
+    X[:,degree-1] = x**(degree)
+
+# We split the data in test and training data
+X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)
+
+param_grid = {'alpha': randuniform()}
+# create and fit a ridge regression model, testing each alpha
+model = Ridge()
+gridsearch = RandomizedSearchCV(estimator=model, param_distributions=param_grid, n_iter=100)
+gridsearch.fit(X_train, y_train)
+print(gridsearch)
+ypredictRidge = gridsearch.predict(X_test)
+# summarize the results of the grid search
+print(f"Best estimated lambda-value: {gridsearch.best_estimator_.alpha}")
+print(f"MSE score: {MSE(y_test,ypredictRidge)}")
+print(f"R2 score: {R2(y_test,ypredictRidge)}")
+
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +

    @@ -334,7 +424,7 @@ MathJax.Hub.Config({

  • 35
  • 36
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs027.html b/doc/pub/week38/html/._week38-bs027.html index d5cd79830..fedfdacad 100644 --- a/doc/pub/week38/html/._week38-bs027.html +++ b/doc/pub/week38/html/._week38-bs027.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,67 +321,36 @@ MathJax.Hub.Config({

     

     

     

    -

    Searching for Optimal Regularization Parameters \( \lambda \)

    +

    Wisconsin Cancer Data

    -

    In project 1, when using Ridge and Lasso regression, we end up -searching for the optimal parameter \( \lambda \) which minimizes our -selected scores (MSE or \( R2 \) values for example). The brute force -approach, as discussed in the code here for Ridge regression, consists -in evaluating the MSE as function of different \( \lambda \) values. -Based on these calculations, one tries then to determine the value of the hyperparameter \( \lambda \) -which results in optimal scores (for example the smallest MSE or an \( R2=1 \)). +

    We show here how we can use a simple regression case on the breast +cancer data using Logistic regression as our algorithm for +classification.

    +
    - 
    import numpy as np
-import pandas as pd
-import matplotlib.pyplot as plt
-from sklearn.model_selection import train_test_split
-from sklearn import linear_model
+  
    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.linear_model import LogisticRegression
 
-def MSE(y_data,y_model):
-    n = np.size(y_model)
-    return np.sum((y_data-y_model)**2)/n
-# A seed just to ensure that the random numbers are the same for every run.
-# Useful for eventual debugging.
-np.random.seed(2021)
+# Load the data
+cancer = load_breast_cancer()
 
-n = 100
-x = np.random.rand(n)
-y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.randn(n)
-
-Maxpolydegree = 5
-X = np.zeros((n,Maxpolydegree-1))
-
-for degree in range(1,Maxpolydegree): #No intercept column
-    X[:,degree-1] = x**(degree)
-
-# We split the data in test and training data
-X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)
-
-# Decide which values of lambda to use
-nlambdas = 500
-MSERidgePredict = np.zeros(nlambdas)
-lambdas = np.logspace(-4, 2, nlambdas)
-for i in range(nlambdas):
-    lmb = lambdas[i]
-    RegRidge = linear_model.Ridge(lmb)
-    RegRidge.fit(X_train,y_train)
-    ypredictRidge = RegRidge.predict(X_test)
-    MSERidgePredict[i] = MSE(y_test,ypredictRidge)
-
-# Now plot the results
-plt.figure()
-plt.plot(np.log10(lambdas), MSERidgePredict, 'g--', label = 'MSE SL Ridge Test')
-plt.xlabel('log10(lambda)')
-plt.ylabel('MSE')
-plt.legend()
-plt.show()
+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)))
    @@ -383,10 +366,6 @@ plt.show()
    -

    Here we have performed a rather data greedy calculation as function of the regularization parameter \( \lambda \). There is no resampling here. The latter can easily be added by employing the function RidgeCV instead of just calling the Ridge function. For RidgeCV we need to pass the array of \( \lambda \) values. -By inspecting the figure we can in turn determine which is the optimal regularization parameter. -This becomes however less functional in the long run. -

    @@ -413,7 +392,7 @@ This becomes however less functional in the long run.

  • 36
  • 37
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs028.html b/doc/pub/week38/html/._week38-bs028.html index 7742ee7b1..8af64bc6b 100644 --- a/doc/pub/week38/html/._week38-bs028.html +++ b/doc/pub/week38/html/._week38-bs028.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,62 +321,52 @@ MathJax.Hub.Config({

     

     

     

    - +

    Using the correlation matrix

    -

    An alternative is to use the so-called grid search functionality -included with the library Scikit-Learn, as demonstrated for the same -example here. +

    In addition to the above scores, we could also study the covariance (and the correlation matrix). +We use Pandas to compute the correlation matrix.

    -
    - 
    import numpy as np
-from sklearn.model_selection import train_test_split
-from sklearn.linear_model import Ridge
-from sklearn.model_selection import GridSearchCV
+  
    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.linear_model import LogisticRegression
+cancer = load_breast_cancer()
+import pandas as pd
+# Making a data frame
+cancerpd = pd.DataFrame(cancer.data, columns=cancer.feature_names)
 
-def R2(y_data, y_model):
-    return 1 - np.sum((y_data - y_model) ** 2) / np.sum((y_data - np.mean(y_data)) ** 2)
+fig, axes = plt.subplots(15,2,figsize=(10,20))
+malignant = cancer.data[cancer.target == 0]
+benign = cancer.data[cancer.target == 1]
+ax = axes.ravel()
 
-def MSE(y_data,y_model):
-    n = np.size(y_model)
-    return np.sum((y_data-y_model)**2)/n
+for i in range(30):
+    _, bins = np.histogram(cancer.data[:,i], bins =50)
+    ax[i].hist(malignant[:,i], bins = bins, alpha = 0.5)
+    ax[i].hist(benign[:,i], bins = bins, alpha = 0.5)
+    ax[i].set_title(cancer.feature_names[i])
+    ax[i].set_yticks(())
+ax[0].set_xlabel("Feature magnitude")
+ax[0].set_ylabel("Frequency")
+ax[0].legend(["Malignant", "Benign"], loc ="best")
+fig.tight_layout()
+plt.show()
 
-# A seed just to ensure that the random numbers are the same for every run.
-# Useful for eventual debugging.
-np.random.seed(2021)
-
-n = 100
-x = np.random.rand(n)
-y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.randn(n)
-
-Maxpolydegree = 5
-X = np.zeros((n,Maxpolydegree-1))
-
-for degree in range(1,Maxpolydegree): #No intercept column
-    X[:,degree-1] = x**(degree)
-
-# We split the data in test and training data
-X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)
-
-# Decide which values of lambda to use
-nlambdas = 10
-lambdas = np.logspace(-4, 2, nlambdas)
-# create and fit a ridge regression model, testing each alpha
-model = Ridge()
-gridsearch = GridSearchCV(estimator=model, param_grid=dict(alpha=lambdas))
-gridsearch.fit(X_train, y_train)
-print(gridsearch)
-ypredictRidge = gridsearch.predict(X_test)
-# summarize the results of the grid search
-print(f"Best estimated lambda-value: {gridsearch.best_estimator_.alpha}")
-print(f"MSE score: {MSE(y_test,ypredictRidge)}")
-print(f"R2 score: {R2(y_test,ypredictRidge)}")
+import seaborn as sns
+correlation_matrix = cancerpd.corr().round(1)
+# use the heatmap function from seaborn to plot the correlation matrix
+# annot = True to print the values inside the square
+plt.figure(figsize=(15,8))
+sns.heatmap(data=correlation_matrix, annot=True)
+plt.show()
    @@ -378,13 +382,6 @@ ypredictRidge = gridsearchScikit-Learn -documentation -contains more information on how to set the different parameters. -

    - -

    If we take out the random noise, running the above codes results in \( \lambda=0 \) yielding the best fit.

    @@ -411,7 +408,7 @@ contains more information on how to set the different parameters.

  • 37
  • 38
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs029.html b/doc/pub/week38/html/._week38-bs029.html index fea4aeda2..9fd43940a 100644 --- a/doc/pub/week38/html/._week38-bs029.html +++ b/doc/pub/week38/html/._week38-bs029.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,18 +321,23 @@ MathJax.Hub.Config({

     

     

     

    - +

    Discussing the correlation data

    -

    An alternative to the above manual grid set up, is to use a random -search where the parameters are tuned from a random distribution -(uniform below) for a fixed number of iterations. A model is -constructed and evaluated for each combination of chosen parameters. -We repeat the previous example but now with a random search. Note -that values of \( \lambda \) are now limited to be within \( x\in -[0,1] \). This domain may not be the most relevant one for the specific -case under study. +

    In the above example we note two things. In the first plot we display +the overlap of benign and malignant tumors as functions of the various +features in the Wisconsing breast cancer data set. We see that for +some of the features we can distinguish clearly the benign and +malignant cases while for other features we cannot. This can point to +us which features may be of greater interest when we wish to classify +a benign or not benign tumour.

    +

    In the second figure we have computed the so-called correlation +matrix, which in our case with thirty features becomes a \( 30\times 30 \) +matrix. +

    + +

    We constructed this matrix using pandas via the statements

    @@ -326,49 +345,7 @@ case under study.
    - 
    import numpy as np
-from sklearn.model_selection import train_test_split
-from sklearn.linear_model import Ridge
-from sklearn.model_selection import GridSearchCV
-from scipy.stats import uniform as randuniform
-from sklearn.model_selection import RandomizedSearchCV
-
-
-def R2(y_data, y_model):
-    return 1 - np.sum((y_data - y_model) ** 2) / np.sum((y_data - np.mean(y_data)) ** 2)
-
-def MSE(y_data,y_model):
-    n = np.size(y_model)
-    return np.sum((y_data-y_model)**2)/n
-
-# A seed just to ensure that the random numbers are the same for every run.
-# Useful for eventual debugging.
-np.random.seed(2021)
-
-n = 100
-x = np.random.rand(n)
-y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.randn(n)
-
-Maxpolydegree = 5
-X = np.zeros((n,Maxpolydegree-1))
-
-for degree in range(1,Maxpolydegree): #No intercept column
-    X[:,degree-1] = x**(degree)
-
-# We split the data in test and training data
-X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)
-
-param_grid = {'alpha': randuniform()}
-# create and fit a ridge regression model, testing each alpha
-model = Ridge()
-gridsearch = RandomizedSearchCV(estimator=model, param_distributions=param_grid, n_iter=100)
-gridsearch.fit(X_train, y_train)
-print(gridsearch)
-ypredictRidge = gridsearch.predict(X_test)
-# summarize the results of the grid search
-print(f"Best estimated lambda-value: {gridsearch.best_estimator_.alpha}")
-print(f"MSE score: {MSE(y_test,ypredictRidge)}")
-print(f"R2 score: {R2(y_test,ypredictRidge)}")
+  
    cancerpd = pd.DataFrame(cancer.data, columns=cancer.feature_names)
    @@ -384,6 +361,35 @@ ypredictRidge = gridsearch +
    +
    +
    +
    +
    + 
    correlation_matrix = cancerpd.corr().round(1)
+
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    + +

    Diagonalizing this matrix we can in turn say something about which +features are of relevance and which are not. This leads us to +the classical Principal Component Analysis (PCA) theorem with +applications. This will be discussed later this semester (week 43). +

    @@ -410,7 +416,7 @@ ypredictRidge = gridsearch38

  • 39
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs030.html b/doc/pub/week38/html/._week38-bs030.html index 2e698108c..cb969c060 100644 --- a/doc/pub/week38/html/._week38-bs030.html +++ b/doc/pub/week38/html/._week38-bs030.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,13 +321,7 @@ MathJax.Hub.Config({

     

     

     

    -

    Wisconsin Cancer Data

    - -

    We show here how we can use a simple regression case on the breast -cancer data using Logistic regression as our algorithm for -classification. -

    - +

    Other measures in classification studies: Cancer Data again

    @@ -336,7 +344,23 @@ X_train, X_test, y_train, y_test = train_tes # Logistic Regression logreg = LogisticRegression(solver='lbfgs') logreg.fit(X_train, y_train) + +from sklearn.preprocessing import LabelEncoder +from sklearn.model_selection import cross_validate +#Cross validation +accuracy = cross_validate(logreg,X_test,y_test,cv=10)['test_score'] +print(accuracy) print("Test set accuracy with Logistic Regression: {:.2f}".format(logreg.score(X_test,y_test))) + +import scikitplot as skplt +y_pred = logreg.predict(X_test) +skplt.metrics.plot_confusion_matrix(y_test, y_pred, normalize=True) +plt.show() +y_probas = logreg.predict_proba(X_test) +skplt.metrics.plot_roc(y_test, y_probas) +plt.show() +skplt.metrics.plot_cumulative_gain(y_test, y_probas) +plt.show()
    @@ -378,7 +402,7 @@ logreg.fit(X_train, y_train)
  • 39
  • 40
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs031.html b/doc/pub/week38/html/._week38-bs031.html index 43ed578c9..2f0818b5a 100644 --- a/doc/pub/week38/html/._week38-bs031.html +++ b/doc/pub/week38/html/._week38-bs031.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,68 +321,20 @@ MathJax.Hub.Config({

     

     

     

    -

    Using the correlation matrix

    +

    Optimization, the central part of any Machine Learning algortithm

    -

    In addition to the above scores, we could also study the covariance (and the correlation matrix). -We use Pandas to compute the correlation matrix. +Overview Video, why do we care about gradient methods? + +

    Almost every problem in machine learning and data science starts with +a dataset \( X \), a model \( g(\beta) \), which is a function of the +parameters \( \beta \) and a cost function \( C(X, g(\beta)) \) that allows +us to judge how well the model \( g(\beta) \) explains the observations +\( X \). The model is fit by finding the values of \( \beta \) that minimize +the cost function. Ideally we would be able to solve for \( \beta \) +analytically, however this is not possible in general and we must use +some approximative/numerical method to compute the minimum.

    - -
    -
    -
    -
    -
    - 
    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.linear_model import LogisticRegression
-cancer = load_breast_cancer()
-import pandas as pd
-# Making a data frame
-cancerpd = pd.DataFrame(cancer.data, columns=cancer.feature_names)
-
-fig, axes = plt.subplots(15,2,figsize=(10,20))
-malignant = cancer.data[cancer.target == 0]
-benign = cancer.data[cancer.target == 1]
-ax = axes.ravel()
-
-for i in range(30):
-    _, bins = np.histogram(cancer.data[:,i], bins =50)
-    ax[i].hist(malignant[:,i], bins = bins, alpha = 0.5)
-    ax[i].hist(benign[:,i], bins = bins, alpha = 0.5)
-    ax[i].set_title(cancer.feature_names[i])
-    ax[i].set_yticks(())
-ax[0].set_xlabel("Feature magnitude")
-ax[0].set_ylabel("Frequency")
-ax[0].legend(["Malignant", "Benign"], loc ="best")
-fig.tight_layout()
-plt.show()
-
-import seaborn as sns
-correlation_matrix = cancerpd.corr().round(1)
-# use the heatmap function from seaborn to plot the correlation matrix
-# annot = True to print the values inside the square
-plt.figure(figsize=(15,8))
-sns.heatmap(data=correlation_matrix, annot=True)
-plt.show()
-
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    - -

    diff --git a/doc/pub/week38/html/._week38-bs032.html b/doc/pub/week38/html/._week38-bs032.html index 6c83d815f..108383ca1 100644 --- a/doc/pub/week38/html/._week38-bs032.html +++ b/doc/pub/week38/html/._week38-bs032.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,75 +321,24 @@ MathJax.Hub.Config({

     

     

     

    -

    Discussing the correlation data

    +

    Revisiting our Logistic Regression case

    -

    In the above example we note two things. In the first plot we display -the overlap of benign and malignant tumors as functions of the various -features in the Wisconsing breast cancer data set. We see that for -some of the features we can distinguish clearly the benign and -malignant cases while for other features we cannot. This can point to -us which features may be of greater interest when we wish to classify -a benign or not benign tumour. +

    In our discussion on Logistic Regression we studied the +case of +two classes, with \( y_i \) either +\( 0 \) or \( 1 \). Furthermore we assumed also that we have only two +parameters \( \beta \) in our fitting, that is we +defined probabilities

    -

    In the second figure we have computed the so-called correlation -matrix, which in our case with thirty features becomes a \( 30\times 30 \) -matrix. -

    +$$ +\begin{align*} +p(y_i=1|x_i,\boldsymbol{\beta}) &= \frac{\exp{(\beta_0+\beta_1x_i)}}{1+\exp{(\beta_0+\beta_1x_i)}},\nonumber\\ +p(y_i=0|x_i,\boldsymbol{\beta}) &= 1 - p(y_i=1|x_i,\boldsymbol{\beta}), +\end{align*} +$$ -

    We constructed this matrix using pandas via the statements

    - - -
    -
    -
    -
    -
    - 
    cancerpd = pd.DataFrame(cancer.data, columns=cancer.feature_names)
-
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    - -

    and then

    - - -
    -
    -
    -
    -
    - 
    correlation_matrix = cancerpd.corr().round(1)
-
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    - -

    Diagonalizing this matrix we can in turn say something about which -features are of relevance and which are not. This leads us to -the classical Principal Component Analysis (PCA) theorem with -applications. This will be discussed later this semester (week 43). -

    +

    where \( \boldsymbol{\beta} \) are the weights we wish to extract from data, in our case \( \beta_0 \) and \( \beta_1 \).

    @@ -402,7 +365,7 @@ applications. This will be discussed later this semester (41

  • 42
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs033.html b/doc/pub/week38/html/._week38-bs033.html index adbfa233f..c030102ca 100644 --- a/doc/pub/week38/html/._week38-bs033.html +++ b/doc/pub/week38/html/._week38-bs033.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,61 +321,28 @@ MathJax.Hub.Config({

     

     

     

    -

    Other measures in classification studies: Cancer Data again

    +

    The equations to solve

    - -
    -
    -
    -
    -
    - 
    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.linear_model import LogisticRegression
+
    Our compact equations used a definition of a vector \( \boldsymbol{y} \) with \( n \)
+elements \( y_i \), an \( n\times p \) matrix \( \boldsymbol{X} \) which contains the
+\( x_i \) values and a vector \( \boldsymbol{p} \) of fitted probabilities
+\( p(y_i\vert x_i,\boldsymbol{\beta}) \). We rewrote in a more compact form
+the first derivative of the cost function as
+
    
 
-# Load the data
-cancer = load_breast_cancer()
+$$
+\frac{\partial \mathcal{C}(\boldsymbol{\beta})}{\partial \boldsymbol{\beta}} = -\boldsymbol{X}^T\left(\boldsymbol{y}-\boldsymbol{p}\right). 
+$$
 
-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)
+
    If we in addition define a diagonal matrix \( \boldsymbol{W} \) with elements 
+\( p(y_i\vert x_i,\boldsymbol{\beta})(1-p(y_i\vert x_i,\boldsymbol{\beta}) \), we can obtain a compact expression of the second derivative as 
+
    
 
-from sklearn.preprocessing import LabelEncoder
-from sklearn.model_selection import cross_validate
-#Cross validation
-accuracy = cross_validate(logreg,X_test,y_test,cv=10)['test_score']
-print(accuracy)
-print("Test set accuracy with Logistic Regression: {:.2f}".format(logreg.score(X_test,y_test)))
-
-import scikitplot as skplt
-y_pred = logreg.predict(X_test)
-skplt.metrics.plot_confusion_matrix(y_test, y_pred, normalize=True)
-plt.show()
-y_probas = logreg.predict_proba(X_test)
-skplt.metrics.plot_roc(y_test, y_probas)
-plt.show()
-skplt.metrics.plot_cumulative_gain(y_test, y_probas)
-plt.show()
-
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    +$$ +\frac{\partial^2 \mathcal{C}(\boldsymbol{\beta})}{\partial \boldsymbol{\beta}\partial \boldsymbol{\beta}^T} = \boldsymbol{X}^T\boldsymbol{W}\boldsymbol{X}. +$$ +

    This defines what is called the Hessian matrix.

    @@ -388,7 +369,7 @@ plt.show()

  • 42
  • 43
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs034.html b/doc/pub/week38/html/._week38-bs034.html index 6d750cce5..4e7c65565 100644 --- a/doc/pub/week38/html/._week38-bs034.html +++ b/doc/pub/week38/html/._week38-bs034.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,19 +321,25 @@ MathJax.Hub.Config({

     

     

     

    -

    Optimization, the central part of any Machine Learning algortithm

    +

    Solving using Newton-Raphson's method

    -Overview Video, why do we care about gradient methods? +

    If we can set up these equations, Newton-Raphson's iterative method is normally the method of choice. It requires however that we can compute in an efficient way the matrices that define the first and second derivatives.

    -

    Almost every problem in machine learning and data science starts with -a dataset \( X \), a model \( g(\beta) \), which is a function of the -parameters \( \beta \) and a cost function \( C(X, g(\beta)) \) that allows -us to judge how well the model \( g(\beta) \) explains the observations -\( X \). The model is fit by finding the values of \( \beta \) that minimize -the cost function. Ideally we would be able to solve for \( \beta \) -analytically, however this is not possible in general and we must use -some approximative/numerical method to compute the minimum. -

    +

    Our iterative scheme is then given by

    + +$$ +\boldsymbol{\beta}^{\mathrm{new}} = \boldsymbol{\beta}^{\mathrm{old}}-\left(\frac{\partial^2 \mathcal{C}(\boldsymbol{\beta})}{\partial \boldsymbol{\beta}\partial \boldsymbol{\beta}^T}\right)^{-1}_{\boldsymbol{\beta}^{\mathrm{old}}}\times \left(\frac{\partial \mathcal{C}(\boldsymbol{\beta})}{\partial \boldsymbol{\beta}}\right)_{\boldsymbol{\beta}^{\mathrm{old}}}, +$$ + +

    or in matrix form as

    + +$$ +\boldsymbol{\beta}^{\mathrm{new}} = \boldsymbol{\beta}^{\mathrm{old}}-\left(\boldsymbol{X}^T\boldsymbol{W}\boldsymbol{X} \right)^{-1}\times \left(-\boldsymbol{X}^T(\boldsymbol{y}-\boldsymbol{p}) \right)_{\boldsymbol{\beta}^{\mathrm{old}}}. +$$ + +

    The right-hand side is computed with the old values of \( \beta \).

    + +

    If we can compute these matrices, in particular the Hessian, the above is often the easiest method to implement.

    @@ -346,7 +366,7 @@ some approximative/numerical method to compute the minimum.

  • 43
  • 44
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs035.html b/doc/pub/week38/html/._week38-bs035.html index f453766ab..5d2f2b04c 100644 --- a/doc/pub/week38/html/._week38-bs035.html +++ b/doc/pub/week38/html/._week38-bs035.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,25 +321,19 @@ MathJax.Hub.Config({

     

     

     

    -

    Revisiting our Logistic Regression case

    +

    Brief reminder on Newton-Raphson's method

    -

    In our discussion on Logistic Regression we studied the -case of -two classes, with \( y_i \) either -\( 0 \) or \( 1 \). Furthermore we assumed also that we have only two -parameters \( \beta \) in our fitting, that is we -defined probabilities +

    Let us quickly remind ourselves how we derive the above method.

    + +

    Perhaps the most celebrated of all one-dimensional root-finding +routines is Newton's method, also called the Newton-Raphson +method. This method requires the evaluation of both the +function \( f \) and its derivative \( f' \) at arbitrary points. +If you can only calculate the derivative +numerically and/or your function is not of the smooth type, we +normally discourage the use of this method.

    -$$ -\begin{align*} -p(y_i=1|x_i,\boldsymbol{\beta}) &= \frac{\exp{(\beta_0+\beta_1x_i)}}{1+\exp{(\beta_0+\beta_1x_i)}},\nonumber\\ -p(y_i=0|x_i,\boldsymbol{\beta}) &= 1 - p(y_i=1|x_i,\boldsymbol{\beta}), -\end{align*} -$$ - -

    where \( \boldsymbol{\beta} \) are the weights we wish to extract from data, in our case \( \beta_0 \) and \( \beta_1 \).

    -

    diff --git a/doc/pub/week38/html/._week38-bs036.html b/doc/pub/week38/html/._week38-bs036.html index 7892f87f8..c65b23175 100644 --- a/doc/pub/week38/html/._week38-bs036.html +++ b/doc/pub/week38/html/._week38-bs036.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,28 +321,38 @@ MathJax.Hub.Config({

     

     

     

    -

    The equations to solve

    +

    The equations

    -

    Our compact equations used a definition of a vector \( \boldsymbol{y} \) with \( n \) -elements \( y_i \), an \( n\times p \) matrix \( \boldsymbol{X} \) which contains the -\( x_i \) values and a vector \( \boldsymbol{p} \) of fitted probabilities -\( p(y_i\vert x_i,\boldsymbol{\beta}) \). We rewrote in a more compact form -the first derivative of the cost function as +

    The Newton-Raphson formula consists geometrically of extending the +tangent line at a current point until it crosses zero, then setting +the next guess to the abscissa of that zero-crossing. The mathematics +behind this method is rather simple. Employing a Taylor expansion for +\( x \) sufficiently close to the solution \( s \), we have

    $$ -\frac{\partial \mathcal{C}(\boldsymbol{\beta})}{\partial \boldsymbol{\beta}} = -\boldsymbol{X}^T\left(\boldsymbol{y}-\boldsymbol{p}\right). + f(s)=0=f(x)+(s-x)f'(x)+\frac{(s-x)^2}{2}f''(x) +\dots. + \tag{2} $$ -

    If we in addition define a diagonal matrix \( \boldsymbol{W} \) with elements -\( p(y_i\vert x_i,\boldsymbol{\beta})(1-p(y_i\vert x_i,\boldsymbol{\beta}) \), we can obtain a compact expression of the second derivative as +

    For small enough values of the function and for well-behaved +functions, the terms beyond linear are unimportant, hence we obtain

    $$ -\frac{\partial^2 \mathcal{C}(\boldsymbol{\beta})}{\partial \boldsymbol{\beta}\partial \boldsymbol{\beta}^T} = \boldsymbol{X}^T\boldsymbol{W}\boldsymbol{X}. + f(x)+(s-x)f'(x)\approx 0, +$$ + +

    yielding

    +$$ + s\approx x-\frac{f(x)}{f'(x)}. +$$ + +

    Having in mind an iterative procedure, it is natural to start iterating with

    +$$ + x_{n+1}=x_n-\frac{f(x_n)}{f'(x_n)}. $$ -

    This defines what is called the Hessian matrix.

    @@ -355,7 +379,7 @@ $$

  • 45
  • 46
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs037.html b/doc/pub/week38/html/._week38-bs037.html index d31f699bf..6599d122b 100644 --- a/doc/pub/week38/html/._week38-bs037.html +++ b/doc/pub/week38/html/._week38-bs037.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,25 +321,20 @@ MathJax.Hub.Config({

     

     

     

    -

    Solving using Newton-Raphson's method

    +

    Simple geometric interpretation

    -

    If we can set up these equations, Newton-Raphson's iterative method is normally the method of choice. It requires however that we can compute in an efficient way the matrices that define the first and second derivatives.

    - -

    Our iterative scheme is then given by

    - -$$ -\boldsymbol{\beta}^{\mathrm{new}} = \boldsymbol{\beta}^{\mathrm{old}}-\left(\frac{\partial^2 \mathcal{C}(\boldsymbol{\beta})}{\partial \boldsymbol{\beta}\partial \boldsymbol{\beta}^T}\right)^{-1}_{\boldsymbol{\beta}^{\mathrm{old}}}\times \left(\frac{\partial \mathcal{C}(\boldsymbol{\beta})}{\partial \boldsymbol{\beta}}\right)_{\boldsymbol{\beta}^{\mathrm{old}}}, -$$ - -

    or in matrix form as

    - -$$ -\boldsymbol{\beta}^{\mathrm{new}} = \boldsymbol{\beta}^{\mathrm{old}}-\left(\boldsymbol{X}^T\boldsymbol{W}\boldsymbol{X} \right)^{-1}\times \left(-\boldsymbol{X}^T(\boldsymbol{y}-\boldsymbol{p}) \right)_{\boldsymbol{\beta}^{\mathrm{old}}}. -$$ - -

    The right-hand side is computed with the old values of \( \beta \).

    - -

    If we can compute these matrices, in particular the Hessian, the above is often the easiest method to implement.

    +

    The above is Newton-Raphson's method. It has a simple geometric +interpretation, namely \( x_{n+1} \) is the point where the tangent from +\( (x_n,f(x_n)) \) crosses the \( x \)-axis. Close to the solution, +Newton-Raphson converges fast to the desired result. However, if we +are far from a root, where the higher-order terms in the series are +important, the Newton-Raphson formula can give grossly inaccurate +results. For instance, the initial guess for the root might be so far +from the true root as to let the search interval include a local +maximum or minimum of the function. If an iteration places a trial +guess near such a local extremum, so that the first derivative nearly +vanishes, then Newton-Raphson may fail totally +

    @@ -352,7 +361,7 @@ $$

  • 46
  • 47
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs038.html b/doc/pub/week38/html/._week38-bs038.html index 05b49a189..8d7769072 100644 --- a/doc/pub/week38/html/._week38-bs038.html +++ b/doc/pub/week38/html/._week38-bs038.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,17 +321,57 @@ MathJax.Hub.Config({

     

     

     

    -

    Brief reminder on Newton-Raphson's method

    +

    Extending to more than one variable

    -

    Let us quickly remind ourselves how we derive the above method.

    +

    Newton's method can be generalized to systems of several non-linear equations +and variables. Consider the case with two equations +

    +$$ + \begin{array}{cc} f_1(x_1,x_2) &=0\\ + f_2(x_1,x_2) &=0,\end{array} +$$ -

    Perhaps the most celebrated of all one-dimensional root-finding -routines is Newton's method, also called the Newton-Raphson -method. This method requires the evaluation of both the -function \( f \) and its derivative \( f' \) at arbitrary points. -If you can only calculate the derivative -numerically and/or your function is not of the smooth type, we -normally discourage the use of this method. +

    which we Taylor expand to obtain

    + +$$ + \begin{array}{cc} 0=f_1(x_1+h_1,x_2+h_2)=&f_1(x_1,x_2)+h_1 + \partial f_1/\partial x_1+h_2 + \partial f_1/\partial x_2+\dots\\ + 0=f_2(x_1+h_1,x_2+h_2)=&f_2(x_1,x_2)+h_1 + \partial f_2/\partial x_1+h_2 + \partial f_2/\partial x_2+\dots + \end{array}. +$$ + +

    Defining the Jacobian matrix \( {\bf \boldsymbol{J}} \) we have

    +$$ + {\bf \boldsymbol{J}}=\left( \begin{array}{cc} + \partial f_1/\partial x_1 & \partial f_1/\partial x_2 \\ + \partial f_2/\partial x_1 &\partial f_2/\partial x_2 + \end{array} \right), +$$ + +

    we can rephrase Newton's method as

    +$$ +\left(\begin{array}{c} x_1^{n+1} \\ x_2^{n+1} \end{array} \right)= +\left(\begin{array}{c} x_1^{n} \\ x_2^{n} \end{array} \right)+ +\left(\begin{array}{c} h_1^{n} \\ h_2^{n} \end{array} \right), +$$ + +

    where we have defined

    +$$ + \left(\begin{array}{c} h_1^{n} \\ h_2^{n} \end{array} \right)= + -{\bf \boldsymbol{J}}^{-1} + \left(\begin{array}{c} f_1(x_1^{n},x_2^{n}) \\ f_2(x_1^{n},x_2^{n}) \end{array} \right). +$$ + +

    We need thus to compute the inverse of the Jacobian matrix and it +is to understand that difficulties may +arise in case \( {\bf \boldsymbol{J}} \) is nearly singular. +

    + +

    It is rather straightforward to extend the above scheme to systems of +more than two non-linear equations. In our case, the Jacobian matrix is given by the Hessian that represents the second derivative of cost function.

    @@ -345,7 +399,7 @@ normally discourage the use of this method.

  • 47
  • 48
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs039.html b/doc/pub/week38/html/._week38-bs039.html index d694dfa3e..92d9df659 100644 --- a/doc/pub/week38/html/._week38-bs039.html +++ b/doc/pub/week38/html/._week38-bs039.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -307,39 +321,26 @@ MathJax.Hub.Config({

     

     

     

    -

    The equations

    +

    Steepest descent

    -

    The Newton-Raphson formula consists geometrically of extending the -tangent line at a current point until it crosses zero, then setting -the next guess to the abscissa of that zero-crossing. The mathematics -behind this method is rather simple. Employing a Taylor expansion for -\( x \) sufficiently close to the solution \( s \), we have +

    The basic idea of gradient descent is +that a function \( F(\mathbf{x}) \), +\( \mathbf{x} \equiv (x_1,\cdots,x_n) \), decreases fastest if one goes from \( \bf {x} \) in the +direction of the negative gradient \( -\nabla F(\mathbf{x}) \).

    +

    It can be shown that if

    $$ - f(s)=0=f(x)+(s-x)f'(x)+\frac{(s-x)^2}{2}f''(x) +\dots. - \tag{2} +\mathbf{x}_{k+1} = \mathbf{x}_k - \gamma_k \nabla F(\mathbf{x}_k), $$ -

    For small enough values of the function and for well-behaved -functions, the terms beyond linear are unimportant, hence we obtain +

    with \( \gamma_k > 0 \).

    + +

    For \( \gamma_k \) small enough, then \( F(\mathbf{x}_{k+1}) \leq +F(\mathbf{x}_k) \). This means that for a sufficiently small \( \gamma_k \) +we are always moving towards smaller function values, i.e a minimum.

    -$$ - f(x)+(s-x)f'(x)\approx 0, -$$ - -

    yielding

    -$$ - s\approx x-\frac{f(x)}{f'(x)}. -$$ - -

    Having in mind an iterative procedure, it is natural to start iterating with

    -$$ - x_{n+1}=x_n-\frac{f(x_n)}{f'(x_n)}. -$$ - -

    diff --git a/doc/pub/week38/html/._week38-bs040.html b/doc/pub/week38/html/._week38-bs040.html index 1b6abc800..c070bdd2e 100644 --- a/doc/pub/week38/html/._week38-bs040.html +++ b/doc/pub/week38/html/._week38-bs040.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -306,20 +320,21 @@ MathJax.Hub.Config({

     

     

     

    - -

    Simple geometric interpretation

    + +

    More on Steepest descent

    -

    The above is Newton-Raphson's method. It has a simple geometric -interpretation, namely \( x_{n+1} \) is the point where the tangent from -\( (x_n,f(x_n)) \) crosses the \( x \)-axis. Close to the solution, -Newton-Raphson converges fast to the desired result. However, if we -are far from a root, where the higher-order terms in the series are -important, the Newton-Raphson formula can give grossly inaccurate -results. For instance, the initial guess for the root might be so far -from the true root as to let the search interval include a local -maximum or minimum of the function. If an iteration places a trial -guess near such a local extremum, so that the first derivative nearly -vanishes, then Newton-Raphson may fail totally +

    The previous observation is the basis of the method of steepest +descent, which is also referred to as just gradient descent (GD). One +starts with an initial guess \( \mathbf{x}_0 \) for a minimum of \( F \) and +computes new approximations according to +

    + +$$ +\mathbf{x}_{k+1} = \mathbf{x}_k - \gamma_k \nabla F(\mathbf{x}_k), \ \ k \geq 0. +$$ + +

    The parameter \( \gamma_k \) is often referred to as the step length or +the learning rate within the context of Machine Learning.

    @@ -347,7 +362,7 @@ vanishes, then Newton-Raphson may fail totally

  • 49
  • 50
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs041.html b/doc/pub/week38/html/._week38-bs041.html index c45a89fc3..5435f9bb7 100644 --- a/doc/pub/week38/html/._week38-bs041.html +++ b/doc/pub/week38/html/._week38-bs041.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -306,58 +320,28 @@ MathJax.Hub.Config({

     

     

     

    - -

    Extending to more than one variable

    + +

    The ideal

    -

    Newton's method can be generalized to systems of several non-linear equations -and variables. Consider the case with two equations -

    -$$ - \begin{array}{cc} f_1(x_1,x_2) &=0\\ - f_2(x_1,x_2) &=0,\end{array} -$$ - -

    which we Taylor expand to obtain

    - -$$ - \begin{array}{cc} 0=f_1(x_1+h_1,x_2+h_2)=&f_1(x_1,x_2)+h_1 - \partial f_1/\partial x_1+h_2 - \partial f_1/\partial x_2+\dots\\ - 0=f_2(x_1+h_1,x_2+h_2)=&f_2(x_1,x_2)+h_1 - \partial f_2/\partial x_1+h_2 - \partial f_2/\partial x_2+\dots - \end{array}. -$$ - -

    Defining the Jacobian matrix \( {\bf \boldsymbol{J}} \) we have

    -$$ - {\bf \boldsymbol{J}}=\left( \begin{array}{cc} - \partial f_1/\partial x_1 & \partial f_1/\partial x_2 \\ - \partial f_2/\partial x_1 &\partial f_2/\partial x_2 - \end{array} \right), -$$ - -

    we can rephrase Newton's method as

    -$$ -\left(\begin{array}{c} x_1^{n+1} \\ x_2^{n+1} \end{array} \right)= -\left(\begin{array}{c} x_1^{n} \\ x_2^{n} \end{array} \right)+ -\left(\begin{array}{c} h_1^{n} \\ h_2^{n} \end{array} \right), -$$ - -

    where we have defined

    -$$ - \left(\begin{array}{c} h_1^{n} \\ h_2^{n} \end{array} \right)= - -{\bf \boldsymbol{J}}^{-1} - \left(\begin{array}{c} f_1(x_1^{n},x_2^{n}) \\ f_2(x_1^{n},x_2^{n}) \end{array} \right). -$$ - -

    We need thus to compute the inverse of the Jacobian matrix and it -is to understand that difficulties may -arise in case \( {\bf \boldsymbol{J}} \) is nearly singular. +

    Ideally the sequence \( \{\mathbf{x}_k \}_{k=0} \) converges to a global +minimum of the function \( F \). In general we do not know if we are in a +global or local minimum. In the special case when \( F \) is a convex +function, all local minima are also global minima, so in this case +gradient descent can converge to the global solution. The advantage of +this scheme is that it is conceptually simple and straightforward to +implement. However the method in this form has some severe +limitations:

    -

    It is rather straightforward to extend the above scheme to systems of -more than two non-linear equations. In our case, the Jacobian matrix is given by the Hessian that represents the second derivative of cost function. +

    In machine learing we are often faced with non-convex high dimensional +cost functions with many local minima. Since GD is deterministic we +will get stuck in a local minimum, if the method converges, unless we +have a very good intial guess. This also implies that the scheme is +sensitive to the chosen initial condition. +

    + +

    Note that the gradient is a function of \( \mathbf{x} = +(x_1,\cdots,x_n) \) which makes it expensive to compute numerically.

    @@ -385,7 +369,7 @@ more than two non-linear equations. In our case, the Jacobian matrix is given by

  • 50
  • 51
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/._week38-bs042.html b/doc/pub/week38/html/._week38-bs042.html index c652126c5..fc8bee609 100644 --- a/doc/pub/week38/html/._week38-bs042.html +++ b/doc/pub/week38/html/._week38-bs042.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -306,25 +320,21 @@ MathJax.Hub.Config({

     

     

     

    - -

    Steepest descent

    + +

    The sensitiveness of the gradient descent

    -

    The basic idea of gradient descent is -that a function \( F(\mathbf{x}) \), -\( \mathbf{x} \equiv (x_1,\cdots,x_n) \), decreases fastest if one goes from \( \bf {x} \) in the -direction of the negative gradient \( -\nabla F(\mathbf{x}) \). +

    The gradient descent method +is sensitive to the choice of learning rate \( \gamma_k \). This is due +to the fact that we are only guaranteed that \( F(\mathbf{x}_{k+1}) \leq +F(\mathbf{x}_k) \) for sufficiently small \( \gamma_k \). The problem is to +determine an optimal learning rate. If the learning rate is chosen too +small the method will take a long time to converge and if it is too +large we can experience erratic behavior.

    -

    It can be shown that if

    -$$ -\mathbf{x}_{k+1} = \mathbf{x}_k - \gamma_k \nabla F(\mathbf{x}_k), -$$ - -

    with \( \gamma_k > 0 \).

    - -

    For \( \gamma_k \) small enough, then \( F(\mathbf{x}_{k+1}) \leq -F(\mathbf{x}_k) \). This means that for a sufficiently small \( \gamma_k \) -we are always moving towards smaller function values, i.e a minimum. +

    Many of these shortcomings can be alleviated by introducing +randomness. One such method is that of Stochastic Gradient Descent +(SGD), to be discussed next week.

    @@ -352,7 +362,7 @@ we are always moving towards smaller function values, i.e a minimum.

  • 51
  • 52
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • diff --git a/doc/pub/week38/html/week38-bs.html b/doc/pub/week38/html/week38-bs.html index d43a9b66b..3c1fdc334 100644 --- a/doc/pub/week38/html/week38-bs.html +++ b/doc/pub/week38/html/week38-bs.html @@ -36,28 +36,18 @@ doconce format html week38.do.txt --html_style=bootstrap --pygments_html_style=d @@ -234,68 +246,70 @@ MathJax.Hub.Config({
  • Contents
    • @@ -325,7 +339,7 @@ MathJax.Hub.Config({
    -

    September 18-22

    +

    September 16-20, 2024


    @@ -350,7 +364,7 @@ MathJax.Hub.Config({
  • 9
  • 10
  • ...
    • -
  • 63
    • +
  • 65
  • »
    • @@ -364,7 +378,7 @@ MathJax.Hub.Config({ -->
    - © 1999-2023, Morten Hjorth-Jensen. Released under CC Attribution-NonCommercial 4.0 license + © 1999-2024, Morten Hjorth-Jensen. Released under CC Attribution-NonCommercial 4.0 license
    diff --git a/doc/pub/week38/html/week38-reveal.html b/doc/pub/week38/html/week38-reveal.html index 6d56a3f14..95a59798f 100644 --- a/doc/pub/week38/html/week38-reveal.html +++ b/doc/pub/week38/html/week38-reveal.html @@ -184,55 +184,48 @@ MathJax.Hub.Config({
    -

    September 18-22

    +

    September 16-20, 2024


    - © 1999-2023, Morten Hjorth-Jensen. Released under CC Attribution-NonCommercial 4.0 license + © 1999-2024, Morten Hjorth-Jensen. Released under CC Attribution-NonCommercial 4.0 license
    -

    Plans for week 38

    +

    Plans for week 38, lecture Monday September 16

    -Material for the active learning sessions on Tuesday and Wednesday -

    -

    -
    - -
    -Material for the lecture on Thursday September 21 +Material for the lecture on Monday September 16

    • Logistic regression as our first encounter of classification methods. From binary cases to several categories.
      • -

    • Start gradient and optimization methods
      • +

    • Start gradient and optimization methods + +
      • +
    +
    +
    -

  • Video of lecture
    • - -

  • Whiteboard notes at https://github.com/CompPhysics/MachineLearning/blob/master/doc/HandWrittenNotes/2023/NotesSep21.pdf
    • +
    +

    Suggested reading and videos

    +
    + +

    +

    • Readings and Videos:

      • Hastie et al 4.1, 4.2 and 4.3 on logistic regression
        • -

      • For a good discussion on gradient methods, see Goodfellow et al section 4.3-4.5 and chapter 8. We will come back to the latter chapter in our discussion of Neural networks as well.
        • +

      • Raschka et al, pages 53-76 on Logistic regression and pages 37-52 on gradient optimization
        • -

      • See also the whiteboard notes from week 37 at https://github.com/CompPhysics/MachineLearning/blob/master/doc/HandWrittenNotes/2023/NotesSep14.pdf for a discussion and derivation of the bias-variance tradeoff.
        • +

      • For a good discussion on gradient methods, see Goodfellow et al section 4.3-4.5 and chapter 8. We will come back to the latter chapter in our discussion of Neural networks as well.

      • Video on Logistic regression
        • @@ -246,189 +239,21 @@ MathJax.Hub.Config({
    -

    Material from last week and relevant for the first project

    -
    +

    Plans for the lab sessions

    -
    -

    Various steps in cross-validation

    - -

    When the repetitive splitting of the data set is done randomly, -samples may accidently end up in a fast majority of the splits in -either training or test set. Such samples may have an unbalanced -influence on either model building or prediction evaluation. To avoid -this \( k \)-fold cross-validation structures the data splitting. The -samples are divided into \( k \) more or less equally sized exhaustive and -mutually exclusive subsets. In turn (at each split) one of these -subsets plays the role of the test set while the union of the -remaining subsets constitutes the training set. Such a splitting -warrants a balanced representation of each sample in both training and -test set over the splits. Still the division into the \( k \) subsets -involves a degree of randomness. This may be fully excluded when -choosing \( k=n \). This particular case is referred to as leave-one-out -cross-validation (LOOCV). -

    -
    - -
    -

    How to set up the cross-validation for Ridge and/or Lasso

    - -
      -

    • Define a range of interest for the penalty parameter.
      • -

    • Divide the data set into training and test set comprising samples \( \{1, \ldots, n\} \setminus i \) and \( \{ i \} \), respectively.
      • -

    • Fit the linear regression model by means of for example Ridge or Lasso regression for each \( \lambda \) in the grid using the training set, and the corresponding estimate of the error variance \( \boldsymbol{\sigma}_{-i}^2(\lambda) \), as
      • -
    +
    +Material for the active learning sessions on Tuesday and Wednesday

    -

     
    -$$ -\begin{align*} -\boldsymbol{\beta}_{-i}(\lambda) & = ( \boldsymbol{X}_{-i, \ast}^{T} -\boldsymbol{X}_{-i, \ast} + \lambda \boldsymbol{I}_{pp})^{-1} -\boldsymbol{X}_{-i, \ast}^{T} \boldsymbol{y}_{-i} -\end{align*} -$$ -

     
    - -

      -

    • Evaluate the prediction performance of these models on the test set by \( C[y_i, \boldsymbol{X}_{i, \ast}; \boldsymbol{\beta}_{-i}(\lambda), \boldsymbol{\sigma}_{-i}^2(\lambda)] \). Or, by the prediction error \( |y_i - \boldsymbol{X}_{i, \ast} \boldsymbol{\beta}_{-i}(\lambda)| \), the relative error, the error squared or the R2 score function.
      • -

    • Repeat the first three steps such that each sample plays the role of the test set once.
      • -

    • Average the prediction performances of the test sets at each grid point of the penalty bias/parameter. It is an estimate of the prediction performance of the model corresponding to this value of the penalty parameter on novel data.
      • + +

    • Repetition from last week on the bias-variance tradeoff
      • + +

    • Resampling techniques, cross-validation examples included here, see also the lectures from last week on the bootstrap method
      • + +

    • Exercise for week 38 on the bias-variance tradeoff, see also the video from the lab session from week 37 at https://youtu.be/omLmp_kkie0
      • + +

    • Work on project 1, in particular resampling methods like cross-validation and bootstrap.
    -
    - -
    -

    Cross-validation in brief

    - -

    For the various values of \( k \)

    - -
      -

    1. shuffle the dataset randomly.
      2. -

    2. Split the dataset into \( k \) groups.
      3. -

    3. For each unique group: -
        -

      1. Decide which group to use as set for test data
        2. -

      2. Take the remaining groups as a training data set
        3. -

      3. Fit a model on the training set and evaluate it on the test set
        4. -

      4. Retain the evaluation score and discard the model
        5. -
      -

      -

    4. Summarize the model using the sample of model evaluation scores
      5. -
    -
    - -
    -

    Code Example for Cross-validation and \( k \)-fold Cross-validation

    - -

    The code here uses Ridge regression with cross-validation (CV) resampling and \( k \)-fold CV in order to fit a specific polynomial.

    - - -
    -
    -
    -
    -
    - 
    import numpy as np
-import matplotlib.pyplot as plt
-from sklearn.model_selection import KFold
-from sklearn.linear_model import Ridge
-from sklearn.model_selection import cross_val_score
-from sklearn.preprocessing import PolynomialFeatures
-
-# A seed just to ensure that the random numbers are the same for every run.
-# Useful for eventual debugging.
-np.random.seed(3155)
-
-# Generate the data.
-nsamples = 100
-x = np.random.randn(nsamples)
-y = 3*x**2 + np.random.randn(nsamples)
-
-## Cross-validation on Ridge regression using KFold only
-
-# Decide degree on polynomial to fit
-poly = PolynomialFeatures(degree = 6)
-
-# Decide which values of lambda to use
-nlambdas = 500
-lambdas = np.logspace(-3, 5, nlambdas)
-
-# Initialize a KFold instance
-k = 5
-kfold = KFold(n_splits = k)
-
-# Perform the cross-validation to estimate MSE
-scores_KFold = np.zeros((nlambdas, k))
-
-i = 0
-for lmb in lambdas:
-    ridge = Ridge(alpha = lmb)
-    j = 0
-    for train_inds, test_inds in kfold.split(x):
-        xtrain = x[train_inds]
-        ytrain = y[train_inds]
-
-        xtest = x[test_inds]
-        ytest = y[test_inds]
-
-        Xtrain = poly.fit_transform(xtrain[:, np.newaxis])
-        ridge.fit(Xtrain, ytrain[:, np.newaxis])
-
-        Xtest = poly.fit_transform(xtest[:, np.newaxis])
-        ypred = ridge.predict(Xtest)
-
-        scores_KFold[i,j] = np.sum((ypred - ytest[:, np.newaxis])**2)/np.size(ypred)
-
-        j += 1
-    i += 1
-
-
-estimated_mse_KFold = np.mean(scores_KFold, axis = 1)
-
-## Cross-validation using cross_val_score from sklearn along with KFold
-
-# kfold is an instance initialized above as:
-# kfold = KFold(n_splits = k)
-
-estimated_mse_sklearn = np.zeros(nlambdas)
-i = 0
-for lmb in lambdas:
-    ridge = Ridge(alpha = lmb)
-
-    X = poly.fit_transform(x[:, np.newaxis])
-    estimated_mse_folds = cross_val_score(ridge, X, y[:, np.newaxis], scoring='neg_mean_squared_error', cv=kfold)
-
-    # cross_val_score return an array containing the estimated negative mse for every fold.
-    # we have to the the mean of every array in order to get an estimate of the mse of the model
-    estimated_mse_sklearn[i] = np.mean(-estimated_mse_folds)
-
-    i += 1
-
-## Plot and compare the slightly different ways to perform cross-validation
-
-plt.figure()
-
-plt.plot(np.log10(lambdas), estimated_mse_sklearn, label = 'cross_val_score')
-plt.plot(np.log10(lambdas), estimated_mse_KFold, 'r--', label = 'KFold')
-
-plt.xlabel('log10(lambda)')
-plt.ylabel('mse')
-
-plt.legend()
-
-plt.show()
-
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    -
    @@ -2349,6 +2174,193 @@ plt.show()

    Write a code which implements gradient descent for a logistic regression example.

    +
    +

    Lab session: Material from last week and relevant for the first project

    +
    + +
    +

    Various steps in cross-validation

    + +

    When the repetitive splitting of the data set is done randomly, +samples may accidently end up in a fast majority of the splits in +either training or test set. Such samples may have an unbalanced +influence on either model building or prediction evaluation. To avoid +this \( k \)-fold cross-validation structures the data splitting. The +samples are divided into \( k \) more or less equally sized exhaustive and +mutually exclusive subsets. In turn (at each split) one of these +subsets plays the role of the test set while the union of the +remaining subsets constitutes the training set. Such a splitting +warrants a balanced representation of each sample in both training and +test set over the splits. Still the division into the \( k \) subsets +involves a degree of randomness. This may be fully excluded when +choosing \( k=n \). This particular case is referred to as leave-one-out +cross-validation (LOOCV). +

    +
    + +
    +

    How to set up the cross-validation for Ridge and/or Lasso

    + +
      +

    • Define a range of interest for the penalty parameter.
      • +

    • Divide the data set into training and test set comprising samples \( \{1, \ldots, n\} \setminus i \) and \( \{ i \} \), respectively.
      • +

    • Fit the linear regression model by means of for example Ridge or Lasso regression for each \( \lambda \) in the grid using the training set, and the corresponding estimate of the error variance \( \boldsymbol{\sigma}_{-i}^2(\lambda) \), as
      • +
    +

    +

     
    +$$ +\begin{align*} +\boldsymbol{\beta}_{-i}(\lambda) & = ( \boldsymbol{X}_{-i, \ast}^{T} +\boldsymbol{X}_{-i, \ast} + \lambda \boldsymbol{I}_{pp})^{-1} +\boldsymbol{X}_{-i, \ast}^{T} \boldsymbol{y}_{-i} +\end{align*} +$$ +

     
    + + +

      +

    • Evaluate the prediction performance of these models on the test set by \( C[y_i, \boldsymbol{X}_{i, \ast}; \boldsymbol{\beta}_{-i}(\lambda), \boldsymbol{\sigma}_{-i}^2(\lambda)] \). Or, by the prediction error \( |y_i - \boldsymbol{X}_{i, \ast} \boldsymbol{\beta}_{-i}(\lambda)| \), the relative error, the error squared or the R2 score function.
      • +

    • Repeat the first three steps such that each sample plays the role of the test set once.
      • +

    • Average the prediction performances of the test sets at each grid point of the penalty bias/parameter. It is an estimate of the prediction performance of the model corresponding to this value of the penalty parameter on novel data.
      • +
    +
    + +
    +

    Cross-validation in brief

    + +

    For the various values of \( k \)

    + +
      +

    1. shuffle the dataset randomly.
      2. +

    2. Split the dataset into \( k \) groups.
      3. +

    3. For each unique group: +
        +

      1. Decide which group to use as set for test data
        2. +

      2. Take the remaining groups as a training data set
        3. +

      3. Fit a model on the training set and evaluate it on the test set
        4. +

      4. Retain the evaluation score and discard the model
        5. +
      +

      +

    4. Summarize the model using the sample of model evaluation scores
      5. +
    +
    + +
    +

    Code Example for Cross-validation and \( k \)-fold Cross-validation

    + +

    The code here uses Ridge regression with cross-validation (CV) resampling and \( k \)-fold CV in order to fit a specific polynomial.

    + + +
    +
    +
    +
    +
    + 
    import numpy as np
+import matplotlib.pyplot as plt
+from sklearn.model_selection import KFold
+from sklearn.linear_model import Ridge
+from sklearn.model_selection import cross_val_score
+from sklearn.preprocessing import PolynomialFeatures
+
+# A seed just to ensure that the random numbers are the same for every run.
+# Useful for eventual debugging.
+np.random.seed(3155)
+
+# Generate the data.
+nsamples = 100
+x = np.random.randn(nsamples)
+y = 3*x**2 + np.random.randn(nsamples)
+
+## Cross-validation on Ridge regression using KFold only
+
+# Decide degree on polynomial to fit
+poly = PolynomialFeatures(degree = 6)
+
+# Decide which values of lambda to use
+nlambdas = 500
+lambdas = np.logspace(-3, 5, nlambdas)
+
+# Initialize a KFold instance
+k = 5
+kfold = KFold(n_splits = k)
+
+# Perform the cross-validation to estimate MSE
+scores_KFold = np.zeros((nlambdas, k))
+
+i = 0
+for lmb in lambdas:
+    ridge = Ridge(alpha = lmb)
+    j = 0
+    for train_inds, test_inds in kfold.split(x):
+        xtrain = x[train_inds]
+        ytrain = y[train_inds]
+
+        xtest = x[test_inds]
+        ytest = y[test_inds]
+
+        Xtrain = poly.fit_transform(xtrain[:, np.newaxis])
+        ridge.fit(Xtrain, ytrain[:, np.newaxis])
+
+        Xtest = poly.fit_transform(xtest[:, np.newaxis])
+        ypred = ridge.predict(Xtest)
+
+        scores_KFold[i,j] = np.sum((ypred - ytest[:, np.newaxis])**2)/np.size(ypred)
+
+        j += 1
+    i += 1
+
+
+estimated_mse_KFold = np.mean(scores_KFold, axis = 1)
+
+## Cross-validation using cross_val_score from sklearn along with KFold
+
+# kfold is an instance initialized above as:
+# kfold = KFold(n_splits = k)
+
+estimated_mse_sklearn = np.zeros(nlambdas)
+i = 0
+for lmb in lambdas:
+    ridge = Ridge(alpha = lmb)
+
+    X = poly.fit_transform(x[:, np.newaxis])
+    estimated_mse_folds = cross_val_score(ridge, X, y[:, np.newaxis], scoring='neg_mean_squared_error', cv=kfold)
+
+    # cross_val_score return an array containing the estimated negative mse for every fold.
+    # we have to the the mean of every array in order to get an estimate of the mse of the model
+    estimated_mse_sklearn[i] = np.mean(-estimated_mse_folds)
+
+    i += 1
+
+## Plot and compare the slightly different ways to perform cross-validation
+
+plt.figure()
+
+plt.plot(np.log10(lambdas), estimated_mse_sklearn, label = 'cross_val_score')
+plt.plot(np.log10(lambdas), estimated_mse_KFold, 'r--', label = 'KFold')
+
+plt.xlabel('log10(lambda)')
+plt.ylabel('mse')
+
+plt.legend()
+
+plt.show()
+
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    +
    diff --git a/doc/pub/week38/html/week38-solarized.html b/doc/pub/week38/html/week38-solarized.html index e06a69a03..c00d535ef 100644 --- a/doc/pub/week38/html/week38-solarized.html +++ b/doc/pub/week38/html/week38-solarized.html @@ -63,28 +63,18 @@ div.toc p,a { @@ -264,37 +276,36 @@ MathJax.Hub.Config({
    -

    September 18-22

    +

    September 16-20, 2024












    -

    Plans for week 38

    +

    Plans for week 38, lecture Monday September 16

    -Material for the active learning sessions on Tuesday and Wednesday -

    -

    -
    - -
    -Material for the lecture on Thursday September 21 +Material for the lecture on Monday September 16

    +
    + + +









    +

    Suggested reading and videos

    +
    + +

    +

    • Readings and Videos:
      • Hastie et al 4.1, 4.2 and 4.3 on logistic regression
        • +
      • Raschka et al, pages 53-76 on Logistic regression and pages 37-52 on gradient optimization
      • For a good discussion on gradient methods, see Goodfellow et al section 4.3-4.5 and chapter 8. We will come back to the latter chapter in our discussion of Neural networks as well.
        • -
      • See also the whiteboard notes from week 37 at https://github.com/CompPhysics/MachineLearning/blob/master/doc/HandWrittenNotes/2023/NotesSep14.pdf for a discussion and derivation of the bias-variance tradeoff.
      • Video on Logistic regression
      • Yet another video on logistic regression
      • Video on gradient descent
        • @@ -304,181 +315,19 @@ MathJax.Hub.Config({









        -

        Material from last week and relevant for the first project

        - - -

        Various steps in cross-validation

        - -

        When the repetitive splitting of the data set is done randomly, -samples may accidently end up in a fast majority of the splits in -either training or test set. Such samples may have an unbalanced -influence on either model building or prediction evaluation. To avoid -this \( k \)-fold cross-validation structures the data splitting. The -samples are divided into \( k \) more or less equally sized exhaustive and -mutually exclusive subsets. In turn (at each split) one of these -subsets plays the role of the test set while the union of the -remaining subsets constitutes the training set. Such a splitting -warrants a balanced representation of each sample in both training and -test set over the splits. Still the division into the \( k \) subsets -involves a degree of randomness. This may be fully excluded when -choosing \( k=n \). This particular case is referred to as leave-one-out -cross-validation (LOOCV). -

        - - -

        How to set up the cross-validation for Ridge and/or Lasso

        +

        Plans for the lab sessions

        +
        +Material for the active learning sessions on Tuesday and Wednesday +

          -
        • Define a range of interest for the penalty parameter.
          • -
        • Divide the data set into training and test set comprising samples \( \{1, \ldots, n\} \setminus i \) and \( \{ i \} \), respectively.
          • -
        • Fit the linear regression model by means of for example Ridge or Lasso regression for each \( \lambda \) in the grid using the training set, and the corresponding estimate of the error variance \( \boldsymbol{\sigma}_{-i}^2(\lambda) \), as
          • +
        • Repetition from last week on the bias-variance tradeoff
          • +
        • Resampling techniques, cross-validation examples included here, see also the lectures from last week on the bootstrap method
          • +
        • Exercise for week 38 on the bias-variance tradeoff, see also the video from the lab session from week 37 at https://youtu.be/omLmp_kkie0
          • +
        • Work on project 1, in particular resampling methods like cross-validation and bootstrap.
        -$$ -\begin{align*} -\boldsymbol{\beta}_{-i}(\lambda) & = ( \boldsymbol{X}_{-i, \ast}^{T} -\boldsymbol{X}_{-i, \ast} + \lambda \boldsymbol{I}_{pp})^{-1} -\boldsymbol{X}_{-i, \ast}^{T} \boldsymbol{y}_{-i} -\end{align*} -$$ - - -
          -
        • Evaluate the prediction performance of these models on the test set by \( C[y_i, \boldsymbol{X}_{i, \ast}; \boldsymbol{\beta}_{-i}(\lambda), \boldsymbol{\sigma}_{-i}^2(\lambda)] \). Or, by the prediction error \( |y_i - \boldsymbol{X}_{i, \ast} \boldsymbol{\beta}_{-i}(\lambda)| \), the relative error, the error squared or the R2 score function.
          • -
        • Repeat the first three steps such that each sample plays the role of the test set once.
          • -
        • Average the prediction performances of the test sets at each grid point of the penalty bias/parameter. It is an estimate of the prediction performance of the model corresponding to this value of the penalty parameter on novel data.
          • -
        -









        -

        Cross-validation in brief

        - -

        For the various values of \( k \)

        - -
          -
        1. shuffle the dataset randomly.
          2. -
        2. Split the dataset into \( k \) groups.
          3. -
        3. For each unique group: -
            -
          1. Decide which group to use as set for test data
            2. -
          2. Take the remaining groups as a training data set
            3. -
          3. Fit a model on the training set and evaluate it on the test set
            4. -
          4. Retain the evaluation score and discard the model
            5. -
          -
        4. Summarize the model using the sample of model evaluation scores
          5. -
        -









        -

        Code Example for Cross-validation and \( k \)-fold Cross-validation

        - -

        The code here uses Ridge regression with cross-validation (CV) resampling and \( k \)-fold CV in order to fit a specific polynomial.

        - - -
        -
        -
        -
        -
        - 
        import numpy as np
-import matplotlib.pyplot as plt
-from sklearn.model_selection import KFold
-from sklearn.linear_model import Ridge
-from sklearn.model_selection import cross_val_score
-from sklearn.preprocessing import PolynomialFeatures
-
-# A seed just to ensure that the random numbers are the same for every run.
-# Useful for eventual debugging.
-np.random.seed(3155)
-
-# Generate the data.
-nsamples = 100
-x = np.random.randn(nsamples)
-y = 3*x**2 + np.random.randn(nsamples)
-
-## Cross-validation on Ridge regression using KFold only
-
-# Decide degree on polynomial to fit
-poly = PolynomialFeatures(degree = 6)
-
-# Decide which values of lambda to use
-nlambdas = 500
-lambdas = np.logspace(-3, 5, nlambdas)
-
-# Initialize a KFold instance
-k = 5
-kfold = KFold(n_splits = k)
-
-# Perform the cross-validation to estimate MSE
-scores_KFold = np.zeros((nlambdas, k))
-
-i = 0
-for lmb in lambdas:
-    ridge = Ridge(alpha = lmb)
-    j = 0
-    for train_inds, test_inds in kfold.split(x):
-        xtrain = x[train_inds]
-        ytrain = y[train_inds]
-
-        xtest = x[test_inds]
-        ytest = y[test_inds]
-
-        Xtrain = poly.fit_transform(xtrain[:, np.newaxis])
-        ridge.fit(Xtrain, ytrain[:, np.newaxis])
-
-        Xtest = poly.fit_transform(xtest[:, np.newaxis])
-        ypred = ridge.predict(Xtest)
-
-        scores_KFold[i,j] = np.sum((ypred - ytest[:, np.newaxis])**2)/np.size(ypred)
-
-        j += 1
-    i += 1
-
-
-estimated_mse_KFold = np.mean(scores_KFold, axis = 1)
-
-## Cross-validation using cross_val_score from sklearn along with KFold
-
-# kfold is an instance initialized above as:
-# kfold = KFold(n_splits = k)
-
-estimated_mse_sklearn = np.zeros(nlambdas)
-i = 0
-for lmb in lambdas:
-    ridge = Ridge(alpha = lmb)
-
-    X = poly.fit_transform(x[:, np.newaxis])
-    estimated_mse_folds = cross_val_score(ridge, X, y[:, np.newaxis], scoring='neg_mean_squared_error', cv=kfold)
-
-    # cross_val_score return an array containing the estimated negative mse for every fold.
-    # we have to the the mean of every array in order to get an estimate of the mse of the model
-    estimated_mse_sklearn[i] = np.mean(-estimated_mse_folds)
-
-    i += 1
-
-## Plot and compare the slightly different ways to perform cross-validation
-
-plt.figure()
-
-plt.plot(np.log10(lambdas), estimated_mse_sklearn, label = 'cross_val_score')
-plt.plot(np.log10(lambdas), estimated_mse_KFold, 'r--', label = 'KFold')
-
-plt.xlabel('log10(lambda)')
-plt.ylabel('mse')
-
-plt.legend()
-
-plt.show()
-
        -
        -
        -
        -
        -
        -
        -
        -
        -
        -
        -
        -
        - +









        Material for lecture Thursday September 21

        @@ -2256,9 +2105,186 @@ plt.show()

        Write a code which implements gradient descent for a logistic regression example.

        +









        +

        Lab session: Material from last week and relevant for the first project

        + + +

        Various steps in cross-validation

        + +

        When the repetitive splitting of the data set is done randomly, +samples may accidently end up in a fast majority of the splits in +either training or test set. Such samples may have an unbalanced +influence on either model building or prediction evaluation. To avoid +this \( k \)-fold cross-validation structures the data splitting. The +samples are divided into \( k \) more or less equally sized exhaustive and +mutually exclusive subsets. In turn (at each split) one of these +subsets plays the role of the test set while the union of the +remaining subsets constitutes the training set. Such a splitting +warrants a balanced representation of each sample in both training and +test set over the splits. Still the division into the \( k \) subsets +involves a degree of randomness. This may be fully excluded when +choosing \( k=n \). This particular case is referred to as leave-one-out +cross-validation (LOOCV). +

        + + +

        How to set up the cross-validation for Ridge and/or Lasso

        + +
          +
        • Define a range of interest for the penalty parameter.
          • +
        • Divide the data set into training and test set comprising samples \( \{1, \ldots, n\} \setminus i \) and \( \{ i \} \), respectively.
          • +
        • Fit the linear regression model by means of for example Ridge or Lasso regression for each \( \lambda \) in the grid using the training set, and the corresponding estimate of the error variance \( \boldsymbol{\sigma}_{-i}^2(\lambda) \), as
          • +
        +$$ +\begin{align*} +\boldsymbol{\beta}_{-i}(\lambda) & = ( \boldsymbol{X}_{-i, \ast}^{T} +\boldsymbol{X}_{-i, \ast} + \lambda \boldsymbol{I}_{pp})^{-1} +\boldsymbol{X}_{-i, \ast}^{T} \boldsymbol{y}_{-i} +\end{align*} +$$ + + +
          +
        • Evaluate the prediction performance of these models on the test set by \( C[y_i, \boldsymbol{X}_{i, \ast}; \boldsymbol{\beta}_{-i}(\lambda), \boldsymbol{\sigma}_{-i}^2(\lambda)] \). Or, by the prediction error \( |y_i - \boldsymbol{X}_{i, \ast} \boldsymbol{\beta}_{-i}(\lambda)| \), the relative error, the error squared or the R2 score function.
          • +
        • Repeat the first three steps such that each sample plays the role of the test set once.
          • +
        • Average the prediction performances of the test sets at each grid point of the penalty bias/parameter. It is an estimate of the prediction performance of the model corresponding to this value of the penalty parameter on novel data.
          • +
        +









        +

        Cross-validation in brief

        + +

        For the various values of \( k \)

        + +
          +
        1. shuffle the dataset randomly.
          2. +
        2. Split the dataset into \( k \) groups.
          3. +
        3. For each unique group: +
            +
          1. Decide which group to use as set for test data
            2. +
          2. Take the remaining groups as a training data set
            3. +
          3. Fit a model on the training set and evaluate it on the test set
            4. +
          4. Retain the evaluation score and discard the model
            5. +
          +
        4. Summarize the model using the sample of model evaluation scores
          5. +
        +









        +

        Code Example for Cross-validation and \( k \)-fold Cross-validation

        + +

        The code here uses Ridge regression with cross-validation (CV) resampling and \( k \)-fold CV in order to fit a specific polynomial.

        + + +
        +
        +
        +
        +
        + 
        import numpy as np
+import matplotlib.pyplot as plt
+from sklearn.model_selection import KFold
+from sklearn.linear_model import Ridge
+from sklearn.model_selection import cross_val_score
+from sklearn.preprocessing import PolynomialFeatures
+
+# A seed just to ensure that the random numbers are the same for every run.
+# Useful for eventual debugging.
+np.random.seed(3155)
+
+# Generate the data.
+nsamples = 100
+x = np.random.randn(nsamples)
+y = 3*x**2 + np.random.randn(nsamples)
+
+## Cross-validation on Ridge regression using KFold only
+
+# Decide degree on polynomial to fit
+poly = PolynomialFeatures(degree = 6)
+
+# Decide which values of lambda to use
+nlambdas = 500
+lambdas = np.logspace(-3, 5, nlambdas)
+
+# Initialize a KFold instance
+k = 5
+kfold = KFold(n_splits = k)
+
+# Perform the cross-validation to estimate MSE
+scores_KFold = np.zeros((nlambdas, k))
+
+i = 0
+for lmb in lambdas:
+    ridge = Ridge(alpha = lmb)
+    j = 0
+    for train_inds, test_inds in kfold.split(x):
+        xtrain = x[train_inds]
+        ytrain = y[train_inds]
+
+        xtest = x[test_inds]
+        ytest = y[test_inds]
+
+        Xtrain = poly.fit_transform(xtrain[:, np.newaxis])
+        ridge.fit(Xtrain, ytrain[:, np.newaxis])
+
+        Xtest = poly.fit_transform(xtest[:, np.newaxis])
+        ypred = ridge.predict(Xtest)
+
+        scores_KFold[i,j] = np.sum((ypred - ytest[:, np.newaxis])**2)/np.size(ypred)
+
+        j += 1
+    i += 1
+
+
+estimated_mse_KFold = np.mean(scores_KFold, axis = 1)
+
+## Cross-validation using cross_val_score from sklearn along with KFold
+
+# kfold is an instance initialized above as:
+# kfold = KFold(n_splits = k)
+
+estimated_mse_sklearn = np.zeros(nlambdas)
+i = 0
+for lmb in lambdas:
+    ridge = Ridge(alpha = lmb)
+
+    X = poly.fit_transform(x[:, np.newaxis])
+    estimated_mse_folds = cross_val_score(ridge, X, y[:, np.newaxis], scoring='neg_mean_squared_error', cv=kfold)
+
+    # cross_val_score return an array containing the estimated negative mse for every fold.
+    # we have to the the mean of every array in order to get an estimate of the mse of the model
+    estimated_mse_sklearn[i] = np.mean(-estimated_mse_folds)
+
+    i += 1
+
+## Plot and compare the slightly different ways to perform cross-validation
+
+plt.figure()
+
+plt.plot(np.log10(lambdas), estimated_mse_sklearn, label = 'cross_val_score')
+plt.plot(np.log10(lambdas), estimated_mse_KFold, 'r--', label = 'KFold')
+
+plt.xlabel('log10(lambda)')
+plt.ylabel('mse')
+
+plt.legend()
+
+plt.show()
+
        +
        +
        +
        +
        +
        +
        +
        +
        +
        +
        +
        +
        +
        + +
        - © 1999-2023, Morten Hjorth-Jensen. Released under CC Attribution-NonCommercial 4.0 license + © 1999-2024, Morten Hjorth-Jensen. Released under CC Attribution-NonCommercial 4.0 license
        diff --git a/doc/pub/week38/html/week38.html b/doc/pub/week38/html/week38.html index 0ea8c3885..e07d0c203 100644 --- a/doc/pub/week38/html/week38.html +++ b/doc/pub/week38/html/week38.html @@ -140,28 +140,18 @@ div.toc p,a { @@ -341,37 +353,36 @@ MathJax.Hub.Config({
        -

        September 18-22

        +

        September 16-20, 2024












        -

        Plans for week 38

        +

        Plans for week 38, lecture Monday September 16

        -Material for the active learning sessions on Tuesday and Wednesday -

        -

        -
        - -
        -Material for the lecture on Thursday September 21 +Material for the lecture on Monday September 16

        +
        + + +









        +

        Suggested reading and videos

        +
        + +

        +

        • Readings and Videos:
          • Hastie et al 4.1, 4.2 and 4.3 on logistic regression
            • +
          • Raschka et al, pages 53-76 on Logistic regression and pages 37-52 on gradient optimization
          • For a good discussion on gradient methods, see Goodfellow et al section 4.3-4.5 and chapter 8. We will come back to the latter chapter in our discussion of Neural networks as well.
            • -
          • See also the whiteboard notes from week 37 at https://github.com/CompPhysics/MachineLearning/blob/master/doc/HandWrittenNotes/2023/NotesSep14.pdf for a discussion and derivation of the bias-variance tradeoff.
          • Video on Logistic regression
          • Yet another video on logistic regression
          • Video on gradient descent
            • @@ -381,181 +392,19 @@ MathJax.Hub.Config({









            -

            Material from last week and relevant for the first project

            - - -

            Various steps in cross-validation

            - -

            When the repetitive splitting of the data set is done randomly, -samples may accidently end up in a fast majority of the splits in -either training or test set. Such samples may have an unbalanced -influence on either model building or prediction evaluation. To avoid -this \( k \)-fold cross-validation structures the data splitting. The -samples are divided into \( k \) more or less equally sized exhaustive and -mutually exclusive subsets. In turn (at each split) one of these -subsets plays the role of the test set while the union of the -remaining subsets constitutes the training set. Such a splitting -warrants a balanced representation of each sample in both training and -test set over the splits. Still the division into the \( k \) subsets -involves a degree of randomness. This may be fully excluded when -choosing \( k=n \). This particular case is referred to as leave-one-out -cross-validation (LOOCV). -

            - - -

            How to set up the cross-validation for Ridge and/or Lasso

            +

            Plans for the lab sessions

            +
            +Material for the active learning sessions on Tuesday and Wednesday +

              -
            • Define a range of interest for the penalty parameter.
              • -
            • Divide the data set into training and test set comprising samples \( \{1, \ldots, n\} \setminus i \) and \( \{ i \} \), respectively.
              • -
            • Fit the linear regression model by means of for example Ridge or Lasso regression for each \( \lambda \) in the grid using the training set, and the corresponding estimate of the error variance \( \boldsymbol{\sigma}_{-i}^2(\lambda) \), as
              • +
            • Repetition from last week on the bias-variance tradeoff
              • +
            • Resampling techniques, cross-validation examples included here, see also the lectures from last week on the bootstrap method
              • +
            • Exercise for week 38 on the bias-variance tradeoff, see also the video from the lab session from week 37 at https://youtu.be/omLmp_kkie0
              • +
            • Work on project 1, in particular resampling methods like cross-validation and bootstrap.
            -$$ -\begin{align*} -\boldsymbol{\beta}_{-i}(\lambda) & = ( \boldsymbol{X}_{-i, \ast}^{T} -\boldsymbol{X}_{-i, \ast} + \lambda \boldsymbol{I}_{pp})^{-1} -\boldsymbol{X}_{-i, \ast}^{T} \boldsymbol{y}_{-i} -\end{align*} -$$ - - -
              -
            • Evaluate the prediction performance of these models on the test set by \( C[y_i, \boldsymbol{X}_{i, \ast}; \boldsymbol{\beta}_{-i}(\lambda), \boldsymbol{\sigma}_{-i}^2(\lambda)] \). Or, by the prediction error \( |y_i - \boldsymbol{X}_{i, \ast} \boldsymbol{\beta}_{-i}(\lambda)| \), the relative error, the error squared or the R2 score function.
              • -
            • Repeat the first three steps such that each sample plays the role of the test set once.
              • -
            • Average the prediction performances of the test sets at each grid point of the penalty bias/parameter. It is an estimate of the prediction performance of the model corresponding to this value of the penalty parameter on novel data.
              • -
            -









            -

            Cross-validation in brief

            - -

            For the various values of \( k \)

            - -
              -
            1. shuffle the dataset randomly.
              2. -
            2. Split the dataset into \( k \) groups.
              3. -
            3. For each unique group: -
                -
              1. Decide which group to use as set for test data
                2. -
              2. Take the remaining groups as a training data set
                3. -
              3. Fit a model on the training set and evaluate it on the test set
                4. -
              4. Retain the evaluation score and discard the model
                5. -
              -
            4. Summarize the model using the sample of model evaluation scores
              5. -
            -









            -

            Code Example for Cross-validation and \( k \)-fold Cross-validation

            - -

            The code here uses Ridge regression with cross-validation (CV) resampling and \( k \)-fold CV in order to fit a specific polynomial.

            - - -
            -
            -
            -
            -
            - 
            import numpy as np
-import matplotlib.pyplot as plt
-from sklearn.model_selection import KFold
-from sklearn.linear_model import Ridge
-from sklearn.model_selection import cross_val_score
-from sklearn.preprocessing import PolynomialFeatures
-
-# A seed just to ensure that the random numbers are the same for every run.
-# Useful for eventual debugging.
-np.random.seed(3155)
-
-# Generate the data.
-nsamples = 100
-x = np.random.randn(nsamples)
-y = 3*x**2 + np.random.randn(nsamples)
-
-## Cross-validation on Ridge regression using KFold only
-
-# Decide degree on polynomial to fit
-poly = PolynomialFeatures(degree = 6)
-
-# Decide which values of lambda to use
-nlambdas = 500
-lambdas = np.logspace(-3, 5, nlambdas)
-
-# Initialize a KFold instance
-k = 5
-kfold = KFold(n_splits = k)
-
-# Perform the cross-validation to estimate MSE
-scores_KFold = np.zeros((nlambdas, k))
-
-i = 0
-for lmb in lambdas:
-    ridge = Ridge(alpha = lmb)
-    j = 0
-    for train_inds, test_inds in kfold.split(x):
-        xtrain = x[train_inds]
-        ytrain = y[train_inds]
-
-        xtest = x[test_inds]
-        ytest = y[test_inds]
-
-        Xtrain = poly.fit_transform(xtrain[:, np.newaxis])
-        ridge.fit(Xtrain, ytrain[:, np.newaxis])
-
-        Xtest = poly.fit_transform(xtest[:, np.newaxis])
-        ypred = ridge.predict(Xtest)
-
-        scores_KFold[i,j] = np.sum((ypred - ytest[:, np.newaxis])**2)/np.size(ypred)
-
-        j += 1
-    i += 1
-
-
-estimated_mse_KFold = np.mean(scores_KFold, axis = 1)
-
-## Cross-validation using cross_val_score from sklearn along with KFold
-
-# kfold is an instance initialized above as:
-# kfold = KFold(n_splits = k)
-
-estimated_mse_sklearn = np.zeros(nlambdas)
-i = 0
-for lmb in lambdas:
-    ridge = Ridge(alpha = lmb)
-
-    X = poly.fit_transform(x[:, np.newaxis])
-    estimated_mse_folds = cross_val_score(ridge, X, y[:, np.newaxis], scoring='neg_mean_squared_error', cv=kfold)
-
-    # cross_val_score return an array containing the estimated negative mse for every fold.
-    # we have to the the mean of every array in order to get an estimate of the mse of the model
-    estimated_mse_sklearn[i] = np.mean(-estimated_mse_folds)
-
-    i += 1
-
-## Plot and compare the slightly different ways to perform cross-validation
-
-plt.figure()
-
-plt.plot(np.log10(lambdas), estimated_mse_sklearn, label = 'cross_val_score')
-plt.plot(np.log10(lambdas), estimated_mse_KFold, 'r--', label = 'KFold')
-
-plt.xlabel('log10(lambda)')
-plt.ylabel('mse')
-
-plt.legend()
-
-plt.show()
-
            -
            -
            -
            -
            -
            -
            -
            -
            -
            -
            -
            -
            - +









            Material for lecture Thursday September 21

            @@ -2333,9 +2182,186 @@ plt.show()

            Write a code which implements gradient descent for a logistic regression example.

            +









            +

            Lab session: Material from last week and relevant for the first project

            + + +

            Various steps in cross-validation

            + +

            When the repetitive splitting of the data set is done randomly, +samples may accidently end up in a fast majority of the splits in +either training or test set. Such samples may have an unbalanced +influence on either model building or prediction evaluation. To avoid +this \( k \)-fold cross-validation structures the data splitting. The +samples are divided into \( k \) more or less equally sized exhaustive and +mutually exclusive subsets. In turn (at each split) one of these +subsets plays the role of the test set while the union of the +remaining subsets constitutes the training set. Such a splitting +warrants a balanced representation of each sample in both training and +test set over the splits. Still the division into the \( k \) subsets +involves a degree of randomness. This may be fully excluded when +choosing \( k=n \). This particular case is referred to as leave-one-out +cross-validation (LOOCV). +

            + + +

            How to set up the cross-validation for Ridge and/or Lasso

            + +
              +
            • Define a range of interest for the penalty parameter.
              • +
            • Divide the data set into training and test set comprising samples \( \{1, \ldots, n\} \setminus i \) and \( \{ i \} \), respectively.
              • +
            • Fit the linear regression model by means of for example Ridge or Lasso regression for each \( \lambda \) in the grid using the training set, and the corresponding estimate of the error variance \( \boldsymbol{\sigma}_{-i}^2(\lambda) \), as
              • +
            +$$ +\begin{align*} +\boldsymbol{\beta}_{-i}(\lambda) & = ( \boldsymbol{X}_{-i, \ast}^{T} +\boldsymbol{X}_{-i, \ast} + \lambda \boldsymbol{I}_{pp})^{-1} +\boldsymbol{X}_{-i, \ast}^{T} \boldsymbol{y}_{-i} +\end{align*} +$$ + + +
              +
            • Evaluate the prediction performance of these models on the test set by \( C[y_i, \boldsymbol{X}_{i, \ast}; \boldsymbol{\beta}_{-i}(\lambda), \boldsymbol{\sigma}_{-i}^2(\lambda)] \). Or, by the prediction error \( |y_i - \boldsymbol{X}_{i, \ast} \boldsymbol{\beta}_{-i}(\lambda)| \), the relative error, the error squared or the R2 score function.
              • +
            • Repeat the first three steps such that each sample plays the role of the test set once.
              • +
            • Average the prediction performances of the test sets at each grid point of the penalty bias/parameter. It is an estimate of the prediction performance of the model corresponding to this value of the penalty parameter on novel data.
              • +
            +









            +

            Cross-validation in brief

            + +

            For the various values of \( k \)

            + +
              +
            1. shuffle the dataset randomly.
              2. +
            2. Split the dataset into \( k \) groups.
              3. +
            3. For each unique group: +
                +
              1. Decide which group to use as set for test data
                2. +
              2. Take the remaining groups as a training data set
                3. +
              3. Fit a model on the training set and evaluate it on the test set
                4. +
              4. Retain the evaluation score and discard the model
                5. +
              +
            4. Summarize the model using the sample of model evaluation scores
              5. +
            +









            +

            Code Example for Cross-validation and \( k \)-fold Cross-validation

            + +

            The code here uses Ridge regression with cross-validation (CV) resampling and \( k \)-fold CV in order to fit a specific polynomial.

            + + +
            +
            +
            +
            +
            + 
            import numpy as np
+import matplotlib.pyplot as plt
+from sklearn.model_selection import KFold
+from sklearn.linear_model import Ridge
+from sklearn.model_selection import cross_val_score
+from sklearn.preprocessing import PolynomialFeatures
+
+# A seed just to ensure that the random numbers are the same for every run.
+# Useful for eventual debugging.
+np.random.seed(3155)
+
+# Generate the data.
+nsamples = 100
+x = np.random.randn(nsamples)
+y = 3*x**2 + np.random.randn(nsamples)
+
+## Cross-validation on Ridge regression using KFold only
+
+# Decide degree on polynomial to fit
+poly = PolynomialFeatures(degree = 6)
+
+# Decide which values of lambda to use
+nlambdas = 500
+lambdas = np.logspace(-3, 5, nlambdas)
+
+# Initialize a KFold instance
+k = 5
+kfold = KFold(n_splits = k)
+
+# Perform the cross-validation to estimate MSE
+scores_KFold = np.zeros((nlambdas, k))
+
+i = 0
+for lmb in lambdas:
+    ridge = Ridge(alpha = lmb)
+    j = 0
+    for train_inds, test_inds in kfold.split(x):
+        xtrain = x[train_inds]
+        ytrain = y[train_inds]
+
+        xtest = x[test_inds]
+        ytest = y[test_inds]
+
+        Xtrain = poly.fit_transform(xtrain[:, np.newaxis])
+        ridge.fit(Xtrain, ytrain[:, np.newaxis])
+
+        Xtest = poly.fit_transform(xtest[:, np.newaxis])
+        ypred = ridge.predict(Xtest)
+
+        scores_KFold[i,j] = np.sum((ypred - ytest[:, np.newaxis])**2)/np.size(ypred)
+
+        j += 1
+    i += 1
+
+
+estimated_mse_KFold = np.mean(scores_KFold, axis = 1)
+
+## Cross-validation using cross_val_score from sklearn along with KFold
+
+# kfold is an instance initialized above as:
+# kfold = KFold(n_splits = k)
+
+estimated_mse_sklearn = np.zeros(nlambdas)
+i = 0
+for lmb in lambdas:
+    ridge = Ridge(alpha = lmb)
+
+    X = poly.fit_transform(x[:, np.newaxis])
+    estimated_mse_folds = cross_val_score(ridge, X, y[:, np.newaxis], scoring='neg_mean_squared_error', cv=kfold)
+
+    # cross_val_score return an array containing the estimated negative mse for every fold.
+    # we have to the the mean of every array in order to get an estimate of the mse of the model
+    estimated_mse_sklearn[i] = np.mean(-estimated_mse_folds)
+
+    i += 1
+
+## Plot and compare the slightly different ways to perform cross-validation
+
+plt.figure()
+
+plt.plot(np.log10(lambdas), estimated_mse_sklearn, label = 'cross_val_score')
+plt.plot(np.log10(lambdas), estimated_mse_KFold, 'r--', label = 'KFold')
+
+plt.xlabel('log10(lambda)')
+plt.ylabel('mse')
+
+plt.legend()
+
+plt.show()
+
            +
            +
            +
            +
            +
            +
            +
            +
            +
            +
            +
            +
            +
            + +
            - © 1999-2023, Morten Hjorth-Jensen. Released under CC Attribution-NonCommercial 4.0 license + © 1999-2024, Morten Hjorth-Jensen. Released under CC Attribution-NonCommercial 4.0 license
            diff --git a/doc/pub/week38/ipynb/ipynb-week38-src.tar.gz b/doc/pub/week38/ipynb/ipynb-week38-src.tar.gz index 6a4453ce4..6f56bfd7e 100644 Binary files a/doc/pub/week38/ipynb/ipynb-week38-src.tar.gz and b/doc/pub/week38/ipynb/ipynb-week38-src.tar.gz differ diff --git a/doc/pub/week38/ipynb/week38.ipynb b/doc/pub/week38/ipynb/week38.ipynb index eef9cea6a..2edb4f733 100644 --- a/doc/pub/week38/ipynb/week38.ipynb +++ b/doc/pub/week38/ipynb/week38.ipynb @@ -2,7 +2,7 @@ "cells": [ { "cell_type": "markdown", - "id": "9cc2eba5", + "id": "a4e5b8f2", "metadata": { "editable": true }, @@ -14,7 +14,7 @@ }, { "cell_type": "markdown", - "id": "a8183a7a", + "id": "93b0d36c", "metadata": { "editable": true }, @@ -22,46 +22,42 @@ "# Week 38: Logistic Regression and Optimization\n", "**Morten Hjorth-Jensen**, Department of Physics and Center for Computing in Science Education, University of Oslo and Department of Physics and Astronomy and Facility for Rare Isotope Beams, Michigan State University\n", "\n", - "Date: **September 18-22**" + "Date: **September 16-20, 2024**" ] }, { "cell_type": "markdown", - "id": "31d68051", + "id": "42899962", "metadata": { "editable": true }, "source": [ - "## Plans for week 38\n", + "## Plans for week 38, lecture Monday September 16\n", "\n", - "**Material for the active learning sessions on Tuesday and Wednesday.**\n", - "\n", - " * Lecture from last week on the bias-variance tradeoff\n", - "\n", - " * Resampling techniques, cross-validation examples included here, see also the lectures from last week on the bootstrap method\n", - "\n", - " * Exercise for week 38, see also the whiteboard notes from week 37 at \n", - "\n", - " * Work on project 1, in particular resampling methods like cross-validation and bootstrap.\n", - "\n", - " \n", - "**Material for the lecture on Thursday September 21.**\n", + "**Material for the lecture on Monday September 16.**\n", "\n", " * Logistic regression as our first encounter of classification methods. From binary cases to several categories.\n", "\n", " * Start gradient and optimization methods\n", - "\n", - " * [Video of lecture](https://youtu.be/cJrSwRsVSGM)\n", - "\n", - " * Whiteboard notes at \n", - "\n", + "\n", + "" + ] + }, + { + "cell_type": "markdown", + "id": "4175fb49", + "metadata": { + "editable": true + }, + "source": [ + "## Suggested reading and videos\n", " * Readings and Videos:\n", "\n", " * Hastie et al 4.1, 4.2 and 4.3 on logistic regression\n", "\n", - " * For a good discussion on gradient methods, see Goodfellow et al section 4.3-4.5 and chapter 8. We will come back to the latter chapter in our discussion of Neural networks as well.\n", + " * Raschka et al, pages 53-76 on Logistic regression and pages 37-52 on gradient optimization\n", "\n", - " * See also the whiteboard notes from week 37 at for a discussion and derivation of the bias-variance tradeoff. \n", + " * For a good discussion on gradient methods, see Goodfellow et al section 4.3-4.5 and chapter 8. We will come back to the latter chapter in our discussion of Neural networks as well.\n", "\n", " * [Video on Logistic regression](https://www.youtube.com/watch?v=C5268D9t9Ak)\n", "\n", @@ -72,17 +68,2756 @@ }, { "cell_type": "markdown", - "id": "1ef593e2", + "id": "cd73082e", "metadata": { "editable": true }, "source": [ - "## Material from last week and relevant for the first project" + "## Plans for the lab sessions\n", + "\n", + "**Material for the active learning sessions on Tuesday and Wednesday.**\n", + "\n", + " * Repetition from last week on the bias-variance tradeoff\n", + "\n", + " * Resampling techniques, cross-validation examples included here, see also the lectures from last week on the bootstrap method\n", + "\n", + " * Exercise for week 38 on the bias-variance tradeoff, see also the video from the lab session from week 37 at \n", + "\n", + " * Work on project 1, in particular resampling methods like cross-validation and bootstrap." ] }, { "cell_type": "markdown", - "id": "d690f648", + "id": "84c47adf", + "metadata": { + "editable": true + }, + "source": [ + "## Material for lecture Thursday September 21" + ] + }, + { + "cell_type": "markdown", + "id": "1299fcbc", + "metadata": { + "editable": true + }, + "source": [ + "## Logistic Regression\n", + "\n", + "In linear regression our main interest was centered on learning the\n", + "coefficients of a functional fit (say a polynomial) in order to be\n", + "able to predict the response of a continuous variable on some unseen\n", + "data. The fit to the continuous variable $y_i$ is based on some\n", + "independent variables $\\boldsymbol{x}_i$. Linear regression resulted in\n", + "analytical expressions for standard ordinary Least Squares or Ridge\n", + "regression (in terms of matrices to invert) for several quantities,\n", + "ranging from the variance and thereby the confidence intervals of the\n", + "parameters $\\boldsymbol{\\beta}$ to the mean squared error. If we can invert\n", + "the product of the design matrices, linear regression gives then a\n", + "simple recipe for fitting our data." + ] + }, + { + "cell_type": "markdown", + "id": "e70ead5f", + "metadata": { + "editable": true + }, + "source": [ + "## Classification problems\n", + "\n", + "Classification problems, however, are concerned with outcomes taking\n", + "the form of discrete variables (i.e. categories). We may for example,\n", + "on the basis of DNA sequencing for a number of patients, like to find\n", + "out which mutations are important for a certain disease; or based on\n", + "scans of various patients' brains, figure out if there is a tumor or\n", + "not; or given a specific physical system, we'd like to identify its\n", + "state, say whether it is an ordered or disordered system (typical\n", + "situation in solid state physics); or classify the status of a\n", + "patient, whether she/he has a stroke or not and many other similar\n", + "situations.\n", + "\n", + "The most common situation we encounter when we apply logistic\n", + "regression is that of two possible outcomes, normally denoted as a\n", + "binary outcome, true or false, positive or negative, success or\n", + "failure etc." + ] + }, + { + "cell_type": "markdown", + "id": "fa1b67be", + "metadata": { + "editable": true + }, + "source": [ + "## Optimization and Deep learning\n", + "\n", + "Logistic regression will also serve as our stepping stone towards\n", + "neural network algorithms and supervised deep learning. For logistic\n", + "learning, the minimization of the cost function leads to a non-linear\n", + "equation in the parameters $\\boldsymbol{\\beta}$. The optimization of the\n", + "problem calls therefore for minimization algorithms. This forms the\n", + "bottle neck of all machine learning algorithms, namely how to find\n", + "reliable minima of a multi-variable function. This leads us to the\n", + "family of gradient descent methods. The latter are the working horses\n", + "of basically all modern machine learning algorithms.\n", + "\n", + "We note also that many of the topics discussed here on logistic \n", + "regression are also commonly used in modern supervised Deep Learning\n", + "models, as we will see later." + ] + }, + { + "cell_type": "markdown", + "id": "7c54a4ff", + "metadata": { + "editable": true + }, + "source": [ + "## Basics\n", + "\n", + "We consider the case where the dependent variables, also called the\n", + "responses or the outcomes, $y_i$ are discrete and only take values\n", + "from $k=0,\\dots,K-1$ (i.e. $K$ classes).\n", + "\n", + "The goal is to predict the\n", + "output classes from the design matrix $\\boldsymbol{X}\\in\\mathbb{R}^{n\\times p}$\n", + "made of $n$ samples, each of which carries $p$ features or predictors. The\n", + "primary goal is to identify the classes to which new unseen samples\n", + "belong.\n", + "\n", + "Let us specialize to the case of two classes only, with outputs\n", + "$y_i=0$ and $y_i=1$. Our outcomes could represent the status of a\n", + "credit card user that could default or not on her/his credit card\n", + "debt. That is" + ] + }, + { + "cell_type": "markdown", + "id": "838d316e", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "y_i = \\begin{bmatrix} 0 & \\mathrm{no}\\\\ 1 & \\mathrm{yes} \\end{bmatrix}.\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "285bf3a2", + "metadata": { + "editable": true + }, + "source": [ + "## Linear classifier\n", + "\n", + "Before moving to the logistic model, let us try to use our linear\n", + "regression model to classify these two outcomes. We could for example\n", + "fit a linear model to the default case if $y_i > 0.5$ and the no\n", + "default case $y_i \\leq 0.5$.\n", + "\n", + "We would then have our \n", + "weighted linear combination, namely" + ] + }, + { + "cell_type": "markdown", + "id": "6b03bb50", + "metadata": { + "editable": true + }, + "source": [ + "\n", + "
            \n", + "\n", + "$$\n", + "\\begin{equation}\n", + "\\boldsymbol{y} = \\boldsymbol{X}^T\\boldsymbol{\\beta} + \\boldsymbol{\\epsilon},\n", + "\\label{_auto1} \\tag{1}\n", + "\\end{equation}\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "f153c89c", + "metadata": { + "editable": true + }, + "source": [ + "where $\\boldsymbol{y}$ is a vector representing the possible outcomes, $\\boldsymbol{X}$ is our\n", + "$n\\times p$ design matrix and $\\boldsymbol{\\beta}$ represents our estimators/predictors." + ] + }, + { + "cell_type": "markdown", + "id": "487ff8fe", + "metadata": { + "editable": true + }, + "source": [ + "## Some selected properties\n", + "\n", + "The main problem with our function is that it takes values on the\n", + "entire real axis. In the case of logistic regression, however, the\n", + "labels $y_i$ are discrete variables. A typical example is the credit\n", + "card data discussed below here, where we can set the state of\n", + "defaulting the debt to $y_i=1$ and not to $y_i=0$ for one the persons\n", + "in the data set (see the full example below).\n", + "\n", + "One simple way to get a discrete output is to have sign\n", + "functions that map the output of a linear regressor to values $\\{0,1\\}$,\n", + "$f(s_i)=sign(s_i)=1$ if $s_i\\ge 0$ and 0 if otherwise. \n", + "We will encounter this model in our first demonstration of neural networks.\n", + "\n", + "Historically it is called the **perceptron** model in the machine learning\n", + "literature. This model is extremely simple. However, in many cases it is more\n", + "favorable to use a ``soft\" classifier that outputs\n", + "the probability of a given category. This leads us to the logistic function." + ] + }, + { + "cell_type": "markdown", + "id": "e64f34ad", + "metadata": { + "editable": true + }, + "source": [ + "## Simple example\n", + "\n", + "The following example on data for coronary heart disease (CHD) as function of age may serve as an illustration. In the code here we read and plot whether a person has had CHD (output = 1) or not (output = 0). This ouput is plotted the person's against age. Clearly, the figure shows that attempting to make a standard linear regression fit may not be very meaningful." + ] + }, + { + "cell_type": "code", + "execution_count": 1, + "id": "eb389ab6", + "metadata": { + "collapsed": false, + "editable": true + }, + "outputs": [], + "source": [ + "%matplotlib inline\n", + "\n", + "# Common imports\n", + "import os\n", + "import numpy as np\n", + "import pandas as pd\n", + "import matplotlib.pyplot as plt\n", + "from sklearn.linear_model import LinearRegression, Ridge, Lasso\n", + "from sklearn.model_selection import train_test_split\n", + "from sklearn.utils import resample\n", + "from sklearn.metrics import mean_squared_error\n", + "from IPython.display import display\n", + "from pylab import plt, mpl\n", + "plt.style.use('seaborn')\n", + "mpl.rcParams['font.family'] = 'serif'\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(\"chddata.csv\"),'r')\n", + "\n", + "# Read the chd data as csv file and organize the data into arrays with age group, age, and chd\n", + "chd = pd.read_csv(infile, names=('ID', 'Age', 'Agegroup', 'CHD'))\n", + "chd.columns = ['ID', 'Age', 'Agegroup', 'CHD']\n", + "output = chd['CHD']\n", + "age = chd['Age']\n", + "agegroup = chd['Agegroup']\n", + "numberID = chd['ID'] \n", + "display(chd)\n", + "\n", + "plt.scatter(age, output, marker='o')\n", + "plt.axis([18,70.0,-0.1, 1.2])\n", + "plt.xlabel(r'Age')\n", + "plt.ylabel(r'CHD')\n", + "plt.title(r'Age distribution and Coronary heart disease')\n", + "plt.show()" + ] + }, + { + "cell_type": "markdown", + "id": "d0f9afe8", + "metadata": { + "editable": true + }, + "source": [ + "## Plotting the mean value for each group\n", + "\n", + "What we could attempt however is to plot the mean value for each group." + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "id": "499179b1", + "metadata": { + "collapsed": false, + "editable": true + }, + "outputs": [], + "source": [ + "agegroupmean = np.array([0.1, 0.133, 0.250, 0.333, 0.462, 0.625, 0.765, 0.800])\n", + "group = np.array([1, 2, 3, 4, 5, 6, 7, 8])\n", + "plt.plot(group, agegroupmean, \"r-\")\n", + "plt.axis([0,9,0, 1.0])\n", + "plt.xlabel(r'Age group')\n", + "plt.ylabel(r'CHD mean values')\n", + "plt.title(r'Mean values for each age group')\n", + "plt.show()" + ] + }, + { + "cell_type": "markdown", + "id": "cf814d3f", + "metadata": { + "editable": true + }, + "source": [ + "We are now trying to find a function $f(y\\vert x)$, that is a function which gives us an expected value for the output $y$ with a given input $x$.\n", + "In standard linear regression with a linear dependence on $x$, we would write this in terms of our model" + ] + }, + { + "cell_type": "markdown", + "id": "28658b61", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "f(y_i\\vert x_i)=\\beta_0+\\beta_1 x_i.\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "dec92579", + "metadata": { + "editable": true + }, + "source": [ + "This expression implies however that $f(y_i\\vert x_i)$ could take any\n", + "value from minus infinity to plus infinity. If we however let\n", + "$f(y\\vert y)$ be represented by the mean value, the above example\n", + "shows us that we can constrain the function to take values between\n", + "zero and one, that is we have $0 \\le f(y_i\\vert x_i) \\le 1$. Looking\n", + "at our last curve we see also that it has an S-shaped form. This leads\n", + "us to a very popular model for the function $f$, namely the so-called\n", + "Sigmoid function or logistic model. We will consider this function as\n", + "representing the probability for finding a value of $y_i$ with a given\n", + "$x_i$." + ] + }, + { + "cell_type": "markdown", + "id": "92ad450b", + "metadata": { + "editable": true + }, + "source": [ + "## The logistic function\n", + "\n", + "Another widely studied model, is the so-called \n", + "perceptron model, which is an example of a \"hard classification\" model. We\n", + "will encounter this model when we discuss neural networks as\n", + "well. Each datapoint is deterministically assigned to a category (i.e\n", + "$y_i=0$ or $y_i=1$). In many cases, and the coronary heart disease data forms one of many such examples, it is favorable to have a \"soft\"\n", + "classifier that outputs the probability of a given category rather\n", + "than a single value. For example, given $x_i$, the classifier\n", + "outputs the probability of being in a category $k$. Logistic regression\n", + "is the most common example of a so-called soft classifier. In logistic\n", + "regression, the probability that a data point $x_i$\n", + "belongs to a category $y_i=\\{0,1\\}$ is given by the so-called logit function (or Sigmoid) which is meant to represent the likelihood for a given event," + ] + }, + { + "cell_type": "markdown", + "id": "35cf0753", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "p(t) = \\frac{1}{1+\\mathrm \\exp{-t}}=\\frac{\\exp{t}}{1+\\mathrm \\exp{t}}.\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "fd212f79", + "metadata": { + "editable": true + }, + "source": [ + "Note that $1-p(t)= p(-t)$." + ] + }, + { + "cell_type": "markdown", + "id": "34d6199f", + "metadata": { + "editable": true + }, + "source": [ + "## Examples of likelihood functions used in logistic regression and nueral networks\n", + "\n", + "The following code plots the logistic function, the step function and other functions we will encounter from here and on." + ] + }, + { + "cell_type": "code", + "execution_count": 3, + "id": "174df351", + "metadata": { + "collapsed": false, + "editable": true + }, + "outputs": [], + "source": [ + "\"\"\"The sigmoid function (or the logistic curve) is a\n", + "function that takes any real number, z, and outputs a number (0,1).\n", + "It is useful in neural networks for assigning weights on a relative scale.\n", + "The value z is the weighted sum of parameters involved in the learning algorithm.\"\"\"\n", + "\n", + "import numpy\n", + "import matplotlib.pyplot as plt\n", + "import math as mt\n", + "\n", + "z = numpy.arange(-5, 5, .1)\n", + "sigma_fn = numpy.vectorize(lambda z: 1/(1+numpy.exp(-z)))\n", + "sigma = sigma_fn(z)\n", + "\n", + "fig = plt.figure()\n", + "ax = fig.add_subplot(111)\n", + "ax.plot(z, sigma)\n", + "ax.set_ylim([-0.1, 1.1])\n", + "ax.set_xlim([-5,5])\n", + "ax.grid(True)\n", + "ax.set_xlabel('z')\n", + "ax.set_title('sigmoid function')\n", + "\n", + "plt.show()\n", + "\n", + "\"\"\"Step Function\"\"\"\n", + "z = numpy.arange(-5, 5, .02)\n", + "step_fn = numpy.vectorize(lambda z: 1.0 if z >= 0.0 else 0.0)\n", + "step = step_fn(z)\n", + "\n", + "fig = plt.figure()\n", + "ax = fig.add_subplot(111)\n", + "ax.plot(z, step)\n", + "ax.set_ylim([-0.5, 1.5])\n", + "ax.set_xlim([-5,5])\n", + "ax.grid(True)\n", + "ax.set_xlabel('z')\n", + "ax.set_title('step function')\n", + "\n", + "plt.show()\n", + "\n", + "\"\"\"tanh Function\"\"\"\n", + "z = numpy.arange(-2*mt.pi, 2*mt.pi, 0.1)\n", + "t = numpy.tanh(z)\n", + "\n", + "fig = plt.figure()\n", + "ax = fig.add_subplot(111)\n", + "ax.plot(z, t)\n", + "ax.set_ylim([-1.0, 1.0])\n", + "ax.set_xlim([-2*mt.pi,2*mt.pi])\n", + "ax.grid(True)\n", + "ax.set_xlabel('z')\n", + "ax.set_title('tanh function')\n", + "\n", + "plt.show()" + ] + }, + { + "cell_type": "markdown", + "id": "89fdf010", + "metadata": { + "editable": true + }, + "source": [ + "## Two parameters\n", + "\n", + "We assume now that we have two classes with $y_i$ either $0$ or $1$. Furthermore we assume also that we have only two parameters $\\beta$ in our fitting of the Sigmoid function, that is we define probabilities" + ] + }, + { + "cell_type": "markdown", + "id": "6216d0ba", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\begin{align*}\n", + "p(y_i=1|x_i,\\boldsymbol{\\beta}) &= \\frac{\\exp{(\\beta_0+\\beta_1x_i)}}{1+\\exp{(\\beta_0+\\beta_1x_i)}},\\nonumber\\\\\n", + "p(y_i=0|x_i,\\boldsymbol{\\beta}) &= 1 - p(y_i=1|x_i,\\boldsymbol{\\beta}),\n", + "\\end{align*}\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "daa87f9a", + "metadata": { + "editable": true + }, + "source": [ + "where $\\boldsymbol{\\beta}$ are the weights we wish to extract from data, in our case $\\beta_0$ and $\\beta_1$. \n", + "\n", + "Note that we used" + ] + }, + { + "cell_type": "markdown", + "id": "85088741", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "p(y_i=0\\vert x_i, \\boldsymbol{\\beta}) = 1-p(y_i=1\\vert x_i, \\boldsymbol{\\beta}).\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "035508f1", + "metadata": { + "editable": true + }, + "source": [ + "## Maximum likelihood\n", + "\n", + "In order to define the total likelihood for all possible outcomes from a \n", + "dataset $\\mathcal{D}=\\{(y_i,x_i)\\}$, with the binary labels\n", + "$y_i\\in\\{0,1\\}$ and where the data points are drawn independently, we use the so-called [Maximum Likelihood Estimation](https://en.wikipedia.org/wiki/Maximum_likelihood_estimation) (MLE) principle. \n", + "We aim thus at maximizing \n", + "the probability of seeing the observed data. We can then approximate the \n", + "likelihood in terms of the product of the individual probabilities of a specific outcome $y_i$, that is" + ] + }, + { + "cell_type": "markdown", + "id": "8972c037", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\begin{align*}\n", + "P(\\mathcal{D}|\\boldsymbol{\\beta})& = \\prod_{i=1}^n \\left[p(y_i=1|x_i,\\boldsymbol{\\beta})\\right]^{y_i}\\left[1-p(y_i=1|x_i,\\boldsymbol{\\beta}))\\right]^{1-y_i}\\nonumber \\\\\n", + "\\end{align*}\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "3eac3b65", + "metadata": { + "editable": true + }, + "source": [ + "from which we obtain the log-likelihood and our **cost/loss** function" + ] + }, + { + "cell_type": "markdown", + "id": "133efea3", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\mathcal{C}(\\boldsymbol{\\beta}) = \\sum_{i=1}^n \\left( y_i\\log{p(y_i=1|x_i,\\boldsymbol{\\beta})} + (1-y_i)\\log\\left[1-p(y_i=1|x_i,\\boldsymbol{\\beta}))\\right]\\right).\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "566aabdb", + "metadata": { + "editable": true + }, + "source": [ + "## The cost function rewritten\n", + "\n", + "Reordering the logarithms, we can rewrite the **cost/loss** function as" + ] + }, + { + "cell_type": "markdown", + "id": "4cf34420", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\mathcal{C}(\\boldsymbol{\\beta}) = \\sum_{i=1}^n \\left(y_i(\\beta_0+\\beta_1x_i) -\\log{(1+\\exp{(\\beta_0+\\beta_1x_i)})}\\right).\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "90ccf024", + "metadata": { + "editable": true + }, + "source": [ + "The maximum likelihood estimator is defined as the set of parameters that maximize the log-likelihood where we maximize with respect to $\\beta$.\n", + "Since the cost (error) function is just the negative log-likelihood, for logistic regression we have that" + ] + }, + { + "cell_type": "markdown", + "id": "824ccda5", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\mathcal{C}(\\boldsymbol{\\beta})=-\\sum_{i=1}^n \\left(y_i(\\beta_0+\\beta_1x_i) -\\log{(1+\\exp{(\\beta_0+\\beta_1x_i)})}\\right).\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "bbfb14d3", + "metadata": { + "editable": true + }, + "source": [ + "This equation is known in statistics as the **cross entropy**. Finally, we note that just as in linear regression, \n", + "in practice we often supplement the cross-entropy with additional regularization terms, usually $L_1$ and $L_2$ regularization as we did for Ridge and Lasso regression." + ] + }, + { + "cell_type": "markdown", + "id": "bfc7e1be", + "metadata": { + "editable": true + }, + "source": [ + "## Minimizing the cross entropy\n", + "\n", + "The cross entropy is a convex function of the weights $\\boldsymbol{\\beta}$ and,\n", + "therefore, any local minimizer is a global minimizer. \n", + "\n", + "Minimizing this\n", + "cost function with respect to the two parameters $\\beta_0$ and $\\beta_1$ we obtain" + ] + }, + { + "cell_type": "markdown", + "id": "bae6c599", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\frac{\\partial \\mathcal{C}(\\boldsymbol{\\beta})}{\\partial \\beta_0} = -\\sum_{i=1}^n \\left(y_i -\\frac{\\exp{(\\beta_0+\\beta_1x_i)}}{1+\\exp{(\\beta_0+\\beta_1x_i)}}\\right),\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "550a7b2b", + "metadata": { + "editable": true + }, + "source": [ + "and" + ] + }, + { + "cell_type": "markdown", + "id": "5b072ab8", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\frac{\\partial \\mathcal{C}(\\boldsymbol{\\beta})}{\\partial \\beta_1} = -\\sum_{i=1}^n \\left(y_ix_i -x_i\\frac{\\exp{(\\beta_0+\\beta_1x_i)}}{1+\\exp{(\\beta_0+\\beta_1x_i)}}\\right).\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "9ed08b1d", + "metadata": { + "editable": true + }, + "source": [ + "## A more compact expression\n", + "\n", + "Let us now define a vector $\\boldsymbol{y}$ with $n$ elements $y_i$, an\n", + "$n\\times p$ matrix $\\boldsymbol{X}$ which contains the $x_i$ values and a\n", + "vector $\\boldsymbol{p}$ of fitted probabilities $p(y_i\\vert x_i,\\boldsymbol{\\beta})$. We can rewrite in a more compact form the first\n", + "derivative of cost function as" + ] + }, + { + "cell_type": "markdown", + "id": "11b32bd3", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\frac{\\partial \\mathcal{C}(\\boldsymbol{\\beta})}{\\partial \\boldsymbol{\\beta}} = -\\boldsymbol{X}^T\\left(\\boldsymbol{y}-\\boldsymbol{p}\\right).\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "47dec6ef", + "metadata": { + "editable": true + }, + "source": [ + "If we in addition define a diagonal matrix $\\boldsymbol{W}$ with elements \n", + "$p(y_i\\vert x_i,\\boldsymbol{\\beta})(1-p(y_i\\vert x_i,\\boldsymbol{\\beta})$, we can obtain a compact expression of the second derivative as" + ] + }, + { + "cell_type": "markdown", + "id": "43e12319", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\frac{\\partial^2 \\mathcal{C}(\\boldsymbol{\\beta})}{\\partial \\boldsymbol{\\beta}\\partial \\boldsymbol{\\beta}^T} = \\boldsymbol{X}^T\\boldsymbol{W}\\boldsymbol{X}.\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "a0c3198a", + "metadata": { + "editable": true + }, + "source": [ + "## Extending to more predictors\n", + "\n", + "Within a binary classification problem, we can easily expand our model to include multiple predictors. Our ratio between likelihoods is then with $p$ predictors" + ] + }, + { + "cell_type": "markdown", + "id": "fb7b011d", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\log{ \\frac{p(\\boldsymbol{\\beta}\\boldsymbol{x})}{1-p(\\boldsymbol{\\beta}\\boldsymbol{x})}} = \\beta_0+\\beta_1x_1+\\beta_2x_2+\\dots+\\beta_px_p.\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "acfaf754", + "metadata": { + "editable": true + }, + "source": [ + "Here we defined $\\boldsymbol{x}=[1,x_1,x_2,\\dots,x_p]$ and $\\boldsymbol{\\beta}=[\\beta_0, \\beta_1, \\dots, \\beta_p]$ leading to" + ] + }, + { + "cell_type": "markdown", + "id": "33e112ba", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "p(\\boldsymbol{\\beta}\\boldsymbol{x})=\\frac{ \\exp{(\\beta_0+\\beta_1x_1+\\beta_2x_2+\\dots+\\beta_px_p)}}{1+\\exp{(\\beta_0+\\beta_1x_1+\\beta_2x_2+\\dots+\\beta_px_p)}}.\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "75173b91", + "metadata": { + "editable": true + }, + "source": [ + "## Including more classes\n", + "\n", + "Till now we have mainly focused on two classes, the so-called binary\n", + "system. Suppose we wish to extend to $K$ classes. Let us for the sake\n", + "of simplicity assume we have only two predictors. We have then following model" + ] + }, + { + "cell_type": "markdown", + "id": "60ea4955", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\log{\\frac{p(C=1\\vert x)}{p(K\\vert x)}} = \\beta_{10}+\\beta_{11}x_1,\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "2ac0c3fe", + "metadata": { + "editable": true + }, + "source": [ + "and" + ] + }, + { + "cell_type": "markdown", + "id": "c5633fe2", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\log{\\frac{p(C=2\\vert x)}{p(K\\vert x)}} = \\beta_{20}+\\beta_{21}x_1,\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "af63c7fc", + "metadata": { + "editable": true + }, + "source": [ + "and so on till the class $C=K-1$ class" + ] + }, + { + "cell_type": "markdown", + "id": "68713442", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\log{\\frac{p(C=K-1\\vert x)}{p(K\\vert x)}} = \\beta_{(K-1)0}+\\beta_{(K-1)1}x_1,\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "ae89a4ec", + "metadata": { + "editable": true + }, + "source": [ + "and the model is specified in term of $K-1$ so-called log-odds or\n", + "**logit** transformations." + ] + }, + { + "cell_type": "markdown", + "id": "dc86b762", + "metadata": { + "editable": true + }, + "source": [ + "## More classes\n", + "\n", + "In our discussion of neural networks we will encounter the above again\n", + "in terms of a slightly modified function, the so-called **Softmax** function.\n", + "\n", + "The softmax function is used in various multiclass classification\n", + "methods, such as multinomial logistic regression (also known as\n", + "softmax regression), multiclass linear discriminant analysis, naive\n", + "Bayes classifiers, and artificial neural networks. Specifically, in\n", + "multinomial logistic regression and linear discriminant analysis, the\n", + "input to the function is the result of $K$ distinct linear functions,\n", + "and the predicted probability for the $k$-th class given a sample\n", + "vector $\\boldsymbol{x}$ and a weighting vector $\\boldsymbol{\\beta}$ is (with two\n", + "predictors):" + ] + }, + { + "cell_type": "markdown", + "id": "4109330d", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "p(C=k\\vert \\mathbf {x} )=\\frac{\\exp{(\\beta_{k0}+\\beta_{k1}x_1)}}{1+\\sum_{l=1}^{K-1}\\exp{(\\beta_{l0}+\\beta_{l1}x_1)}}.\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "dcf3f4c4", + "metadata": { + "editable": true + }, + "source": [ + "It is easy to extend to more predictors. The final class is" + ] + }, + { + "cell_type": "markdown", + "id": "148d45e3", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "p(C=K\\vert \\mathbf {x} )=\\frac{1}{1+\\sum_{l=1}^{K-1}\\exp{(\\beta_{l0}+\\beta_{l1}x_1)}},\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "5c082186", + "metadata": { + "editable": true + }, + "source": [ + "and they sum to one. Our earlier discussions were all specialized to\n", + "the case with two classes only. It is easy to see from the above that\n", + "what we derived earlier is compatible with these equations.\n", + "\n", + "To find the optimal parameters we would typically use a gradient\n", + "descent method. Newton's method and gradient descent methods are\n", + "discussed in the material on [optimization\n", + "methods](https://compphysics.github.io/MachineLearning/doc/pub/Splines/html/Splines-bs.html)." + ] + }, + { + "cell_type": "markdown", + "id": "d5068fec", + "metadata": { + "editable": true + }, + "source": [ + "## Friday September 23" + ] + }, + { + "cell_type": "markdown", + "id": "d33abe73", + "metadata": { + "editable": true + }, + "source": [ + "## Searching for Optimal Regularization Parameters $\\lambda$\n", + "\n", + "In project 1, when using Ridge and Lasso regression, we end up\n", + "searching for the optimal parameter $\\lambda$ which minimizes our\n", + "selected scores (MSE or $R2$ values for example). The brute force\n", + "approach, as discussed in the code here for Ridge regression, consists\n", + "in evaluating the MSE as function of different $\\lambda$ values.\n", + "Based on these calculations, one tries then to determine the value of the hyperparameter $\\lambda$\n", + "which results in optimal scores (for example the smallest MSE or an $R2=1$)." + ] + }, + { + "cell_type": "code", + "execution_count": 4, + "id": "d445125f", + "metadata": { + "collapsed": false, + "editable": true + }, + "outputs": [], + "source": [ + "import numpy as np\n", + "import pandas as pd\n", + "import matplotlib.pyplot as plt\n", + "from sklearn.model_selection import train_test_split\n", + "from sklearn import linear_model\n", + "\n", + "def MSE(y_data,y_model):\n", + " n = np.size(y_model)\n", + " return np.sum((y_data-y_model)**2)/n\n", + "# A seed just to ensure that the random numbers are the same for every run.\n", + "# Useful for eventual debugging.\n", + "np.random.seed(2021)\n", + "\n", + "n = 100\n", + "x = np.random.rand(n)\n", + "y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.randn(n)\n", + "\n", + "Maxpolydegree = 5\n", + "X = np.zeros((n,Maxpolydegree-1))\n", + "\n", + "for degree in range(1,Maxpolydegree): #No intercept column\n", + " X[:,degree-1] = x**(degree)\n", + "\n", + "# We split the data in test and training data\n", + "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)\n", + "\n", + "# Decide which values of lambda to use\n", + "nlambdas = 500\n", + "MSERidgePredict = np.zeros(nlambdas)\n", + "lambdas = np.logspace(-4, 2, nlambdas)\n", + "for i in range(nlambdas):\n", + " lmb = lambdas[i]\n", + " RegRidge = linear_model.Ridge(lmb)\n", + " RegRidge.fit(X_train,y_train)\n", + " ypredictRidge = RegRidge.predict(X_test)\n", + " MSERidgePredict[i] = MSE(y_test,ypredictRidge)\n", + "\n", + "# Now plot the results\n", + "plt.figure()\n", + "plt.plot(np.log10(lambdas), MSERidgePredict, 'g--', label = 'MSE SL Ridge Test')\n", + "plt.xlabel('log10(lambda)')\n", + "plt.ylabel('MSE')\n", + "plt.legend()\n", + "plt.show()" + ] + }, + { + "cell_type": "markdown", + "id": "3ad4a826", + "metadata": { + "editable": true + }, + "source": [ + "Here we have performed a rather data greedy calculation as function of the regularization parameter $\\lambda$. There is no resampling here. The latter can easily be added by employing the function **RidgeCV** instead of just calling the **Ridge** function. For **RidgeCV** we need to pass the array of $\\lambda$ values.\n", + "By inspecting the figure we can in turn determine which is the optimal regularization parameter.\n", + "This becomes however less functional in the long run." + ] + }, + { + "cell_type": "markdown", + "id": "ff84b2e7", + "metadata": { + "editable": true + }, + "source": [ + "## Grid Search\n", + "\n", + "An alternative is to use the so-called grid search functionality\n", + "included with the library **Scikit-Learn**, as demonstrated for the same\n", + "example here." + ] + }, + { + "cell_type": "code", + "execution_count": 5, + "id": "4b834234", + "metadata": { + "collapsed": false, + "editable": true + }, + "outputs": [], + "source": [ + "import numpy as np\n", + "from sklearn.model_selection import train_test_split\n", + "from sklearn.linear_model import Ridge\n", + "from sklearn.model_selection import GridSearchCV\n", + "\n", + "def R2(y_data, y_model):\n", + " return 1 - np.sum((y_data - y_model) ** 2) / np.sum((y_data - np.mean(y_data)) ** 2)\n", + "\n", + "def MSE(y_data,y_model):\n", + " n = np.size(y_model)\n", + " return np.sum((y_data-y_model)**2)/n\n", + "\n", + "# A seed just to ensure that the random numbers are the same for every run.\n", + "# Useful for eventual debugging.\n", + "np.random.seed(2021)\n", + "\n", + "n = 100\n", + "x = np.random.rand(n)\n", + "y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.randn(n)\n", + "\n", + "Maxpolydegree = 5\n", + "X = np.zeros((n,Maxpolydegree-1))\n", + "\n", + "for degree in range(1,Maxpolydegree): #No intercept column\n", + " X[:,degree-1] = x**(degree)\n", + "\n", + "# We split the data in test and training data\n", + "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)\n", + "\n", + "# Decide which values of lambda to use\n", + "nlambdas = 10\n", + "lambdas = np.logspace(-4, 2, nlambdas)\n", + "# create and fit a ridge regression model, testing each alpha\n", + "model = Ridge()\n", + "gridsearch = GridSearchCV(estimator=model, param_grid=dict(alpha=lambdas))\n", + "gridsearch.fit(X_train, y_train)\n", + "print(gridsearch)\n", + "ypredictRidge = gridsearch.predict(X_test)\n", + "# summarize the results of the grid search\n", + "print(f\"Best estimated lambda-value: {gridsearch.best_estimator_.alpha}\")\n", + "print(f\"MSE score: {MSE(y_test,ypredictRidge)}\")\n", + "print(f\"R2 score: {R2(y_test,ypredictRidge)}\")" + ] + }, + { + "cell_type": "markdown", + "id": "7514764b", + "metadata": { + "editable": true + }, + "source": [ + "By default the grid search function includes cross validation with\n", + "five folds. The [Scikit-Learn\n", + "documentation](https://scikit-learn.org/stable/modules/generated/sklearn.model_selection.GridSearchCV.html#sklearn.model_selection.GridSearchCV)\n", + "contains more information on how to set the different parameters.\n", + "\n", + "If we take out the random noise, running the above codes results in $\\lambda=0$ yielding the best fit." + ] + }, + { + "cell_type": "markdown", + "id": "7991ea9f", + "metadata": { + "editable": true + }, + "source": [ + "## Randomized Grid Search\n", + "\n", + "An alternative to the above manual grid set up, is to use a random\n", + "search where the parameters are tuned from a random distribution\n", + "(uniform below) for a fixed number of iterations. A model is\n", + "constructed and evaluated for each combination of chosen parameters.\n", + "We repeat the previous example but now with a random search. Note\n", + "that values of $\\lambda$ are now limited to be within $x\\in\n", + "[0,1]$. This domain may not be the most relevant one for the specific\n", + "case under study." + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "id": "ae4f99a1", + "metadata": { + "collapsed": false, + "editable": true + }, + "outputs": [], + "source": [ + "import numpy as np\n", + "from sklearn.model_selection import train_test_split\n", + "from sklearn.linear_model import Ridge\n", + "from sklearn.model_selection import GridSearchCV\n", + "from scipy.stats import uniform as randuniform\n", + "from sklearn.model_selection import RandomizedSearchCV\n", + "\n", + "\n", + "def R2(y_data, y_model):\n", + " return 1 - np.sum((y_data - y_model) ** 2) / np.sum((y_data - np.mean(y_data)) ** 2)\n", + "\n", + "def MSE(y_data,y_model):\n", + " n = np.size(y_model)\n", + " return np.sum((y_data-y_model)**2)/n\n", + "\n", + "# A seed just to ensure that the random numbers are the same for every run.\n", + "# Useful for eventual debugging.\n", + "np.random.seed(2021)\n", + "\n", + "n = 100\n", + "x = np.random.rand(n)\n", + "y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.randn(n)\n", + "\n", + "Maxpolydegree = 5\n", + "X = np.zeros((n,Maxpolydegree-1))\n", + "\n", + "for degree in range(1,Maxpolydegree): #No intercept column\n", + " X[:,degree-1] = x**(degree)\n", + "\n", + "# We split the data in test and training data\n", + "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)\n", + "\n", + "param_grid = {'alpha': randuniform()}\n", + "# create and fit a ridge regression model, testing each alpha\n", + "model = Ridge()\n", + "gridsearch = RandomizedSearchCV(estimator=model, param_distributions=param_grid, n_iter=100)\n", + "gridsearch.fit(X_train, y_train)\n", + "print(gridsearch)\n", + "ypredictRidge = gridsearch.predict(X_test)\n", + "# summarize the results of the grid search\n", + "print(f\"Best estimated lambda-value: {gridsearch.best_estimator_.alpha}\")\n", + "print(f\"MSE score: {MSE(y_test,ypredictRidge)}\")\n", + "print(f\"R2 score: {R2(y_test,ypredictRidge)}\")" + ] + }, + { + "cell_type": "markdown", + "id": "91278700", + "metadata": { + "editable": true + }, + "source": [ + "## Wisconsin Cancer Data\n", + "\n", + "We show here how we can use a simple regression case on the breast\n", + "cancer data using Logistic regression as our algorithm for\n", + "classification." + ] + }, + { + "cell_type": "code", + "execution_count": 7, + "id": "8f006bb8", + "metadata": { + "collapsed": false, + "editable": true + }, + "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.linear_model import LogisticRegression\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)))" + ] + }, + { + "cell_type": "markdown", + "id": "015f8534", + "metadata": { + "editable": true + }, + "source": [ + "## Using the correlation matrix\n", + "\n", + "In addition to the above scores, we could also study the covariance (and the correlation matrix).\n", + "We use **Pandas** to compute the correlation matrix." + ] + }, + { + "cell_type": "code", + "execution_count": 8, + "id": "fa6d3421", + "metadata": { + "collapsed": false, + "editable": true + }, + "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.linear_model import LogisticRegression\n", + "cancer = load_breast_cancer()\n", + "import pandas as pd\n", + "# Making a data frame\n", + "cancerpd = pd.DataFrame(cancer.data, columns=cancer.feature_names)\n", + "\n", + "fig, axes = plt.subplots(15,2,figsize=(10,20))\n", + "malignant = cancer.data[cancer.target == 0]\n", + "benign = cancer.data[cancer.target == 1]\n", + "ax = axes.ravel()\n", + "\n", + "for i in range(30):\n", + " _, bins = np.histogram(cancer.data[:,i], bins =50)\n", + " ax[i].hist(malignant[:,i], bins = bins, alpha = 0.5)\n", + " ax[i].hist(benign[:,i], bins = bins, alpha = 0.5)\n", + " ax[i].set_title(cancer.feature_names[i])\n", + " ax[i].set_yticks(())\n", + "ax[0].set_xlabel(\"Feature magnitude\")\n", + "ax[0].set_ylabel(\"Frequency\")\n", + "ax[0].legend([\"Malignant\", \"Benign\"], loc =\"best\")\n", + "fig.tight_layout()\n", + "plt.show()\n", + "\n", + "import seaborn as sns\n", + "correlation_matrix = cancerpd.corr().round(1)\n", + "# use the heatmap function from seaborn to plot the correlation matrix\n", + "# annot = True to print the values inside the square\n", + "plt.figure(figsize=(15,8))\n", + "sns.heatmap(data=correlation_matrix, annot=True)\n", + "plt.show()" + ] + }, + { + "cell_type": "markdown", + "id": "49ff3f87", + "metadata": { + "editable": true + }, + "source": [ + "## Discussing the correlation data\n", + "\n", + "In the above example we note two things. In the first plot we display\n", + "the overlap of benign and malignant tumors as functions of the various\n", + "features in the Wisconsing breast cancer data set. We see that for\n", + "some of the features we can distinguish clearly the benign and\n", + "malignant cases while for other features we cannot. This can point to\n", + "us which features may be of greater interest when we wish to classify\n", + "a benign or not benign tumour.\n", + "\n", + "In the second figure we have computed the so-called correlation\n", + "matrix, which in our case with thirty features becomes a $30\\times 30$\n", + "matrix.\n", + "\n", + "We constructed this matrix using **pandas** via the statements" + ] + }, + { + "cell_type": "code", + "execution_count": 9, + "id": "3e800820", + "metadata": { + "collapsed": false, + "editable": true + }, + "outputs": [], + "source": [ + "cancerpd = pd.DataFrame(cancer.data, columns=cancer.feature_names)" + ] + }, + { + "cell_type": "markdown", + "id": "9f831f22", + "metadata": { + "editable": true + }, + "source": [ + "and then" + ] + }, + { + "cell_type": "code", + "execution_count": 10, + "id": "e663ae38", + "metadata": { + "collapsed": false, + "editable": true + }, + "outputs": [], + "source": [ + "correlation_matrix = cancerpd.corr().round(1)" + ] + }, + { + "cell_type": "markdown", + "id": "737193a2", + "metadata": { + "editable": true + }, + "source": [ + "Diagonalizing this matrix we can in turn say something about which\n", + "features are of relevance and which are not. This leads us to\n", + "the classical Principal Component Analysis (PCA) theorem with\n", + "applications. This will be discussed later this semester ([week 43](https://compphysics.github.io/MachineLearning/doc/pub/week43/html/week43-bs.html))." + ] + }, + { + "cell_type": "markdown", + "id": "25104113", + "metadata": { + "editable": true + }, + "source": [ + "## Other measures in classification studies: Cancer Data again" + ] + }, + { + "cell_type": "code", + "execution_count": 11, + "id": "6783b76e", + "metadata": { + "collapsed": false, + "editable": true + }, + "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.linear_model import LogisticRegression\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", + "\n", + "from sklearn.preprocessing import LabelEncoder\n", + "from sklearn.model_selection import cross_validate\n", + "#Cross validation\n", + "accuracy = cross_validate(logreg,X_test,y_test,cv=10)['test_score']\n", + "print(accuracy)\n", + "print(\"Test set accuracy with Logistic Regression: {:.2f}\".format(logreg.score(X_test,y_test)))\n", + "\n", + "import scikitplot as skplt\n", + "y_pred = logreg.predict(X_test)\n", + "skplt.metrics.plot_confusion_matrix(y_test, y_pred, normalize=True)\n", + "plt.show()\n", + "y_probas = logreg.predict_proba(X_test)\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", + "id": "bdb4928a", + "metadata": { + "editable": true + }, + "source": [ + "## Optimization, the central part of any Machine Learning algortithm\n", + "\n", + "[Overview Video, why do we care about gradient methods?](https://www.uio.no/studier/emner/matnat/fys/FYS-STK3155/h20/forelesningsvideoer/OverarchingAimsWeek39.mp4?vrtx=view-as-webpage)\n", + "\n", + "Almost every problem in machine learning and data science starts with\n", + "a dataset $X$, a model $g(\\beta)$, which is a function of the\n", + "parameters $\\beta$ and a cost function $C(X, g(\\beta))$ that allows\n", + "us to judge how well the model $g(\\beta)$ explains the observations\n", + "$X$. The model is fit by finding the values of $\\beta$ that minimize\n", + "the cost function. Ideally we would be able to solve for $\\beta$\n", + "analytically, however this is not possible in general and we must use\n", + "some approximative/numerical method to compute the minimum." + ] + }, + { + "cell_type": "markdown", + "id": "8b73ff85", + "metadata": { + "editable": true + }, + "source": [ + "## Revisiting our Logistic Regression case\n", + "\n", + "In our discussion on Logistic Regression we studied the \n", + "case of\n", + "two classes, with $y_i$ either\n", + "$0$ or $1$. Furthermore we assumed also that we have only two\n", + "parameters $\\beta$ in our fitting, that is we\n", + "defined probabilities" + ] + }, + { + "cell_type": "markdown", + "id": "2b771369", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\begin{align*}\n", + "p(y_i=1|x_i,\\boldsymbol{\\beta}) &= \\frac{\\exp{(\\beta_0+\\beta_1x_i)}}{1+\\exp{(\\beta_0+\\beta_1x_i)}},\\nonumber\\\\\n", + "p(y_i=0|x_i,\\boldsymbol{\\beta}) &= 1 - p(y_i=1|x_i,\\boldsymbol{\\beta}),\n", + "\\end{align*}\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "10c01ae1", + "metadata": { + "editable": true + }, + "source": [ + "where $\\boldsymbol{\\beta}$ are the weights we wish to extract from data, in our case $\\beta_0$ and $\\beta_1$." + ] + }, + { + "cell_type": "markdown", + "id": "b205858e", + "metadata": { + "editable": true + }, + "source": [ + "## The equations to solve\n", + "\n", + "Our compact equations used a definition of a vector $\\boldsymbol{y}$ with $n$\n", + "elements $y_i$, an $n\\times p$ matrix $\\boldsymbol{X}$ which contains the\n", + "$x_i$ values and a vector $\\boldsymbol{p}$ of fitted probabilities\n", + "$p(y_i\\vert x_i,\\boldsymbol{\\beta})$. We rewrote in a more compact form\n", + "the first derivative of the cost function as" + ] + }, + { + "cell_type": "markdown", + "id": "5f48f306", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\frac{\\partial \\mathcal{C}(\\boldsymbol{\\beta})}{\\partial \\boldsymbol{\\beta}} = -\\boldsymbol{X}^T\\left(\\boldsymbol{y}-\\boldsymbol{p}\\right).\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "deff0dec", + "metadata": { + "editable": true + }, + "source": [ + "If we in addition define a diagonal matrix $\\boldsymbol{W}$ with elements \n", + "$p(y_i\\vert x_i,\\boldsymbol{\\beta})(1-p(y_i\\vert x_i,\\boldsymbol{\\beta})$, we can obtain a compact expression of the second derivative as" + ] + }, + { + "cell_type": "markdown", + "id": "4a3eee34", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\frac{\\partial^2 \\mathcal{C}(\\boldsymbol{\\beta})}{\\partial \\boldsymbol{\\beta}\\partial \\boldsymbol{\\beta}^T} = \\boldsymbol{X}^T\\boldsymbol{W}\\boldsymbol{X}.\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "ac2953d0", + "metadata": { + "editable": true + }, + "source": [ + "This defines what is called the Hessian matrix." + ] + }, + { + "cell_type": "markdown", + "id": "fcacb96c", + "metadata": { + "editable": true + }, + "source": [ + "## Solving using Newton-Raphson's method\n", + "\n", + "If we can set up these equations, Newton-Raphson's iterative method is normally the method of choice. It requires however that we can compute in an efficient way the matrices that define the first and second derivatives. \n", + "\n", + "Our iterative scheme is then given by" + ] + }, + { + "cell_type": "markdown", + "id": "d8946939", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\boldsymbol{\\beta}^{\\mathrm{new}} = \\boldsymbol{\\beta}^{\\mathrm{old}}-\\left(\\frac{\\partial^2 \\mathcal{C}(\\boldsymbol{\\beta})}{\\partial \\boldsymbol{\\beta}\\partial \\boldsymbol{\\beta}^T}\\right)^{-1}_{\\boldsymbol{\\beta}^{\\mathrm{old}}}\\times \\left(\\frac{\\partial \\mathcal{C}(\\boldsymbol{\\beta})}{\\partial \\boldsymbol{\\beta}}\\right)_{\\boldsymbol{\\beta}^{\\mathrm{old}}},\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "1cb3f7cd", + "metadata": { + "editable": true + }, + "source": [ + "or in matrix form as" + ] + }, + { + "cell_type": "markdown", + "id": "5707329b", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\boldsymbol{\\beta}^{\\mathrm{new}} = \\boldsymbol{\\beta}^{\\mathrm{old}}-\\left(\\boldsymbol{X}^T\\boldsymbol{W}\\boldsymbol{X} \\right)^{-1}\\times \\left(-\\boldsymbol{X}^T(\\boldsymbol{y}-\\boldsymbol{p}) \\right)_{\\boldsymbol{\\beta}^{\\mathrm{old}}}.\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "996511e9", + "metadata": { + "editable": true + }, + "source": [ + "The right-hand side is computed with the old values of $\\beta$. \n", + "\n", + "If we can compute these matrices, in particular the Hessian, the above is often the easiest method to implement." + ] + }, + { + "cell_type": "markdown", + "id": "a35fcbf1", + "metadata": { + "editable": true + }, + "source": [ + "## Brief reminder on Newton-Raphson's method\n", + "\n", + "Let us quickly remind ourselves how we derive the above method.\n", + "\n", + "Perhaps the most celebrated of all one-dimensional root-finding\n", + "routines is Newton's method, also called the Newton-Raphson\n", + "method. This method requires the evaluation of both the\n", + "function $f$ and its derivative $f'$ at arbitrary points. \n", + "If you can only calculate the derivative\n", + "numerically and/or your function is not of the smooth type, we\n", + "normally discourage the use of this method." + ] + }, + { + "cell_type": "markdown", + "id": "080ea026", + "metadata": { + "editable": true + }, + "source": [ + "## The equations\n", + "\n", + "The Newton-Raphson formula consists geometrically of extending the\n", + "tangent line at a current point until it crosses zero, then setting\n", + "the next guess to the abscissa of that zero-crossing. The mathematics\n", + "behind this method is rather simple. Employing a Taylor expansion for\n", + "$x$ sufficiently close to the solution $s$, we have" + ] + }, + { + "cell_type": "markdown", + "id": "49636667", + "metadata": { + "editable": true + }, + "source": [ + "\n", + "
            \n", + "\n", + "$$\n", + "f(s)=0=f(x)+(s-x)f'(x)+\\frac{(s-x)^2}{2}f''(x) +\\dots.\n", + " \\label{eq:taylornr} \\tag{2}\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "3bd16905", + "metadata": { + "editable": true + }, + "source": [ + "For small enough values of the function and for well-behaved\n", + "functions, the terms beyond linear are unimportant, hence we obtain" + ] + }, + { + "cell_type": "markdown", + "id": "68cc347e", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "f(x)+(s-x)f'(x)\\approx 0,\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "3ba3e954", + "metadata": { + "editable": true + }, + "source": [ + "yielding" + ] + }, + { + "cell_type": "markdown", + "id": "18a71133", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "s\\approx x-\\frac{f(x)}{f'(x)}.\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "cebe1598", + "metadata": { + "editable": true + }, + "source": [ + "Having in mind an iterative procedure, it is natural to start iterating with" + ] + }, + { + "cell_type": "markdown", + "id": "262fc2e7", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "x_{n+1}=x_n-\\frac{f(x_n)}{f'(x_n)}.\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "d8ac2615", + "metadata": { + "editable": true + }, + "source": [ + "## Simple geometric interpretation\n", + "\n", + "The above is Newton-Raphson's method. It has a simple geometric\n", + "interpretation, namely $x_{n+1}$ is the point where the tangent from\n", + "$(x_n,f(x_n))$ crosses the $x$-axis. Close to the solution,\n", + "Newton-Raphson converges fast to the desired result. However, if we\n", + "are far from a root, where the higher-order terms in the series are\n", + "important, the Newton-Raphson formula can give grossly inaccurate\n", + "results. For instance, the initial guess for the root might be so far\n", + "from the true root as to let the search interval include a local\n", + "maximum or minimum of the function. If an iteration places a trial\n", + "guess near such a local extremum, so that the first derivative nearly\n", + "vanishes, then Newton-Raphson may fail totally" + ] + }, + { + "cell_type": "markdown", + "id": "1637b673", + "metadata": { + "editable": true + }, + "source": [ + "## Extending to more than one variable\n", + "\n", + "Newton's method can be generalized to systems of several non-linear equations\n", + "and variables. Consider the case with two equations" + ] + }, + { + "cell_type": "markdown", + "id": "efb3a153", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\begin{array}{cc} f_1(x_1,x_2) &=0\\\\\n", + " f_2(x_1,x_2) &=0,\\end{array}\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "cd1d0b0b", + "metadata": { + "editable": true + }, + "source": [ + "which we Taylor expand to obtain" + ] + }, + { + "cell_type": "markdown", + "id": "1409701b", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\begin{array}{cc} 0=f_1(x_1+h_1,x_2+h_2)=&f_1(x_1,x_2)+h_1\n", + " \\partial f_1/\\partial x_1+h_2\n", + " \\partial f_1/\\partial x_2+\\dots\\\\\n", + " 0=f_2(x_1+h_1,x_2+h_2)=&f_2(x_1,x_2)+h_1\n", + " \\partial f_2/\\partial x_1+h_2\n", + " \\partial f_2/\\partial x_2+\\dots\n", + " \\end{array}.\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "969cdccb", + "metadata": { + "editable": true + }, + "source": [ + "Defining the Jacobian matrix ${\\bf \\boldsymbol{J}}$ we have" + ] + }, + { + "cell_type": "markdown", + "id": "1d26f373", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "{\\bf \\boldsymbol{J}}=\\left( \\begin{array}{cc}\n", + " \\partial f_1/\\partial x_1 & \\partial f_1/\\partial x_2 \\\\\n", + " \\partial f_2/\\partial x_1 &\\partial f_2/\\partial x_2\n", + " \\end{array} \\right),\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "b6fa2c0e", + "metadata": { + "editable": true + }, + "source": [ + "we can rephrase Newton's method as" + ] + }, + { + "cell_type": "markdown", + "id": "c26cca81", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\left(\\begin{array}{c} x_1^{n+1} \\\\ x_2^{n+1} \\end{array} \\right)=\n", + "\\left(\\begin{array}{c} x_1^{n} \\\\ x_2^{n} \\end{array} \\right)+\n", + "\\left(\\begin{array}{c} h_1^{n} \\\\ h_2^{n} \\end{array} \\right),\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "e6813f1b", + "metadata": { + "editable": true + }, + "source": [ + "where we have defined" + ] + }, + { + "cell_type": "markdown", + "id": "f53cce6d", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\left(\\begin{array}{c} h_1^{n} \\\\ h_2^{n} \\end{array} \\right)=\n", + " -{\\bf \\boldsymbol{J}}^{-1}\n", + " \\left(\\begin{array}{c} f_1(x_1^{n},x_2^{n}) \\\\ f_2(x_1^{n},x_2^{n}) \\end{array} \\right).\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "47a097ab", + "metadata": { + "editable": true + }, + "source": [ + "We need thus to compute the inverse of the Jacobian matrix and it\n", + "is to understand that difficulties may\n", + "arise in case ${\\bf \\boldsymbol{J}}$ is nearly singular.\n", + "\n", + "It is rather straightforward to extend the above scheme to systems of\n", + "more than two non-linear equations. In our case, the Jacobian matrix is given by the Hessian that represents the second derivative of cost function." + ] + }, + { + "cell_type": "markdown", + "id": "682046e4", + "metadata": { + "editable": true + }, + "source": [ + "## Steepest descent\n", + "\n", + "The basic idea of gradient descent is\n", + "that a function $F(\\mathbf{x})$, \n", + "$\\mathbf{x} \\equiv (x_1,\\cdots,x_n)$, decreases fastest if one goes from $\\bf {x}$ in the\n", + "direction of the negative gradient $-\\nabla F(\\mathbf{x})$.\n", + "\n", + "It can be shown that if" + ] + }, + { + "cell_type": "markdown", + "id": "91f0fc21", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\mathbf{x}_{k+1} = \\mathbf{x}_k - \\gamma_k \\nabla F(\\mathbf{x}_k),\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "7e415ba2", + "metadata": { + "editable": true + }, + "source": [ + "with $\\gamma_k > 0$.\n", + "\n", + "For $\\gamma_k$ small enough, then $F(\\mathbf{x}_{k+1}) \\leq\n", + "F(\\mathbf{x}_k)$. This means that for a sufficiently small $\\gamma_k$\n", + "we are always moving towards smaller function values, i.e a minimum." + ] + }, + { + "cell_type": "markdown", + "id": "d67a88be", + "metadata": { + "editable": true + }, + "source": [ + "## More on Steepest descent\n", + "\n", + "The previous observation is the basis of the method of steepest\n", + "descent, which is also referred to as just gradient descent (GD). One\n", + "starts with an initial guess $\\mathbf{x}_0$ for a minimum of $F$ and\n", + "computes new approximations according to" + ] + }, + { + "cell_type": "markdown", + "id": "ee068302", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\mathbf{x}_{k+1} = \\mathbf{x}_k - \\gamma_k \\nabla F(\\mathbf{x}_k), \\ \\ k \\geq 0.\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "d63dc655", + "metadata": { + "editable": true + }, + "source": [ + "The parameter $\\gamma_k$ is often referred to as the step length or\n", + "the learning rate within the context of Machine Learning." + ] + }, + { + "cell_type": "markdown", + "id": "fc1fc57d", + "metadata": { + "editable": true + }, + "source": [ + "## The ideal\n", + "\n", + "Ideally the sequence $\\{\\mathbf{x}_k \\}_{k=0}$ converges to a global\n", + "minimum of the function $F$. In general we do not know if we are in a\n", + "global or local minimum. In the special case when $F$ is a convex\n", + "function, all local minima are also global minima, so in this case\n", + "gradient descent can converge to the global solution. The advantage of\n", + "this scheme is that it is conceptually simple and straightforward to\n", + "implement. However the method in this form has some severe\n", + "limitations:\n", + "\n", + "In machine learing we are often faced with non-convex high dimensional\n", + "cost functions with many local minima. Since GD is deterministic we\n", + "will get stuck in a local minimum, if the method converges, unless we\n", + "have a very good intial guess. This also implies that the scheme is\n", + "sensitive to the chosen initial condition.\n", + "\n", + "Note that the gradient is a function of $\\mathbf{x} =\n", + "(x_1,\\cdots,x_n)$ which makes it expensive to compute numerically." + ] + }, + { + "cell_type": "markdown", + "id": "b97eed70", + "metadata": { + "editable": true + }, + "source": [ + "## The sensitiveness of the gradient descent\n", + "\n", + "The gradient descent method \n", + "is sensitive to the choice of learning rate $\\gamma_k$. This is due\n", + "to the fact that we are only guaranteed that $F(\\mathbf{x}_{k+1}) \\leq\n", + "F(\\mathbf{x}_k)$ for sufficiently small $\\gamma_k$. The problem is to\n", + "determine an optimal learning rate. If the learning rate is chosen too\n", + "small the method will take a long time to converge and if it is too\n", + "large we can experience erratic behavior.\n", + "\n", + "Many of these shortcomings can be alleviated by introducing\n", + "randomness. One such method is that of Stochastic Gradient Descent\n", + "(SGD), to be discussed next week." + ] + }, + { + "cell_type": "markdown", + "id": "c35084e9", + "metadata": { + "editable": true + }, + "source": [ + "## Convex functions\n", + "\n", + "Ideally we want our cost/loss function to be convex(concave).\n", + "\n", + "First we give the definition of a convex set: A set $C$ in\n", + "$\\mathbb{R}^n$ is said to be convex if, for all $x$ and $y$ in $C$ and\n", + "all $t \\in (0,1)$ , the point $(1 − t)x + ty$ also belongs to\n", + "C. Geometrically this means that every point on the line segment\n", + "connecting $x$ and $y$ is in $C$ as discussed below.\n", + "\n", + "The convex subsets of $\\mathbb{R}$ are the intervals of\n", + "$\\mathbb{R}$. Examples of convex sets of $\\mathbb{R}^2$ are the\n", + "regular polygons (triangles, rectangles, pentagons, etc...)." + ] + }, + { + "cell_type": "markdown", + "id": "a11b1893", + "metadata": { + "editable": true + }, + "source": [ + "## Convex function\n", + "\n", + "**Convex function**: Let $X \\subset \\mathbb{R}^n$ be a convex set. Assume that the function $f: X \\rightarrow \\mathbb{R}$ is continuous, then $f$ is said to be convex if $$f(tx_1 + (1-t)x_2) \\leq tf(x_1) + (1-t)f(x_2) $$ for all $x_1, x_2 \\in X$ and for all $t \\in [0,1]$. If $\\leq$ is replaced with a strict inequaltiy in the definition, we demand $x_1 \\neq x_2$ and $t\\in(0,1)$ then $f$ is said to be strictly convex. For a single variable function, convexity means that if you draw a straight line connecting $f(x_1)$ and $f(x_2)$, the value of the function on the interval $[x_1,x_2]$ is always below the line as illustrated below." + ] + }, + { + "cell_type": "markdown", + "id": "f54f8d27", + "metadata": { + "editable": true + }, + "source": [ + "## Conditions on convex functions\n", + "\n", + "In the following we state first and second-order conditions which\n", + "ensures convexity of a function $f$. We write $D_f$ to denote the\n", + "domain of $f$, i.e the subset of $R^n$ where $f$ is defined. For more\n", + "details and proofs we refer to: [S. Boyd and L. Vandenberghe. Convex Optimization. Cambridge University Press](http://stanford.edu/boyd/cvxbook/, 2004).\n", + "\n", + "**First order condition.**\n", + "\n", + "Suppose $f$ is differentiable (i.e $\\nabla f(x)$ is well defined for\n", + "all $x$ in the domain of $f$). Then $f$ is convex if and only if $D_f$\n", + "is a convex set and $$f(y) \\geq f(x) + \\nabla f(x)^T (y-x) $$ holds\n", + "for all $x,y \\in D_f$. This condition means that for a convex function\n", + "the first order Taylor expansion (right hand side above) at any point\n", + "a global under estimator of the function. To convince yourself you can\n", + "make a drawing of $f(x) = x^2+1$ and draw the tangent line to $f(x)$ and\n", + "note that it is always below the graph.\n", + "\n", + "**Second order condition.**\n", + "\n", + "Assume that $f$ is twice\n", + "differentiable, i.e the Hessian matrix exists at each point in\n", + "$D_f$. Then $f$ is convex if and only if $D_f$ is a convex set and its\n", + "Hessian is positive semi-definite for all $x\\in D_f$. For a\n", + "single-variable function this reduces to $f''(x) \\geq 0$. Geometrically this means that $f$ has nonnegative curvature\n", + "everywhere.\n", + "\n", + "This condition is particularly useful since it gives us an procedure for determining if the function under consideration is convex, apart from using the definition." + ] + }, + { + "cell_type": "markdown", + "id": "a8d7420c", + "metadata": { + "editable": true + }, + "source": [ + "## More on convex functions\n", + "\n", + "The next result is of great importance to us and the reason why we are\n", + "going on about convex functions. In machine learning we frequently\n", + "have to minimize a loss/cost function in order to find the best\n", + "parameters for the model we are considering. \n", + "\n", + "Ideally we want the\n", + "global minimum (for high-dimensional models it is hard to know\n", + "if we have local or global minimum). However, if the cost/loss function\n", + "is convex the following result provides invaluable information:\n", + "\n", + "**Any minimum is global for convex functions.**\n", + "\n", + "Consider the problem of finding $x \\in \\mathbb{R}^n$ such that $f(x)$\n", + "is minimal, where $f$ is convex and differentiable. Then, any point\n", + "$x^*$ that satisfies $\\nabla f(x^*) = 0$ is a global minimum.\n", + "\n", + "This result means that if we know that the cost/loss function is convex and we are able to find a minimum, we are guaranteed that it is a global minimum." + ] + }, + { + "cell_type": "markdown", + "id": "877d661b", + "metadata": { + "editable": true + }, + "source": [ + "## Some simple problems\n", + "\n", + "1. Show that $f(x)=x^2$ is convex for $x \\in \\mathbb{R}$ using the definition of convexity. Hint: If you re-write the definition, $f$ is convex if the following holds for all $x,y \\in D_f$ and any $\\lambda \\in [0,1]$ $\\lambda f(x)+(1-\\lambda)f(y)-f(\\lambda x + (1-\\lambda) y ) \\geq 0$.\n", + "\n", + "2. Using the second order condition show that the following functions are convex on the specified domain.\n", + "\n", + " * $f(x) = e^x$ is convex for $x \\in \\mathbb{R}$.\n", + "\n", + " * $g(x) = -\\ln(x)$ is convex for $x \\in (0,\\infty)$.\n", + "\n", + "3. Let $f(x) = x^2$ and $g(x) = e^x$. Show that $f(g(x))$ and $g(f(x))$ is convex for $x \\in \\mathbb{R}$. Also show that if $f(x)$ is any convex function than $h(x) = e^{f(x)}$ is convex.\n", + "\n", + "4. A norm is any function that satisfy the following properties\n", + "\n", + " * $f(\\alpha x) = |\\alpha| f(x)$ for all $\\alpha \\in \\mathbb{R}$.\n", + "\n", + " * $f(x+y) \\leq f(x) + f(y)$\n", + "\n", + " * $f(x) \\leq 0$ for all $x \\in \\mathbb{R}^n$ with equality if and only if $x = 0$\n", + "\n", + "Using the definition of convexity, try to show that a function satisfying the properties above is convex (the third condition is not needed to show this)." + ] + }, + { + "cell_type": "markdown", + "id": "74823208", + "metadata": { + "editable": true + }, + "source": [ + "## Revisiting our first homework\n", + "\n", + "We will use linear regression as a case study for the gradient descent\n", + "methods. Linear regression is a great test case for the gradient\n", + "descent methods discussed in the lectures since it has several\n", + "desirable properties such as:\n", + "\n", + "1. An analytical solution (recall homework set 1).\n", + "\n", + "2. The gradient can be computed analytically.\n", + "\n", + "3. The cost function is convex which guarantees that gradient descent converges for small enough learning rates\n", + "\n", + "We revisit an example similar to what we had in the first homework set. We had a function of the type" + ] + }, + { + "cell_type": "code", + "execution_count": 12, + "id": "2bafe07c", + "metadata": { + "collapsed": false, + "editable": true + }, + "outputs": [], + "source": [ + "x = 2*np.random.rand(m,1)\n", + "y = 4+3*x+np.random.randn(m,1)" + ] + }, + { + "cell_type": "markdown", + "id": "a18bdeb4", + "metadata": { + "editable": true + }, + "source": [ + "with $x_i \\in [0,1] $ is chosen randomly using a uniform distribution. Additionally we have a stochastic noise chosen according to a normal distribution $\\cal {N}(0,1)$. \n", + "The linear regression model is given by" + ] + }, + { + "cell_type": "markdown", + "id": "16415b8c", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "h_\\beta(x) = \\boldsymbol{y} = \\beta_0 + \\beta_1 x,\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "dc273c04", + "metadata": { + "editable": true + }, + "source": [ + "such that" + ] + }, + { + "cell_type": "markdown", + "id": "4332c289", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\boldsymbol{y}_i = \\beta_0 + \\beta_1 x_i.\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "e96ba3ef", + "metadata": { + "editable": true + }, + "source": [ + "## Gradient descent example\n", + "\n", + "Let $\\mathbf{y} = (y_1,\\cdots,y_n)^T$, $\\mathbf{\\boldsymbol{y}} = (\\boldsymbol{y}_1,\\cdots,\\boldsymbol{y}_n)^T$ and $\\beta = (\\beta_0, \\beta_1)^T$\n", + "\n", + "It is convenient to write $\\mathbf{\\boldsymbol{y}} = X\\beta$ where $X \\in \\mathbb{R}^{100 \\times 2} $ is the design matrix given by (we keep the intercept here)" + ] + }, + { + "cell_type": "markdown", + "id": "deb14fe7", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "X \\equiv \\begin{bmatrix}\n", + "1 & x_1 \\\\\n", + "\\vdots & \\vdots \\\\\n", + "1 & x_{100} & \\\\\n", + "\\end{bmatrix}.\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "e468ef7d", + "metadata": { + "editable": true + }, + "source": [ + "The cost/loss/risk function is given by (" + ] + }, + { + "cell_type": "markdown", + "id": "97154cec", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "C(\\beta) = \\frac{1}{n}||X\\beta-\\mathbf{y}||_{2}^{2} = \\frac{1}{n}\\sum_{i=1}^{100}\\left[ (\\beta_0 + \\beta_1 x_i)^2 - 2 y_i (\\beta_0 + \\beta_1 x_i) + y_i^2\\right]\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "6e30b09a", + "metadata": { + "editable": true + }, + "source": [ + "and we want to find $\\beta$ such that $C(\\beta)$ is minimized." + ] + }, + { + "cell_type": "markdown", + "id": "d5dc7377", + "metadata": { + "editable": true + }, + "source": [ + "## The derivative of the cost/loss function\n", + "\n", + "Computing $\\partial C(\\beta) / \\partial \\beta_0$ and $\\partial C(\\beta) / \\partial \\beta_1$ we can show that the gradient can be written as" + ] + }, + { + "cell_type": "markdown", + "id": "49d2621f", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\nabla_{\\beta} C(\\beta) = \\frac{2}{n}\\begin{bmatrix} \\sum_{i=1}^{100} \\left(\\beta_0+\\beta_1x_i-y_i\\right) \\\\\n", + "\\sum_{i=1}^{100}\\left( x_i (\\beta_0+\\beta_1x_i)-y_ix_i\\right) \\\\\n", + "\\end{bmatrix} = \\frac{2}{n}X^T(X\\beta - \\mathbf{y}),\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "f2950e6b", + "metadata": { + "editable": true + }, + "source": [ + "where $X$ is the design matrix defined above." + ] + }, + { + "cell_type": "markdown", + "id": "00be7184", + "metadata": { + "editable": true + }, + "source": [ + "## The Hessian matrix\n", + "The Hessian matrix of $C(\\beta)$ is given by" + ] + }, + { + "cell_type": "markdown", + "id": "97b443f2", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\boldsymbol{H} \\equiv \\begin{bmatrix}\n", + "\\frac{\\partial^2 C(\\beta)}{\\partial \\beta_0^2} & \\frac{\\partial^2 C(\\beta)}{\\partial \\beta_0 \\partial \\beta_1} \\\\\n", + "\\frac{\\partial^2 C(\\beta)}{\\partial \\beta_0 \\partial \\beta_1} & \\frac{\\partial^2 C(\\beta)}{\\partial \\beta_1^2} & \\\\\n", + "\\end{bmatrix} = \\frac{2}{n}X^T X.\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "0676ebaf", + "metadata": { + "editable": true + }, + "source": [ + "This result implies that $C(\\beta)$ is a convex function since the matrix $X^T X$ always is positive semi-definite." + ] + }, + { + "cell_type": "markdown", + "id": "4ad98c2d", + "metadata": { + "editable": true + }, + "source": [ + "## Simple program\n", + "\n", + "We can now write a program that minimizes $C(\\beta)$ using the gradient descent method with a constant learning rate $\\gamma$ according to" + ] + }, + { + "cell_type": "markdown", + "id": "7405f7a9", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\beta_{k+1} = \\beta_k - \\gamma \\nabla_\\beta C(\\beta_k), \\ k=0,1,\\cdots\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "ca70147e", + "metadata": { + "editable": true + }, + "source": [ + "We can use the expression we computed for the gradient and let use a\n", + "$\\beta_0$ be chosen randomly and let $\\gamma = 0.001$. Stop iterating\n", + "when $||\\nabla_\\beta C(\\beta_k) || \\leq \\epsilon = 10^{-8}$. **Note that the code below does not include the latter stop criterion**.\n", + "\n", + "And finally we can compare our solution for $\\beta$ with the analytic result given by \n", + "$\\beta= (X^TX)^{-1} X^T \\mathbf{y}$." + ] + }, + { + "cell_type": "markdown", + "id": "d07fdcda", + "metadata": { + "editable": true + }, + "source": [ + "## Gradient Descent Example\n", + "\n", + "Here our simple example" + ] + }, + { + "cell_type": "code", + "execution_count": 13, + "id": "2483d42a", + "metadata": { + "collapsed": false, + "editable": true + }, + "outputs": [], + "source": [ + "\n", + "# Importing various packages\n", + "from random import random, seed\n", + "import numpy as np\n", + "import matplotlib.pyplot as plt\n", + "from mpl_toolkits.mplot3d import Axes3D\n", + "from matplotlib import cm\n", + "from matplotlib.ticker import LinearLocator, FormatStrFormatter\n", + "import sys\n", + "\n", + "# the number of datapoints\n", + "n = 100\n", + "x = 2*np.random.rand(n,1)\n", + "y = 4+3*x+np.random.randn(n,1)\n", + "\n", + "X = np.c_[np.ones((n,1)), x]\n", + "# Hessian matrix\n", + "H = (2.0/n)* X.T @ X\n", + "# Get the eigenvalues\n", + "EigValues, EigVectors = np.linalg.eig(H)\n", + "print(f\"Eigenvalues of Hessian Matrix:{EigValues}\")\n", + "\n", + "beta_linreg = np.linalg.inv(X.T @ X) @ X.T @ y\n", + "print(beta_linreg)\n", + "beta = np.random.randn(2,1)\n", + "\n", + "eta = 1.0/np.max(EigValues)\n", + "Niterations = 1000\n", + "\n", + "for iter in range(Niterations):\n", + " gradient = (2.0/n)*X.T @ (X @ beta-y)\n", + " beta -= eta*gradient\n", + "\n", + "print(beta)\n", + "xnew = np.array([[0],[2]])\n", + "xbnew = np.c_[np.ones((2,1)), xnew]\n", + "ypredict = xbnew.dot(beta)\n", + "ypredict2 = xbnew.dot(beta_linreg)\n", + "plt.plot(xnew, ypredict, \"r-\")\n", + "plt.plot(xnew, ypredict2, \"b-\")\n", + "plt.plot(x, y ,'ro')\n", + "plt.axis([0,2.0,0, 15.0])\n", + "plt.xlabel(r'$x$')\n", + "plt.ylabel(r'$y$')\n", + "plt.title(r'Gradient descent example')\n", + "plt.show()" + ] + }, + { + "cell_type": "markdown", + "id": "6071973b", + "metadata": { + "editable": true + }, + "source": [ + "## And a corresponding example using **scikit-learn**" + ] + }, + { + "cell_type": "code", + "execution_count": 14, + "id": "f31a67da", + "metadata": { + "collapsed": false, + "editable": true + }, + "outputs": [], + "source": [ + "# Importing various packages\n", + "from random import random, seed\n", + "import numpy as np\n", + "import matplotlib.pyplot as plt\n", + "from sklearn.linear_model import SGDRegressor\n", + "\n", + "n = 100\n", + "x = 2*np.random.rand(n,1)\n", + "y = 4+3*x+np.random.randn(n,1)\n", + "\n", + "X = np.c_[np.ones((n,1)), x]\n", + "beta_linreg = np.linalg.inv(X.T @ X) @ (X.T @ y)\n", + "print(beta_linreg)\n", + "sgdreg = SGDRegressor(max_iter = 50, penalty=None, eta0=0.1)\n", + "sgdreg.fit(x,y.ravel())\n", + "print(sgdreg.intercept_, sgdreg.coef_)" + ] + }, + { + "cell_type": "markdown", + "id": "c6ae27ed", + "metadata": { + "editable": true + }, + "source": [ + "## Gradient descent and Ridge\n", + "\n", + "We have also discussed Ridge regression where the loss function contains a regularized term given by the $L_2$ norm of $\\beta$," + ] + }, + { + "cell_type": "markdown", + "id": "55efcb99", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "C_{\\text{ridge}}(\\beta) = \\frac{1}{n}||X\\beta -\\mathbf{y}||^2 + \\lambda ||\\beta||^2, \\ \\lambda \\geq 0.\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "aedf6cb2", + "metadata": { + "editable": true + }, + "source": [ + "In order to minimize $C_{\\text{ridge}}(\\beta)$ using GD we adjust the gradient as follows" + ] + }, + { + "cell_type": "markdown", + "id": "7bb7c811", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\nabla_\\beta C_{\\text{ridge}}(\\beta) = \\frac{2}{n}\\begin{bmatrix} \\sum_{i=1}^{100} \\left(\\beta_0+\\beta_1x_i-y_i\\right) \\\\\n", + "\\sum_{i=1}^{100}\\left( x_i (\\beta_0+\\beta_1x_i)-y_ix_i\\right) \\\\\n", + "\\end{bmatrix} + 2\\lambda\\begin{bmatrix} \\beta_0 \\\\ \\beta_1\\end{bmatrix} = 2 (\\frac{1}{n}X^T(X\\beta - \\mathbf{y})+\\lambda \\beta).\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "42227947", + "metadata": { + "editable": true + }, + "source": [ + "We can easily extend our program to minimize $C_{\\text{ridge}}(\\beta)$ using gradient descent and compare with the analytical solution given by" + ] + }, + { + "cell_type": "markdown", + "id": "9c164deb", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\beta_{\\text{ridge}} = \\left(X^T X + n\\lambda I_{2 \\times 2} \\right)^{-1} X^T \\mathbf{y}.\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "eb38ae48", + "metadata": { + "editable": true + }, + "source": [ + "## The Hessian matrix for Ridge Regression\n", + "The Hessian matrix of Ridge Regression for our simple example is given by" + ] + }, + { + "cell_type": "markdown", + "id": "a6fcf56c", + "metadata": { + "editable": true + }, + "source": [ + "$$\n", + "\\boldsymbol{H} \\equiv \\begin{bmatrix}\n", + "\\frac{\\partial^2 C(\\beta)}{\\partial \\beta_0^2} & \\frac{\\partial^2 C(\\beta)}{\\partial \\beta_0 \\partial \\beta_1} \\\\\n", + "\\frac{\\partial^2 C(\\beta)}{\\partial \\beta_0 \\partial \\beta_1} & \\frac{\\partial^2 C(\\beta)}{\\partial \\beta_1^2} & \\\\\n", + "\\end{bmatrix} = \\frac{2}{n}X^T X+2\\lambda\\boldsymbol{I}.\n", + "$$" + ] + }, + { + "cell_type": "markdown", + "id": "3a2b5ae4", + "metadata": { + "editable": true + }, + "source": [ + "This implies that the Hessian matrix is positive definite, hence the stationary point is a\n", + "minimum.\n", + "Note that the Ridge cost function is convex being a sum of two convex\n", + "functions. Therefore, the stationary point is a global\n", + "minimum of this function." + ] + }, + { + "cell_type": "markdown", + "id": "1a7eb7e6", + "metadata": { + "editable": true + }, + "source": [ + "## Program example for gradient descent with Ridge Regression" + ] + }, + { + "cell_type": "code", + "execution_count": 15, + "id": "ccb7a9b1", + "metadata": { + "collapsed": false, + "editable": true + }, + "outputs": [], + "source": [ + "from random import random, seed\n", + "import numpy as np\n", + "import matplotlib.pyplot as plt\n", + "from mpl_toolkits.mplot3d import Axes3D\n", + "from matplotlib import cm\n", + "from matplotlib.ticker import LinearLocator, FormatStrFormatter\n", + "import sys\n", + "\n", + "# the number of datapoints\n", + "n = 100\n", + "x = 2*np.random.rand(n,1)\n", + "y = 4+3*x+np.random.randn(n,1)\n", + "\n", + "X = np.c_[np.ones((n,1)), x]\n", + "XT_X = X.T @ X\n", + "\n", + "#Ridge parameter lambda\n", + "lmbda = 0.001\n", + "Id = n*lmbda* np.eye(XT_X.shape[0])\n", + "\n", + "# Hessian matrix\n", + "H = (2.0/n)* XT_X+2*lmbda* np.eye(XT_X.shape[0])\n", + "# Get the eigenvalues\n", + "EigValues, EigVectors = np.linalg.eig(H)\n", + "print(f\"Eigenvalues of Hessian Matrix:{EigValues}\")\n", + "\n", + "\n", + "beta_linreg = np.linalg.inv(XT_X+Id) @ X.T @ y\n", + "print(beta_linreg)\n", + "# Start plain gradient descent\n", + "beta = np.random.randn(2,1)\n", + "\n", + "eta = 1.0/np.max(EigValues)\n", + "Niterations = 100\n", + "\n", + "for iter in range(Niterations):\n", + " gradients = 2.0/n*X.T @ (X @ (beta)-y)+2*lmbda*beta\n", + " beta -= eta*gradients\n", + "\n", + "print(beta)\n", + "ypredict = X @ beta\n", + "ypredict2 = X @ beta_linreg\n", + "plt.plot(x, ypredict, \"r-\")\n", + "plt.plot(x, ypredict2, \"b-\")\n", + "plt.plot(x, y ,'ro')\n", + "plt.axis([0,2.0,0, 15.0])\n", + "plt.xlabel(r'$x$')\n", + "plt.ylabel(r'$y$')\n", + "plt.title(r'Gradient descent example for Ridge')\n", + "plt.show()" + ] + }, + { + "cell_type": "markdown", + "id": "95a71108", + "metadata": { + "editable": true + }, + "source": [ + "## Using gradient descent methods, limitations\n", + "\n", + "* **Gradient descent (GD) finds local minima of our function**. Since the GD algorithm is deterministic, if it converges, it will converge to a local minimum of our cost/loss/risk function. Because in ML we are often dealing with extremely rugged landscapes with many local minima, this can lead to poor performance.\n", + "\n", + "* **GD is sensitive to initial conditions**. One consequence of the local nature of GD is that initial conditions matter. Depending on where one starts, one will end up at a different local minima. Therefore, it is very important to think about how one initializes the training process. This is true for GD as well as more complicated variants of GD.\n", + "\n", + "* **Gradients are computationally expensive to calculate for large datasets**. In many cases in statistics and ML, the cost/loss/risk function is a sum of terms, with one term for each data point. For example, in linear regression, $E \\propto \\sum_{i=1}^n (y_i - \\mathbf{w}^T\\cdot\\mathbf{x}_i)^2$; for logistic regression, the square error is replaced by the cross entropy. To calculate the gradient we have to sum over *all* $n$ data points. Doing this at every GD step becomes extremely computationally expensive. An ingenious solution to this, is to calculate the gradients using small subsets of the data called \"mini batches\". This has the added benefit of introducing stochasticity into our algorithm.\n", + "\n", + "* **GD is very sensitive to choices of learning rates**. GD is extremely sensitive to the choice of learning rates. If the learning rate is very small, the training process take an extremely long time. For larger learning rates, GD can diverge and give poor results. Furthermore, depending on what the local landscape looks like, we have to modify the learning rates to ensure convergence. Ideally, we would *adaptively* choose the learning rates to match the landscape.\n", + "\n", + "* **GD treats all directions in parameter space uniformly.** Another major drawback of GD is that unlike Newton's method, the learning rate for GD is the same in all directions in parameter space. For this reason, the maximum learning rate is set by the behavior of the steepest direction and this can significantly slow down training. Ideally, we would like to take large steps in flat directions and small steps in steep directions. Since we are exploring rugged landscapes where curvatures change, this requires us to keep track of not only the gradient but second derivatives. The ideal scenario would be to calculate the Hessian but this proves to be too computationally expensive. \n", + "\n", + "* GD can take exponential time to escape saddle points, even with random initialization. As we mentioned, GD is extremely sensitive to initial condition since it determines the particular local minimum GD would eventually reach. However, even with a good initialization scheme, through the introduction of randomness, GD can still take exponential time to escape saddle points." + ] + }, + { + "cell_type": "markdown", + "id": "33f25532", + "metadata": { + "editable": true + }, + "source": [ + "## Challenge yourself the coming weekend\n", + "\n", + "Write a code which implements gradient descent for a logistic regression example." + ] + }, + { + "cell_type": "markdown", + "id": "13ee68fe", + "metadata": { + "editable": true + }, + "source": [ + "## Lab session: Material from last week and relevant for the first project" + ] + }, + { + "cell_type": "markdown", + "id": "e3e607f1", "metadata": { "editable": true }, @@ -107,7 +2842,7 @@ }, { "cell_type": "markdown", - "id": "3c88f88a", + "id": "42b264a0", "metadata": { "editable": true }, @@ -123,7 +2858,7 @@ }, { "cell_type": "markdown", - "id": "d5927384", + "id": "ff85b5e7", "metadata": { "editable": true }, @@ -139,7 +2874,7 @@ }, { "cell_type": "markdown", - "id": "1e8e6d71", + "id": "5d0533bb", "metadata": { "editable": true }, @@ -153,7 +2888,7 @@ }, { "cell_type": "markdown", - "id": "e43296fa", + "id": "3ab3921a", "metadata": { "editable": true }, @@ -181,7 +2916,7 @@ }, { "cell_type": "markdown", - "id": "a2cb346d", + "id": "0595f2f0", "metadata": { "editable": true }, @@ -193,16 +2928,14 @@ }, { "cell_type": "code", - "execution_count": 1, - "id": "1e3f279d", + "execution_count": 16, + "id": "2902bb22", "metadata": { "collapsed": false, "editable": true }, "outputs": [], "source": [ - "%matplotlib inline\n", - "\n", "import numpy as np\n", "import matplotlib.pyplot as plt\n", "from sklearn.model_selection import KFold\n", @@ -293,2723 +3026,6 @@ "\n", "plt.show()" ] - }, - { - "cell_type": "markdown", - "id": "1de41ebb", - "metadata": { - "editable": true - }, - "source": [ - "## Material for lecture Thursday September 21" - ] - }, - { - "cell_type": "markdown", - "id": "c7c59b25", - "metadata": { - "editable": true - }, - "source": [ - "## Logistic Regression\n", - "\n", - "In linear regression our main interest was centered on learning the\n", - "coefficients of a functional fit (say a polynomial) in order to be\n", - "able to predict the response of a continuous variable on some unseen\n", - "data. The fit to the continuous variable $y_i$ is based on some\n", - "independent variables $\\boldsymbol{x}_i$. Linear regression resulted in\n", - "analytical expressions for standard ordinary Least Squares or Ridge\n", - "regression (in terms of matrices to invert) for several quantities,\n", - "ranging from the variance and thereby the confidence intervals of the\n", - "parameters $\\boldsymbol{\\beta}$ to the mean squared error. If we can invert\n", - "the product of the design matrices, linear regression gives then a\n", - "simple recipe for fitting our data." - ] - }, - { - "cell_type": "markdown", - "id": "dffe4c20", - "metadata": { - "editable": true - }, - "source": [ - "## Classification problems\n", - "\n", - "Classification problems, however, are concerned with outcomes taking\n", - "the form of discrete variables (i.e. categories). We may for example,\n", - "on the basis of DNA sequencing for a number of patients, like to find\n", - "out which mutations are important for a certain disease; or based on\n", - "scans of various patients' brains, figure out if there is a tumor or\n", - "not; or given a specific physical system, we'd like to identify its\n", - "state, say whether it is an ordered or disordered system (typical\n", - "situation in solid state physics); or classify the status of a\n", - "patient, whether she/he has a stroke or not and many other similar\n", - "situations.\n", - "\n", - "The most common situation we encounter when we apply logistic\n", - "regression is that of two possible outcomes, normally denoted as a\n", - "binary outcome, true or false, positive or negative, success or\n", - "failure etc." - ] - }, - { - "cell_type": "markdown", - "id": "4622a0ef", - "metadata": { - "editable": true - }, - "source": [ - "## Optimization and Deep learning\n", - "\n", - "Logistic regression will also serve as our stepping stone towards\n", - "neural network algorithms and supervised deep learning. For logistic\n", - "learning, the minimization of the cost function leads to a non-linear\n", - "equation in the parameters $\\boldsymbol{\\beta}$. The optimization of the\n", - "problem calls therefore for minimization algorithms. This forms the\n", - "bottle neck of all machine learning algorithms, namely how to find\n", - "reliable minima of a multi-variable function. This leads us to the\n", - "family of gradient descent methods. The latter are the working horses\n", - "of basically all modern machine learning algorithms.\n", - "\n", - "We note also that many of the topics discussed here on logistic \n", - "regression are also commonly used in modern supervised Deep Learning\n", - "models, as we will see later." - ] - }, - { - "cell_type": "markdown", - "id": "98aee52d", - "metadata": { - "editable": true - }, - "source": [ - "## Basics\n", - "\n", - "We consider the case where the dependent variables, also called the\n", - "responses or the outcomes, $y_i$ are discrete and only take values\n", - "from $k=0,\\dots,K-1$ (i.e. $K$ classes).\n", - "\n", - "The goal is to predict the\n", - "output classes from the design matrix $\\boldsymbol{X}\\in\\mathbb{R}^{n\\times p}$\n", - "made of $n$ samples, each of which carries $p$ features or predictors. The\n", - "primary goal is to identify the classes to which new unseen samples\n", - "belong.\n", - "\n", - "Let us specialize to the case of two classes only, with outputs\n", - "$y_i=0$ and $y_i=1$. Our outcomes could represent the status of a\n", - "credit card user that could default or not on her/his credit card\n", - "debt. That is" - ] - }, - { - "cell_type": "markdown", - "id": "8c5ed826", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "y_i = \\begin{bmatrix} 0 & \\mathrm{no}\\\\ 1 & \\mathrm{yes} \\end{bmatrix}.\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "9d206548", - "metadata": { - "editable": true - }, - "source": [ - "## Linear classifier\n", - "\n", - "Before moving to the logistic model, let us try to use our linear\n", - "regression model to classify these two outcomes. We could for example\n", - "fit a linear model to the default case if $y_i > 0.5$ and the no\n", - "default case $y_i \\leq 0.5$.\n", - "\n", - "We would then have our \n", - "weighted linear combination, namely" - ] - }, - { - "cell_type": "markdown", - "id": "0d15e902", - "metadata": { - "editable": true - }, - "source": [ - "\n", - "
            \n", - "\n", - "$$\n", - "\\begin{equation}\n", - "\\boldsymbol{y} = \\boldsymbol{X}^T\\boldsymbol{\\beta} + \\boldsymbol{\\epsilon},\n", - "\\label{_auto1} \\tag{1}\n", - "\\end{equation}\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "80839355", - "metadata": { - "editable": true - }, - "source": [ - "where $\\boldsymbol{y}$ is a vector representing the possible outcomes, $\\boldsymbol{X}$ is our\n", - "$n\\times p$ design matrix and $\\boldsymbol{\\beta}$ represents our estimators/predictors." - ] - }, - { - "cell_type": "markdown", - "id": "0d6274e7", - "metadata": { - "editable": true - }, - "source": [ - "## Some selected properties\n", - "\n", - "The main problem with our function is that it takes values on the\n", - "entire real axis. In the case of logistic regression, however, the\n", - "labels $y_i$ are discrete variables. A typical example is the credit\n", - "card data discussed below here, where we can set the state of\n", - "defaulting the debt to $y_i=1$ and not to $y_i=0$ for one the persons\n", - "in the data set (see the full example below).\n", - "\n", - "One simple way to get a discrete output is to have sign\n", - "functions that map the output of a linear regressor to values $\\{0,1\\}$,\n", - "$f(s_i)=sign(s_i)=1$ if $s_i\\ge 0$ and 0 if otherwise. \n", - "We will encounter this model in our first demonstration of neural networks.\n", - "\n", - "Historically it is called the **perceptron** model in the machine learning\n", - "literature. This model is extremely simple. However, in many cases it is more\n", - "favorable to use a ``soft\" classifier that outputs\n", - "the probability of a given category. This leads us to the logistic function." - ] - }, - { - "cell_type": "markdown", - "id": "a334d17f", - "metadata": { - "editable": true - }, - "source": [ - "## Simple example\n", - "\n", - "The following example on data for coronary heart disease (CHD) as function of age may serve as an illustration. In the code here we read and plot whether a person has had CHD (output = 1) or not (output = 0). This ouput is plotted the person's against age. Clearly, the figure shows that attempting to make a standard linear regression fit may not be very meaningful." - ] - }, - { - "cell_type": "code", - "execution_count": 2, - "id": "389b3fd0", - "metadata": { - "collapsed": false, - "editable": true - }, - "outputs": [], - "source": [ - "# Common imports\n", - "import os\n", - "import numpy as np\n", - "import pandas as pd\n", - "import matplotlib.pyplot as plt\n", - "from sklearn.linear_model import LinearRegression, Ridge, Lasso\n", - "from sklearn.model_selection import train_test_split\n", - "from sklearn.utils import resample\n", - "from sklearn.metrics import mean_squared_error\n", - "from IPython.display import display\n", - "from pylab import plt, mpl\n", - "plt.style.use('seaborn')\n", - "mpl.rcParams['font.family'] = 'serif'\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(\"chddata.csv\"),'r')\n", - "\n", - "# Read the chd data as csv file and organize the data into arrays with age group, age, and chd\n", - "chd = pd.read_csv(infile, names=('ID', 'Age', 'Agegroup', 'CHD'))\n", - "chd.columns = ['ID', 'Age', 'Agegroup', 'CHD']\n", - "output = chd['CHD']\n", - "age = chd['Age']\n", - "agegroup = chd['Agegroup']\n", - "numberID = chd['ID'] \n", - "display(chd)\n", - "\n", - "plt.scatter(age, output, marker='o')\n", - "plt.axis([18,70.0,-0.1, 1.2])\n", - "plt.xlabel(r'Age')\n", - "plt.ylabel(r'CHD')\n", - "plt.title(r'Age distribution and Coronary heart disease')\n", - "plt.show()" - ] - }, - { - "cell_type": "markdown", - "id": "79264b3f", - "metadata": { - "editable": true - }, - "source": [ - "## Plotting the mean value for each group\n", - "\n", - "What we could attempt however is to plot the mean value for each group." - ] - }, - { - "cell_type": "code", - "execution_count": 3, - "id": "363298b9", - "metadata": { - "collapsed": false, - "editable": true - }, - "outputs": [], - "source": [ - "agegroupmean = np.array([0.1, 0.133, 0.250, 0.333, 0.462, 0.625, 0.765, 0.800])\n", - "group = np.array([1, 2, 3, 4, 5, 6, 7, 8])\n", - "plt.plot(group, agegroupmean, \"r-\")\n", - "plt.axis([0,9,0, 1.0])\n", - "plt.xlabel(r'Age group')\n", - "plt.ylabel(r'CHD mean values')\n", - "plt.title(r'Mean values for each age group')\n", - "plt.show()" - ] - }, - { - "cell_type": "markdown", - "id": "8726b22c", - "metadata": { - "editable": true - }, - "source": [ - "We are now trying to find a function $f(y\\vert x)$, that is a function which gives us an expected value for the output $y$ with a given input $x$.\n", - "In standard linear regression with a linear dependence on $x$, we would write this in terms of our model" - ] - }, - { - "cell_type": "markdown", - "id": "4cae8892", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "f(y_i\\vert x_i)=\\beta_0+\\beta_1 x_i.\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "067b39b0", - "metadata": { - "editable": true - }, - "source": [ - "This expression implies however that $f(y_i\\vert x_i)$ could take any\n", - "value from minus infinity to plus infinity. If we however let\n", - "$f(y\\vert y)$ be represented by the mean value, the above example\n", - "shows us that we can constrain the function to take values between\n", - "zero and one, that is we have $0 \\le f(y_i\\vert x_i) \\le 1$. Looking\n", - "at our last curve we see also that it has an S-shaped form. This leads\n", - "us to a very popular model for the function $f$, namely the so-called\n", - "Sigmoid function or logistic model. We will consider this function as\n", - "representing the probability for finding a value of $y_i$ with a given\n", - "$x_i$." - ] - }, - { - "cell_type": "markdown", - "id": "64df3b8b", - "metadata": { - "editable": true - }, - "source": [ - "## The logistic function\n", - "\n", - "Another widely studied model, is the so-called \n", - "perceptron model, which is an example of a \"hard classification\" model. We\n", - "will encounter this model when we discuss neural networks as\n", - "well. Each datapoint is deterministically assigned to a category (i.e\n", - "$y_i=0$ or $y_i=1$). In many cases, and the coronary heart disease data forms one of many such examples, it is favorable to have a \"soft\"\n", - "classifier that outputs the probability of a given category rather\n", - "than a single value. For example, given $x_i$, the classifier\n", - "outputs the probability of being in a category $k$. Logistic regression\n", - "is the most common example of a so-called soft classifier. In logistic\n", - "regression, the probability that a data point $x_i$\n", - "belongs to a category $y_i=\\{0,1\\}$ is given by the so-called logit function (or Sigmoid) which is meant to represent the likelihood for a given event," - ] - }, - { - "cell_type": "markdown", - "id": "395ebb35", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "p(t) = \\frac{1}{1+\\mathrm \\exp{-t}}=\\frac{\\exp{t}}{1+\\mathrm \\exp{t}}.\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "1664d57c", - "metadata": { - "editable": true - }, - "source": [ - "Note that $1-p(t)= p(-t)$." - ] - }, - { - "cell_type": "markdown", - "id": "a9d24fae", - "metadata": { - "editable": true - }, - "source": [ - "## Examples of likelihood functions used in logistic regression and nueral networks\n", - "\n", - "The following code plots the logistic function, the step function and other functions we will encounter from here and on." - ] - }, - { - "cell_type": "code", - "execution_count": 4, - "id": "9f45ea41", - "metadata": { - "collapsed": false, - "editable": true - }, - "outputs": [], - "source": [ - "\"\"\"The sigmoid function (or the logistic curve) is a\n", - "function that takes any real number, z, and outputs a number (0,1).\n", - "It is useful in neural networks for assigning weights on a relative scale.\n", - "The value z is the weighted sum of parameters involved in the learning algorithm.\"\"\"\n", - "\n", - "import numpy\n", - "import matplotlib.pyplot as plt\n", - "import math as mt\n", - "\n", - "z = numpy.arange(-5, 5, .1)\n", - "sigma_fn = numpy.vectorize(lambda z: 1/(1+numpy.exp(-z)))\n", - "sigma = sigma_fn(z)\n", - "\n", - "fig = plt.figure()\n", - "ax = fig.add_subplot(111)\n", - "ax.plot(z, sigma)\n", - "ax.set_ylim([-0.1, 1.1])\n", - "ax.set_xlim([-5,5])\n", - "ax.grid(True)\n", - "ax.set_xlabel('z')\n", - "ax.set_title('sigmoid function')\n", - "\n", - "plt.show()\n", - "\n", - "\"\"\"Step Function\"\"\"\n", - "z = numpy.arange(-5, 5, .02)\n", - "step_fn = numpy.vectorize(lambda z: 1.0 if z >= 0.0 else 0.0)\n", - "step = step_fn(z)\n", - "\n", - "fig = plt.figure()\n", - "ax = fig.add_subplot(111)\n", - "ax.plot(z, step)\n", - "ax.set_ylim([-0.5, 1.5])\n", - "ax.set_xlim([-5,5])\n", - "ax.grid(True)\n", - "ax.set_xlabel('z')\n", - "ax.set_title('step function')\n", - "\n", - "plt.show()\n", - "\n", - "\"\"\"tanh Function\"\"\"\n", - "z = numpy.arange(-2*mt.pi, 2*mt.pi, 0.1)\n", - "t = numpy.tanh(z)\n", - "\n", - "fig = plt.figure()\n", - "ax = fig.add_subplot(111)\n", - "ax.plot(z, t)\n", - "ax.set_ylim([-1.0, 1.0])\n", - "ax.set_xlim([-2*mt.pi,2*mt.pi])\n", - "ax.grid(True)\n", - "ax.set_xlabel('z')\n", - "ax.set_title('tanh function')\n", - "\n", - "plt.show()" - ] - }, - { - "cell_type": "markdown", - "id": "5523d26d", - "metadata": { - "editable": true - }, - "source": [ - "## Two parameters\n", - "\n", - "We assume now that we have two classes with $y_i$ either $0$ or $1$. Furthermore we assume also that we have only two parameters $\\beta$ in our fitting of the Sigmoid function, that is we define probabilities" - ] - }, - { - "cell_type": "markdown", - "id": "2acaf89d", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\begin{align*}\n", - "p(y_i=1|x_i,\\boldsymbol{\\beta}) &= \\frac{\\exp{(\\beta_0+\\beta_1x_i)}}{1+\\exp{(\\beta_0+\\beta_1x_i)}},\\nonumber\\\\\n", - "p(y_i=0|x_i,\\boldsymbol{\\beta}) &= 1 - p(y_i=1|x_i,\\boldsymbol{\\beta}),\n", - "\\end{align*}\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "a18fd4cf", - "metadata": { - "editable": true - }, - "source": [ - "where $\\boldsymbol{\\beta}$ are the weights we wish to extract from data, in our case $\\beta_0$ and $\\beta_1$. \n", - "\n", - "Note that we used" - ] - }, - { - "cell_type": "markdown", - "id": "b082e7fd", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "p(y_i=0\\vert x_i, \\boldsymbol{\\beta}) = 1-p(y_i=1\\vert x_i, \\boldsymbol{\\beta}).\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "db385c32", - "metadata": { - "editable": true - }, - "source": [ - "## Maximum likelihood\n", - "\n", - "In order to define the total likelihood for all possible outcomes from a \n", - "dataset $\\mathcal{D}=\\{(y_i,x_i)\\}$, with the binary labels\n", - "$y_i\\in\\{0,1\\}$ and where the data points are drawn independently, we use the so-called [Maximum Likelihood Estimation](https://en.wikipedia.org/wiki/Maximum_likelihood_estimation) (MLE) principle. \n", - "We aim thus at maximizing \n", - "the probability of seeing the observed data. We can then approximate the \n", - "likelihood in terms of the product of the individual probabilities of a specific outcome $y_i$, that is" - ] - }, - { - "cell_type": "markdown", - "id": "274e8ace", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\begin{align*}\n", - "P(\\mathcal{D}|\\boldsymbol{\\beta})& = \\prod_{i=1}^n \\left[p(y_i=1|x_i,\\boldsymbol{\\beta})\\right]^{y_i}\\left[1-p(y_i=1|x_i,\\boldsymbol{\\beta}))\\right]^{1-y_i}\\nonumber \\\\\n", - "\\end{align*}\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "13b0bc0c", - "metadata": { - "editable": true - }, - "source": [ - "from which we obtain the log-likelihood and our **cost/loss** function" - ] - }, - { - "cell_type": "markdown", - "id": "c5e6ed58", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\mathcal{C}(\\boldsymbol{\\beta}) = \\sum_{i=1}^n \\left( y_i\\log{p(y_i=1|x_i,\\boldsymbol{\\beta})} + (1-y_i)\\log\\left[1-p(y_i=1|x_i,\\boldsymbol{\\beta}))\\right]\\right).\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "05da3d11", - "metadata": { - "editable": true - }, - "source": [ - "## The cost function rewritten\n", - "\n", - "Reordering the logarithms, we can rewrite the **cost/loss** function as" - ] - }, - { - "cell_type": "markdown", - "id": "197b7a03", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\mathcal{C}(\\boldsymbol{\\beta}) = \\sum_{i=1}^n \\left(y_i(\\beta_0+\\beta_1x_i) -\\log{(1+\\exp{(\\beta_0+\\beta_1x_i)})}\\right).\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "04c5a708", - "metadata": { - "editable": true - }, - "source": [ - "The maximum likelihood estimator is defined as the set of parameters that maximize the log-likelihood where we maximize with respect to $\\beta$.\n", - "Since the cost (error) function is just the negative log-likelihood, for logistic regression we have that" - ] - }, - { - "cell_type": "markdown", - "id": "433c8b68", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\mathcal{C}(\\boldsymbol{\\beta})=-\\sum_{i=1}^n \\left(y_i(\\beta_0+\\beta_1x_i) -\\log{(1+\\exp{(\\beta_0+\\beta_1x_i)})}\\right).\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "1dd419f9", - "metadata": { - "editable": true - }, - "source": [ - "This equation is known in statistics as the **cross entropy**. Finally, we note that just as in linear regression, \n", - "in practice we often supplement the cross-entropy with additional regularization terms, usually $L_1$ and $L_2$ regularization as we did for Ridge and Lasso regression." - ] - }, - { - "cell_type": "markdown", - "id": "f6da85f3", - "metadata": { - "editable": true - }, - "source": [ - "## Minimizing the cross entropy\n", - "\n", - "The cross entropy is a convex function of the weights $\\boldsymbol{\\beta}$ and,\n", - "therefore, any local minimizer is a global minimizer. \n", - "\n", - "Minimizing this\n", - "cost function with respect to the two parameters $\\beta_0$ and $\\beta_1$ we obtain" - ] - }, - { - "cell_type": "markdown", - "id": "b218ca83", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\frac{\\partial \\mathcal{C}(\\boldsymbol{\\beta})}{\\partial \\beta_0} = -\\sum_{i=1}^n \\left(y_i -\\frac{\\exp{(\\beta_0+\\beta_1x_i)}}{1+\\exp{(\\beta_0+\\beta_1x_i)}}\\right),\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "9df81637", - "metadata": { - "editable": true - }, - "source": [ - "and" - ] - }, - { - "cell_type": "markdown", - "id": "2c6810d9", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\frac{\\partial \\mathcal{C}(\\boldsymbol{\\beta})}{\\partial \\beta_1} = -\\sum_{i=1}^n \\left(y_ix_i -x_i\\frac{\\exp{(\\beta_0+\\beta_1x_i)}}{1+\\exp{(\\beta_0+\\beta_1x_i)}}\\right).\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "cf887b17", - "metadata": { - "editable": true - }, - "source": [ - "## A more compact expression\n", - "\n", - "Let us now define a vector $\\boldsymbol{y}$ with $n$ elements $y_i$, an\n", - "$n\\times p$ matrix $\\boldsymbol{X}$ which contains the $x_i$ values and a\n", - "vector $\\boldsymbol{p}$ of fitted probabilities $p(y_i\\vert x_i,\\boldsymbol{\\beta})$. We can rewrite in a more compact form the first\n", - "derivative of cost function as" - ] - }, - { - "cell_type": "markdown", - "id": "b8d00a49", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\frac{\\partial \\mathcal{C}(\\boldsymbol{\\beta})}{\\partial \\boldsymbol{\\beta}} = -\\boldsymbol{X}^T\\left(\\boldsymbol{y}-\\boldsymbol{p}\\right).\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "2f792720", - "metadata": { - "editable": true - }, - "source": [ - "If we in addition define a diagonal matrix $\\boldsymbol{W}$ with elements \n", - "$p(y_i\\vert x_i,\\boldsymbol{\\beta})(1-p(y_i\\vert x_i,\\boldsymbol{\\beta})$, we can obtain a compact expression of the second derivative as" - ] - }, - { - "cell_type": "markdown", - "id": "eb2085a0", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\frac{\\partial^2 \\mathcal{C}(\\boldsymbol{\\beta})}{\\partial \\boldsymbol{\\beta}\\partial \\boldsymbol{\\beta}^T} = \\boldsymbol{X}^T\\boldsymbol{W}\\boldsymbol{X}.\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "79468f06", - "metadata": { - "editable": true - }, - "source": [ - "## Extending to more predictors\n", - "\n", - "Within a binary classification problem, we can easily expand our model to include multiple predictors. Our ratio between likelihoods is then with $p$ predictors" - ] - }, - { - "cell_type": "markdown", - "id": "6c8c7cf8", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\log{ \\frac{p(\\boldsymbol{\\beta}\\boldsymbol{x})}{1-p(\\boldsymbol{\\beta}\\boldsymbol{x})}} = \\beta_0+\\beta_1x_1+\\beta_2x_2+\\dots+\\beta_px_p.\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "1c371a59", - "metadata": { - "editable": true - }, - "source": [ - "Here we defined $\\boldsymbol{x}=[1,x_1,x_2,\\dots,x_p]$ and $\\boldsymbol{\\beta}=[\\beta_0, \\beta_1, \\dots, \\beta_p]$ leading to" - ] - }, - { - "cell_type": "markdown", - "id": "bddcf893", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "p(\\boldsymbol{\\beta}\\boldsymbol{x})=\\frac{ \\exp{(\\beta_0+\\beta_1x_1+\\beta_2x_2+\\dots+\\beta_px_p)}}{1+\\exp{(\\beta_0+\\beta_1x_1+\\beta_2x_2+\\dots+\\beta_px_p)}}.\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "130af732", - "metadata": { - "editable": true - }, - "source": [ - "## Including more classes\n", - "\n", - "Till now we have mainly focused on two classes, the so-called binary\n", - "system. Suppose we wish to extend to $K$ classes. Let us for the sake\n", - "of simplicity assume we have only two predictors. We have then following model" - ] - }, - { - "cell_type": "markdown", - "id": "e221378d", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\log{\\frac{p(C=1\\vert x)}{p(K\\vert x)}} = \\beta_{10}+\\beta_{11}x_1,\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "10ec286a", - "metadata": { - "editable": true - }, - "source": [ - "and" - ] - }, - { - "cell_type": "markdown", - "id": "233ef2db", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\log{\\frac{p(C=2\\vert x)}{p(K\\vert x)}} = \\beta_{20}+\\beta_{21}x_1,\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "4cc7a244", - "metadata": { - "editable": true - }, - "source": [ - "and so on till the class $C=K-1$ class" - ] - }, - { - "cell_type": "markdown", - "id": "15d596d5", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\log{\\frac{p(C=K-1\\vert x)}{p(K\\vert x)}} = \\beta_{(K-1)0}+\\beta_{(K-1)1}x_1,\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "66cd8be9", - "metadata": { - "editable": true - }, - "source": [ - "and the model is specified in term of $K-1$ so-called log-odds or\n", - "**logit** transformations." - ] - }, - { - "cell_type": "markdown", - "id": "e5d7c12a", - "metadata": { - "editable": true - }, - "source": [ - "## More classes\n", - "\n", - "In our discussion of neural networks we will encounter the above again\n", - "in terms of a slightly modified function, the so-called **Softmax** function.\n", - "\n", - "The softmax function is used in various multiclass classification\n", - "methods, such as multinomial logistic regression (also known as\n", - "softmax regression), multiclass linear discriminant analysis, naive\n", - "Bayes classifiers, and artificial neural networks. Specifically, in\n", - "multinomial logistic regression and linear discriminant analysis, the\n", - "input to the function is the result of $K$ distinct linear functions,\n", - "and the predicted probability for the $k$-th class given a sample\n", - "vector $\\boldsymbol{x}$ and a weighting vector $\\boldsymbol{\\beta}$ is (with two\n", - "predictors):" - ] - }, - { - "cell_type": "markdown", - "id": "ac38c1a6", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "p(C=k\\vert \\mathbf {x} )=\\frac{\\exp{(\\beta_{k0}+\\beta_{k1}x_1)}}{1+\\sum_{l=1}^{K-1}\\exp{(\\beta_{l0}+\\beta_{l1}x_1)}}.\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "42a1c905", - "metadata": { - "editable": true - }, - "source": [ - "It is easy to extend to more predictors. The final class is" - ] - }, - { - "cell_type": "markdown", - "id": "a0cf54f7", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "p(C=K\\vert \\mathbf {x} )=\\frac{1}{1+\\sum_{l=1}^{K-1}\\exp{(\\beta_{l0}+\\beta_{l1}x_1)}},\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "c36dab59", - "metadata": { - "editable": true - }, - "source": [ - "and they sum to one. Our earlier discussions were all specialized to\n", - "the case with two classes only. It is easy to see from the above that\n", - "what we derived earlier is compatible with these equations.\n", - "\n", - "To find the optimal parameters we would typically use a gradient\n", - "descent method. Newton's method and gradient descent methods are\n", - "discussed in the material on [optimization\n", - "methods](https://compphysics.github.io/MachineLearning/doc/pub/Splines/html/Splines-bs.html)." - ] - }, - { - "cell_type": "markdown", - "id": "976a80db", - "metadata": { - "editable": true - }, - "source": [ - "## Friday September 23" - ] - }, - { - "cell_type": "markdown", - "id": "c1e05822", - "metadata": { - "editable": true - }, - "source": [ - "## Searching for Optimal Regularization Parameters $\\lambda$\n", - "\n", - "In project 1, when using Ridge and Lasso regression, we end up\n", - "searching for the optimal parameter $\\lambda$ which minimizes our\n", - "selected scores (MSE or $R2$ values for example). The brute force\n", - "approach, as discussed in the code here for Ridge regression, consists\n", - "in evaluating the MSE as function of different $\\lambda$ values.\n", - "Based on these calculations, one tries then to determine the value of the hyperparameter $\\lambda$\n", - "which results in optimal scores (for example the smallest MSE or an $R2=1$)." - ] - }, - { - "cell_type": "code", - "execution_count": 5, - "id": "2718e48d", - "metadata": { - "collapsed": false, - "editable": true - }, - "outputs": [], - "source": [ - "import numpy as np\n", - "import pandas as pd\n", - "import matplotlib.pyplot as plt\n", - "from sklearn.model_selection import train_test_split\n", - "from sklearn import linear_model\n", - "\n", - "def MSE(y_data,y_model):\n", - " n = np.size(y_model)\n", - " return np.sum((y_data-y_model)**2)/n\n", - "# A seed just to ensure that the random numbers are the same for every run.\n", - "# Useful for eventual debugging.\n", - "np.random.seed(2021)\n", - "\n", - "n = 100\n", - "x = np.random.rand(n)\n", - "y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.randn(n)\n", - "\n", - "Maxpolydegree = 5\n", - "X = np.zeros((n,Maxpolydegree-1))\n", - "\n", - "for degree in range(1,Maxpolydegree): #No intercept column\n", - " X[:,degree-1] = x**(degree)\n", - "\n", - "# We split the data in test and training data\n", - "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)\n", - "\n", - "# Decide which values of lambda to use\n", - "nlambdas = 500\n", - "MSERidgePredict = np.zeros(nlambdas)\n", - "lambdas = np.logspace(-4, 2, nlambdas)\n", - "for i in range(nlambdas):\n", - " lmb = lambdas[i]\n", - " RegRidge = linear_model.Ridge(lmb)\n", - " RegRidge.fit(X_train,y_train)\n", - " ypredictRidge = RegRidge.predict(X_test)\n", - " MSERidgePredict[i] = MSE(y_test,ypredictRidge)\n", - "\n", - "# Now plot the results\n", - "plt.figure()\n", - "plt.plot(np.log10(lambdas), MSERidgePredict, 'g--', label = 'MSE SL Ridge Test')\n", - "plt.xlabel('log10(lambda)')\n", - "plt.ylabel('MSE')\n", - "plt.legend()\n", - "plt.show()" - ] - }, - { - "cell_type": "markdown", - "id": "a2e5858b", - "metadata": { - "editable": true - }, - "source": [ - "Here we have performed a rather data greedy calculation as function of the regularization parameter $\\lambda$. There is no resampling here. The latter can easily be added by employing the function **RidgeCV** instead of just calling the **Ridge** function. For **RidgeCV** we need to pass the array of $\\lambda$ values.\n", - "By inspecting the figure we can in turn determine which is the optimal regularization parameter.\n", - "This becomes however less functional in the long run." - ] - }, - { - "cell_type": "markdown", - "id": "98b90510", - "metadata": { - "editable": true - }, - "source": [ - "## Grid Search\n", - "\n", - "An alternative is to use the so-called grid search functionality\n", - "included with the library **Scikit-Learn**, as demonstrated for the same\n", - "example here." - ] - }, - { - "cell_type": "code", - "execution_count": 6, - "id": "5eac16a1", - "metadata": { - "collapsed": false, - "editable": true - }, - "outputs": [], - "source": [ - "import numpy as np\n", - "from sklearn.model_selection import train_test_split\n", - "from sklearn.linear_model import Ridge\n", - "from sklearn.model_selection import GridSearchCV\n", - "\n", - "def R2(y_data, y_model):\n", - " return 1 - np.sum((y_data - y_model) ** 2) / np.sum((y_data - np.mean(y_data)) ** 2)\n", - "\n", - "def MSE(y_data,y_model):\n", - " n = np.size(y_model)\n", - " return np.sum((y_data-y_model)**2)/n\n", - "\n", - "# A seed just to ensure that the random numbers are the same for every run.\n", - "# Useful for eventual debugging.\n", - "np.random.seed(2021)\n", - "\n", - "n = 100\n", - "x = np.random.rand(n)\n", - "y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.randn(n)\n", - "\n", - "Maxpolydegree = 5\n", - "X = np.zeros((n,Maxpolydegree-1))\n", - "\n", - "for degree in range(1,Maxpolydegree): #No intercept column\n", - " X[:,degree-1] = x**(degree)\n", - "\n", - "# We split the data in test and training data\n", - "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)\n", - "\n", - "# Decide which values of lambda to use\n", - "nlambdas = 10\n", - "lambdas = np.logspace(-4, 2, nlambdas)\n", - "# create and fit a ridge regression model, testing each alpha\n", - "model = Ridge()\n", - "gridsearch = GridSearchCV(estimator=model, param_grid=dict(alpha=lambdas))\n", - "gridsearch.fit(X_train, y_train)\n", - "print(gridsearch)\n", - "ypredictRidge = gridsearch.predict(X_test)\n", - "# summarize the results of the grid search\n", - "print(f\"Best estimated lambda-value: {gridsearch.best_estimator_.alpha}\")\n", - "print(f\"MSE score: {MSE(y_test,ypredictRidge)}\")\n", - "print(f\"R2 score: {R2(y_test,ypredictRidge)}\")" - ] - }, - { - "cell_type": "markdown", - "id": "1ca2c928", - "metadata": { - "editable": true - }, - "source": [ - "By default the grid search function includes cross validation with\n", - "five folds. The [Scikit-Learn\n", - "documentation](https://scikit-learn.org/stable/modules/generated/sklearn.model_selection.GridSearchCV.html#sklearn.model_selection.GridSearchCV)\n", - "contains more information on how to set the different parameters.\n", - "\n", - "If we take out the random noise, running the above codes results in $\\lambda=0$ yielding the best fit." - ] - }, - { - "cell_type": "markdown", - "id": "2b23985a", - "metadata": { - "editable": true - }, - "source": [ - "## Randomized Grid Search\n", - "\n", - "An alternative to the above manual grid set up, is to use a random\n", - "search where the parameters are tuned from a random distribution\n", - "(uniform below) for a fixed number of iterations. A model is\n", - "constructed and evaluated for each combination of chosen parameters.\n", - "We repeat the previous example but now with a random search. Note\n", - "that values of $\\lambda$ are now limited to be within $x\\in\n", - "[0,1]$. This domain may not be the most relevant one for the specific\n", - "case under study." - ] - }, - { - "cell_type": "code", - "execution_count": 7, - "id": "959dfd40", - "metadata": { - "collapsed": false, - "editable": true - }, - "outputs": [], - "source": [ - "import numpy as np\n", - "from sklearn.model_selection import train_test_split\n", - "from sklearn.linear_model import Ridge\n", - "from sklearn.model_selection import GridSearchCV\n", - "from scipy.stats import uniform as randuniform\n", - "from sklearn.model_selection import RandomizedSearchCV\n", - "\n", - "\n", - "def R2(y_data, y_model):\n", - " return 1 - np.sum((y_data - y_model) ** 2) / np.sum((y_data - np.mean(y_data)) ** 2)\n", - "\n", - "def MSE(y_data,y_model):\n", - " n = np.size(y_model)\n", - " return np.sum((y_data-y_model)**2)/n\n", - "\n", - "# A seed just to ensure that the random numbers are the same for every run.\n", - "# Useful for eventual debugging.\n", - "np.random.seed(2021)\n", - "\n", - "n = 100\n", - "x = np.random.rand(n)\n", - "y = np.exp(-x**2) + 1.5 * np.exp(-(x-2)**2)+ np.random.randn(n)\n", - "\n", - "Maxpolydegree = 5\n", - "X = np.zeros((n,Maxpolydegree-1))\n", - "\n", - "for degree in range(1,Maxpolydegree): #No intercept column\n", - " X[:,degree-1] = x**(degree)\n", - "\n", - "# We split the data in test and training data\n", - "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)\n", - "\n", - "param_grid = {'alpha': randuniform()}\n", - "# create and fit a ridge regression model, testing each alpha\n", - "model = Ridge()\n", - "gridsearch = RandomizedSearchCV(estimator=model, param_distributions=param_grid, n_iter=100)\n", - "gridsearch.fit(X_train, y_train)\n", - "print(gridsearch)\n", - "ypredictRidge = gridsearch.predict(X_test)\n", - "# summarize the results of the grid search\n", - "print(f\"Best estimated lambda-value: {gridsearch.best_estimator_.alpha}\")\n", - "print(f\"MSE score: {MSE(y_test,ypredictRidge)}\")\n", - "print(f\"R2 score: {R2(y_test,ypredictRidge)}\")" - ] - }, - { - "cell_type": "markdown", - "id": "5d99762f", - "metadata": { - "editable": true - }, - "source": [ - "## Wisconsin Cancer Data\n", - "\n", - "We show here how we can use a simple regression case on the breast\n", - "cancer data using Logistic regression as our algorithm for\n", - "classification." - ] - }, - { - "cell_type": "code", - "execution_count": 8, - "id": "ebb174de", - "metadata": { - "collapsed": false, - "editable": true - }, - "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.linear_model import LogisticRegression\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)))" - ] - }, - { - "cell_type": "markdown", - "id": "d274920d", - "metadata": { - "editable": true - }, - "source": [ - "## Using the correlation matrix\n", - "\n", - "In addition to the above scores, we could also study the covariance (and the correlation matrix).\n", - "We use **Pandas** to compute the correlation matrix." - ] - }, - { - "cell_type": "code", - "execution_count": 9, - "id": "3449f395", - "metadata": { - "collapsed": false, - "editable": true - }, - "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.linear_model import LogisticRegression\n", - "cancer = load_breast_cancer()\n", - "import pandas as pd\n", - "# Making a data frame\n", - "cancerpd = pd.DataFrame(cancer.data, columns=cancer.feature_names)\n", - "\n", - "fig, axes = plt.subplots(15,2,figsize=(10,20))\n", - "malignant = cancer.data[cancer.target == 0]\n", - "benign = cancer.data[cancer.target == 1]\n", - "ax = axes.ravel()\n", - "\n", - "for i in range(30):\n", - " _, bins = np.histogram(cancer.data[:,i], bins =50)\n", - " ax[i].hist(malignant[:,i], bins = bins, alpha = 0.5)\n", - " ax[i].hist(benign[:,i], bins = bins, alpha = 0.5)\n", - " ax[i].set_title(cancer.feature_names[i])\n", - " ax[i].set_yticks(())\n", - "ax[0].set_xlabel(\"Feature magnitude\")\n", - "ax[0].set_ylabel(\"Frequency\")\n", - "ax[0].legend([\"Malignant\", \"Benign\"], loc =\"best\")\n", - "fig.tight_layout()\n", - "plt.show()\n", - "\n", - "import seaborn as sns\n", - "correlation_matrix = cancerpd.corr().round(1)\n", - "# use the heatmap function from seaborn to plot the correlation matrix\n", - "# annot = True to print the values inside the square\n", - "plt.figure(figsize=(15,8))\n", - "sns.heatmap(data=correlation_matrix, annot=True)\n", - "plt.show()" - ] - }, - { - "cell_type": "markdown", - "id": "664bfafd", - "metadata": { - "editable": true - }, - "source": [ - "## Discussing the correlation data\n", - "\n", - "In the above example we note two things. In the first plot we display\n", - "the overlap of benign and malignant tumors as functions of the various\n", - "features in the Wisconsing breast cancer data set. We see that for\n", - "some of the features we can distinguish clearly the benign and\n", - "malignant cases while for other features we cannot. This can point to\n", - "us which features may be of greater interest when we wish to classify\n", - "a benign or not benign tumour.\n", - "\n", - "In the second figure we have computed the so-called correlation\n", - "matrix, which in our case with thirty features becomes a $30\\times 30$\n", - "matrix.\n", - "\n", - "We constructed this matrix using **pandas** via the statements" - ] - }, - { - "cell_type": "code", - "execution_count": 10, - "id": "361ab50d", - "metadata": { - "collapsed": false, - "editable": true - }, - "outputs": [], - "source": [ - "cancerpd = pd.DataFrame(cancer.data, columns=cancer.feature_names)" - ] - }, - { - "cell_type": "markdown", - "id": "615cd236", - "metadata": { - "editable": true - }, - "source": [ - "and then" - ] - }, - { - "cell_type": "code", - "execution_count": 11, - "id": "20df5011", - "metadata": { - "collapsed": false, - "editable": true - }, - "outputs": [], - "source": [ - "correlation_matrix = cancerpd.corr().round(1)" - ] - }, - { - "cell_type": "markdown", - "id": "748bf832", - "metadata": { - "editable": true - }, - "source": [ - "Diagonalizing this matrix we can in turn say something about which\n", - "features are of relevance and which are not. This leads us to\n", - "the classical Principal Component Analysis (PCA) theorem with\n", - "applications. This will be discussed later this semester ([week 43](https://compphysics.github.io/MachineLearning/doc/pub/week43/html/week43-bs.html))." - ] - }, - { - "cell_type": "markdown", - "id": "e7ffdb9e", - "metadata": { - "editable": true - }, - "source": [ - "## Other measures in classification studies: Cancer Data again" - ] - }, - { - "cell_type": "code", - "execution_count": 12, - "id": "f98af390", - "metadata": { - "collapsed": false, - "editable": true - }, - "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.linear_model import LogisticRegression\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", - "\n", - "from sklearn.preprocessing import LabelEncoder\n", - "from sklearn.model_selection import cross_validate\n", - "#Cross validation\n", - "accuracy = cross_validate(logreg,X_test,y_test,cv=10)['test_score']\n", - "print(accuracy)\n", - "print(\"Test set accuracy with Logistic Regression: {:.2f}\".format(logreg.score(X_test,y_test)))\n", - "\n", - "import scikitplot as skplt\n", - "y_pred = logreg.predict(X_test)\n", - "skplt.metrics.plot_confusion_matrix(y_test, y_pred, normalize=True)\n", - "plt.show()\n", - "y_probas = logreg.predict_proba(X_test)\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", - "id": "07bcaf40", - "metadata": { - "editable": true - }, - "source": [ - "## Optimization, the central part of any Machine Learning algortithm\n", - "\n", - "[Overview Video, why do we care about gradient methods?](https://www.uio.no/studier/emner/matnat/fys/FYS-STK3155/h20/forelesningsvideoer/OverarchingAimsWeek39.mp4?vrtx=view-as-webpage)\n", - "\n", - "Almost every problem in machine learning and data science starts with\n", - "a dataset $X$, a model $g(\\beta)$, which is a function of the\n", - "parameters $\\beta$ and a cost function $C(X, g(\\beta))$ that allows\n", - "us to judge how well the model $g(\\beta)$ explains the observations\n", - "$X$. The model is fit by finding the values of $\\beta$ that minimize\n", - "the cost function. Ideally we would be able to solve for $\\beta$\n", - "analytically, however this is not possible in general and we must use\n", - "some approximative/numerical method to compute the minimum." - ] - }, - { - "cell_type": "markdown", - "id": "861a7704", - "metadata": { - "editable": true - }, - "source": [ - "## Revisiting our Logistic Regression case\n", - "\n", - "In our discussion on Logistic Regression we studied the \n", - "case of\n", - "two classes, with $y_i$ either\n", - "$0$ or $1$. Furthermore we assumed also that we have only two\n", - "parameters $\\beta$ in our fitting, that is we\n", - "defined probabilities" - ] - }, - { - "cell_type": "markdown", - "id": "9b924e09", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\begin{align*}\n", - "p(y_i=1|x_i,\\boldsymbol{\\beta}) &= \\frac{\\exp{(\\beta_0+\\beta_1x_i)}}{1+\\exp{(\\beta_0+\\beta_1x_i)}},\\nonumber\\\\\n", - "p(y_i=0|x_i,\\boldsymbol{\\beta}) &= 1 - p(y_i=1|x_i,\\boldsymbol{\\beta}),\n", - "\\end{align*}\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "2b49dd7b", - "metadata": { - "editable": true - }, - "source": [ - "where $\\boldsymbol{\\beta}$ are the weights we wish to extract from data, in our case $\\beta_0$ and $\\beta_1$." - ] - }, - { - "cell_type": "markdown", - "id": "09c1f325", - "metadata": { - "editable": true - }, - "source": [ - "## The equations to solve\n", - "\n", - "Our compact equations used a definition of a vector $\\boldsymbol{y}$ with $n$\n", - "elements $y_i$, an $n\\times p$ matrix $\\boldsymbol{X}$ which contains the\n", - "$x_i$ values and a vector $\\boldsymbol{p}$ of fitted probabilities\n", - "$p(y_i\\vert x_i,\\boldsymbol{\\beta})$. We rewrote in a more compact form\n", - "the first derivative of the cost function as" - ] - }, - { - "cell_type": "markdown", - "id": "f904df15", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\frac{\\partial \\mathcal{C}(\\boldsymbol{\\beta})}{\\partial \\boldsymbol{\\beta}} = -\\boldsymbol{X}^T\\left(\\boldsymbol{y}-\\boldsymbol{p}\\right).\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "70c2631f", - "metadata": { - "editable": true - }, - "source": [ - "If we in addition define a diagonal matrix $\\boldsymbol{W}$ with elements \n", - "$p(y_i\\vert x_i,\\boldsymbol{\\beta})(1-p(y_i\\vert x_i,\\boldsymbol{\\beta})$, we can obtain a compact expression of the second derivative as" - ] - }, - { - "cell_type": "markdown", - "id": "8ace89bc", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\frac{\\partial^2 \\mathcal{C}(\\boldsymbol{\\beta})}{\\partial \\boldsymbol{\\beta}\\partial \\boldsymbol{\\beta}^T} = \\boldsymbol{X}^T\\boldsymbol{W}\\boldsymbol{X}.\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "77a52027", - "metadata": { - "editable": true - }, - "source": [ - "This defines what is called the Hessian matrix." - ] - }, - { - "cell_type": "markdown", - "id": "f64c6f88", - "metadata": { - "editable": true - }, - "source": [ - "## Solving using Newton-Raphson's method\n", - "\n", - "If we can set up these equations, Newton-Raphson's iterative method is normally the method of choice. It requires however that we can compute in an efficient way the matrices that define the first and second derivatives. \n", - "\n", - "Our iterative scheme is then given by" - ] - }, - { - "cell_type": "markdown", - "id": "e9f0ab43", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\boldsymbol{\\beta}^{\\mathrm{new}} = \\boldsymbol{\\beta}^{\\mathrm{old}}-\\left(\\frac{\\partial^2 \\mathcal{C}(\\boldsymbol{\\beta})}{\\partial \\boldsymbol{\\beta}\\partial \\boldsymbol{\\beta}^T}\\right)^{-1}_{\\boldsymbol{\\beta}^{\\mathrm{old}}}\\times \\left(\\frac{\\partial \\mathcal{C}(\\boldsymbol{\\beta})}{\\partial \\boldsymbol{\\beta}}\\right)_{\\boldsymbol{\\beta}^{\\mathrm{old}}},\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "41bd2b63", - "metadata": { - "editable": true - }, - "source": [ - "or in matrix form as" - ] - }, - { - "cell_type": "markdown", - "id": "fdbab7f4", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\boldsymbol{\\beta}^{\\mathrm{new}} = \\boldsymbol{\\beta}^{\\mathrm{old}}-\\left(\\boldsymbol{X}^T\\boldsymbol{W}\\boldsymbol{X} \\right)^{-1}\\times \\left(-\\boldsymbol{X}^T(\\boldsymbol{y}-\\boldsymbol{p}) \\right)_{\\boldsymbol{\\beta}^{\\mathrm{old}}}.\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "33f46848", - "metadata": { - "editable": true - }, - "source": [ - "The right-hand side is computed with the old values of $\\beta$. \n", - "\n", - "If we can compute these matrices, in particular the Hessian, the above is often the easiest method to implement." - ] - }, - { - "cell_type": "markdown", - "id": "ad2a7d0a", - "metadata": { - "editable": true - }, - "source": [ - "## Brief reminder on Newton-Raphson's method\n", - "\n", - "Let us quickly remind ourselves how we derive the above method.\n", - "\n", - "Perhaps the most celebrated of all one-dimensional root-finding\n", - "routines is Newton's method, also called the Newton-Raphson\n", - "method. This method requires the evaluation of both the\n", - "function $f$ and its derivative $f'$ at arbitrary points. \n", - "If you can only calculate the derivative\n", - "numerically and/or your function is not of the smooth type, we\n", - "normally discourage the use of this method." - ] - }, - { - "cell_type": "markdown", - "id": "f716cd22", - "metadata": { - "editable": true - }, - "source": [ - "## The equations\n", - "\n", - "The Newton-Raphson formula consists geometrically of extending the\n", - "tangent line at a current point until it crosses zero, then setting\n", - "the next guess to the abscissa of that zero-crossing. The mathematics\n", - "behind this method is rather simple. Employing a Taylor expansion for\n", - "$x$ sufficiently close to the solution $s$, we have" - ] - }, - { - "cell_type": "markdown", - "id": "e5210d0f", - "metadata": { - "editable": true - }, - "source": [ - "\n", - "
            \n", - "\n", - "$$\n", - "f(s)=0=f(x)+(s-x)f'(x)+\\frac{(s-x)^2}{2}f''(x) +\\dots.\n", - " \\label{eq:taylornr} \\tag{2}\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "4f3aa8ed", - "metadata": { - "editable": true - }, - "source": [ - "For small enough values of the function and for well-behaved\n", - "functions, the terms beyond linear are unimportant, hence we obtain" - ] - }, - { - "cell_type": "markdown", - "id": "da20d4ff", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "f(x)+(s-x)f'(x)\\approx 0,\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "ffe32653", - "metadata": { - "editable": true - }, - "source": [ - "yielding" - ] - }, - { - "cell_type": "markdown", - "id": "63c73cd9", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "s\\approx x-\\frac{f(x)}{f'(x)}.\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "f2c4c201", - "metadata": { - "editable": true - }, - "source": [ - "Having in mind an iterative procedure, it is natural to start iterating with" - ] - }, - { - "cell_type": "markdown", - "id": "a43b3360", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "x_{n+1}=x_n-\\frac{f(x_n)}{f'(x_n)}.\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "1de9199a", - "metadata": { - "editable": true - }, - "source": [ - "## Simple geometric interpretation\n", - "\n", - "The above is Newton-Raphson's method. It has a simple geometric\n", - "interpretation, namely $x_{n+1}$ is the point where the tangent from\n", - "$(x_n,f(x_n))$ crosses the $x$-axis. Close to the solution,\n", - "Newton-Raphson converges fast to the desired result. However, if we\n", - "are far from a root, where the higher-order terms in the series are\n", - "important, the Newton-Raphson formula can give grossly inaccurate\n", - "results. For instance, the initial guess for the root might be so far\n", - "from the true root as to let the search interval include a local\n", - "maximum or minimum of the function. If an iteration places a trial\n", - "guess near such a local extremum, so that the first derivative nearly\n", - "vanishes, then Newton-Raphson may fail totally" - ] - }, - { - "cell_type": "markdown", - "id": "be74493c", - "metadata": { - "editable": true - }, - "source": [ - "## Extending to more than one variable\n", - "\n", - "Newton's method can be generalized to systems of several non-linear equations\n", - "and variables. Consider the case with two equations" - ] - }, - { - "cell_type": "markdown", - "id": "c9403406", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\begin{array}{cc} f_1(x_1,x_2) &=0\\\\\n", - " f_2(x_1,x_2) &=0,\\end{array}\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "de21ed85", - "metadata": { - "editable": true - }, - "source": [ - "which we Taylor expand to obtain" - ] - }, - { - "cell_type": "markdown", - "id": "981503d6", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\begin{array}{cc} 0=f_1(x_1+h_1,x_2+h_2)=&f_1(x_1,x_2)+h_1\n", - " \\partial f_1/\\partial x_1+h_2\n", - " \\partial f_1/\\partial x_2+\\dots\\\\\n", - " 0=f_2(x_1+h_1,x_2+h_2)=&f_2(x_1,x_2)+h_1\n", - " \\partial f_2/\\partial x_1+h_2\n", - " \\partial f_2/\\partial x_2+\\dots\n", - " \\end{array}.\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "0748abb5", - "metadata": { - "editable": true - }, - "source": [ - "Defining the Jacobian matrix ${\\bf \\boldsymbol{J}}$ we have" - ] - }, - { - "cell_type": "markdown", - "id": "d868ebd3", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "{\\bf \\boldsymbol{J}}=\\left( \\begin{array}{cc}\n", - " \\partial f_1/\\partial x_1 & \\partial f_1/\\partial x_2 \\\\\n", - " \\partial f_2/\\partial x_1 &\\partial f_2/\\partial x_2\n", - " \\end{array} \\right),\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "1abc622c", - "metadata": { - "editable": true - }, - "source": [ - "we can rephrase Newton's method as" - ] - }, - { - "cell_type": "markdown", - "id": "0ade8cd3", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\left(\\begin{array}{c} x_1^{n+1} \\\\ x_2^{n+1} \\end{array} \\right)=\n", - "\\left(\\begin{array}{c} x_1^{n} \\\\ x_2^{n} \\end{array} \\right)+\n", - "\\left(\\begin{array}{c} h_1^{n} \\\\ h_2^{n} \\end{array} \\right),\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "ccf6c19a", - "metadata": { - "editable": true - }, - "source": [ - "where we have defined" - ] - }, - { - "cell_type": "markdown", - "id": "04910c1c", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\left(\\begin{array}{c} h_1^{n} \\\\ h_2^{n} \\end{array} \\right)=\n", - " -{\\bf \\boldsymbol{J}}^{-1}\n", - " \\left(\\begin{array}{c} f_1(x_1^{n},x_2^{n}) \\\\ f_2(x_1^{n},x_2^{n}) \\end{array} \\right).\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "9ccc0fdd", - "metadata": { - "editable": true - }, - "source": [ - "We need thus to compute the inverse of the Jacobian matrix and it\n", - "is to understand that difficulties may\n", - "arise in case ${\\bf \\boldsymbol{J}}$ is nearly singular.\n", - "\n", - "It is rather straightforward to extend the above scheme to systems of\n", - "more than two non-linear equations. In our case, the Jacobian matrix is given by the Hessian that represents the second derivative of cost function." - ] - }, - { - "cell_type": "markdown", - "id": "7a2a5b20", - "metadata": { - "editable": true - }, - "source": [ - "## Steepest descent\n", - "\n", - "The basic idea of gradient descent is\n", - "that a function $F(\\mathbf{x})$, \n", - "$\\mathbf{x} \\equiv (x_1,\\cdots,x_n)$, decreases fastest if one goes from $\\bf {x}$ in the\n", - "direction of the negative gradient $-\\nabla F(\\mathbf{x})$.\n", - "\n", - "It can be shown that if" - ] - }, - { - "cell_type": "markdown", - "id": "3db0a74b", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\mathbf{x}_{k+1} = \\mathbf{x}_k - \\gamma_k \\nabla F(\\mathbf{x}_k),\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "af2b0746", - "metadata": { - "editable": true - }, - "source": [ - "with $\\gamma_k > 0$.\n", - "\n", - "For $\\gamma_k$ small enough, then $F(\\mathbf{x}_{k+1}) \\leq\n", - "F(\\mathbf{x}_k)$. This means that for a sufficiently small $\\gamma_k$\n", - "we are always moving towards smaller function values, i.e a minimum." - ] - }, - { - "cell_type": "markdown", - "id": "8bdb9863", - "metadata": { - "editable": true - }, - "source": [ - "## More on Steepest descent\n", - "\n", - "The previous observation is the basis of the method of steepest\n", - "descent, which is also referred to as just gradient descent (GD). One\n", - "starts with an initial guess $\\mathbf{x}_0$ for a minimum of $F$ and\n", - "computes new approximations according to" - ] - }, - { - "cell_type": "markdown", - "id": "80223e09", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\mathbf{x}_{k+1} = \\mathbf{x}_k - \\gamma_k \\nabla F(\\mathbf{x}_k), \\ \\ k \\geq 0.\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "ae46c661", - "metadata": { - "editable": true - }, - "source": [ - "The parameter $\\gamma_k$ is often referred to as the step length or\n", - "the learning rate within the context of Machine Learning." - ] - }, - { - "cell_type": "markdown", - "id": "dc9af2ce", - "metadata": { - "editable": true - }, - "source": [ - "## The ideal\n", - "\n", - "Ideally the sequence $\\{\\mathbf{x}_k \\}_{k=0}$ converges to a global\n", - "minimum of the function $F$. In general we do not know if we are in a\n", - "global or local minimum. In the special case when $F$ is a convex\n", - "function, all local minima are also global minima, so in this case\n", - "gradient descent can converge to the global solution. The advantage of\n", - "this scheme is that it is conceptually simple and straightforward to\n", - "implement. However the method in this form has some severe\n", - "limitations:\n", - "\n", - "In machine learing we are often faced with non-convex high dimensional\n", - "cost functions with many local minima. Since GD is deterministic we\n", - "will get stuck in a local minimum, if the method converges, unless we\n", - "have a very good intial guess. This also implies that the scheme is\n", - "sensitive to the chosen initial condition.\n", - "\n", - "Note that the gradient is a function of $\\mathbf{x} =\n", - "(x_1,\\cdots,x_n)$ which makes it expensive to compute numerically." - ] - }, - { - "cell_type": "markdown", - "id": "873ccd0e", - "metadata": { - "editable": true - }, - "source": [ - "## The sensitiveness of the gradient descent\n", - "\n", - "The gradient descent method \n", - "is sensitive to the choice of learning rate $\\gamma_k$. This is due\n", - "to the fact that we are only guaranteed that $F(\\mathbf{x}_{k+1}) \\leq\n", - "F(\\mathbf{x}_k)$ for sufficiently small $\\gamma_k$. The problem is to\n", - "determine an optimal learning rate. If the learning rate is chosen too\n", - "small the method will take a long time to converge and if it is too\n", - "large we can experience erratic behavior.\n", - "\n", - "Many of these shortcomings can be alleviated by introducing\n", - "randomness. One such method is that of Stochastic Gradient Descent\n", - "(SGD), to be discussed next week." - ] - }, - { - "cell_type": "markdown", - "id": "0ecc1560", - "metadata": { - "editable": true - }, - "source": [ - "## Convex functions\n", - "\n", - "Ideally we want our cost/loss function to be convex(concave).\n", - "\n", - "First we give the definition of a convex set: A set $C$ in\n", - "$\\mathbb{R}^n$ is said to be convex if, for all $x$ and $y$ in $C$ and\n", - "all $t \\in (0,1)$ , the point $(1 − t)x + ty$ also belongs to\n", - "C. Geometrically this means that every point on the line segment\n", - "connecting $x$ and $y$ is in $C$ as discussed below.\n", - "\n", - "The convex subsets of $\\mathbb{R}$ are the intervals of\n", - "$\\mathbb{R}$. Examples of convex sets of $\\mathbb{R}^2$ are the\n", - "regular polygons (triangles, rectangles, pentagons, etc...)." - ] - }, - { - "cell_type": "markdown", - "id": "999efd0b", - "metadata": { - "editable": true - }, - "source": [ - "## Convex function\n", - "\n", - "**Convex function**: Let $X \\subset \\mathbb{R}^n$ be a convex set. Assume that the function $f: X \\rightarrow \\mathbb{R}$ is continuous, then $f$ is said to be convex if $$f(tx_1 + (1-t)x_2) \\leq tf(x_1) + (1-t)f(x_2) $$ for all $x_1, x_2 \\in X$ and for all $t \\in [0,1]$. If $\\leq$ is replaced with a strict inequaltiy in the definition, we demand $x_1 \\neq x_2$ and $t\\in(0,1)$ then $f$ is said to be strictly convex. For a single variable function, convexity means that if you draw a straight line connecting $f(x_1)$ and $f(x_2)$, the value of the function on the interval $[x_1,x_2]$ is always below the line as illustrated below." - ] - }, - { - "cell_type": "markdown", - "id": "dbe04f76", - "metadata": { - "editable": true - }, - "source": [ - "## Conditions on convex functions\n", - "\n", - "In the following we state first and second-order conditions which\n", - "ensures convexity of a function $f$. We write $D_f$ to denote the\n", - "domain of $f$, i.e the subset of $R^n$ where $f$ is defined. For more\n", - "details and proofs we refer to: [S. Boyd and L. Vandenberghe. Convex Optimization. Cambridge University Press](http://stanford.edu/boyd/cvxbook/, 2004).\n", - "\n", - "**First order condition.**\n", - "\n", - "Suppose $f$ is differentiable (i.e $\\nabla f(x)$ is well defined for\n", - "all $x$ in the domain of $f$). Then $f$ is convex if and only if $D_f$\n", - "is a convex set and $$f(y) \\geq f(x) + \\nabla f(x)^T (y-x) $$ holds\n", - "for all $x,y \\in D_f$. This condition means that for a convex function\n", - "the first order Taylor expansion (right hand side above) at any point\n", - "a global under estimator of the function. To convince yourself you can\n", - "make a drawing of $f(x) = x^2+1$ and draw the tangent line to $f(x)$ and\n", - "note that it is always below the graph.\n", - "\n", - "**Second order condition.**\n", - "\n", - "Assume that $f$ is twice\n", - "differentiable, i.e the Hessian matrix exists at each point in\n", - "$D_f$. Then $f$ is convex if and only if $D_f$ is a convex set and its\n", - "Hessian is positive semi-definite for all $x\\in D_f$. For a\n", - "single-variable function this reduces to $f''(x) \\geq 0$. Geometrically this means that $f$ has nonnegative curvature\n", - "everywhere.\n", - "\n", - "This condition is particularly useful since it gives us an procedure for determining if the function under consideration is convex, apart from using the definition." - ] - }, - { - "cell_type": "markdown", - "id": "cd2756b4", - "metadata": { - "editable": true - }, - "source": [ - "## More on convex functions\n", - "\n", - "The next result is of great importance to us and the reason why we are\n", - "going on about convex functions. In machine learning we frequently\n", - "have to minimize a loss/cost function in order to find the best\n", - "parameters for the model we are considering. \n", - "\n", - "Ideally we want the\n", - "global minimum (for high-dimensional models it is hard to know\n", - "if we have local or global minimum). However, if the cost/loss function\n", - "is convex the following result provides invaluable information:\n", - "\n", - "**Any minimum is global for convex functions.**\n", - "\n", - "Consider the problem of finding $x \\in \\mathbb{R}^n$ such that $f(x)$\n", - "is minimal, where $f$ is convex and differentiable. Then, any point\n", - "$x^*$ that satisfies $\\nabla f(x^*) = 0$ is a global minimum.\n", - "\n", - "This result means that if we know that the cost/loss function is convex and we are able to find a minimum, we are guaranteed that it is a global minimum." - ] - }, - { - "cell_type": "markdown", - "id": "45ce738a", - "metadata": { - "editable": true - }, - "source": [ - "## Some simple problems\n", - "\n", - "1. Show that $f(x)=x^2$ is convex for $x \\in \\mathbb{R}$ using the definition of convexity. Hint: If you re-write the definition, $f$ is convex if the following holds for all $x,y \\in D_f$ and any $\\lambda \\in [0,1]$ $\\lambda f(x)+(1-\\lambda)f(y)-f(\\lambda x + (1-\\lambda) y ) \\geq 0$.\n", - "\n", - "2. Using the second order condition show that the following functions are convex on the specified domain.\n", - "\n", - " * $f(x) = e^x$ is convex for $x \\in \\mathbb{R}$.\n", - "\n", - " * $g(x) = -\\ln(x)$ is convex for $x \\in (0,\\infty)$.\n", - "\n", - "3. Let $f(x) = x^2$ and $g(x) = e^x$. Show that $f(g(x))$ and $g(f(x))$ is convex for $x \\in \\mathbb{R}$. Also show that if $f(x)$ is any convex function than $h(x) = e^{f(x)}$ is convex.\n", - "\n", - "4. A norm is any function that satisfy the following properties\n", - "\n", - " * $f(\\alpha x) = |\\alpha| f(x)$ for all $\\alpha \\in \\mathbb{R}$.\n", - "\n", - " * $f(x+y) \\leq f(x) + f(y)$\n", - "\n", - " * $f(x) \\leq 0$ for all $x \\in \\mathbb{R}^n$ with equality if and only if $x = 0$\n", - "\n", - "Using the definition of convexity, try to show that a function satisfying the properties above is convex (the third condition is not needed to show this)." - ] - }, - { - "cell_type": "markdown", - "id": "dfb5ef6d", - "metadata": { - "editable": true - }, - "source": [ - "## Revisiting our first homework\n", - "\n", - "We will use linear regression as a case study for the gradient descent\n", - "methods. Linear regression is a great test case for the gradient\n", - "descent methods discussed in the lectures since it has several\n", - "desirable properties such as:\n", - "\n", - "1. An analytical solution (recall homework set 1).\n", - "\n", - "2. The gradient can be computed analytically.\n", - "\n", - "3. The cost function is convex which guarantees that gradient descent converges for small enough learning rates\n", - "\n", - "We revisit an example similar to what we had in the first homework set. We had a function of the type" - ] - }, - { - "cell_type": "code", - "execution_count": 13, - "id": "9dfbaf09", - "metadata": { - "collapsed": false, - "editable": true - }, - "outputs": [], - "source": [ - "x = 2*np.random.rand(m,1)\n", - "y = 4+3*x+np.random.randn(m,1)" - ] - }, - { - "cell_type": "markdown", - "id": "35a79da9", - "metadata": { - "editable": true - }, - "source": [ - "with $x_i \\in [0,1] $ is chosen randomly using a uniform distribution. Additionally we have a stochastic noise chosen according to a normal distribution $\\cal {N}(0,1)$. \n", - "The linear regression model is given by" - ] - }, - { - "cell_type": "markdown", - "id": "06967409", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "h_\\beta(x) = \\boldsymbol{y} = \\beta_0 + \\beta_1 x,\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "ddbd2526", - "metadata": { - "editable": true - }, - "source": [ - "such that" - ] - }, - { - "cell_type": "markdown", - "id": "1733a15b", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\boldsymbol{y}_i = \\beta_0 + \\beta_1 x_i.\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "bee0148d", - "metadata": { - "editable": true - }, - "source": [ - "## Gradient descent example\n", - "\n", - "Let $\\mathbf{y} = (y_1,\\cdots,y_n)^T$, $\\mathbf{\\boldsymbol{y}} = (\\boldsymbol{y}_1,\\cdots,\\boldsymbol{y}_n)^T$ and $\\beta = (\\beta_0, \\beta_1)^T$\n", - "\n", - "It is convenient to write $\\mathbf{\\boldsymbol{y}} = X\\beta$ where $X \\in \\mathbb{R}^{100 \\times 2} $ is the design matrix given by (we keep the intercept here)" - ] - }, - { - "cell_type": "markdown", - "id": "f8bf3934", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "X \\equiv \\begin{bmatrix}\n", - "1 & x_1 \\\\\n", - "\\vdots & \\vdots \\\\\n", - "1 & x_{100} & \\\\\n", - "\\end{bmatrix}.\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "a14b7c83", - "metadata": { - "editable": true - }, - "source": [ - "The cost/loss/risk function is given by (" - ] - }, - { - "cell_type": "markdown", - "id": "255750cb", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "C(\\beta) = \\frac{1}{n}||X\\beta-\\mathbf{y}||_{2}^{2} = \\frac{1}{n}\\sum_{i=1}^{100}\\left[ (\\beta_0 + \\beta_1 x_i)^2 - 2 y_i (\\beta_0 + \\beta_1 x_i) + y_i^2\\right]\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "c67da34c", - "metadata": { - "editable": true - }, - "source": [ - "and we want to find $\\beta$ such that $C(\\beta)$ is minimized." - ] - }, - { - "cell_type": "markdown", - "id": "f5f3d460", - "metadata": { - "editable": true - }, - "source": [ - "## The derivative of the cost/loss function\n", - "\n", - "Computing $\\partial C(\\beta) / \\partial \\beta_0$ and $\\partial C(\\beta) / \\partial \\beta_1$ we can show that the gradient can be written as" - ] - }, - { - "cell_type": "markdown", - "id": "3d1ede19", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\nabla_{\\beta} C(\\beta) = \\frac{2}{n}\\begin{bmatrix} \\sum_{i=1}^{100} \\left(\\beta_0+\\beta_1x_i-y_i\\right) \\\\\n", - "\\sum_{i=1}^{100}\\left( x_i (\\beta_0+\\beta_1x_i)-y_ix_i\\right) \\\\\n", - "\\end{bmatrix} = \\frac{2}{n}X^T(X\\beta - \\mathbf{y}),\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "0685b6b7", - "metadata": { - "editable": true - }, - "source": [ - "where $X$ is the design matrix defined above." - ] - }, - { - "cell_type": "markdown", - "id": "e3c9410a", - "metadata": { - "editable": true - }, - "source": [ - "## The Hessian matrix\n", - "The Hessian matrix of $C(\\beta)$ is given by" - ] - }, - { - "cell_type": "markdown", - "id": "fcde03f8", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\boldsymbol{H} \\equiv \\begin{bmatrix}\n", - "\\frac{\\partial^2 C(\\beta)}{\\partial \\beta_0^2} & \\frac{\\partial^2 C(\\beta)}{\\partial \\beta_0 \\partial \\beta_1} \\\\\n", - "\\frac{\\partial^2 C(\\beta)}{\\partial \\beta_0 \\partial \\beta_1} & \\frac{\\partial^2 C(\\beta)}{\\partial \\beta_1^2} & \\\\\n", - "\\end{bmatrix} = \\frac{2}{n}X^T X.\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "bd7c7145", - "metadata": { - "editable": true - }, - "source": [ - "This result implies that $C(\\beta)$ is a convex function since the matrix $X^T X$ always is positive semi-definite." - ] - }, - { - "cell_type": "markdown", - "id": "273862c8", - "metadata": { - "editable": true - }, - "source": [ - "## Simple program\n", - "\n", - "We can now write a program that minimizes $C(\\beta)$ using the gradient descent method with a constant learning rate $\\gamma$ according to" - ] - }, - { - "cell_type": "markdown", - "id": "ff091180", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\beta_{k+1} = \\beta_k - \\gamma \\nabla_\\beta C(\\beta_k), \\ k=0,1,\\cdots\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "510ded4a", - "metadata": { - "editable": true - }, - "source": [ - "We can use the expression we computed for the gradient and let use a\n", - "$\\beta_0$ be chosen randomly and let $\\gamma = 0.001$. Stop iterating\n", - "when $||\\nabla_\\beta C(\\beta_k) || \\leq \\epsilon = 10^{-8}$. **Note that the code below does not include the latter stop criterion**.\n", - "\n", - "And finally we can compare our solution for $\\beta$ with the analytic result given by \n", - "$\\beta= (X^TX)^{-1} X^T \\mathbf{y}$." - ] - }, - { - "cell_type": "markdown", - "id": "b465897d", - "metadata": { - "editable": true - }, - "source": [ - "## Gradient Descent Example\n", - "\n", - "Here our simple example" - ] - }, - { - "cell_type": "code", - "execution_count": 14, - "id": "55de11e8", - "metadata": { - "collapsed": false, - "editable": true - }, - "outputs": [], - "source": [ - "\n", - "# Importing various packages\n", - "from random import random, seed\n", - "import numpy as np\n", - "import matplotlib.pyplot as plt\n", - "from mpl_toolkits.mplot3d import Axes3D\n", - "from matplotlib import cm\n", - "from matplotlib.ticker import LinearLocator, FormatStrFormatter\n", - "import sys\n", - "\n", - "# the number of datapoints\n", - "n = 100\n", - "x = 2*np.random.rand(n,1)\n", - "y = 4+3*x+np.random.randn(n,1)\n", - "\n", - "X = np.c_[np.ones((n,1)), x]\n", - "# Hessian matrix\n", - "H = (2.0/n)* X.T @ X\n", - "# Get the eigenvalues\n", - "EigValues, EigVectors = np.linalg.eig(H)\n", - "print(f\"Eigenvalues of Hessian Matrix:{EigValues}\")\n", - "\n", - "beta_linreg = np.linalg.inv(X.T @ X) @ X.T @ y\n", - "print(beta_linreg)\n", - "beta = np.random.randn(2,1)\n", - "\n", - "eta = 1.0/np.max(EigValues)\n", - "Niterations = 1000\n", - "\n", - "for iter in range(Niterations):\n", - " gradient = (2.0/n)*X.T @ (X @ beta-y)\n", - " beta -= eta*gradient\n", - "\n", - "print(beta)\n", - "xnew = np.array([[0],[2]])\n", - "xbnew = np.c_[np.ones((2,1)), xnew]\n", - "ypredict = xbnew.dot(beta)\n", - "ypredict2 = xbnew.dot(beta_linreg)\n", - "plt.plot(xnew, ypredict, \"r-\")\n", - "plt.plot(xnew, ypredict2, \"b-\")\n", - "plt.plot(x, y ,'ro')\n", - "plt.axis([0,2.0,0, 15.0])\n", - "plt.xlabel(r'$x$')\n", - "plt.ylabel(r'$y$')\n", - "plt.title(r'Gradient descent example')\n", - "plt.show()" - ] - }, - { - "cell_type": "markdown", - "id": "fa823fff", - "metadata": { - "editable": true - }, - "source": [ - "## And a corresponding example using **scikit-learn**" - ] - }, - { - "cell_type": "code", - "execution_count": 15, - "id": "cd62db07", - "metadata": { - "collapsed": false, - "editable": true - }, - "outputs": [], - "source": [ - "# Importing various packages\n", - "from random import random, seed\n", - "import numpy as np\n", - "import matplotlib.pyplot as plt\n", - "from sklearn.linear_model import SGDRegressor\n", - "\n", - "n = 100\n", - "x = 2*np.random.rand(n,1)\n", - "y = 4+3*x+np.random.randn(n,1)\n", - "\n", - "X = np.c_[np.ones((n,1)), x]\n", - "beta_linreg = np.linalg.inv(X.T @ X) @ (X.T @ y)\n", - "print(beta_linreg)\n", - "sgdreg = SGDRegressor(max_iter = 50, penalty=None, eta0=0.1)\n", - "sgdreg.fit(x,y.ravel())\n", - "print(sgdreg.intercept_, sgdreg.coef_)" - ] - }, - { - "cell_type": "markdown", - "id": "748b7943", - "metadata": { - "editable": true - }, - "source": [ - "## Gradient descent and Ridge\n", - "\n", - "We have also discussed Ridge regression where the loss function contains a regularized term given by the $L_2$ norm of $\\beta$," - ] - }, - { - "cell_type": "markdown", - "id": "0d71b09d", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "C_{\\text{ridge}}(\\beta) = \\frac{1}{n}||X\\beta -\\mathbf{y}||^2 + \\lambda ||\\beta||^2, \\ \\lambda \\geq 0.\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "42d5847c", - "metadata": { - "editable": true - }, - "source": [ - "In order to minimize $C_{\\text{ridge}}(\\beta)$ using GD we adjust the gradient as follows" - ] - }, - { - "cell_type": "markdown", - "id": "ea239249", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\nabla_\\beta C_{\\text{ridge}}(\\beta) = \\frac{2}{n}\\begin{bmatrix} \\sum_{i=1}^{100} \\left(\\beta_0+\\beta_1x_i-y_i\\right) \\\\\n", - "\\sum_{i=1}^{100}\\left( x_i (\\beta_0+\\beta_1x_i)-y_ix_i\\right) \\\\\n", - "\\end{bmatrix} + 2\\lambda\\begin{bmatrix} \\beta_0 \\\\ \\beta_1\\end{bmatrix} = 2 (\\frac{1}{n}X^T(X\\beta - \\mathbf{y})+\\lambda \\beta).\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "fbc34af7", - "metadata": { - "editable": true - }, - "source": [ - "We can easily extend our program to minimize $C_{\\text{ridge}}(\\beta)$ using gradient descent and compare with the analytical solution given by" - ] - }, - { - "cell_type": "markdown", - "id": "add0ed1b", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\beta_{\\text{ridge}} = \\left(X^T X + n\\lambda I_{2 \\times 2} \\right)^{-1} X^T \\mathbf{y}.\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "dff6e4c5", - "metadata": { - "editable": true - }, - "source": [ - "## The Hessian matrix for Ridge Regression\n", - "The Hessian matrix of Ridge Regression for our simple example is given by" - ] - }, - { - "cell_type": "markdown", - "id": "a8ac4e98", - "metadata": { - "editable": true - }, - "source": [ - "$$\n", - "\\boldsymbol{H} \\equiv \\begin{bmatrix}\n", - "\\frac{\\partial^2 C(\\beta)}{\\partial \\beta_0^2} & \\frac{\\partial^2 C(\\beta)}{\\partial \\beta_0 \\partial \\beta_1} \\\\\n", - "\\frac{\\partial^2 C(\\beta)}{\\partial \\beta_0 \\partial \\beta_1} & \\frac{\\partial^2 C(\\beta)}{\\partial \\beta_1^2} & \\\\\n", - "\\end{bmatrix} = \\frac{2}{n}X^T X+2\\lambda\\boldsymbol{I}.\n", - "$$" - ] - }, - { - "cell_type": "markdown", - "id": "6503aeb0", - "metadata": { - "editable": true - }, - "source": [ - "This implies that the Hessian matrix is positive definite, hence the stationary point is a\n", - "minimum.\n", - "Note that the Ridge cost function is convex being a sum of two convex\n", - "functions. Therefore, the stationary point is a global\n", - "minimum of this function." - ] - }, - { - "cell_type": "markdown", - "id": "68447ce5", - "metadata": { - "editable": true - }, - "source": [ - "## Program example for gradient descent with Ridge Regression" - ] - }, - { - "cell_type": "code", - "execution_count": 16, - "id": "65927570", - "metadata": { - "collapsed": false, - "editable": true - }, - "outputs": [], - "source": [ - "from random import random, seed\n", - "import numpy as np\n", - "import matplotlib.pyplot as plt\n", - "from mpl_toolkits.mplot3d import Axes3D\n", - "from matplotlib import cm\n", - "from matplotlib.ticker import LinearLocator, FormatStrFormatter\n", - "import sys\n", - "\n", - "# the number of datapoints\n", - "n = 100\n", - "x = 2*np.random.rand(n,1)\n", - "y = 4+3*x+np.random.randn(n,1)\n", - "\n", - "X = np.c_[np.ones((n,1)), x]\n", - "XT_X = X.T @ X\n", - "\n", - "#Ridge parameter lambda\n", - "lmbda = 0.001\n", - "Id = n*lmbda* np.eye(XT_X.shape[0])\n", - "\n", - "# Hessian matrix\n", - "H = (2.0/n)* XT_X+2*lmbda* np.eye(XT_X.shape[0])\n", - "# Get the eigenvalues\n", - "EigValues, EigVectors = np.linalg.eig(H)\n", - "print(f\"Eigenvalues of Hessian Matrix:{EigValues}\")\n", - "\n", - "\n", - "beta_linreg = np.linalg.inv(XT_X+Id) @ X.T @ y\n", - "print(beta_linreg)\n", - "# Start plain gradient descent\n", - "beta = np.random.randn(2,1)\n", - "\n", - "eta = 1.0/np.max(EigValues)\n", - "Niterations = 100\n", - "\n", - "for iter in range(Niterations):\n", - " gradients = 2.0/n*X.T @ (X @ (beta)-y)+2*lmbda*beta\n", - " beta -= eta*gradients\n", - "\n", - "print(beta)\n", - "ypredict = X @ beta\n", - "ypredict2 = X @ beta_linreg\n", - "plt.plot(x, ypredict, \"r-\")\n", - "plt.plot(x, ypredict2, \"b-\")\n", - "plt.plot(x, y ,'ro')\n", - "plt.axis([0,2.0,0, 15.0])\n", - "plt.xlabel(r'$x$')\n", - "plt.ylabel(r'$y$')\n", - "plt.title(r'Gradient descent example for Ridge')\n", - "plt.show()" - ] - }, - { - "cell_type": "markdown", - "id": "81a4c315", - "metadata": { - "editable": true - }, - "source": [ - "## Using gradient descent methods, limitations\n", - "\n", - "* **Gradient descent (GD) finds local minima of our function**. Since the GD algorithm is deterministic, if it converges, it will converge to a local minimum of our cost/loss/risk function. Because in ML we are often dealing with extremely rugged landscapes with many local minima, this can lead to poor performance.\n", - "\n", - "* **GD is sensitive to initial conditions**. One consequence of the local nature of GD is that initial conditions matter. Depending on where one starts, one will end up at a different local minima. Therefore, it is very important to think about how one initializes the training process. This is true for GD as well as more complicated variants of GD.\n", - "\n", - "* **Gradients are computationally expensive to calculate for large datasets**. In many cases in statistics and ML, the cost/loss/risk function is a sum of terms, with one term for each data point. For example, in linear regression, $E \\propto \\sum_{i=1}^n (y_i - \\mathbf{w}^T\\cdot\\mathbf{x}_i)^2$; for logistic regression, the square error is replaced by the cross entropy. To calculate the gradient we have to sum over *all* $n$ data points. Doing this at every GD step becomes extremely computationally expensive. An ingenious solution to this, is to calculate the gradients using small subsets of the data called \"mini batches\". This has the added benefit of introducing stochasticity into our algorithm.\n", - "\n", - "* **GD is very sensitive to choices of learning rates**. GD is extremely sensitive to the choice of learning rates. If the learning rate is very small, the training process take an extremely long time. For larger learning rates, GD can diverge and give poor results. Furthermore, depending on what the local landscape looks like, we have to modify the learning rates to ensure convergence. Ideally, we would *adaptively* choose the learning rates to match the landscape.\n", - "\n", - "* **GD treats all directions in parameter space uniformly.** Another major drawback of GD is that unlike Newton's method, the learning rate for GD is the same in all directions in parameter space. For this reason, the maximum learning rate is set by the behavior of the steepest direction and this can significantly slow down training. Ideally, we would like to take large steps in flat directions and small steps in steep directions. Since we are exploring rugged landscapes where curvatures change, this requires us to keep track of not only the gradient but second derivatives. The ideal scenario would be to calculate the Hessian but this proves to be too computationally expensive. \n", - "\n", - "* GD can take exponential time to escape saddle points, even with random initialization. As we mentioned, GD is extremely sensitive to initial condition since it determines the particular local minimum GD would eventually reach. However, even with a good initialization scheme, through the introduction of randomness, GD can still take exponential time to escape saddle points." - ] - }, - { - "cell_type": "markdown", - "id": "5c2b7530", - "metadata": { - "editable": true - }, - "source": [ - "## Challenge yourself the coming weekend\n", - "\n", - "Write a code which implements gradient descent for a logistic regression example." - ] } ], "metadata": {}, diff --git a/doc/src/week38/week38.do.txt b/doc/src/week38/week38.do.txt index 518e5da84..d1427ede4 100644 --- a/doc/src/week38/week38.do.txt +++ b/doc/src/week38/week38.do.txt @@ -1,191 +1,44 @@ TITLE: Week 38: Logistic Regression and Optimization AUTHOR: Morten Hjorth-Jensen {copyright, 1999-present|CC BY-NC} at Department of Physics and Center for Computing in Science Education, University of Oslo & Department of Physics and Astronomy and Facility for Rare Isotope Beams, Michigan State University -DATE: September 18-22 +DATE: September 16-20, 2024 !split -===== Plans for week 38 ===== +===== Plans for week 38, lecture Monday September 16 ===== -!bblock Material for the active learning sessions on Tuesday and Wednesday - * Lecture from last week on the bias-variance tradeoff - * Resampling techniques, cross-validation examples included here, see also the lectures from last week on the bootstrap method - * Exercise for week 38, see also the whiteboard notes from week 37 at URL:"https://github.com/CompPhysics/MachineLearning/blob/master/doc/HandWrittenNotes/2023/NotesSep14.pdf" - * Work on project 1, in particular resampling methods like cross-validation and bootstrap. -!eblock -!bblock Material for the lecture on Thursday September 21 +!bblock Material for the lecture on Monday September 16 * Logistic regression as our first encounter of classification methods. From binary cases to several categories. * Start gradient and optimization methods - * "Video of lecture":"https://youtu.be/cJrSwRsVSGM" - * Whiteboard notes at URL:"https://github.com/CompPhysics/MachineLearning/blob/master/doc/HandWrittenNotes/2023/NotesSep21.pdf" +# * "Video of lecture":"https://youtu.be/cJrSwRsVSGM" +# * Whiteboard notes at URL:"https://github.com/CompPhysics/MachineLearning/blob/master/doc/HandWrittenNotes/2023/NotesSep21.pdf" +!eblock + +!split +===== Suggested reading and videos ===== +!bblock * Readings and Videos: * Hastie et al 4.1, 4.2 and 4.3 on logistic regression + * Raschka et al, pages 53-76 on Logistic regression and pages 37-52 on gradient optimization * For a good discussion on gradient methods, see Goodfellow et al section 4.3-4.5 and chapter 8. We will come back to the latter chapter in our discussion of Neural networks as well. - * See also the whiteboard notes from week 37 at URL:"https://github.com/CompPhysics/MachineLearning/blob/master/doc/HandWrittenNotes/2023/NotesSep14.pdf" for a discussion and derivation of the bias-variance tradeoff. * "Video on Logistic regression":"https://www.youtube.com/watch?v=C5268D9t9Ak" * "Yet another video on logistic regression":"https://www.youtube.com/watch?v=yIYKR4sgzI8" * "Video on gradient descent":"https://www.youtube.com/watch?v=sDv4f4s2SB8" !eblock -!split -===== Material from last week and relevant for the first project ===== - - -!split -===== Various steps in cross-validation ===== - -When the repetitive splitting of the data set is done randomly, -samples may accidently end up in a fast majority of the splits in -either training or test set. Such samples may have an unbalanced -influence on either model building or prediction evaluation. To avoid -this $k$-fold cross-validation structures the data splitting. The -samples are divided into $k$ more or less equally sized exhaustive and -mutually exclusive subsets. In turn (at each split) one of these -subsets plays the role of the test set while the union of the -remaining subsets constitutes the training set. Such a splitting -warrants a balanced representation of each sample in both training and -test set over the splits. Still the division into the $k$ subsets -involves a degree of randomness. This may be fully excluded when -choosing $k=n$. This particular case is referred to as leave-one-out -cross-validation (LOOCV). - -!split -===== How to set up the cross-validation for Ridge and/or Lasso ===== - -* Define a range of interest for the penalty parameter. - -* Divide the data set into training and test set comprising samples $\{1, \ldots, n\} \setminus i$ and $\{ i \}$, respectively. - -* Fit the linear regression model by means of for example Ridge or Lasso regression for each $\lambda$ in the grid using the training set, and the corresponding estimate of the error variance $\bm{\sigma}_{-i}^2(\lambda)$, as -!bt -\begin{align*} -\bm{\beta}_{-i}(\lambda) & = ( \bm{X}_{-i, \ast}^{T} -\bm{X}_{-i, \ast} + \lambda \bm{I}_{pp})^{-1} -\bm{X}_{-i, \ast}^{T} \bm{y}_{-i} -\end{align*} -!et - -* Evaluate the prediction performance of these models on the test set by $C[y_i, \bm{X}_{i, \ast}; \bm{\beta}_{-i}(\lambda), \bm{\sigma}_{-i}^2(\lambda)]$. Or, by the prediction error $|y_i - \bm{X}_{i, \ast} \bm{\beta}_{-i}(\lambda)|$, the relative error, the error squared or the R2 score function. - -* Repeat the first three steps such that each sample plays the role of the test set once. - -* Average the prediction performances of the test sets at each grid point of the penalty bias/parameter. It is an estimate of the prediction performance of the model corresponding to this value of the penalty parameter on novel data. - !split -===== Cross-validation in brief ===== +===== Plans for the lab sessions ===== -For the various values of $k$ +!bblock Material for the active learning sessions on Tuesday and Wednesday + * Repetition from last week on the bias-variance tradeoff + * Resampling techniques, cross-validation examples included here, see also the lectures from last week on the bootstrap method + * Exercise for week 38 on the bias-variance tradeoff, see also the video from the lab session from week 37 at URL:"https://youtu.be/omLmp_kkie0" + * Work on project 1, in particular resampling methods like cross-validation and bootstrap. +!eblock -o shuffle the dataset randomly. -o Split the dataset into $k$ groups. -o For each unique group: - o Decide which group to use as set for test data - o Take the remaining groups as a training data set - o Fit a model on the training set and evaluate it on the test set - o Retain the evaluation score and discard the model -o Summarize the model using the sample of model evaluation scores - - - -!split -===== Code Example for Cross-validation and $k$-fold Cross-validation ===== - -The code here uses Ridge regression with cross-validation (CV) resampling and $k$-fold CV in order to fit a specific polynomial. -!bc pycod -import numpy as np -import matplotlib.pyplot as plt -from sklearn.model_selection import KFold -from sklearn.linear_model import Ridge -from sklearn.model_selection import cross_val_score -from sklearn.preprocessing import PolynomialFeatures - -# A seed just to ensure that the random numbers are the same for every run. -# Useful for eventual debugging. -np.random.seed(3155) - -# Generate the data. -nsamples = 100 -x = np.random.randn(nsamples) -y = 3*x**2 + np.random.randn(nsamples) - -## Cross-validation on Ridge regression using KFold only - -# Decide degree on polynomial to fit -poly = PolynomialFeatures(degree = 6) - -# Decide which values of lambda to use -nlambdas = 500 -lambdas = np.logspace(-3, 5, nlambdas) - -# Initialize a KFold instance -k = 5 -kfold = KFold(n_splits = k) - -# Perform the cross-validation to estimate MSE -scores_KFold = np.zeros((nlambdas, k)) - -i = 0 -for lmb in lambdas: - ridge = Ridge(alpha = lmb) - j = 0 - for train_inds, test_inds in kfold.split(x): - xtrain = x[train_inds] - ytrain = y[train_inds] - - xtest = x[test_inds] - ytest = y[test_inds] - - Xtrain = poly.fit_transform(xtrain[:, np.newaxis]) - ridge.fit(Xtrain, ytrain[:, np.newaxis]) - - Xtest = poly.fit_transform(xtest[:, np.newaxis]) - ypred = ridge.predict(Xtest) - - scores_KFold[i,j] = np.sum((ypred - ytest[:, np.newaxis])**2)/np.size(ypred) - - j += 1 - i += 1 - - -estimated_mse_KFold = np.mean(scores_KFold, axis = 1) - -## Cross-validation using cross_val_score from sklearn along with KFold - -# kfold is an instance initialized above as: -# kfold = KFold(n_splits = k) - -estimated_mse_sklearn = np.zeros(nlambdas) -i = 0 -for lmb in lambdas: - ridge = Ridge(alpha = lmb) - - X = poly.fit_transform(x[:, np.newaxis]) - estimated_mse_folds = cross_val_score(ridge, X, y[:, np.newaxis], scoring='neg_mean_squared_error', cv=kfold) - - # cross_val_score return an array containing the estimated negative mse for every fold. - # we have to the the mean of every array in order to get an estimate of the mse of the model - estimated_mse_sklearn[i] = np.mean(-estimated_mse_folds) - - i += 1 - -## Plot and compare the slightly different ways to perform cross-validation - -plt.figure() - -plt.plot(np.log10(lambdas), estimated_mse_sklearn, label = 'cross_val_score') -plt.plot(np.log10(lambdas), estimated_mse_KFold, 'r--', label = 'KFold') - -plt.xlabel('log10(lambda)') -plt.ylabel('mse') - -plt.legend() - -plt.show() - -!ec !split @@ -1701,3 +1554,164 @@ plt.show() Write a code which implements gradient descent for a logistic regression example. + + +!split +===== Lab session: Material from last week and relevant for the first project ===== + + +!split +===== Various steps in cross-validation ===== + +When the repetitive splitting of the data set is done randomly, +samples may accidently end up in a fast majority of the splits in +either training or test set. Such samples may have an unbalanced +influence on either model building or prediction evaluation. To avoid +this $k$-fold cross-validation structures the data splitting. The +samples are divided into $k$ more or less equally sized exhaustive and +mutually exclusive subsets. In turn (at each split) one of these +subsets plays the role of the test set while the union of the +remaining subsets constitutes the training set. Such a splitting +warrants a balanced representation of each sample in both training and +test set over the splits. Still the division into the $k$ subsets +involves a degree of randomness. This may be fully excluded when +choosing $k=n$. This particular case is referred to as leave-one-out +cross-validation (LOOCV). + +!split +===== How to set up the cross-validation for Ridge and/or Lasso ===== + +* Define a range of interest for the penalty parameter. + +* Divide the data set into training and test set comprising samples $\{1, \ldots, n\} \setminus i$ and $\{ i \}$, respectively. + +* Fit the linear regression model by means of for example Ridge or Lasso regression for each $\lambda$ in the grid using the training set, and the corresponding estimate of the error variance $\bm{\sigma}_{-i}^2(\lambda)$, as +!bt +\begin{align*} +\bm{\beta}_{-i}(\lambda) & = ( \bm{X}_{-i, \ast}^{T} +\bm{X}_{-i, \ast} + \lambda \bm{I}_{pp})^{-1} +\bm{X}_{-i, \ast}^{T} \bm{y}_{-i} +\end{align*} +!et + +* Evaluate the prediction performance of these models on the test set by $C[y_i, \bm{X}_{i, \ast}; \bm{\beta}_{-i}(\lambda), \bm{\sigma}_{-i}^2(\lambda)]$. Or, by the prediction error $|y_i - \bm{X}_{i, \ast} \bm{\beta}_{-i}(\lambda)|$, the relative error, the error squared or the R2 score function. + +* Repeat the first three steps such that each sample plays the role of the test set once. + +* Average the prediction performances of the test sets at each grid point of the penalty bias/parameter. It is an estimate of the prediction performance of the model corresponding to this value of the penalty parameter on novel data. + + +!split +===== Cross-validation in brief ===== + +For the various values of $k$ + +o shuffle the dataset randomly. +o Split the dataset into $k$ groups. +o For each unique group: + o Decide which group to use as set for test data + o Take the remaining groups as a training data set + o Fit a model on the training set and evaluate it on the test set + o Retain the evaluation score and discard the model +o Summarize the model using the sample of model evaluation scores + + + +!split +===== Code Example for Cross-validation and $k$-fold Cross-validation ===== + +The code here uses Ridge regression with cross-validation (CV) resampling and $k$-fold CV in order to fit a specific polynomial. +!bc pycod +import numpy as np +import matplotlib.pyplot as plt +from sklearn.model_selection import KFold +from sklearn.linear_model import Ridge +from sklearn.model_selection import cross_val_score +from sklearn.preprocessing import PolynomialFeatures + +# A seed just to ensure that the random numbers are the same for every run. +# Useful for eventual debugging. +np.random.seed(3155) + +# Generate the data. +nsamples = 100 +x = np.random.randn(nsamples) +y = 3*x**2 + np.random.randn(nsamples) + +## Cross-validation on Ridge regression using KFold only + +# Decide degree on polynomial to fit +poly = PolynomialFeatures(degree = 6) + +# Decide which values of lambda to use +nlambdas = 500 +lambdas = np.logspace(-3, 5, nlambdas) + +# Initialize a KFold instance +k = 5 +kfold = KFold(n_splits = k) + +# Perform the cross-validation to estimate MSE +scores_KFold = np.zeros((nlambdas, k)) + +i = 0 +for lmb in lambdas: + ridge = Ridge(alpha = lmb) + j = 0 + for train_inds, test_inds in kfold.split(x): + xtrain = x[train_inds] + ytrain = y[train_inds] + + xtest = x[test_inds] + ytest = y[test_inds] + + Xtrain = poly.fit_transform(xtrain[:, np.newaxis]) + ridge.fit(Xtrain, ytrain[:, np.newaxis]) + + Xtest = poly.fit_transform(xtest[:, np.newaxis]) + ypred = ridge.predict(Xtest) + + scores_KFold[i,j] = np.sum((ypred - ytest[:, np.newaxis])**2)/np.size(ypred) + + j += 1 + i += 1 + + +estimated_mse_KFold = np.mean(scores_KFold, axis = 1) + +## Cross-validation using cross_val_score from sklearn along with KFold + +# kfold is an instance initialized above as: +# kfold = KFold(n_splits = k) + +estimated_mse_sklearn = np.zeros(nlambdas) +i = 0 +for lmb in lambdas: + ridge = Ridge(alpha = lmb) + + X = poly.fit_transform(x[:, np.newaxis]) + estimated_mse_folds = cross_val_score(ridge, X, y[:, np.newaxis], scoring='neg_mean_squared_error', cv=kfold) + + # cross_val_score return an array containing the estimated negative mse for every fold. + # we have to the the mean of every array in order to get an estimate of the mse of the model + estimated_mse_sklearn[i] = np.mean(-estimated_mse_folds) + + i += 1 + +## Plot and compare the slightly different ways to perform cross-validation + +plt.figure() + +plt.plot(np.log10(lambdas), estimated_mse_sklearn, label = 'cross_val_score') +plt.plot(np.log10(lambdas), estimated_mse_KFold, 'r--', label = 'KFold') + +plt.xlabel('log10(lambda)') +plt.ylabel('mse') + +plt.legend() + +plt.show() + +!ec + +