Cherry pick pretty plots

src/main.ipynb

@@ -53,6 +53,7 @@
     "x_train, x_test, y_train, y_test = train_test_split(x, y, test_size=0.2)\n",
     "\n",
     "fig, ax = plotting.scatter_dataset(x_train, x_test, y_train, y_test)\n",
+    "fig.set_layout_engine(\"compressed\")\n",
     "fig.savefig(os.path.join(FIG_DIR, \"data_scatter.png\"), dpi=300)"
    ]
   },
@@ -101,6 +102,7 @@
     "fig, ax = plotting.mse_r2_plot(\n",
     "    polynomial_degrees, train_mse_list, mse_list, train_r2_list, r2_list\n",
     ")\n",
+    "fig.set_layout_engine(\"compressed\")\n",
     "fig.savefig(os.path.join(FIG_DIR, \"ols_mse_r2.pdf\"))"
    ]
   },
@@ -112,6 +114,7 @@
    "outputs": [],
    "source": [
     "fig, ax = plotting.parameter_plot(polynomial_degrees, beta_OLS_list)\n",
+    "# fig.set_layout_engine(\"compressed\")\n",
     "fig.savefig(os.path.join(FIG_DIR, \"ols_parameter_plot.pdf\"))"
    ]
   },
@@ -165,6 +168,7 @@
     "    train_r2_ridge_list,\n",
     "    r2_ridge_list,\n",
     ")\n",
+    "fig.set_layout_engine(\"compressed\")\n",
     "fig.savefig(os.path.join(FIG_DIR, \"ridge_mse_r2.pdf\"))"
    ]
   },
@@ -222,6 +226,7 @@
     ")\n",
     "for ax in axs:\n",
     "    ax.set_xscale(\"log\")\n",
+    "fig.set_layout_engine(\"compressed\")\n",
     "fig.savefig(os.path.join(FIG_DIR, \"ridge_mse_r2_lambda.pdf\"))"
    ]
   },
@@ -249,6 +254,7 @@
     "ax.set_xticklabels(\n",
     "    [f\"${format_number(tick)}$\" for tick in lambda_values[::3]], rotation=45\n",
     ")\n",
+    "\n",
     "fig.savefig(os.path.join(FIG_DIR, \"ridge_parameter_plot.pdf\"))"
    ]
   },
@@ -260,7 +266,7 @@
    "outputs": [],
    "source": [
     "lambda_values = np.logspace(-5, 3, 60)\n",
-    "polynomial_degrees = np.arange(1, 31, dtype=int)\n",
+    "polynomial_degrees = np.arange(3, 20, dtype=int)\n",
     "\n",
     "test_mse_ridge_list = np.zeros((len(polynomial_degrees), len(lambda_values)))\n",
     "\n",
@@ -312,7 +318,7 @@
     "X_tr, X_te = datamanip.scale_data(X_train, X_test)\n",
     "y_tr, y_te = datamanip.scale_data(y_train, y_test)\n",
     "\n",
-    "fig, (ax1, ax2) = plt.subplots(1, 2, figsize=plotting.get_figsize(0.5))\n",
+    "fig, (ax1, ax2) = plt.subplots(2, 1, figsize=plotting.get_figsize(1.0))\n",
     "num_iters = 1_000\n",
     "\n",
     "learning_rates = np.logspace(-4, 0, 5)\n",
@@ -339,7 +345,7 @@
     "plotting.plot_optimizers(ax2, ridge_optimizers, X_tr, y_tr, \"Ridge Cost\", labels=labels)\n",
     "ax1.set_ylim(bottom=0.465, top=0.505)\n",
     "ax2.set_ylim(bottom=0.4775, top=0.505)\n",
-    "fig.tight_layout()\n",
+    "fig.set_layout_engine(\"compressed\")\n",
     "fig.savefig(os.path.join(FIG_DIR, \"gradient_descent_convergence.pdf\"))"
    ]
   },
@@ -350,7 +356,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "fig, (ax1, ax2) = plt.subplots(2, 1, figsize=plotting.get_figsize(0.8))\n",
+    "fig, (ax1, ax2) = plt.subplots(2, 1, figsize=plotting.get_figsize(1.0))\n",
     "\n",
     "X_train = datamanip.polynomial_features(x_train, 10, False)\n",
     "X_test = datamanip.polynomial_features(x_test, 10, False)\n",
@@ -393,7 +399,7 @@
     "\n",
     "ax1.set_ylim(bottom=0.465, top=0.505)\n",
     "ax2.set_ylim(bottom=0.4775, top=0.505)\n",
-    "fig.tight_layout()\n",
+    "fig.set_layout_engine(\"compressed\")\n",
     "fig.savefig(os.path.join(FIG_DIR, \"optimizer_comparison.pdf\"))"
    ]
   },
@@ -404,7 +410,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "fig, ax = plt.subplots(figsize=plotting.get_figsize(0.5))\n",
+    "fig, ax = plt.subplots(figsize=plotting.get_figsize(0.6))\n",
     "\n",
     "ols_adam = optimizers.OLSAdam(learning_rate=0.01, beta1=0.9, beta2=0.999)\n",
     "ridge_adam = optimizers.RidgeAdam(learning_rate=0.01, beta1=0.9, beta2=0.999, lam=0.1)\n",
@@ -435,7 +441,7 @@
     "]\n",
     "labels += [\"Adam Opt.\", \"Grad. Desc.\"]\n",
     "ax.legend(handles, labels)\n",
-    "fig.tight_layout()\n",
+    "fig.set_layout_engine(\"compressed\")\n",
     "fig.savefig(os.path.join(FIG_DIR, \"cost_function_comparison.pdf\"))"
    ]
   },
@@ -488,7 +494,7 @@
     "fig, ax = plotting.optimization_performance_evaluation(\n",
     "    optimizer_list, labels, X_tr, y_tr\n",
     ")\n",
-    "fig.tight_layout()\n",
+    "fig.set_layout_engine(\"compressed\")\n",
     "fig.savefig(os.path.join(FIG_DIR, \"optimization_performance.pdf\"))"
    ]
   },
@@ -500,7 +506,7 @@
    "outputs": [],
    "source": [
     "X_size = 1_000_000\n",
-    "num_epochs = 1000\n",
+    "num_epochs = 500\n",
     "batches_per_epoch = 100\n",
     "batch_size = 128\n",
     "\n",
@@ -566,23 +572,27 @@
     "\n",
     "labels = [\"SGD\", \"Mom. SGD\", \"AdaGrad\", \"RMSProp\", \"Adam\"]\n",
     "\n",
-    "fig, axs = plt.subplots(1, 3, figsize=plotting.get_figsize(0.5), sharey=True)\n",
+    "fig, axs = plt.subplots(\n",
-    "plotting.plot_optimizers(\n",
+    "    3, 1, figsize=plotting.get_figsize(1.0, full_width=False), sharex=True\n",
-    "    axs[0], optimizers_ols, X_tr, y_tr, ylabel=\"Average Cost per Epoch\", labels=labels\n",
     ")\n",
     "plotting.plot_optimizers(\n",
-    "    axs[1], optimizers_ridge, X_tr, y_tr, ylabel=None, labels=labels\n",
+    "    axs[0], optimizers_ols, X_tr, y_tr, ylabel=\"OLS Cost\", labels=labels\n",
     ")\n",
     "plotting.plot_optimizers(\n",
-    "    axs[2], optimizers_lasso, X_tr, y_tr, ylabel=None, labels=labels\n",
+    "    axs[1], optimizers_ridge, X_tr, y_tr, ylabel=\"Ridge Cost\", labels=labels\n",
+    ")\n",
+    "plotting.plot_optimizers(\n",
+    "    axs[2], optimizers_lasso, X_tr, y_tr, ylabel=\"LASSO Cost\", labels=labels\n",
     ")\n",
     "for ax in axs:\n",
-    "    ax.set_xlabel(\"Epoch\")\n",
+    "    ax.set_xlabel(None)\n",
     "\n",
-    "axs[0].set_title(\"OLS\")\n",
+    "axs[2].set_xlabel(\"Epoch\")\n",
-    "axs[1].set_title(\"Ridge\")\n",
+    "\n",
-    "axs[2].set_title(\"LASSO\")\n",
+    "# axs[0].set_title(\"OLS\")\n",
-    "fig.tight_layout()\n",
+    "# axs[1].set_title(\"Ridge\")\n",
+    "# axs[2].set_title(\"LASSO\")\n",
+    "fig.set_layout_engine(\"compressed\")\n",
     "fig.savefig(os.path.join(FIG_DIR, \"stochastic_gradient_descent_convergence.pdf\"))"
    ]
   },
@@ -674,7 +684,7 @@
     "ax.set_ylabel(\"Error\")\n",
     "ax.legend()\n",
     "\n",
-    "fig.tight_layout()\n",
+    "fig.set_layout_engine(\"compressed\")\n",
     "fig.savefig(os.path.join(FIG_DIR, \"bias_variance_tradeoff.pdf\"))"
    ]
   },
@@ -709,7 +719,9 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "fig, (ax, ax2) = plt.subplots(1, 2, figsize=plotting.get_figsize(0.5), sharey=True)\n",
+    "fig, (ax, ax2) = plt.subplots(\n",
+    "    1, 2, figsize=plotting.get_figsize(0.35, True), sharey=True\n",
+    ")\n",
     "mse_mean = np.mean(mses, axis=1)\n",
     "bias_mean = np.mean(biases, axis=1)\n",
     "var_mean = np.mean(variances, axis=1)\n",
@@ -768,7 +780,7 @@
     "ax2.set_xlabel(\"Polynomial Degree\")\n",
     "# ax2.set_ylabel(\"Mean Squared Error\")\n",
     "ax2.legend()\n",
-    "fig.tight_layout()\n",
+    "fig.set_layout_engine(\"compressed\")\n",
     "fig.savefig(os.path.join(FIG_DIR, \"bias_variance_tradeoff_combined_plot.pdf\"))"
    ]
   },
@@ -838,9 +850,9 @@
     "        )\n",
     "\n",
     "ax.set_xlabel(\"Polynomial Degree\")\n",
-    "ax.set_ylabel(f\"MSE ({k_folds}-fold validation)\")\n",
+    "ax.set_ylabel(f\"MSE ({k_folds}-fold)\")\n",
     "ax.legend()\n",
-    "fig.tight_layout()\n",
+    "fig.set_layout_engine(\"compressed\")\n",
     "fig.savefig(os.path.join(FIG_DIR, \"kfold_mse_comparison_per_cost_function.pdf\"))"
    ]
   },

src/pyoptim/plotting.py

@@ -95,7 +95,6 @@ def mse_r2_plot(
     ax2.set_xlabel(labels["xlabel"])
     ax2.set_ylabel(labels["ylabel2"])
 
-    fig.tight_layout()
     return fig, (ax1, ax2)
 
 
@@ -136,7 +135,6 @@ def parameter_plot(
     ax.set_xticklabels(polynomial_degrees[::3])
     ax.set_yticks(np.arange(0, beta_OLS.shape[0], 5))
     ax.set_yticklabels(np.arange(1, beta_OLS.shape[0] + 1, step=5))
-    fig.tight_layout()
     return fig, ax
 
 
@@ -149,7 +147,6 @@ def scatter_dataset(
     ax.set_xlabel("$x$")
     ax.set_ylabel("$y$")
     ax.legend()
-    fig.tight_layout()
     return fig, ax
 
 
@@ -188,7 +185,7 @@ def optimization_performance_evaluation(
     X_tr: np.ndarray,
     y_tr: np.ndarray,
 ):
-    fig, ax = plt.subplots(figsize=get_figsize(0.5))
+    fig, ax = plt.subplots(figsize=get_figsize(0.6))
 
     for optimizer, label in zip(optimizers, labels):
         start_time = time.time()
@@ -215,5 +212,4 @@ def optimization_performance_evaluation(
     ax.set_xlabel("Epoch")
     ax.set_ylabel("Cost")
     ax.legend()
-    fig.tight_layout()
     return fig, ax