{ "cells": [ { "cell_type": "markdown", "metadata": {}, "source": [ "# Lab 05: Regularized linear regression" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "The goal of this lab is to explore and understand l1 and l2 regularization of linear models." ] }, { "cell_type": "code", "execution_count": 1, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Populating the interactive namespace from numpy and matplotlib\n" ] } ], "source": [ "import pandas as pd\n", "%pylab inline" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# 1. Classification data\n", "\n", "We will use the same data as in Lab 4: the samples are tumors, each described by the expression (= the abundance) of 3,000 genes. The goal is to separate the endometrium tumors from the uterine ones." ] }, { "cell_type": "code", "execution_count": 2, "metadata": {}, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
ID_REF1554530_at1553185_at1554340_a_at1556202_at1553957_at1555469_a_at1553660_at1554681_a_at1554938_a_at...1553967_at1553362_at1553002_at1556194_a_at1556420_s_at1555855_at1554508_at1555097_a_at1556371_atTissue
011772210.813233.727.2167.8450.7283.86.48.626.7...165.243.777.042.2154.8266.6444.066.950.6Endometrium
17663812.64986.81.7221.1380.8394.3121.28.0153.8...190.73.284.0183.0288.020.699.36.412.2Endometrium
28895216.66053.8121.4342.7217.6367.9159.710.8124.4...95.917.172.3292.9209.511.651.333.833.4Endometrium
3766329.96109.123.0139.3501.8289.9101.79.7204.8...235.137.981.5109.3537.758.773.958.915.4Endometrium
48896613.18430.917.429.4449.1248.2104.111.294.5...125.059.9186.8122.5355.265.1139.914.111.2Endometrium
\n", "

5 rows × 3002 columns

\n", "
" ], "text/plain": [ " ID_REF 1554530_at 1553185_at 1554340_a_at 1556202_at 1553957_at \\\n", "0 117722 10.8 13233.7 27.2 167.8 450.7 \n", "1 76638 12.6 4986.8 1.7 221.1 380.8 \n", "2 88952 16.6 6053.8 121.4 342.7 217.6 \n", "3 76632 9.9 6109.1 23.0 139.3 501.8 \n", "4 88966 13.1 8430.9 17.4 29.4 449.1 \n", "\n", " 1555469_a_at 1553660_at 1554681_a_at 1554938_a_at ... 1553967_at \\\n", "0 283.8 6.4 8.6 26.7 ... 165.2 \n", "1 394.3 121.2 8.0 153.8 ... 190.7 \n", "2 367.9 159.7 10.8 124.4 ... 95.9 \n", "3 289.9 101.7 9.7 204.8 ... 235.1 \n", "4 248.2 104.1 11.2 94.5 ... 125.0 \n", "\n", " 1553362_at 1553002_at 1556194_a_at 1556420_s_at 1555855_at 1554508_at \\\n", "0 43.7 77.0 42.2 154.8 266.6 444.0 \n", "1 3.2 84.0 183.0 288.0 20.6 99.3 \n", "2 17.1 72.3 292.9 209.5 11.6 51.3 \n", "3 37.9 81.5 109.3 537.7 58.7 73.9 \n", "4 59.9 186.8 122.5 355.2 65.1 139.9 \n", "\n", " 1555097_a_at 1556371_at Tissue \n", "0 66.9 50.6 Endometrium \n", "1 6.4 12.2 Endometrium \n", "2 33.8 33.4 Endometrium \n", "3 58.9 15.4 Endometrium \n", "4 14.1 11.2 Endometrium \n", "\n", "[5 rows x 3002 columns]" ] }, "execution_count": 2, "metadata": {}, "output_type": "execute_result" } ], "source": [ "# load the endometrium vs. uterus tumor data\n", "endometrium_data = pd.read_csv('data/small_Endometrium_Uterus.csv', sep=\",\") # load data\n", "endometrium_data.head(n=5) # adjust n to view more data" ] }, { "cell_type": "code", "execution_count": 3, "metadata": {}, "outputs": [], "source": [ "# Create the design matrix and target vector\n", "X_clf = endometrium_data.drop(['ID_REF', 'Tissue'], axis=1).values\n", "y_clf = pd.get_dummies(endometrium_data['Tissue']).values[:,1]" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "__Question:__ Split the data in a train set containing 70% of the data and a test set containing the remaining 30%. We use [model_selection.train_test_split](http://scikit-learn.org/stable/modules/generated/sklearn.model_selection.train_test_split.html#sklearn.model_selection.train_test_split_)" ] }, { "cell_type": "code", "execution_count": 4, "metadata": {}, "outputs": [], "source": [ "from sklearn import model_selection\n", "Xtr, Xte, ytr, yte = model_selection.train_test_split(X_clf, y_clf, \n", " test_size=0.3, \n", " random_state=42,\n", " stratify=y_clf)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "Let us also compute a scaled version of the data. The data is scaled on the _train_ set, and the scaling parameters (mean, standard deviation) are applied to the test set. __Question:__ Why?" ] }, { "cell_type": "code", "execution_count": 5, "metadata": {}, "outputs": [], "source": [ "from sklearn import preprocessing\n", "scaler = preprocessing.StandardScaler()\n", "Xtr_scaled = scaler.fit_transform(Xtr)\n", "Xte_scaled = scaler.transform(Xte)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# 2. Logistic regression, not regularized \n", "\n", "Let us train a logisitic regression _without regularization_ on our train set, and evaluate it on the test set. This is similar to Lab 4 and will serve as a comparison point." ] }, { "cell_type": "code", "execution_count": 6, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "No regularization: accuracy = 0.691\n", "AUC = 0.651\n" ] }, { "name": "stderr", "output_type": "stream", "text": [ "/home/wbader/miniconda3/envs/tp-ml/lib/python3.6/site-packages/sklearn/linear_model/_logistic.py:765: ConvergenceWarning: lbfgs failed to converge (status=1):\n", "STOP: TOTAL NO. of ITERATIONS REACHED LIMIT.\n", "\n", "Increase the number of iterations (max_iter) or scale the data as shown in:\n", " https://scikit-learn.org/stable/modules/preprocessing.html\n", "Please also refer to the documentation for alternative solver options:\n", " https://scikit-learn.org/stable/modules/linear_model.html#logistic-regression\n", " extra_warning_msg=_LOGISTIC_SOLVER_CONVERGENCE_MSG)\n" ] } ], "source": [ "# Logistic regression (no regularization, no scaling)\n", "from sklearn import linear_model\n", "clf_logreg = linear_model.LogisticRegression(C=1e6) # large C = no regularization\n", "\n", "# Train the model\n", "clf_logreg.fit(Xtr, ytr)\n", "\n", "# Predict on the test set\n", "# Predicted probabilities of belonging to the positive class\n", "pos_idx = list(clf_logreg.classes_).index(1)\n", "ypred_logreg = clf_logreg.predict_proba(Xte)[:, pos_idx]\n", "\n", "# Predicted binary labels\n", "ypred_logreg_b = np.where(ypred_logreg > 0.5, 1, 0)\n", "\n", "from sklearn import metrics\n", "print(\"No regularization: accuracy = %.3f\" % metrics.accuracy_score(yte, ypred_logreg_b))\n", "print(\"AUC = %.3f\" % (metrics.roc_auc_score(yte, ypred_logreg)))" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "__Question:__ Repeat the experiment on the scaled data. What do you observe in terms of performance?" ] }, { "cell_type": "code", "execution_count": 7, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Scaled, no regularization: accuracy = 0.691\n", "AUC = 0.784\n" ] } ], "source": [ "# Logistic regression (no regularization, scaling)\n", "clf_logreg_s = linear_model.LogisticRegression(C=1e6)\n", "\n", "# Train the model\n", "clf_logreg_s.fit(Xtr_scaled, ytr)\n", "\n", "# Predict on the test set\n", "# Predicted probabilities of belonging to the positive class\n", "pos_idx = list(clf_logreg_s.classes_).index(1)\n", "ypred_logreg_s = clf_logreg_s.predict_proba(Xte_scaled)[:, pos_idx]\n", "# Predicted binary labels\n", "ypred_logreg_s_b = np.where(ypred_logreg_s > 0.5, 1, 0)\n", "\n", "print(\"Scaled, no regularization: accuracy = %.3f\" % metrics.accuracy_score(yte, ypred_logreg_s_b))\n", "print(\"AUC = %.3f\" % (metrics.roc_auc_score(yte, ypred_logreg_s)))" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# 3. L2-regularized logistic regression \n", "\n", "__Question:__ What is the role of L2 regularization?" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "Let us use an l2-regularized logistic regression with the parameter `C` set to 0.01. \n", "\n", "__Question:__ What is the role of `C`? How does it relate to the `lambda` regularization parameter we have seen in class?\n", "\n", "__Question:__ Train the l2-regularized logistic regression initialized below on the scaled training data, and evaluate it on the sclaed test set (as above). How does the performance evolve?" ] }, { "cell_type": "code", "execution_count": 8, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Scaled, l2 regularization (C=1.00e-02): accuracy = 0.727\n", "AUC = 0.817\n" ] } ], "source": [ "cvalue = 0.01\n", "clf_logreg_l2_s = linear_model.LogisticRegression(C=cvalue, penalty='l2')\n", "\n", "# Train the model\n", "clf_logreg_l2_s.fit(Xtr_scaled, ytr)\n", "\n", "# index of positive class\n", "pos_idx = list(clf_logreg_l2_s.classes_).index(1)\n", "# predict probability of being positive\n", "ypred_logreg_l2_s = clf_logreg_l2_s.predict_proba(Xte_scaled)[:, pos_idx]\n", "# predict binary labels\n", "ypred_logreg_l2_s_b = np.where(ypred_logreg_l2_s > 0.5, 1, 0)\n", "\n", "print(\"Scaled, l2 regularization (C=%.2e): accuracy = %.3f\" % (cvalue, \n", " metrics.accuracy_score(yte, ypred_logreg_l2_s_b)))\n", "print(\"AUC = %.3f\" % (metrics.roc_auc_score(yte, ypred_logreg_l2_s)))" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### 3.1 Effect of L2-regularization on the logistic regression coefficients\n", "\n", "We will now look at how the regression coefficients have evolved between the non-regularized and the regularized versions of the logistic regression.\n", "\n", "__Question:__ Fill in the blanks below to plot the regression coefficients of both the trained `clf_logreg_l2_s` and `clf_logreg_s` models. Use the [documentation](http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.LogisticRegression.html) to figure out how to access these coefficients. What do you observe?" ] }, { "cell_type": "markdown", "metadata": {}, "source": [] }, { "cell_type": "code", "execution_count": 9, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "(0.0, 3000.0)" ] }, "execution_count": 9, "metadata": {}, "output_type": "execute_result" }, { "data": { "image/png": "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", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "# Effect of l2-regularization on the weights\n", "num_features = X_clf.shape[1]\n", "plt.scatter(range(num_features), clf_logreg_s.coef_, \n", " color='blue', marker='x', label='Logistic regression')\n", "plt.scatter(range(num_features), clf_logreg_l2_s.coef_, \n", " color='orange', marker='+', label='L2-regularized logistic regression')\n", "\n", "plt.xlabel('Genes', fontsize=16)\n", "plt.ylabel('Weights', fontsize=16)\n", "plt.title('Logistic regression weights', fontsize=16)\n", "plt.legend(fontsize=14, loc=(1.05, 0))\n", "plt.xlim([0, num_features])" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 3.2 Optimization of the regularization parameter\n", "\n", "We will now use a 3-fold cross-validation on the training set to optimize the value of C. Scikit-learn makes it really easy to use a cross-validation to choose a good value for $\\alpha$ among a grid of several choices. Check the [GridSearchCV class](http://scikit-learn.org/0.17/modules/generated/sklearn.grid_search.GridSearchCV.html#sklearn.grid_search.GridSearchCV)." ] }, { "cell_type": "code", "execution_count": 10, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "[1.00000000e-05 2.06913808e-05 4.28133240e-05 8.85866790e-05\n", " 1.83298071e-04 3.79269019e-04 7.84759970e-04 1.62377674e-03\n", " 3.35981829e-03 6.95192796e-03 1.43844989e-02 2.97635144e-02\n", " 6.15848211e-02 1.27427499e-01 2.63665090e-01 5.45559478e-01\n", " 1.12883789e+00 2.33572147e+00 4.83293024e+00 1.00000000e+01]\n" ] } ], "source": [ "# Create a range of values to test for the parameter C\n", "cvalues_list = np.logspace(-5, 1, 20)\n", "print(cvalues_list)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "__Question:__ Fill in the blanks below to find the optimal value of the parameter C.\n", "\n", "Use the `.best_estimator_` attribute of a `GridSearchCV`. " ] }, { "cell_type": "code", "execution_count": 11, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Scaled, l2 regularization (C=5.46e-01): accuracy = 0.709\n", "AUC = 0.806\n" ] } ], "source": [ "# Optimize cvalue\n", "classifier = linear_model.LogisticRegression(penalty='l2')\n", "param_grid = {'C': cvalues_list\n", " }\n", "clf_logreg_l2_s_opt = model_selection.GridSearchCV(classifier, \n", " param_grid, \n", " cv=3) \n", "\n", "# Train the model\n", "clf_logreg_l2_s_opt.fit(Xtr_scaled, ytr)\n", "\n", "# index of positive class\n", "pos_idx = list(clf_logreg_l2_s_opt.best_estimator_.classes_).index(1)\n", "# predict probability of being positive\n", "ypred_logreg_l2_s_opt = clf_logreg_l2_s_opt.best_estimator_.predict_proba(Xte_scaled)[:, pos_idx]\n", "# predict binary label\n", "ypred_logreg_l2_s_opt_b = np.where(ypred_logreg_l2_s_opt > 0.5, 1, 0)\n", "\n", "# optimal value of C\n", "cvalue_opt = clf_logreg_l2_s_opt.best_params_['C']\n", "print(\"Scaled, l2 regularization (C=%.2e): accuracy = %.3f\" % (cvalue_opt, \n", " metrics.accuracy_score(yte, ypred_logreg_l2_s_opt_b)))\n", "print(\"AUC = %.3f\" % (metrics.roc_auc_score(yte, ypred_logreg_l2_s_opt)))" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "__Question:__ Fill in the code below to compare the ROC curves of the non-regularized and l2-regularized logistic regressions." ] }, { "cell_type": "code", "execution_count": 12, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "" ] }, "execution_count": 12, "metadata": {}, "output_type": "execute_result" }, { "data": { "image/png": "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", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "fpr_logreg_s, tpr_logreg_s, t = metrics.roc_curve(yte, ypred_logreg_s , pos_label=1)\n", "auc_logreg_s = metrics.auc(fpr_logreg_s, tpr_logreg_s)\n", "plt.plot(fpr_logreg_s, tpr_logreg_s, color='blue', \n", " label='No regularization: AUC = %0.3f' % auc_logreg_s)\n", "\n", "fpr_logreg_l2_s_opt, tpr_logreg_l2_s_opt, t = metrics.roc_curve(yte, ypred_logreg_l2_s_opt , pos_label=1)\n", "auc_logreg_l2_s_opt = metrics.auc(fpr_logreg_l2_s_opt, tpr_logreg_l2_s_opt)\n", "plt.plot(fpr_logreg_l2_s_opt, tpr_logreg_l2_s_opt, color='orange', \n", " label='L2 regularization: AUC = %0.3f' % auc_logreg_l2_s_opt)\n", "\n", "plt.xlabel('False Positive Rate', fontsize=16)\n", "plt.ylabel('True Positive Rate', fontsize=16)\n", "plt.title('ROC curve: Logistic regression', fontsize=16)\n", "plt.legend(fontsize=14)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "__Question:__ Is the optimal C larger or smaller than the one we tried before? Does this mean more or less regularization? Do you expect larger or smaller regularization coefficients?" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "__Question:__ Fill in the blanks to compare the regularization weights of the different methods." ] }, { "cell_type": "code", "execution_count": 13, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "(0.0, 3000.0)" ] }, "execution_count": 13, "metadata": {}, "output_type": "execute_result" }, { "data": { "image/png": "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", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "# Effect of l2-regularization on the weights\n", "num_features = X_clf.shape[1]\n", "plt.scatter(range(num_features), clf_logreg_s.coef_, \n", " color='blue', marker='x', label='Logistic regression')\n", "plt.scatter(range(num_features), clf_logreg_l2_s.coef_, \n", " color='orange', marker='+', label='L2-regularized logistic regression')\n", "plt.scatter(range(num_features), clf_logreg_l2_s_opt.best_estimator_.coef_, \n", " color='magenta', marker='.', label='L2-regularized logistic regression (opt)')\n", "\n", "plt.xlabel('Genes', fontsize=16)\n", "plt.ylabel('Weights', fontsize=16)\n", "plt.title('Logistic regression weights', fontsize=16)\n", "plt.legend(fontsize=14, loc=(1.05, 0))\n", "plt.xlim([0, num_features])" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "# 4. L1-regularized logistic regression\n", "\n", "__Question:__ What is the role of the l1-regularized logistic regression?" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "__Question:__ Instead of a l2-regularized logistic regression with `C=0.01`, now train and evaluate a __l1__-regularized logistic regression with __`C=10.0`__." ] }, { "cell_type": "code", "execution_count": 14, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Scaled, l1 regularization (C=1.00e+01): accuracy = 0.709\n", "AUC = 0.757\n" ] } ], "source": [ "cvalue = 10\n", "clf_logreg_l1_s = linear_model.LogisticRegression(C=cvalue, penalty='l1' ,solver='liblinear')\n", "clf_logreg_l1_s.fit(Xtr_scaled, ytr)\n", "\n", "# index of the positive class\n", "pos_idx = list(clf_logreg_l1_s.classes_).index(1)\n", "# predict the probability of belonging to the positive class\n", "ypred_logreg_l1_s = clf_logreg_l1_s.predict_proba(Xte_scaled)[:, pos_idx]\n", "# predict binary labels\n", "ypred_logreg_l1_s_b = np.where(ypred_logreg_l1_s > 0.5, 1, 0)\n", "\n", "print(\"Scaled, l1 regularization (C=%.2e): accuracy = %.3f\" % (cvalue, \n", " metrics.accuracy_score(yte, ypred_logreg_l1_s_b)))\n", "print(\"AUC = %.3f\" % (metrics.roc_auc_score(yte, ypred_logreg_l1_s)))" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "__Question__: How did the performance evolve?" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 4.1 Effect of regularization on the regression coefficients\n", "\n", "__Question:__ Plot the weights that were given to each feature in your data." ] }, { "cell_type": "code", "execution_count": 15, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "(0.0, 3000.0)" ] }, "execution_count": 15, "metadata": {}, "output_type": "execute_result" }, { "data": { "image/png": "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", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "num_features = X_clf.shape[1]\n", "plt.scatter(range(num_features), clf_logreg_s.coef_, \n", " color='blue', marker='x', label='Logistic regression')\n", "plt.scatter(range(num_features), clf_logreg_l1_s.coef_, \n", " color='orange', marker='+', label='L1-regularized logistic regression')\n", "\n", "plt.xlabel('Genes', fontsize=16)\n", "plt.ylabel('Weights', fontsize=16)\n", "plt.title('Logistic regression weights', fontsize=16)\n", "plt.legend(fontsize=14, loc=(1.05, 0))\n", "plt.xlim([0, num_features])" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "__Question:__ What do you observe? How does this differ from l2-regularization?" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "__Question:__ How many weights are different from zero? How many features are _not_ used by the l1-regularized model? " ] }, { "cell_type": "code", "execution_count": 16, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "The non-regularized logistic regression uses 3000 features\n", "The L2-regularized logistic regression uses 3000 features\n", "The L1-regularized logistic regression uses 138 features\n", "Number of features discarded by the L1-regularization: 2862\n" ] } ], "source": [ "# Number of selected features\n", "print(\"The non-regularized logistic regression uses %d features\" % len(np.where(clf_logreg_s.coef_ !=0)[1]))\n", "print(\"The L2-regularized logistic regression uses %d features\" % sum(clf_logreg_l2_s.coef_!=0)\n", " )\n", "print(\"The L1-regularized logistic regression uses %d features\" % sum(clf_logreg_l1_s.coef_!=0)\n", " )\n", "\n", "print(\"Number of features discarded by the L1-regularization: %d\" % sum(clf_logreg_l1_s.coef_ == 0))" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 4.2 Optimization of the regularization parameter\n", "\n", "__Question:__ Fill in the blanks to optimize the value of `C` for l1-regularized logistic regression." ] }, { "cell_type": "code", "execution_count": 17, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "[1.00000000e-02 1.83298071e-02 3.35981829e-02 6.15848211e-02\n", " 1.12883789e-01 2.06913808e-01 3.79269019e-01 6.95192796e-01\n", " 1.27427499e+00 2.33572147e+00 4.28133240e+00 7.84759970e+00\n", " 1.43844989e+01 2.63665090e+01 4.83293024e+01 8.85866790e+01\n", " 1.62377674e+02 2.97635144e+02 5.45559478e+02 1.00000000e+03]\n", "Scaled, l1 regularization (C=4.83e+01): accuracy = 0.709\n", "AUC = 0.760\n" ] } ], "source": [ "# Optimize cvalue\n", "cvalues_list = np.logspace(-2, 3, 20)\n", "print(cvalues_list)\n", "\n", "classifier = linear_model.LogisticRegression(penalty='l1',solver='liblinear')\n", "param_grid = {'C': cvalues_list\n", " }\n", "clf_logreg_l1_s_opt = model_selection.GridSearchCV(classifier, \n", " param_grid, \n", " cv=3) \n", "# Train the model\n", "clf_logreg_l1_s_opt.fit(Xtr_scaled, ytr)\n", "\n", "# index of positive class\n", "pos_idx = list(clf_logreg_l1_s_opt.best_estimator_.classes_).index(1)\n", "# predict probability of being positive\n", "ypred_logreg_l1_s_opt = clf_logreg_l1_s_opt.best_estimator_.predict_proba(Xte_scaled)[:, pos_idx]\n", "# predict binary label\n", "ypred_logreg_l1_s_opt_b = np.where(ypred_logreg_l1_s_opt > 0.5, 1, 0)\n", "\n", "# optimal value of C\n", "cvalue_opt = clf_logreg_l1_s_opt.best_params_['C'];\n", "\n", "print(\"Scaled, l1 regularization (C=%.2e): accuracy = %.3f\" % (cvalue_opt, \n", " metrics.accuracy_score(yte, ypred_logreg_l1_s_b)))\n", "print(\"AUC = %.3f\" % (metrics.roc_auc_score(yte, ypred_logreg_l1_s_opt)))" ] }, { "cell_type": "code", "execution_count": 18, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "" ] }, "execution_count": 18, "metadata": {}, "output_type": "execute_result" }, { "data": { "image/png": "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", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "# ROC curve\n", "fpr_logreg_s, tpr_logreg_s, t = metrics.roc_curve(yte, ypred_logreg_s , pos_label=1)\n", "auc_logreg_s = metrics.auc(fpr_logreg_s, tpr_logreg_s)\n", "plt.plot(fpr_logreg_s, tpr_logreg_s, color='blue', \n", " label='No regularization: AUC = %0.3f' % auc_logreg_s)\n", "\n", "fpr_logreg_l2_s_opt, tpr_logreg_l2_s_opt, t = metrics.roc_curve(yte, ypred_logreg_l2_s_opt , pos_label=1)\n", "auc_logreg_l2_s_opt = metrics.auc(fpr_logreg_l2_s_opt, tpr_logreg_l2_s_opt)\n", "plt.plot(fpr_logreg_l2_s_opt, tpr_logreg_l2_s_opt, color='orange', \n", " label='L2 regularization: AUC = %0.3f' % auc_logreg_l2_s_opt)\n", "\n", "fpr_logreg_l1_s_opt, tpr_logreg_l1_s_opt, t = metrics.roc_curve(yte, ypred_logreg_l1_s_opt, pos_label=1)\n", "auc_logreg_l1_s_opt = metrics.auc(fpr_logreg_l1_s_opt, tpr_logreg_l1_s_opt)\n", "plt.plot(fpr_logreg_l1_s_opt, tpr_logreg_l1_s_opt, color='magenta', \n", " label='L1 regularization: AUC = %0.3f' % auc_logreg_l1_s_opt)\n", "\n", "plt.xlabel('False Positive Rate', fontsize=16)\n", "plt.ylabel('True Positive Rate', fontsize=16)\n", "plt.title('ROC curve: Logistic regression', fontsize=16)\n", "plt.legend(fontsize=14)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "__Question:__ How many genes does the l1-regularized approach select? How does this affect the performance, compared to `C=10.0` or the l2-regularized approach?" ] }, { "cell_type": "code", "execution_count": 19, "metadata": { "scrolled": true }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "The L1-regularized logistic regression uses 138 features\n", "The optimized L1-regularized logistic regression uses 153 features\n" ] } ], "source": [ "print(\"The L1-regularized logistic regression uses %d features\" % sum(clf_logreg_l1_s.coef_!=0)\n", " )\n", "print(\"The optimized L1-regularized logistic regression uses %d features\" % \\\n", " sum(clf_logreg_l1_s_opt.best_estimator_.coef_!=0)\n", " )" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "__Question:__ Compare the features selected with `C=0.01` and your optimal `C` value. What do you observe?" ] }, { "cell_type": "code", "execution_count": 20, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "(0.0, 3000.0)" ] }, "execution_count": 20, "metadata": {}, "output_type": "execute_result" }, { "data": { "image/png": "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", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "num_features = X_clf.shape[1]\n", "plt.scatter(range(num_features), clf_logreg_l1_s.coef_, \n", " color='orange', marker='+', label='L1-regularized logistic regression')\n", "plt.scatter(range(num_features), clf_logreg_l1_s_opt.best_estimator_.coef_, \n", " color='magenta', marker='x', label='L1-regularized logistic regression (optimal C)')\n", "\n", "plt.xlabel('Genes', fontsize=16)\n", "plt.ylabel('Weights', fontsize=16)\n", "plt.title('Logistic regression weights', fontsize=16)\n", "plt.legend(fontsize=14, loc=(1.05, 0))\n", "plt.xlim([0, num_features])" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [] } ], "metadata": { "kernelspec": { "display_name": "Python 3", "language": "python", "name": "python3" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.6.13" } }, "nbformat": 4, "nbformat_minor": 2 }