Logistic regression और regularization

Python में Linear Classifiers

Michael (Mike) Gelbart

Instructor, The University of British Columbia

Regularized logistic regression

Regularization training accuracy को कैसे प्रभावित करता है?

lr_weak_reg = LogisticRegression(C=100)
lr_strong_reg = LogisticRegression(C=0.01)


lr_weak_reg.fit(X_train, y_train)
lr_strong_reg.fit(X_train, y_train)


lr_weak_reg.score(X_train, y_train)
lr_strong_reg.score(X_train, y_train)

1.0
0.92

$\text{regularized loss = original loss + large coefficient penalty}$

अधिक regularization: training accuracy कम

Regularization test accuracy को कैसे प्रभावित करता है?

lr_weak_reg.score(X_test, y_test)

0.86

lr_strong_reg.score(X_test, y_test)

0.88

$\text{regularized loss = original loss + large coefficient penalty}$

अधिक regularization: training accuracy कम
अधिक regularization: (लगभग हमेशा) test accuracy ज्यादा

L1 बनाम L2 regularization

Lasso = linear regression के साथ L1 regularization
Ridge = linear regression के साथ L2 regularization
अन्य मॉडलों जैसे logistic regression के लिए हम बस L1, L2, आदि कहते हैं.

lr_L1 = LogisticRegression(solver='liblinear', penalty='l1')
lr_L2 = LogisticRegression() # penalty='l2' by default

lr_L1.fit(X_train, y_train)
lr_L2.fit(X_train, y_train)

plt.plot(lr_L1.coef_.flatten())
plt.plot(lr_L2.coef_.flatten())

L2 बनाम L1 regularization

अभ्यास करते हैं!

Python में Linear Classifiers