Explain L1 vs L2 regularization in Machine Learning.

Question

Accepted Answer

L1 and L2 regularization are techniques to prevent overfitting in machine learning models by adding penalty terms to the loss function.

L1 regularization (Lasso) adds the absolute value of coefficients to the loss function, promoting sparsity.
L2 regularization (Ridge) adds the square of coefficients to the loss function, which discourages large weights.
L1 can lead to feature selection by driving some coefficients to zero, while L2 tends to distribute weights more evenly.
Example: In a linear regression model, L1 might eliminate less important features, while L2 keeps all features but reduces their impact.
L1 is useful when we suspect that many features are irrelevant, while L2 is better when we believe all features contribute to the outcome.

Accepted Answer

L1 regularization, also known as Lasso regularization, adds a penalty equal to absolute value of magnitude of coefficients.
L2 relgularization (Ridge Regression) adds a penalty equal to square of the magnitude of coefficients. Useful for dealing with multicolinearity by keeping the weights as small as possible.

Asked in Fonada and 4 others

Explain L1 vs L2 regularization in Machine Learning.

Interview Questions from Popular Companies