L1 vs L2 Regularization

Medium

machine-learning Google Netflix Two Sigma

You are building a linear model with 500 features, but you suspect only 20 are actually predictive. Which regularization technique is best suited for automatic feature selection?

Multiple Choice

L2 (Ridge) regularization, because it shrinks all coefficients equally L1 (Lasso) regularization, because it can drive coefficients to exactly zero Dropout regularization, because it randomly removes features during training No regularization is needed with only 500 features

Correct!

Incorrect — try again next time!

Upgrade to Pro to unlock solutions and explanations

L1 vs L2 Regularization

Multiple Choice

Related problems

Deep-dive guides