What is the difference between L1 and L2 regularization?

Updated May 17, 2026

Short answer

L1 regularization promotes sparsity, while L2 penalizes large weights smoothly.

Deep explanation

L1 adds absolute value penalty leading to feature selection by shrinking some weights to zero. L2 adds squared penalty reducing magnitude of all weights but rarely making them zero. L1 is useful for interpretability, L2 for stability.

Real-world example

L1 is used in gene selection in bioinformatics to identify important genes.

Common mistakes

  • Assuming both regularizations behave the same way.

Follow-up questions

  • What is elastic net?
  • When should L1 be preferred?

More Supervised Learning interview questions

View all →