SESKStuff Everyone Should Know
SESK: Stuff Everyone Should Know · sesk.info/ai-l1-regularization-zero

Which regularization can push useless features' weights to exactly zero?

L1 regularization, which penalizes the absolute values of weights.

Start playing free

  1. L1 regularization ✔
  2. L2 regularization
  3. Dropout regularization

Why: L1 penalizes the sum of the absolute values of the weights. It drives weights of irrelevant features to exactly 0, removing them from the model. L2 only shrinks them close to 0.

Huh, didn't know that: Because it zeroes out features, L1 doubles as a simple form of feature selection.

The dad joke
L1 regularization is the strictest coach. If you do not help the team, your weight goes to zero.
Like this card?
Play the game

References

Answer: L1 regularization. L1 penalizes the sum of the absolute values of the weights. It drives weights of irrelevant features to exactly 0, removing them from the model. L2 only shrinks them close to 0.

  1. Google Machine Learning Glossary: L1 regularization developers.google.com/machine-learning/glossary
    L1 regularization helps drive the weights of irrelevant or barely relevant features to exactly 0.
    Checked 2026-10-10.

Think this answer is wrong? Tap "Challenge this answer" on the card in the game and tell us why.

Test yourself in the game

Artificial IntelligenceWhat does L2 regularization penalize?Artificial IntelligenceWhat did AlexNet's authors use to reduce overfitting in its fully connected layers?Artificial IntelligenceWhat does dropout randomly remove during a training step?

That's stuff everyone should know.