Regularization Checkpoint

1. Overfitting means the model:

Memorizes training data, fails on new data Performs poorly on all data Is too simple Needs more parameters

2. L2 regularization adds to the loss:

Sum of squared weights Sum of absolute weights Number of weights Maximum weight

3. L1 regularization tends to produce:

Sparse weights (some exactly zero) All small weights Large weights Negative weights only

4. Name one technique besides regularization to prevent overfitting.

5. The regularization strength (lambda) that is too high causes:

Underfitting Overfitting Perfect fit No change