Regularization adds a penalty to the model's loss function to prevent overfitting by discouraging overly complex models (e.g., L1 and L2 regularization).
Key concept to improve model generalization.
Used in almost all modern ML configurations.