Deep Learning · MT2 · guide

Regularization and generalization

1. Recognizing overfitting and underfitting
2. L1 and L2 add a cost for large weights
3. Model answer: regularization and the L1–L2 difference
4. Dropout removes activations during training
5. Early stopping uses the best validation point
6. Model answer: generalization and early stopping
7. Choosing a remedy
8. Inverted dropout during training and inference
9. Model answer: dropout with training and inference explained