# Regularization and generalization

Subject: Deep Learning

Exam: MT2

Reader: https://deep-learning-study-room.vercel.app/#mt2/guide/6

Canonical: https://deep-learning-study-room.vercel.app/read/dl/mt2/guide/6/

1. [Recognizing overfitting and underfitting](https://deep-learning-study-room.vercel.app/read/dl/mt2/section/mt2-c6-s0/)
2. [L1 and L2 add a cost for large weights](https://deep-learning-study-room.vercel.app/read/dl/mt2/section/mt2-c6-s1/)
3. [Model answer: regularization and the L1–L2 difference](https://deep-learning-study-room.vercel.app/read/dl/mt2/section/mt2-c6-s2/)
4. [Dropout removes activations during training](https://deep-learning-study-room.vercel.app/read/dl/mt2/section/mt2-c6-s3/)
5. [Early stopping uses the best validation point](https://deep-learning-study-room.vercel.app/read/dl/mt2/section/mt2-c6-s4/)
6. [Model answer: generalization and early stopping](https://deep-learning-study-room.vercel.app/read/dl/mt2/section/mt2-c6-s5/)
7. [Choosing a remedy](https://deep-learning-study-room.vercel.app/read/dl/mt2/section/mt2-c6-s6/)
8. [Inverted dropout during training and inference](https://deep-learning-study-room.vercel.app/read/dl/mt2/section/mt2-c6-s7/)
9. [Model answer: dropout with training and inference explained](https://deep-learning-study-room.vercel.app/read/dl/mt2/section/mt2-c6-s8/)