Regularization
Umberto Michelucci · Apress eBooks · 2022
This chapter explains a very important technique often used when training deep networks: regularization . We look at techniques such as the ℓ 1 and ℓ 2 methods , dropout, and early stopping. You learn how these methods help prevent the problem of overfitting and help you achieve much better results from your models when applied correctly. We look at the mathematics behind the methods and at how to implement them correctly in Python and Keras.