VI Lezione Koroliouk - Introduction into Neural Networks and Deep Learnig

Lecture 6. Deep learning. Regularization
Example: handwritten digit recognition, 2nd solution. Augmentaiton. VGG-16 neural network architecture. Regularization, its purpose. Regularization in linear regression analysis. Normal equations of linear regression analysis. Adding a regularization term to normal equations. The special role of a free member. Example: polynomial approximation of points. Sample validation. Variants of the regularization term (ridge regression, lasso, elastic net). Why Lasso allows you to reduce the number of predictors.