深度学习入门2——Linear & DeepLearning Intro

Revision

KNN

Difficiencies:

1. Train $O(1)$ Predict $O(n)$
2. ignore the position information, as well as the connection between pixels

K-fold validation

Linear Classifier

We need activation functions, and the most common one is ReLU.

Loss funtions

1. SVM loss: ${\sum }_{j\ne y_i}\max (0,s_j-s$