机器学习入门理论基础|莫烦

自负 = 过拟合

方法一增加数据量

方法二正规化

cost = （Wx - real y) 2 

L1 regularization

cost = （Wx - real y) 2 + abs(W)

L2 regularization

cost = （Wx - real y) 2 + (W) 2

L3、L4 regularization

Dropout regularization

Overfitting 过拟合

L1 regularization:

L2 regularization:

Dropout reularization: