Classification(李宏毅课程)

Classification as Regression

if class 1/2/3 have some relations, we can use this method. ex: age and grade

Class as one-hot vector

two-dimension softmax is same as sigmoid

Loss of Classification

cross-entropy is more useful for classification