Where is softmax used?
First, softmax is used in the classification model, whether it is deep learning or machine learning.
Is the last layer of multi-class activation function! Remember, it is the activation function!
Logistic regression relationship
Also talk about activation functions Sigmoid, Tanh, ReLu, softplus, softmax
The output of multiple neurons is mapped into the (0,1) interval, which can be understood as a probability, so as to carry out multi-classification!
Why is it called softmax?
First look at what does max mean? We know max, if I have two numbers, a and b, and a> b, if max is taken, then just take a, there is no second possibility
When we improve the categorized Loss, we have to use gradient descent to optimize one step size gradient at a time . At this time, we require Loss to partial derivate each weight matrix, and then apply the chain rule . Then the first step in this process is to pass back the softmax derivation. Don't worry, I will give a very detailed explanation with examples later. In this process, you will find that after using the softmax function, the gradient derivation process is very very convenient!
