01-basics - Logistic regression example

[int8]

hi there,

First of all, This is really good tutorial, thank you for your work!

I am having some troubles understanding Logistic Regression example here. What bothers me is lack of clear activation function at the very end of the graph. From what I see I conclude this is in fact softmax classifier rather than logistic regression. Am I missing smth?

[yunjey]

You can see here. nn.CrossEntropy computes softmax internally.This is same as when we use softmax with nn.NLLLoss.

[int8]

sure, I understand softmax exists there implicitly via error function formulation - what I do not see is logistic part of logistic regresstion

[yunjey]

logistic sigmoid is just special case of softmax (when the number of class is 2). In multi-class classification task, we use softmax instead of sigmoid. This will help you.

[int8]

thanks a lot - helped indeed, take care