dvlab-research / LBGAT

Learnable Boundary Guided Adversarial Training (ICCV2021)
https://arxiv.org/abs/2011.11164
MIT License
33 stars 2 forks source link

Do you use ReLu instead of Softmax? #12

Open headscott opened 5 days ago

headscott commented 5 days ago

In your paper I saw, that you used Softmax, but I can't find it in your code. I only see ReLu. Did I unsterstand that wrong, or don't you even use Softmax?

headscott commented 5 days ago

Oh wait, an Update: I found Softmax. But you only use it for loss functions, right? Not as activation function?