How to use Categorical Cross Entropy for Multi-Label Classification?

Question

Say my target with classes A, B, C, D, E is [0, 1, 1, 0, 0].

And my output layer is of B x N where N is the number of classes.

How do I use Categorical Cross Entropy for this?

Answer 1

You want to treat the problem as binary classification, and of course use binary cross-entropy.

The only requirement is to ensure to use a sigmoid activation in your model's last layer.