is it technically wrong to use simple accuracy in keras model metrics for multi-class classification? should we use categoricalaccuracy?

You should use CategoricalAccuracy. It works with one-hot labels and predicted logits of classes.

I think that you should use CategoricalCrossentropy as loss function. BinaryCrossentropy is used when you have two classes.

CLICK HERE to find out more related problems solutions.

Leave a Comment

Your email address will not be published.

Scroll to Top