How to implement weighted cross entropy loss in Keras?

Question

I would like to know how to add in custom weights for the loss function in a binary or multiclass classifier in Keras. I am using binary_crossentropy or sparse_categorical_crossentropy as the baseline and I want to be able to choose what weight to give incorrect predictions for each class.

Answer 1

For multiple classes one should use not binary but categorical crossentropy.

Consider using custom loss function as described here: Custom loss function in Keras