I’m using multi class classification but for some instances choosing one class is a “bad” decision than other (e.g., there are weights per-instance-per class for “how bad” is this class).
I’m currently using cross entropy loss, is there a better loss function for it?