I’ve been working on my Zero to GANS final project and I am not able to find a suitable loss function, what would you suggest for this task? Also what extra can I do to make the model better?
Neural networks can decrease the loss of this model, i.e adding extra layers, As for the loss function, I think cross_entropy/binary_cross_entropy works better in the case where you have to predict among a set of classes. But that is where trial and error comes into play, you can try different layers of the model, also different loss function, and get an idea which works better, also I would suggest adding an accuracy tester with the validation loss it gives a better evaluation of the model.
I’ll try this stuff out