Suggestion-Box
Suggestion-Box copied to clipboard
Cross entropy loss implementation on the Toy NN
Currently, the Toy NN uses a very primitive loss function (Label - Output). Maybe an implementation of a categorical loss function could improve its performance?