Loss function + data imbalance


(Ashwin) #1

I’m currently training a MLP for classification with weighted cross entropy loss layer and evaluating the model based on mean per class accuracy.

Is there a better loss function?