Breaking the Conventional Forward-Backward Tie in Neural Networks: Activation Functions

Open in new window