Basically I'm trying to get an array of 1s and 0s as an output from my sequential keras model. This is so that I can unflatten the array into a one-hot encoded array. I assume the best way to do this would be to code a binary step function as a custom activation function. The layer before has a tanh activation but I could change it to a sigmoid activation.
I tried but I just couldn't code the binary step function and I have no other clue as to what to try. Please help...
