anny
anny copied to clipboard
Normalize training data before training
As of #83, activation ranges are now specified for all activation functions. Using this we can now scale and translate the training data to the optimal range for the activation functions being used in the network.