math - Neural Network 0 vs -1 -
I use -1
for 0
input data How is this better when working with the neural network and it implements any math to implement it?
Edit: Using Feedword and Backprop
Edit2: I let go of it but stop learning the network So, I think mathematics has to change somewhere?
Edit 3: Finally got the answer. Mathematics for binary is different from bilateral.
Recently found that sigmoid and sigmoid-derived derivatives while using bipolar at biogas Need to change the formula.
Bipolar seamless function: f (x) = -1 + 2 / (1 + e ^ -x)
Bipolar sigmoid derivative: F '(x) = 0.5 * (1 + f (x)) * (1 - f (x))
Comments
Post a Comment