is using the fmin functions from scipy optimize a good way to train neural nets. And in neural nets, why are multipliers used in each weight of the perceptron - has anyone tried powers or more complicated thing on each weight of the perceptron?
[link] [10 comments]