You will find many rules of thumb to say exactly how many layers or
units to put into a network. Don't believe them. The best answer is to
search the parameter space and use cross-validation. A helpful
discussion is at
On Tuesday, June 3, 2003, at 12:33 PM, Alberto Vera wrote:
> Could you tell me about any method to answer this questions?
> a) How many hidden layers do I set in a neural network?
> b) How many neurons do I set for each hidden layer ?
> c) how many layers is it better to set in a neural network?