Re: Overflow

"Krzysztof Kolago" <[EMAIL PROTECTED]> wrote in message 
news:<[EMAIL PROTECTED]>...
 
>     I prepare some neural networks using our program, but in my last
> projects I've to use very long list of training data. When I've add 504
> example for training there was an error: "exp: OVERFLOW error". What should
> I do, to skip this error?

I assume this is occuring in the activation calculation step. One
cure is to replace the theoretical tanh or logsig containing 
exponentials with a piecewise differentiable low order polynomial.
Other models are given in Tveter's home page.
 
Consider simple odd gain(g>0)/threshold(T>0) functions with 
properties 

1. f(x >= T) = 1,
2. f'(T)     = 0,
3. f'(0)     = g.

Let z = x/T and G = g*T. Then one example for |x| <= T is

f(z) = z*[G+(3-2*G)*z-(2-G)*z^2]

Hope this helps.

Greg