How i can configure a neural network with hidden layer and output layer function transfer logsig?

I have to design a neural network with n input and one output. The value input are between zero and 10000 and the output limits are zero or one.

Answer 1

[Greg Heath]

Use patternnet for classifying into two categories. It automatically normalizes input and output to [-1,1] and uses tansig hidden and output transfer functions. Then it will unnormalize the output to your original scaling.

This is optimal for your data.

help patternnet

doc patternnet

All you have to do is

1. Determine a reasonable value for the number of hidden nodes, H, if the default of H = 10 is unreasonable.

2. Find a good starting set of initial weights. I tend to use 10 trials of weight initializations for each candidate value of H.

3. You can find some of my double loop code by searching NEWSGROUP and ANSWERS using the keywords

greg Hub Ntrials

Hope this helps.

Thank you for formally accepting my answer

Greg