How to design a nonlinear fitting neural networks with inequility constraints using MATLAB ANN Toolbox ?

I have some input and output samples, X and Y, and I want to model the relationship between them using ANN. I have two inputs x1 x2, and they have roughly opposite relationship with the output Y. That is Y=F(X), and F(x1>0,x2=0)>0, and F(x1=0,x2>0)<0. Though I know this inequlity constraint, I cannot get real samples like (x1>0,x2=0) or (x1=0,x2>0), is there some tools in ANN Toolbox to deal with this problem? Thank you!

Answer 1

[Greg Heath]

NN design is based on a large sample of I/O pairs of data.

If you neither have, nor can generate, representative I/O pairs I don't see how you can design a net to do what you want.

Thank you for formally accepting my answer

Greg