Is there an easy way to perform a Gamma Test within the Neural Network Toolbox? Or are there other ways to determine significant inputs for neural networks within the toolbox?

Hello everyone!

I am still pretty new to matlab and neural networks but so far I have succeeded in creating a code for a neural network within the neural network toolbox. Now I want to determine which inputs I want to include in the neural network and which inputs are not significant. I read that it is possible to perform a Gamma Test to determine the best inputs for a neural network. Is there any possibility or possibly an easy way to perform a Gamma Test within the neural network toolbox or in any other toolboxes? If not, do you know of any other easy ways to determine the best inputs for my neural network within matlab??

Thank you a lot!!

Answer 1

[Greg Heath]

I have never heard of using the gamma test to rank NN inputs.

The most effective methods I have used are a non-optimal succession of backwards stepwise searches with linear, quadratic and minimum-hidden-node NN models using standardized (zero-mean/unit-variance) variables.

In the quadratic case, products and squares are considered as separate variables.

Hope this helps.

Thank you for formally accepting my answer

Greg