How to select inputs for my neural network

Hi,

I have, currently, 36 inputs for my neural network that detects if a movie is violent. I would like to know which of the inputs have more weight for the problem. Do you know any technic or algorithm to do that?

Thanks in advance

Answer 1

[Greg Heath]

1. Standardize inputs to zero-mean/unit-variance

 xn = zscore(x);

You can rank inputs by replacing, one at a time, input rows of length N with one of the following:

 xn(i,:) = xn(i, randperm(N));% Scrambled
 input(i,:) = randn(1,N);% Random, unit-variance
 input(i,:) = zeros(1,N);% Zero mean and variance

You can also get fancy by using one of the above in a sequential backward or forward search

In general, all will yield different rankings. However, the exceedingly weak and exceedingly strong should reveal themselves.

Hope this helps

Thank you for formally accepting my answer

Greg