How could I optimize structure of ANN using GA?

I have written matlab code as follows:

inputs = X(:,:)'; five neurons for input targets = y(:,:)'; one neuron for output

net=newff(minmax(inputs),targets, (Number of Neurons in Hidden Layer),{'tansig','purelin'},'trainlm');

[net,tr]=train(net,inputs,targets); and ........

I want to find the number of neurons in the hidden layers. I have always used trail an errors method to find the must appropriate numbers of neurons. I wonder if any one could help me to find this number using GA.

Cheers Saman