I have never heard of using a neural network as a fitness function. This wording implies that the network is fully trained and implements the estimate of a nonnegative function of the inputs that will be minimized for one or more optimal subsets of inputs.
Or are you just trying to use GA to design a net by finding one of the optimal sets of weights?
If so, the NNTBX does not have that option. You will have to design your own code which, ideally, will include GA functions of other Toolboxes. To find these type
lookfor genetic
and
ver
Hope this helps.
Thank you for formally accepting my answer
Greg
