How to change Crossover Function in GA midway through solving?

Hi all,

I have a GA setup to search for a minimum of a problem. I think my problem is well defined, except it has many local optima. I read in the documentation that changing the Crossover and Mutation Functions midway through iterations can be very useful in aiding of finding global optima. I understand it needs to be modified by the 'OutputFcn' settings, but all the documentation I see include cases where multiple things are being changed. Say if I just want to change the Crossover from 0.2 at the beginning to 0.8 after a certain number of iterations, what would the script for OutputFcn look like?

All the best.

Answer 1

[Alan Weiss]

I think that this will work, but I have not tested it, so use at your own risk.

Suppose that you want to change the crossover probability at generation 27.

function [state,options,optchanged] = gaoutfun(options,state,flag)
optchanged = false;
if state.Generation == 27
    options.CrossoverFraction = 0.8;
    optchanged = true;
end

Pass @gaoutfun as the OutputFcn option (I assume that you use optimoptions to set your options).

Alan Weiss

MATLAB mathematical toolbox documentation