Call GA in a loop to customize stop criteria

Hi:

I'm trying to run a GA optimization and I want to stop the GA when a mean fitness for the last population is met.

I don't see how to use such a custom stop criteria, so I'm thinking I could do this:

1. Generate a first random population

while stopFuncCriteria is FALSE 2. Call GA for just 1 generation using the previous population as a start point 3. Evaluate stopFuncCriteria

end

Is this possible? Will it work the way I want?

Thank you.

Answer 1

[Alan Weiss]

If you want to create your own stopping criterion, I suggest you disable the existing criteria and insert your own via an output function.

 opts = gaoptimset('Generations',Inf,'StallGenLimit',Inf,'TolFun',0); % disable
 opts = gaoptimset(opts,'OutputFcns',@yourOutputFcnNameHere);

You have to write yourOutputFcnNameHere.m first, of course. You might want to set the Generations limit lower while you are testing your output function, and you might want to enable iterative display or a plot function so you can keep track of progress.

Note that the way to stop an optimization in the output function is to set state.StopFlag to a nonempty string.

Alan Weiss

MATLAB mathematical toolbox documentation