Restricting members to choose in population selection

Hello all,

I'd like to be able to define the list of possibilities that my population can be selected from.

For example, if using the optimization tool, I can define my bounds to be [1,1] to [400,400] and in this case any value is chosen between this range on each iteration of the ga.

If I want to constraint this as follows so that I only use steps of 5 from the population. How can I do this. Here is what I mean by steps.

popsize = 20;
steps = [1;(5:5:400)';]
[X,Y] = ndgrid(steps,steps);
A = [X(:),Y(:)];
A(randperm(length(A),popsize),:)

Answer 1

[Alan Weiss]

It sounds to me as if you want to restrict your population to be multiples of 5, from 5 to 400. You can do this by restricting your population to be integers from 1 to 400/5=80, and scaling your results afterwards.

To restrict your population to integers from 1 to 80, see the documentation of mixed integer programming in ga.

Alan Weiss

MATLAB mathematical toolbox documentation

[David]

I see what you are saying. Thanks. I assume one could also do:

function [c, ceq] = simple_constraint(x)
    c = [mod(x(:,1),5) mod(x(:,2),5)];
    ceq = [];

[Alan Weiss]

I would not try to use a nonlinear constraint to enforce integrality. If it were that easy, we would not have made a special integer solver!

Please try the integer method before wasting time on a nonlinear constraint method.

Alan Weiss

MATLAB mathematical toolbox documentation