Using genetic algorithm to minimize standard deviation

Hello,

I am trying to optimize a function and its standard deviation, when the input variables are changing a little bit (Monte Carlo).

function out = fitnessfcn(Y)
disp('MonteCarlo')
X_MCS = MonteCarlo(Y); 
t = @(x) x(1)^X(2)-sin(x(2)); %just an example function

temp = zeros(length(X_MCS),1);
for i=1:length(X_MCS)
    temp(i,1) = t(X_MCS(i,:));
end

out = [std(temp), t(Y)];
end

[solution,Fval] = gamultiobj(@fitnessfcn,n,A,b,[],[],lb,ub,@(X)nonlin(X),optsmulti);

So I thought now in every generation the whole population runs through my Monte Carlo simulation. PopulationSize is 50, so I should see 50 times my disp 'Monte Carlo' in the command window before a new generations starts, but it looks like this:

(5x) MonteCarlo
Generation   Func-count    Pareto distance    Pareto spread
    1            51                 1                 1
(2x) MonteCarlo
    2           101                 0                 1
(13x) MonteCarlo
    3           151                 0                 1

How can I achieve that the whole population runs through my fitnessfcn, not just some random parts?

Answer 1

[Matt J]

The fitnesfcn will be called only on those population members that satisfy your constraints. That is undoubtedly why you are seeing only a subset of your population processed. But isn't that what you want?

[Stephan Freitag]

I didn't think about this, you are so right, thank you!!