Stop mutation in GAMULTIOBj

Hi, I tried this simple problem with an initial population. I want to find the best set of combinations of these variables without altering their values. This I'm trying to get by having only crossover children. As I understand from the documentation this can be done by setting crossoverfraction to 1. But I don't get the expected result.

e.g.

    P_s=randi(20,3,9);
    numberOfVariables=9;
    FitFunc = @multiobjectiveFit;
    options = gaoptimset('CrossoverFraction',1,'InitialPopulation',P_s,'CrossoverFcn',@crossovertwopoint)
    [x,fval,exitflag]=gamultiobj(FitFunc,numberOfVariables,[],[],[],[],[],[],options);

function y=multiobjectiveFit(x)

y = zeros(2,1);   
y(1) =-sum(x);
y(2) =var(x);

gives me a new set of populations x with non-integer values whereas I started with integer values. Please help me. (FYI: My original problem has a very large initial population with non-integer values)

[Dilini]

May be I mis-interpreted the document. In crossover it doesn't simply swap the elements it combines them whereas in mutation it alters the value, isn't it? May be I should write my own crossover function?

Answer 1

[Zhangxi Feng]

Sad no one answers this question. I am trying to understand the whole process more also. I believe making your own creation, selection, crossover, and mutation functions are definitely the way to go if you have a specific need.