Linear constrains meet problem when using parallel GA optimization

zexi wang

2020 6 12

0 个回答

7 次查看（30 天）

0 个投票

I was using following code to do a parallel GA optimization, and it works fine without setting A and b, which is the linear constrains. But when I add A and B, the variable chorm will be 0xn (n is the number of design variables). It's quite wired and I don't know why. Could any one help me with it? Many thanks!

spmd
  cd(sprintf('%d', labindex));
end
options = gaoptimset('Generations', control_param.generation_number,... %       
    'PopulationSize', control_param.population_size,... % 
    'UseParallel', true,... % 
    'Vectorized', 'on'); %     
[bestchrom,~,~,~,~,~] = ga(@FitnessFcn, nVars, A,b,[],[],Lb,Ub,[],IntCon, options);
function FitVal = FitnessFcn(chrom)
    parfor i = 1:size(chrom,1)
        FitVal(i,1) = fitness_parallel_f2(chrom(i,:)); %  
    end
end

3 个评论
显示 1更早的评论隐藏 1更早的评论

zexi wang 2020-6-13

Well, the problem now is that the variable 'chrom' is totally empty, but when I do this in a non-parallel version, it's fine. So I think I need to change something when I want to set up this to a parallel version. But I don't know what to do. Maybe I need to change the formulation of A and b.

zexi wang 2020-6-13

It seems the problem caused by using 'Vectorized', 'on' and linear constrains at same time. But I need to do a parallel optimization, without 'Vectorized', 'on' parameter, I don't know how to achieve the parallel computation.

请先登录，再进行评论。

请先登录，再回答此问题。

Follow Question