Provide gradient for Fmincon

Question

Dat Tran 2016-2-16

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/268458-provide-gradient-for-fmincon

回答： Walter Roberson 2016-2-16

在 MATLAB Online 中打开

Dear all,

How can I include Gradient for Objective Function in code below?

Thank you so much for your help!

Dat

 function [p,fval] = MC_NT(p0,Aeq,beq,N)
    if nargin < 5
        opts = optimoptions('fmincon','Algorithm','interior-point');
    end
    M=length(p0);
    pr1=[0.3185  0.0001  0.1574  0.2902  0.0003  0.00001  0.8426  0.7098  0.6804  0.0822  0.00001  0.00001  0.0008  0.9177  0.00001  0.00001];
    pr=horzcat(pr1,(0.25*ones(48,1)'))';
    p=nan(M,N);
    fval=nan(1,N);
   for i=1:N
    fun=@(p) sum(p.*log(p./pr));
    [p(:,i),fval(i)] = fmincon(fun,p0,[],[],Aeq,beq,[],[],[],opts);
    pr=p(:,i);
   end

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Walter Roberson 2016-2-16

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/268458-provide-gradient-for-fmincon#answer_210053

You would have to recode fun as a real function instead of anonymous function, as it is difficult for an anonymous function to return multiple outputs.

The function you provide would need to return the gradient in the second output.

You would need to specify the 'GradObj', 'on' option in your optimoptions()

There is an example showing a gradient calculation at http://www.mathworks.com/help/optim/ug/fmincon.html#busxd7j-1

As you have linear equality constraints but no bounds constraints and no non-linear constraints, you would have the option of switching to trust-region-reflective algorithm once you provide the gradient.