Interior-point and sequential quadratic programming give me the same answer. Is there an error with my code?

2 次查看（过去 30 天）
Aisel
Aisel 2023-11-13
编辑： Aisel 2023-11-13
I am attempting to minimise a least squares regression model and compare answers given by SQP and IP, and the system is constrained by x(1) + x(2) + x(3) = 1, and each variable cannot be smaller than 0 or bigger than 1. However, I consistently get the same numerical answers for both methods, which should not be the case. I am wondering if I missed a step or something along those lines? My code is as follows;
fun=@(x) (x(1)-147/505).^2+(x(1)-167/530).^2+(x(2)-504/2525).^2+(x(2)-471/2120).^2+(x(3)-1011/2120).^2+(x(3)-1256/2525).^2;
x0=[0.3, 0.5, 0.2];
Aeq=[1, 1, 1];
beq=1;
lb=[0, 0, 0];
ub=[1, 1, 1];
IP=fmincon(fun,x0,[],[],Aeq,beq,lb,ub);
fprintf('%.13f\n',IP)
options = optimoptions('fmincon','Algorithm','sqp');
SQP=fmincon(fun,x0,[],[],Aeq,beq,lb,ub);
fprintf('%.13f\n',SQP)
These are my results;
IP:
0.3027134331491
0.2105086010671
0.4867779657839
SQP:
0.3027134331491
0.2105086010671
0.4867779657839
Any help or pointers would be appreciated. Thank you in advance.
  4 个评论
显示 2更早的评论
Torsten
Torsten 2023-11-13
Maybe one needs more iterations with one of the methods (use 'Display','iter' in the options structure). But if you prescribe the accuracy of the solution to be equal for both methods (as you implicitly do in your code because you didn't change them in the optimoptions structure), in theory both solutions should be equally precise, shouldn't they ?

请先登录，再进行评论。

请先登录，再回答此问题。

采纳的回答

Torsten
Torsten 2023-11-13
编辑：Torsten 2023-11-13
Ran in:
I just saw in your code that you don't pass the options structure to fmincon. Thus in both calls, the same solver is used.
Use
fun=@(x) (x(1)-147/505).^2+(x(1)-167/530).^2+(x(2)-504/2525).^2+(x(2)-471/2120).^2+(x(3)-1011/2120).^2+(x(3)-1256/2525).^2;
x0=[0.3, 0.5, 0.2];
Aeq=[1, 1, 1];
beq=1;
lb=[0, 0, 0];
ub=[1, 1, 1];
options = optimoptions('fmincon','Algorithm','interior-point','Display','iter');
IP=fmincon(fun,x0,[],[],Aeq,beq,lb,ub,[],options);
First-order Norm of Iter F-count f(x) Feasibility optimality step 0 4 3.328630e-01 0.000e+00 1.455e+00 1 8 9.097657e-02 0.000e+00 1.369e+00 6.098e-01 2 12 2.579621e-03 0.000e+00 9.931e-02 1.857e-01 3 16 1.181553e-03 0.000e+00 7.284e-02 3.823e-02 4 20 1.098100e-03 0.000e+00 4.139e-02 2.558e-02 5 24 7.552129e-04 0.000e+00 2.454e-03 1.287e-02 6 28 7.552951e-04 0.000e+00 1.000e-03 8.613e-04 7 32 7.545440e-04 0.000e+00 2.015e-04 5.086e-04 8 36 7.545172e-04 0.000e+00 2.077e-06 1.144e-04 9 40 7.545172e-04 0.000e+00 2.001e-08 1.263e-06 Local minimum found that satisfies the constraints. Optimization completed because the objective function is non-decreasing in feasible directions, to within the value of the optimality tolerance, and constraints are satisfied to within the value of the constraint tolerance.
fprintf('%.13f\n',IP)
0.3027134331492 0.2105086010669 0.4867779657839
options = optimoptions('fmincon','Algorithm','sqp','Display','iter');
SQP=fmincon(fun,x0,[],[],Aeq,beq,lb,ub,[],options);
Iter Func-count Fval Feasibility Step Length Norm of First-order step optimality 0 4 3.328630e-01 0.000e+00 1.000e+00 0.000e+00 1.156e+00 1 9 2.478717e-01 1.110e-16 7.000e-01 6.930e-01 1.335e+00 2 13 2.415543e-01 0.000e+00 1.000e+00 6.044e-01 1.139e+00 3 17 5.519559e-02 1.110e-16 1.000e+00 4.532e-01 5.134e-01 4 21 4.792452e-03 0.000e+00 1.000e+00 1.976e-01 1.459e-01 5 25 7.959527e-04 1.110e-16 1.000e+00 4.792e-02 1.485e-02 6 29 7.546824e-04 2.220e-16 1.000e+00 4.766e-03 8.850e-04 7 33 7.545173e-04 2.220e-16 1.000e+00 2.919e-04 1.947e-05 8 37 7.545172e-04 2.220e-16 1.000e+00 6.953e-06 3.359e-07 Local minimum found that satisfies the constraints. Optimization completed because the objective function is non-decreasing in feasible directions, to within the value of the optimality tolerance, and constraints are satisfied to within the value of the constraint tolerance.
fprintf('%.13f\n',SQP)
0.3027134886159 0.2105086203988 0.4867778909853

更多回答（0 个）

类别

在 Help Center 和 File Exchange 中查找有关 Solver Outputs and Iterative Display 的更多信息

标签

产品


版本

R2022b

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by