When solving the global optimal solution with MATLAB fmincon function, is it to find several local optimal solutions and then compare them to choose the optimal solution?

When solving the global optimal solution with MATLAB fmincon function, is it to find several local optimal solutions and then compare them to choose the optimal solution?

Answer 1

[Walter Roberson]

fmincon() does not do that automatically. fmincon() stops at the first local minima it finds that satisfies the constraints -- but the process of determining whether it is a local minima involves probing other locations to see if it can find an improvement.

If you need to have several local minima compared automatically then you would use the Global Optimization Toolbox, probably using multistart() or similar.

[jessie sun]

I used the following command：

gs = GlobalSearch;

problem = createOptimProblem('fmincon','x0',dl_power,'objective',@fun,'Aineq',A,'bineq',b,'Aeq',Aeq,'beq',beq,'lb',lb,'ub',ub,'nonlcon',@mycon,'options',options);

[X,fval] = run(gs,problem)；

Can I automatically compare several local minima to obtain the global optimal solution?

[Walter Roberson]

Yes, that code would automatically generate multiple starting points and would compare the results to select the best of them.

[jessie sun]

OK,thanks for your answer.