Fmincon fobj is increasing. All the iteration are feasible.

Hi to everyone,

I'm trying to solve a non linear opt problem with fmincon.

After some iteration the optimizer keeps going up in terms of fobj value. All the constraints are satisfied.

How is this possible?

Also, I have 2009 variables to optimize, and severals non linear constraints. Is there a different solver to use in order to speed up the convergence? Both

@fobj and @nlcon are at their minimun in terms of time.

Thanks in advance

Answer 1

[Pranav Verma]

Hi Luigi,

From your question I understand that the objective function is not converging. I would suggest you to check if you have formulated your problem in the standard format as expected by fmincon.

If your objective function is sum of least squares, I would suggest you to use lsqnonlin instead of fmincon.

Also please refer to the below discussions threads on the similar lines:

https://www.mathworks.com/matlabcentral/answers/577537-fmincon-does-not-converge-to-the-best-solution

https://www.mathworks.com/matlabcentral/answers/358500-fmincon-with-nonlcon-does-not-converge-to-optimum-value

You can try to change the algorithm in the fmincon. To choose the algorithm, refer to this link.

In case you want to opt for other solvers, refer to this link which mentions all the solvers which expect non linear constraints.

Thanks

[Luigi Emanuel di Grazia]

Thanks for your help,

The problem is well formulated.

Yes I'm using a sum of least squares as fobj, I'll try as you suggested.

Thanks also for all the previous discussions you linked.