Why fmincon cannot find the local minumum back when it is supplied as initial point

Hi all,

I'm using fmincon to solve a contrained optimization problem. Fmincon finds different local minima for different initial points, which is normal given that it's not a global technique. However, I'm confused why it cannot find the same local optimum once I supply it as the inital point. It says it converged to an infeasible point, with exitflag -2, which is the local optimum that it found before. Any idea why?

Many thanks!

Answer 1

[Nachiket Katakkar]

The error message indicates that the given set of constraints are impossible to solve due to an inconsistency. To troubleshoot this, it is helpful to remove these constraints one at a time to get a general idea of where the problem may originate.

Some other things to check:

1. If the local optimum (supplied by the first optimization) is actually a feasible point according to the set of constraints
2. The number of iterations does fmincon runs in both cases. You may need to increase the number of iterations.
3. The tolerance values might need to be changed using optimoptions

Have you considered global searching algorithms like "ga" or "patternsearch", considering that the problem you wish to solve is non-convex?