How to use fmincon's constraintviolation option

Hello there,

I'm trying to use fmincon for an optimization problem.

I use Nonlinear constraints and my function returns a ceq value. This however is not "0" it is 0.0014693, hence fmincon will give me an exitflag of -2 (no feasible point was found). I then try to use the 'ConstraintTolerance' option (which by default is 1e-6) and put it to 0.01. However, still no feasible solution could be found (now the ceq value is 0.0012396).

Can you tell me what I'm doing wrong?

Answer 1

[Torsten]

Can you tell me what I'm doing wrong?

You try to betray yourself that your problem has a solution. But it has none.

[xone92_]

Hello there,

that's true it has none with this contraints. But actually the solution which gives me ceq=0.0014693 is a good solution for me. I just want to increase the Tolerance in a way that fmincon gives me back a 1 or 2 as exitflag. In fact if i set 'ConstraintTolerance' to 0.1 I get what i want. But fot this reason i don't understand the 'ConstraintTolerance' option, that's why I'm writing here

To clarify my question from above. The ceq constraints has some Tolerance, since its hard to have a solution ceq==0. I though it is adjusted by the 'ConstraintTolerance' option. If I have for example the case as above ceq=0.0014693 i want this still to be a valid tolerance.

[Torsten]

ConstraintTolerance is defined as:

ConstraintTolerance is an upper bound on the magnitude of any constraint functions. If a solver returns a point x with c(x) > ConstraintTolerance or |ceq(x)| > ConstraintTolerance, the solver reports that the constraints are violated at x. ConstraintTolerance can also be a relative bound. See Tolerance Details.

We cannot tell you more than this.

To diagnose why you get an exitflag of -2 although you set the constraint tolerance to 0.01 and the real error in the constraint is only 0.0012396 - we are not able to without your code. Maybe another constraint is now violated. Or the solver takes the relative error instead of the absolute error.