[SciPy-user] Problem with NLP in OpenOpt

Dominique Orban dominique.orban@gmail....
Thu Nov 15 10:54:44 CST 2007


On 11/15/07, Dark Wind <darkwind.87@gmail.com> wrote:
> The initial value x0 satisfies the constraints hi's but i am still
> getting that there is no feasible solution.
>
> Can anyone help me with this?

Cobyla is an infeasible method, meaning that in order to identifiy a
solution to your problem, it will venture outside of the feasible set.
In your example, it wasn't able to reach a feasible solution within
the given limits (max number of iterations, tolerances, ...) It is not
because their is a feasible point that the code will be able to find a
solution; in this case, it got stuck before reaching the feasible set
again.

You can either re-solve your problem with looser tolerances, which is
not very satisfactory, you can re-solve from a different starting
point (which doesn't have to be feasible), or you can try another
solver (as Dmitrey showed in his response).

Dominique


More information about the SciPy-user mailing list