[Scip] unresolved numerical troubles in LP

[Timo Berthold]

Hi Luca,

this warning can be seen frequently in numerically challenging problems
(as LP relaxations of nonconvex MINLPs often tend to be) and does not
necessarily indicate a problem. It basically means that the LP relaxation
at that node could not be solved to proven optimality (within required
tolerances) and SCIP thus has to continue without further cutting (at that
node) and by branching without LP information. The hope is that the
troubles get resolved in subsequent nodes. This may lead to bigger search
trees but not to wrong results.

If this warning comes without further errors or unintended behavior (like
the solver hanging after this warning), you can ignore it (or improve the
conditioning of your model, if possible).

Hope that helps.

Kind regards,
Timo

> Dear Scip community,
>
> I'm solving a nonconvex minlp, using Scip through ampl and at some point
> I've have the following problem:
>
> (node 4710) unresolved numerical troubles in LP 15734
>
> Best,
> Luca
>
> --
> "A mathematician is a machine for turning coffee into theorems."
> Alfréd Rényi (erroneously attributed to Paul Erd?s)
> ---------------------------------------------------------
> Luca Mencarelli
> PhD Candidate at LIX, Ecole Polytechnique
> 91128 Palaiseau CEDEX, France
> Office: 2072 (Bâtiment Alan Turing, second floor)
> LIX e-mail address: mencarelli at lix.polytechnique.fr
> INRIA e-mail address:  luca.mencarelli at inria.fr
>
>
>
>
>
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> http://listserv.zib.de/mailman/listinfo/scip
>