[Scip] How good are SCIP solutions?
Stefan Vigerske
stefan at math.hu-berlin.de
Sat Nov 24 14:59:43 MET 2012
Hi,
as long as solutions are feasible, the one with the smaller objective
value are better.
The points provided in MINLPLib should all be feasible. For easy and
medium difficult instances, they correspond to an optimal solution. For
difficult ones, they should correspond at least to a good feasible one.
Solutions you get from SCIP are usually feasible w.r.t. tolerances. In
the case of MINLP, we sometimes observed violations of nonlinear
constraints that are slightly above SCIP's feasiblity tolerance. We are
working on fixing this problem.
Since we have run SCIP also on MINLPLib instances and submitted
improving solutions to this library, it should be hard to find solutions
for MINLPLib instances with SCIP that have a smaller objective value
than the one given at the library.
Best,
Stefan
On 11/24/2012 02:39 PM, Liffey1986 at gmx.de wrote:
> Dear All,
>
> I'm using SCIP to solve MINLPs. I tried to solve some instances from libraries such as the MINLPLib.
> In these libraries the best solutions are usually given. But how do I know which solution is the better one, the given solution or the solution I got from solving the instance with SCIP? Suppose I have a minimization problem, can I then always assume that the smaller objective value is the better one?
>
> Regards,
> Ina
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> http://listserv.zib.de/mailman/listinfo/scip
>
More information about the Scip
mailing list