[Scip] early branching?
Timo Berthold
berthold at zib.de
Wed Jan 14 19:46:19 MET 2009
Hi Markus.
Unfortunately, it is currently not supported to stop pricing without solving
the LP to optimality.
However, we are currently working on integrating this support (and setting own
lower bounds while pricing) and we will probably include this feature in the
next release.
All the best,
Kati and Timo
Am Dienstag 13 Januar 2009 13:27:04 schrieb Markus Leitner:
> Hi all,
>
> Since I am experiencing a significant tail-off effect in my B&P application
> I wanted to use early branching, i.e. stopping column generation even if
> columns with negative reduced costs do still exist. However, if I branch
> due to stagnation I cannot be sure if the correct LP value has already been
> found. It might even happen (and it really does happen in my case) that the
> current LP value is larger than the optimal integer solution leading to
> wrong pruning of nodes and therefore a non-optimal solution at the end.
>
> However, I would like to ask if SCIP supports early branching by means of
> either having some parameters indicating that the lower bound found is not
> necessarily a real lower bound or if there is any other way to avoid wrong
> pruning of nodes. Is it for example possible to manually set the current
> lower bound to some value which is for sure below the LP value of my
> formulation - even if this value cannot be found in the current restricted
> master problem due to missing columns? Will SCIP automatically overwrite
> this manually introduced lower bound?
>
> Any ideas / suggestions on this topic are highly appreciated!
>
> Best Regards,
> Markus Leitner
>
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> http://listserv.zib.de/mailman/listinfo/scip
More information about the Scip
mailing list