[Scip] questions on GUB cut of SCIP

[Tobias Achterberg]

Hi Jinghui,

sorry to answer so late, I am really busy these days...

As far as I know, there is still no GUB cover cut in SCIP. There is a corresponding code
in the presolving methods of the knapsack constraint hander, but not in the separation
procedures.

The way to integrate this separator would be to extend the cut separation procedures in
cons_knapsack.c. You should ask the SCIP development team (which I am no longer a member
of), in this case Kati Wolter (wolter at zib.de), how to proceed.

Regarding your second question, the answer is "yes". It should be pretty easy to include
your primal heuristic. I don't know exactly what you mean with "variable reduction".
Variable selection rules for branching as well as presolving methods can be added easily
to SCIP.


Tobias


Jinghui Zhang wrote:
> Dear Dr Tobias,
> 
> Sorry for distrubing you for questions on SCIP.
> 
> I am gonna implement a customized branch and cut algorithm based on the
> existing branch and cut framework implementations. After reading a few
> survey papers and relevant working papers by you and your colleagues at
> ZIB, I realize SCIP probably will be a good candidate
> for adpating my customized branch and cut procedure.
> 
> However, since my branch and cut procedure relies heavily on Generalized
> Upper Bound cut, which seems currently not found in SCIP framework, I am curious about that how GUB cut
> procedure can be integrated into SCIP. As far as I know, in open
> source software Minto and commercial software CPLEX, XPRESS-MP, GUB cut
> is implemented and adopted as default cut generation options.
> 
> To summrize, my questions are as follows:
> 1) is there a default GUB cut procedure within SCIP? If not, what's
> the easiest way to integrate GUB cut into SCIP.
> 2) is it easy to adapt my customized primal heuristics and variable reduction into SCIP?
> 
> I am looking forward for your reply!
> 
> Cheers,
> Jinghui
>