[Scip] Cutpool

[michael.winkler@zib.de]

Hi Jonas,

you also need to disable propagation in the root node to derive only cuts
that do not cut off any feasible solution. This is done by "set
propagating maxroundsroot 0" in the interactive shell or the parameter
"propagating/maxroundsroot = 0". This only works for solving your subscip
until the end of the root node.

But, can you explain the idea behind only separating your sub-problem(s)
without solving them to optimality?

Best, Michael

> Hi Michael,
>
> is it possible to get only the cuts which are valid for all feasible
> points or to generate only these cuts?
>
> Best regards,
> Jonas
>
> Am 18.06.2013 23:38, schrieb michael.winkler at zib.de:
>> Hi Jonas,
>>
>> in your case the optimal solution is found and afterwards the cmir
>> seperator creates such a cut rendering the problem to be infeasible,
>> which
>> is correct.
>>
>> Best, Michael
>>
>>> I forgot to mention that I disabled presolving. Otherwise the problem
>>> is
>>> solved due to presolving.
>>>
>>> Best regards
>>> Jonas
>>>
>>> Am 18.06.2013 15:41, schrieb Jonas Witt:
>>>> Hi all,
>>>>
>>>> I use a sub-SCIP and get all cuts of the cutpool via SCIPgetPoolCuts()
>>>> to use these cuts for my SCIP-instance. The problem is that one of the
>>>> cuts in the cutpool seemes to be not valid for my sub-SCIP.
>>>>
>>>> The sub-SCIP is some kind of knapsack of the form max{c^T x: w^T x <=
>>>> B*y, x_i and y binary} (so x is a vector, but y is just a binary
>>>> variable) where w_i >= 0 for all i and B >= 0.  The cut of the cutpool
>>>> which is not valid has the form x_i - y <= -1 (for some i) and was
>>>> generated by the cmir separator. Since the zero vector is feasible for
>>>> this problem the cut is not valid, right? Or have I overlooked
>>>> something?
>>>>
>>>> I attached the problem file to this mail.
>>>>
>>>> Best regards
>>>> Jonas
>>>>
>>>>
>>>> _______________________________________________
>>>> Scip mailing list
>>>> Scip at zib.de
>>>> http://listserv.zib.de/mailman/listinfo/scip
>>> _______________________________________________
>>> Scip mailing list
>>> Scip at zib.de
>>> http://listserv.zib.de/mailman/listinfo/scip
>>>
>>
>
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> http://listserv.zib.de/mailman/listinfo/scip
>