[SCIP] branch-and-price-and-cut
Gerald Gamrath
gamrath at zib.de
Mon Jun 20 10:24:01 CEST 2022
Hi Marko,
you should just set the initial flag of the constraint to TRUE when
creating the constraint. If I remember correctly, it should then
automatically be added to the LP relaxation of nodes where it is valid.
Best,
Gerald
On 6/20/22 09:01, Horváth Markó wrote:
> Hi,
>
> if I set it to TRUE, also nothing is happened. Should I use the
> callback scip_active for this purpose? Anyway, which is the best way
> to add new rows to the problem for branching purpose?
>
> Best regards,
> Marko
>
> 2022. 06. 20. 8:44 keltezéssel, Gerald Gamrath írta:
>> Hi Marko,
>>
>> about 2): did you set the constraints to initial?
>>
>> Best,
>>
>> Gerald
>>
>> On 6/20/22 08:18, Horváth Markó wrote:
>>> Dear all,
>>>
>>> none of my attempts were succesful. I have two further questions.
>>>
>>> 1) Is it possible to add new rows to the problem of the new child
>>> nodes during a branching? When using scip_execlp, I could add
>>> (stick) new constraints to the child nodes, but I cannot find
>>> similar functionality for new rows.
>>>
>>> 2) Assume I added new constraints to the child nodes during
>>> branching. When a node is focused, is it possible to add new rows to
>>> the problem (based on these constraints) before the first LP is
>>> solved? I tried to modify the LP inside callback scip_active,
>>> however, nothing happened.
>>>
>>> Best regards,
>>> Marko
>>>
>>> 2022. 06. 13. 9:42 keltezéssel, Horváth Markó írta:
>>>> Dear all,
>>>>
>>>> I'm working on a branch-and-price-and-cut approach. Actually, it is
>>>> a branch-and-price approach, where until now, branching decisions
>>>> were handled by variable fixing (i.e., variables were fixed by the
>>>> propagation callback of the corresponding constraint handler).
>>>> However, I want to improve this branching, that is, in each child
>>>> node I want to add multiple rows to the problem. Of course, I need
>>>> to store these decisions to take them into consideration during
>>>> variable pricing. (So, yet it is a branch-and-price-and-cut
>>>> approach, but cuts are not added by separation procedure, but
>>>> branching). I would like to ask, what is the best way to do that?
>>>> (My best guess, is that in a branching rule I add the rows to the
>>>> problem, and I also create a constraint sticked to child nodes. The
>>>> corresponding constraint handler do nothing (may be propagation
>>>> only), but could be used in variable pricing).
>>>>
>>>> Thanks, and best regards,
>>>> Marko
>>>>
>>>>
>>>> _______________________________________________
>>>> Scip mailing list
>>>> Scip at zib.de
>>>> https://listserv.zib.de/mailman/listinfo/scip
>>>
More information about the Scip
mailing list