[SCIP] Pyscipopt conshdlr enforcement

[Mark Turner]

Dear Kody,


For further Python clarification: Have you ever explicitly added a 
constraint of the type you've created, or have you set the kwarg 
`needscons=False`? A constraint handler will not perform its enforcement 
routines unless at least one of those criteria are satisfied.


Cheers and hope your bug is fixed,

Mark

On 11/17/24 23:04, Besancon, Mathieu wrote:
>
> Dear Kody,
>
>
> Just to clarify one thing in case this induced some confusion.
>
> Constraint handlers are defined to add new constraints to the model, 
> while separators would be the suitable mechanism for strengthening the 
> LP relaxation (separators define the algorithms that produce cuts).
>
>
> If a constraint handler is still what you think is suitable for your 
> use case, did you implement all corresponding methods?
>
>
> Best,
> Mathieu
>
> ------------------------------------------------------------------------
> *From:* Scip <scip-bounces at zib.de> on behalf of Kody Kazda 
> <17kk18 at queensu.ca>
> *Sent:* Sunday, November 17, 2024 4:37:27 PM
> *To:* scip at zib.de
> *Subject:* [SCIP] Pyscipopt conshdlr enforcement
>
> Hi all,
>
> I am looking to implement a constraint handler in pyscipopt to add 
> strengthening cuts to tighten the LP relaxation before branching.
>
> I have successfully implemented this on other models, but for some 
> unknown reason on the model I am working with now I can’t get 
> pyscipopt to enforce my constraint handler before branching.
>
> SCIP will get to over 25,000 nodes and still not run my constraint 
> handler.
>
> I have tried all sorts of different enforcement priorities, going as 
> high as 1E8, so that does not seem to be the issue.
>
> Does anyone have any thoughts on what might be causing this issue?
>
> Is there some algorithmic property that I am not considering that may 
> cause SCIP to not enforce a high priority constraint handler prior to 
> branching?
>
> Thank you!
>
> Kody
>
>
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> https://listserv.zib.de/mailman/listinfo/scip
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.zib.de/pipermail/scip/attachments/20241118/93b0aa1e/attachment.html>