[SCIP] cuts.c

[Leona Gottwald]

Hello Tony,


about your first point. It is the responsibility of the separators to
check the minefficacy because some constraint handlers might want to add
cuts to the sepastore that are necessary to enforce their constraints
but are violated below minefficacy.


Best,
Leona


On 8/14/19 4:56 PM, Wang, Tony Congqian wrote:
>
> Hello,
>
>
> Sorry, I missed something in the code about the second point. The
> question has been resolved.
>
>
> Best Regards,
>
> Tony
>
>
> ------------------------------------------------------------------------
> *Von:* Scip <scip-bounces at zib.de> im Auftrag von Wang, Tony Congqian
> <tony.wang at rwth-aachen.de>
> *Gesendet:* Mittwoch, 14. August 2019 16:49:48
> *An:* scip at zib.de
> *Betreff:* Re: [SCIP] cuts.c
>  
>
> Hello Leona,
>
>
> thanks for the quick reply. Thats also what I was thinking. I have two
> other questions:
>
>
>  1. Why is minefficacy not used in SCIPsepastoreApplyCuts?
>     SCIPsetIsFeasPositive(set, SCIProwGetLPEfficacy(cut, set, stat,
>     lp)) is called anyway, so I guess it isn't the separators task to
>     check for minefficacy, so shouldn't there be
>     SCIPsetIsFeasPositive(set, SCIProwGetLPEfficacy(cut, set, stat,
>     lp) - "*minefficacy*" )
>  2. Objectparallelism is not calculated when not using debug-mode
>     since checkRowObjprod() is deactivated in non-debug-mode. Is this
>     intended (since objectparallelism seems to not give a lot of
>     improvement and is appearently slow to calculate)? What makes me
>     wonder though is that SCIProwGetObjParallelism() is the only
>     function that calls checkRowObjprod() and
>     SCIProwGetObjParallelism() is only referenced by selectCuts(). So
>     why even bother disabling the function when the user can disable
>     the use by setting objparalweight = 0
>
>
> Best Regards,
> Tony
> ------------------------------------------------------------------------
> *Von:* Scip <scip-bounces at zib.de> im Auftrag von Leona Gottwald
> <gottwald at zib.de>
> *Gesendet:* Mittwoch, 14. August 2019 12:43:51
> *An:* scip at zib.de
> *Betreff:* Re: [SCIP] cuts.c
>  
>
> Hello Tony,
>
>
> thanks for your question, indeed that does not make sense and was not
> intended.
>
>
> The code will be changed to add the dircutoffdistweight instead of the
> objparalweight in the next release which is consistent with the
> behavior for locally valid cuts.
>
>
> Best,
>
> Leona
>
>
> On 8/12/19 12:48 AM, Wang, Tony Congqian wrote:
>>
>> Hello community,
>>
>>
>> I have a question regarding cuts.c:2563 -- efficacyfac +=
>> objparalweight; --. I couldn't find the reasoning of this addition
>> anywhere. It is done, when dircutoffdistance is not calculated
>> (either due to missing incumbent or dircutoffdistweight == 0). So I
>> would think, that in the case that the incucmbent is missing,
>> efficacy would get more weight, but I can't yet see the reasoning
>> behind adding objparalweight on it.
>>
>>
>> Best Regards,
>>
>> Tony
>>
>>
>> _______________________________________________
>> Scip mailing list
>> Scip at zib.de
>> https://listserv.zib.de/mailman/listinfo/scip
>
> _______________________________________________
> 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/20190826/0bd1ded9/attachment.html>