[Cbc] [CBC] Preprocess causing infeasibility
John Forrest
john.forrest at fastercoin.com
Tue Sep 29 10:44:09 EDT 2015
Marco,
Yes - I don't think other presolve functions change bounds in same way.
John
On 29/09/15 14:57, mg wrote:
> Ok, so if I understood correctly you've only changed
> CoinPresolveDoubleton.cpp, am I right ?
>
> Thanks a lot for the quick fix !
>
> Marco
>
> 2015-09-29 15:41 GMT+02:00 John Forrest <john.forrest at fastercoin.com
> <mailto:john.forrest at fastercoin.com>>:
>
> Marco,
>
> Should be fixed.
>
> The dual code correctly thought it should push a variable all the
> way to its upper bound. Unhappily the upper bound was 1.3. Then
> a later part of presolve worked out that that was infeasible for
> an integer variable.
>
> The error occurred in CoinPresolveDoubleton which changed bounds
> from 0,infinity to 0,1.3 without checking whether it was integer.
> I have corrected code so bounds would be 0,1.
>
> John Forrest
>
>
> On 28/09/15 23:46, Haroldo Gambini Santos wrote:
>> Hi Marco,
>>
>> I found out that the problem is related to a presolving procedure
>> which tries to fix bounds using dual information.
>>
>> A /quick and dirty/ fix is to change in
>>
>> Osi/src/OsiPresolve.cpp line 988
>>
>> if (dual) {
>> to
>> if ((dual)&&0) {
>>
>> to temporarily disable this processing...
>>
>> Hoping that a better fix comes soon,
>>
>> Cheers,
>>
>> Haroldo
>>
>> On 28-09-2015 06:22, mg wrote:
>>> Hi all,
>>> I've found a case where preprocess says the problem is
>>> infeasible, but if I turn preprocess off, the problem is
>>> correctly solved.
>>>
>>> Attached you can find the LP file (m3.lp). To reproduce the
>>> problem you can use the following code (I use callCbc):
>>>
>>> OsiClpSolverInterface solver;
>>> CbcModel model(solver);
>>> model.solver()->readLp("m3.lp");
>>> std::string args("-heur off -cuts off -solve");
>>> // using these arguments instead it works
>>> // std::string args("-heur off -cuts off -preprocess off -solve");
>>> callCbc(args, model);
>>>
>>> The problem happens in both 2.9.5 releaseand 2.9.6 release versions.
>>>
>>> Thanks in advance,
>>> Marco
>>>
>>>
>>> _______________________________________________
>>> Cbc mailing list
>>> Cbc at list.coin-or.org <mailto:Cbc at list.coin-or.org>
>>> http://list.coin-or.org/mailman/listinfo/cbc
>>
>> --
>> ==================================================
>> Haroldo Gambini Santos
>> D.Sc, Computer Science
>> Universidade Federal de Ouro Preto
>> http://www.decom.ufop.br/haroldo/
>>
>>
>> _______________________________________________
>> Cbc mailing list
>> Cbc at list.coin-or.org <mailto:Cbc at list.coin-or.org>
>> http://list.coin-or.org/mailman/listinfo/cbc
>
>
> _______________________________________________
> Cbc mailing list
> Cbc at list.coin-or.org <mailto:Cbc at list.coin-or.org>
> http://list.coin-or.org/mailman/listinfo/cbc
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/cbc/attachments/20150929/4f59455d/attachment.html>
More information about the Cbc
mailing list