[Scip] non valid solution using SCIP in the pricing

[Hélène Toussaint]

Hi Stefan,

Indeed this error no longer arises with SCIP 3.0.1.
Thank you for your answer.

Best,
Hélène

Le 14/02/2013 11:07, Stefan Heinz a écrit :
> Hi Hélène,
>
> can you please upgrade to SCIP 3.0.1 and check if the error still 
> arises. The upgrade should be easy since the interface did not change. 
> The 3.0.1 contains basically only bug fixes compared to 3.0.0
>
> Thanks Stefan
>
> On 02/14/2013 08:58 AM, Hélène Toussaint wrote:
>> Dear all,
>>
>> I use SCIP 3.0.0 to run a column generation program. I have sometimes 
>> (very rarely) an issue when solving my sub problem with SCIP during 
>> the pricing: the solution doesn't respect all the constraints. I 
>> enclose a LP ("subPb.lp") for which SCIP gives the solution y11=1; 
>> y18=1; y25=1 (all other variables are 0). The constraint cR_0 is 
>> violated by this solution (since it involves " cR_0 : 18 <= 15"). 
>> This problem only occurs when I use a subSCIP: if I use SCIPreadProb 
>> directly on the LP enclosed then SCIP gives the exact result (there 
>> exists several solutions whose objective = 1) .
>>
>> I had a look to the transformed problem (enclosed as 
>> "transSubPb.lp"). In this transformed problem, the bounds force 
>> variables t_y7, t_y17 and t_26 to be 1. But this violate the cR_2 
>> constraint, so this problem is infeasible (while the original problem 
>> is feasible...).
>>
>> I have encountered a very similar problem few months ago (solutions 
>> violating constraints on instances for which the objective function 
>> is parallel to a ranged row) and it was solved by the enclosed patch 
>> (given by Stefan). But this time the violated constraint appears in 
>> the transformed problem.
>>
>> Could you please help me with this issue?
>>
>> Best regards,
>>
>> Hélène
>>
>>
>>
>>
>> _______________________________________________
>> Scip mailing list
>> Scip at zib.de
>> http://listserv.zib.de/mailman/listinfo/scip
>
>