[Scip] optimal solution is infeasible

Timo Berthold berthold at zib.de
Thu Jan 29 16:34:35 MET 2009


Hi Tamas.

Taking the large coefficients into account (e+06), this sounds a bit like a 
numerical issue.
I could imagine, that SCIP presolving tightens those coefficients, SCIP finds 
a solutions which is feasible +- epsilon for the tightened constraint and 
after reversing the tightening, the solution is (numerically) infeasible in 
the original problem.

However, a violation by 319 (hence 10-03 relative to the right hand side) is 
definitely too much.
Is it possible that you send me the instance s.t. I can have a closer look?

I suggest to just send it to berthold at zib.de and not to the mailing list.

Best regards, Timo


Am Donnerstag 29 Januar 2009 11:55:50 schrieb Mahr Tamas:
> Dear all,
>
> I get infeasible solutions from scip as optimal. (See text below!) I
> am not exactly sure what is happening, the only thing we could come up
> is that the optimal solution is optimal in the relaxed problem, and
> not feasible in the original problem. Is there any way to retrieve a
> feasible (optimal) solution?
>
>
> SCIP Status        : problem is solved [optimal solution found]
> Solving Time (sec) : 2593.57
> Solving Nodes      : 33816
> Primal Bound       : +8.87969904899597e+04 (37 solutions)
> Dual Bound         : +8.87969904899597e+04
> Gap                : 0.00 %
>    [linear] <c1758>: +13165.9711303711<x6572>[B]
> -136248.229980469<x6603>[B] -1<x11047>[C] +1<x11078>[C] >=
> -123082.258850098
> violation: left hand side is violated by -319.481872558245
> best solution is not feasible in original problem
>
> Thanks,
> Tamas
>
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> http://listserv.zib.de/mailman/listinfo/scip




More information about the Scip mailing list