[SCIP] Fwd: Sub-NLP solution is infeasible

[Stefan Vigerske]

Hi,

sometimes solutions that SCIP computes for the presolved problem are 
slightly infeasible when transferring them back into the original 
problem. In case of using SCIP through GAMS for solving a (MI)NLP, the 
GAMS/SCIP interface then attempts to fix these infeasibilities by 
resolving the original problem with Ipopt with all integer variables 
fixed and starting from the SCIP solution point. In your case, this has 
failed for all solutions that SCIP has computed. You can disable this 
postprocessing step by setting gams/resolvenlp = FALSE in a GAMS/SCIP 
options file.
You will still have the problem that the solution reported back to GAMS 
will be slightly infeasible w.r.t. the used feasibility tolerance, 
though. If you find that the violation is too large, then you can set 
option numerics/feastol to have SCIP using a smaller feasibility 
tolerance when solving the presolved problem.

Stefan

On 4/15/19 3:47 PM, Haitham Kanakri wrote:
> Hello SCIP Technical Team;
> 
> Trust this email finds you well!
> 
> 
> 
> I am trying to solve MINLP model using SCIP within GAMS interface and I am
> facing the following message: subnlp solution is infeasible. Any ideas to
> solve the problem would be appreciated.  I am attaching my GAMS code along
> with the data files for your reference.
> 
> 
> 
> Looking forward to hear back from you!
> 
> 
> 
> Best Regards,
> 
> Haitham
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> Sent from Mail <https://go.microsoft.com/fwlink/?LinkId=550986> for Windows
> 10
> 
> 
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> https://listserv.zib.de/mailman/listinfo/scip
>