[SCIP] Question about SCIP performance in a MIQCP model

[Stefan Vigerske]

Hi,

if the solution reported by SCIP is indeed feasible, preferably without 
any constraint violations, then there seems to be a problem with CPLEX. 
You can try to feed the solution to CPLEX and see if it is accepted.

If the solution reported by SCIP is not feasible, then either there was 
something wrong in setting up the problem in SCIP (that is, SCIP didn't 
solve the problem you thought it would), or there were some wrong 
reformulations happening in SCIP (so the feasible space that SCIP sees 
was changed in an invalid way), or the solution is only feasible w.r.t. 
tolerances and there is a no solution that satisfies all constraints 
without any violations. For the last case, both infeasible and feasible 
are valid results.

Stefan

On 7/15/22 16:13, behnam alizadeh wrote:
> Dear SCIP,
> I have solved a MIQCP model both by SCIP and CPLEX.
> CPLEX immediately report the model as infeasible. However, SCIP solves
> the model to optimality.
> I guessed this may be because of the feasibility tolerance in CPLEX.
> But after I increased the feasibility tolerance to relatively a large
> value, there was no change in CPLEX performance (i.e. CPLEX insists on
> infeasible model).
> What else can be the cause of this strange behavior?
> Thanks in advance,
> Behnam Alizadeh
> 
> 
> Sent from Mail<https://go.microsoft.com/fwlink/?LinkId=550986> for Windows
> 
> 
> 
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> https://listserv.zib.de/mailman/listinfo/scip