Originally posted by: SystemAdmin

I am dealing with a problem with multiple optimal values which might also be unbounded.
This a simple LP model. I use one single thread, IloCplex::RootAlg set to IloCplex::Barrier and the cross over algorithm set to IloCplex::NoCrosAlg.
I expect that using these options barrier does not "hopefully" get me an extreme point but a point on a optimal face for the bounded problems.

I need to know whether the problem is unbounded XOR is feasible optimal.

I get the following status "AbortPrimObjLim" or "AbortDualObjLim" or "InfOrUnbd".
The last one is clear to me because my problem can only be unbounded but the first two are less clear.
I did not understand much from the documentation.
Does anyone like to shed some light on them?

Any comment is appreciated.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Hi Shahin,

The nature of the barrier algorithm makes it harder to declare that a problem is unbounded than when using the simplex method. As a safeguard against running into numerical difficulties, CPLEX uses a parameter to limit the magnitude of the objective function values in the barrier algorithm. In Concert, this parameter is BarObjRng; it has a default value of 1.0e20. When the absolute value of the objective exceeds the value of this parameter, one of the statuses AbortPrimObjLim or AbortDualObjLim is returned. This is effectively a conclusion of unboundedness for the problem. On the other hand, if you know that your problem might have an optimal objective value with magnitude greater than 1.0e20, then you should either consider rescaling your problem or increase the value of this parameter to be greater than the known bound on your objective function.

Philip Starhill
CPLEX Research Engineer
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Hi Philip, many thanks for your comment.
That is clearly understood.
But, now the question is which of those statuses are actually an indication of optimality or almost optimality.
E.g. I dont really know how to interpret "NumBest" or other statuses from Barrier output.
Is that documented anywhere in more details?

thanks
#CPLEXOptimizers
#DecisionOptimization