Originally posted by: mangotree

Dear All:

We are observing strange log information from a model. CPLEX (V12.2) barrier optimizer claims
"best solution found", however, also says "primal infeasible". What is the meaning of this message
"Barrier cannot determine infeasibility" and how to overcome it?

Please advise!

Thank you,

---

Refinement - orig 5.72e-08, refined 5.72e-08, target 2.81e+00, 0 iter
46 -2.2136059e+08 -2.2136059e+08 2.58e+00 3.59e-10 3.28e+00
0.37s for iteration (0.25s, 1759 Mflops for lin. solve)
Refinement - orig 1.53e+05, refined 1.53e+05, target 5.16e-01, 1 iter
47 -2.2136059e+08 -2.2136059e+08 5.10e+01 3.71e-10 3.31e+00
Barrier cannot determine infeasibility.
* -2.2136059e+08 -2.2136059e+08 7.02e-01 1.86e-09 4.93e-01
CPLEX 12.2.0.0: best solution found, primal infeasible; objective -221360589.6
47 separable QP barrier iterations
No basis.

1. /product/abb/sable/apprun/ver/bin/ampl_uc_dam_solve_QP.run:247(6083) display ...

##genmod times:
##seq seconds cum. sec. mem. inc. name
solve_result_num = 206
solve_result = infeasible

---

#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

The barrier method ran into numerical issues. I suggest to upgrade to CPLEX 12.5. Since 12.2 we improved barrier numerics, in particular for infeasible problems, quite a bit.

Tobias
#CPLEXOptimizers
#DecisionOptimization