Originally posted by: SystemAdmin

Hello,

I observe a strange behavior for the attached model (problem.lp).

When it is solved with default settings, CPLEX gives 3.567885 as the optimal value. If it is solved by turning off presolve, it finds 3.558526 which is the true optimum.

Is something wrong with presolve or the model?

Thank you,

Onur
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: T_O

I have no obvious answer to your question, but I found out that it does not happen if I disable primal reductions.

Thomas
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Thanks for reporting that. It looks like a bug in CPLEX.
We are investigating and I will get back to you once we have a definitive answer.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Thanks again for reporting this problem. This is a numerical bug that we have in the code since ages and which is triggered for some weird reason by your problem instance. The issue is inside probing, and will be fixed in the next release. You should be able to get the correct answer by disabling probing. This can be done using

set mip strategy probing -1

in the interactive CPLEX shell.
Tobias
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Thank you for clarifying the issue.

(Just a note for future readers of this thread, my CPLEX version is 12.4.)

Onur
#CPLEXOptimizers
#DecisionOptimization