Originally posted by: bqt

Hi all,

I am solving a MIP problem, where binary decision variables are declared as follows

IloArray<IloIntVarArray> x(env, n); //Variables
for(i = 0; i < n; i++){
x[i] = IloIntVarArray(env, n, 0, 1);
}
}

In the UserCutCallback, I show assigned value of the binary variables in using the method getValues(x[i]). As the results, some variables are assigned negative values.

Someone could please give me an advice to fix this bug ?

Thanks,

bqt
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

What is the magnitude (absolute value) of the negative variables?

Paul

Mathematicians are like Frenchmen: whenever you say something to them, they translate it into their own language, and at once it is something entirely different. (Goethe)
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: bqt

Hi Paul,

Some variables are signed to -1.4, -0.65 ..

bqt
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

That looks like a bug in either CPLEX or your model. Could you call cplex.exportModel("model.sav") right before doing cplex.solve() and attach the generated .sav file here?
Maybe also try calling getLB() and getUB() on the variables that have negative values in the callback to double check that the bounds are as expected.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: bqt

Thanks Daniel,

I am doing a research, posting source code on the forum is not good. I have just tried calling getLB() and getUB() on the variables that have negative values, as result, they are always 0 and 1. It means that their bounds are correct.

bqt
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: bqt

I tested my code on some different instances. My code run well on almost of instances. Few of instances, it presents the bug
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Try collecting kappa statistics to see if CPLEX is encountering unstable bases (which would indicate numerical instability).

Paul

Mathematicians are like Frenchmen: whenever you say something to them, they translate it into their own language, and at once it is something entirely different. (Goethe)
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: bqt

Thank you for your reply. I found the reason, it is a floating error.
bqt
#CPLEXOptimizers
#DecisionOptimization