Originally posted by: SergeyPolyakovskiy

Dear Colleagues,

I wonder about memory or numeric limitations on the number of variables of the objective function. I am using CPLEX 12.4. Academic Version, Visual C++ with access through the Concert Technology. The objective function is generated dynamically term by term using the command like obj = obj + koef[j]*X[j][i], where coefficients are of the double-type. Finally, I obtain error of “Out of memory” once I reach approx. 400000 terms.
I wonder if such kind of generation is too recourse consuming and there are better ways to generate large objective functions?

Thanks in advance.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Yes, there are better ways to do that. First of all, instead of doing

obj = obj + koef[j]*X[j][i]

use

obj += koef[j]*X[j][i]

If that does not help try to first collect all coefficients in a double and an IloNumVar array and the use IloScalProd.
#CPLEXOptimizers
#DecisionOptimization