Originally posted by: SystemAdmin

I have a model with two binary variables x[i][j] an y[i]; converted x to continuous, in the range of 0.0 - 1.0 using IloConversion, as follows, for all variables x:

cplex.add(cplex.conversion(x[i][j], IloNumVarType.Float));

however, the computational time rises with a model relaxed (15.49 sec), in comparison of the model not relaxed (1.73 sec). Is normal this in relaxation model in CPLEX? or missing something in the code?

--
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Could you post the logs of the two solves here?
For the relaxed model CPLEX can only apply LP presolve.
For the non-relaxed model CPLEX can apply MIP presolve which exploits the integrality conditions on variables and may therefore find much more reductions than LP presolve. So the presolved non-relaxed model may be much smaller than the presolved relaxed model.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Thanks for the answers. Attach file of the output, relaxed and not relaxed.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

(Ignore the previous post, the file is wrong; I'm sorry...)
Thanks for the answers. File attached of the output, relaxed and not relaxed.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

I think what you observe is expected. You have two different models and having less integral variables does not guarantee that solve times are smaller in general.
You can see in the log that the additional integrality constraints in the non-relaxed model allow CPLEX to perform more presolve reductions. Also branching works much better in the non-relaxed case.
#CPLEXOptimizers
#DecisionOptimization