Originally posted by: SystemAdmin

[irisabril said:]

Hi!

I'm programing a code in C++ with cplex concert tecnology in order to solve the following problem.

Having the parameter [b]D[/b], where 0<=[b]D[/b]<=1<br />
Lets say the objective function for a mathematical model is:

Max: [b]D[/b] (2[i]x1[/i]+3[i]x3[/i] + .....) + (1-[b]D[/b]) ([i]X2[/i]-5[i]X3[/i] + .....)

And the model has to be solved for several values of [b]D[/b], considering everything else in the model (constraints, other values in the objective function, etc) stays exactly the same (except the [b]D[/b] value).

For this purpose, the program considers a Loop in which the value of [b]D[/b] gradually decrease or increase in a certain fixed value[i] y[/i].


Which is the better way to change the value of [b]D[/b] in the objective function of the model?

I know one option is to solve the model for one value of [b]D[/b], lets say [b]D[/b]=0.5, erase the objective function and add a new one for the next value of [b]D[/b], but I was wondering if there is an efficient way to fulfill this matter.


#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

[TheEngineer said:]

I do not know about a possible performance loss but you could also model D as a CPLEX variable and set it fixed at your desired value y via a constraint.
During your loop you can change this single constraint but it is not necessary to change the objective function.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

[prubin said:]

Using the Java API (C++ is similar) and omitting try/catch blocks:

IloCplex cpx;
IloNumVar firstTerm;  // stuff multiplied by D
IloNumVar secondTerm;  // stuff multiplied by (1-D)
// build model -- add constraints defining firstTerm and secondTerm using other variables.
IloObjective objExpr;
// create the expression for the objective: D*firstTerm + (1-D)*secondTerm
IloObjective obj = cpx.minimize(objExpr);  // or maximize
cpx.solve();
// do something with solution and change D
cpx.setLinearCoef(obj, firstTerm, D);
cpx.setLinearCoef(obj, secondTerm, 1-D);
// iterate ad nauseum
#CPLEXOptimizers
#DecisionOptimization