Originally posted by: SystemAdmin

Hi,
Is it possible in cplex 12 using java concert technology for a value to have a multiple range.
I want a constraint where I want a IloNumVar to have values as either zero or number in range say 3 to 11, but nothing in between 0 and 3.
Could anyone suggest me how this can be done.

Regards
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

> surajjung wrote:
> I want a constraint where I want a IloNumVar to have values as either zero or number in range say 3 to 11, but nothing in between 0 and 3.

IloCplex cplex = new IloCplex();
IloSemiContVar x = cplex.semiContVar(3., 11., IloNumVarType.Float);

/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: SystemAdmin

Hi,
Thank you for the reply.
I have one another problem. Is it possible to have the constraints which satisfies the range, -10 to -3 and 0 and +3 to +10.
i.e I want a variable which can take the value from -10 to -3 and 0 and +3 to +10 but nothing in between -3 to +3 except 0.

Regards
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

I'm afraid not.
What you can do is:

• Use two variables x1 and x2. x1 is 0 or in 3,10, x2 is 0 or in -10,-3.
• Replace x by x=x1+x2.
• Add an additional binary variable b with constraints x1<=10*b, x2>=-10*(1-b). If b=0 then this says x1<=0 (i.e. x1=0) and x2>=-10, if b=1 then this says x1<=10 and x2>=0 (i.e. x2=0).

Instead of using a binary variable b and the constraints I mentioned you can also use indicator variables/constraints.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Just curious , but can it be done in the way
define X as -10 to +10
define X_abs as 0 or +3 to +10
force X_abs=|X|
where |X| is absolute value of X.

Is it possible?
If it is could you help me with the implementation of it. AM not sure how to do it.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Yes, it can also be done in this way The difficult part is to force X_abs = |X|. One direction is simple:
X_abs >= X
X_abs >= -X
This ensures X_abs >= |X|. Now if X_abs has a positive objective function coefficient and you have a minimization problem then you will automatically have X_abs=|X| (otherwise the solution is non-optimal.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

IloCplex model = new IloCplex();
    IloNumVar X[] = model.numVarArray(noOfVar, -10, 10);
    IloSemiContVar X1[] = model.semiContVarArray(noOfVar, 3, 10, IloNumVarType.Float);
    IloSemiContVar X2[] = model.semiContVarArray(noOfVar, -3, -10, IloNumVarType.Float);
    
for (int i = 0; i < X.length; ++i) {
            model.addEq(model.sum(X1[i], X2[i]), X[i]);
            model.addLe(X1[i],model.prod(10, b));
            model.addGe(X2[i],model.prod(-10, model.sum(1, model.prod(-1, b))));
        }

Could you please check if this is the correct implementation of your method..
Cplex 12 java concert tech.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Looks correct except for the constructor if X2. In this constructor you must swap -3 and -10. -10 is the lower bound and -3 is the upper bound.
And you probably need a b-variable for each variable in X, don't you? Or are all your X[i] positive/negative simultaneously?
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Thanks,

IloCplex model = null;
    IloNumVar X[] = model.numVarArray(noOfVar, -11, 11);
    IloSemiContVar X1[] = model.semiContVarArray(noOfVar, 3, 11, IloNumVarType.Float);
    IloSemiContVar X2[] = model.semiContVarArray(noOfVar, -11, -3, IloNumVarType.Float);
    IloNumVar E[] = model.numVarArray(noOfRev, -100, 100);
    IloNumVar E_abs[] = model.numVarArray(noOfRev, 0, 100);
    model = new IloCplex();
    IloNumVar b[]=model.boolVarArray(noOfVar);
 
      
       for (int i = 0; i < E.length; ++i) {
            model.addLe(model.prod(1.0, E[i]), E_abs[i]);
            model.addLe(model.prod(-1.0, E[i]), E_abs[i]);
        }
 
        for (int i = 0; i < X.length; ++i) {
            model.addEq(model.sum(X1[i], X2[i]), X[i]);
            model.addLe(X1[i],model.prod(10, b[i]));
            model.addGe(X2[i],model.prod(-10, model.sum(1, model.prod(-1, b[i]))));
        }
 
        IloNumExpr exprn1 = model.scalProd(X, X);
 
        IloNumExpr exprn2 = model.sum(E_abs);
 
        model.addMinimize(model.sum(model.prod(0.7, exprn1),
                model.prod(0.3, exprn2)));

Here I have added my code with the optimization equation...
But I get the warning, " Bound infeasibility column 'IloX386'."

The error is persistent even if i increase the bounds to very higer ranges.
Is there anything wrong I am doing?

Regards
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Sorry, my fault. I didn't remember that semi-continuous variables must not have negative bounds (from the docs):

A semi-continuous variable is a variable that by default can take the value 0 (zero) or any value between its semi-continuous lower bound (sclb) and its upper bound (ub). The semi-continuous lower bound (sclb) must be finite. The upper bound (ub) need not be finite. The semi-continuous lower bound (sclb) must be greater than or equal to 0 (zero). An attempt to use a negative value for the semi-continuous lower bound (sclb) will result in that bound being treated as 0 (zero).

So instead of using X, X1 and X2 as described before you should use

• X1 is 0 or in 3,10
• X2 is 0 or in 3,10
• X = X1 - X2
• X1 <= 10*b
• -X2 >= -10*(1-b)

This should do the trick. I changed your code accordingly and no longer get the error:

import ilog.cplex.*;
import ilog.concert.*;
 
public class Semi {
   public static void main(String[] args) {
      try {
         int noOfVar = 10;
         int noOfRev = 5;
         IloCplex model = new IloCplex();
         IloNumVar X[] = model.numVarArray(noOfVar, -11, 11);
         IloSemiContVar X1[] = model.semiContVarArray(noOfVar, 3, 11, IloNumVarType.Float);
         IloSemiContVar X2[] = model.semiContVarArray(noOfVar, 3, 11, IloNumVarType.Float);
         IloNumVar E[] = model.numVarArray(noOfRev, -100, 100);
         IloNumVar E_abs[] = model.numVarArray(noOfRev, 0, 100);
         IloNumVar b[]=model.boolVarArray(noOfVar);
 
 
         for (int i = 0; i < E.length; ++i) {
            model.addLe(model.prod(1.0, E[i]), E_abs[i]);
            model.addLe(model.prod(-1.0, E[i]), E_abs[i]);
         }
 
         for (int i = 0; i < X.length; ++i) {
            model.addEq(model.diff(X1[i], X2[i]), X[i]);
            model.addLe(X1[i],model.prod(10, b[i]));
            model.addGe(model.prod(-1., X2[i]),model.prod(-10, model.sum(1, model.prod(-1, b[i]))));
         }
 
         IloNumExpr exprn1 = model.scalProd(X, X);
 
         IloNumExpr exprn2 = model.sum(E_abs);
 
         model.addMinimize(model.sum(model.prod(0.7, exprn1),
                                     model.prod(0.3, exprn2)));
 
         model.solve();
      } catch (IloException e) {
         System.err.println("IloException: " + e);
      }
   }
}

#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Thanks a lot...
Highly appreciated...
But suddenly I started getting lots of multiple solutions and running out of memory...
Regards
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Don't know if these multiple solutions and the out of memory are a problem with CPLEX or with your model? Anyway, if you want to discuss this it would be nice if you opened a new thread here.
#CPLEXOptimizers
#DecisionOptimization