Originally posted by: renatocuri

Hello,

I wrote a time indexed model for a scheduling problem and I'm trying to solve it with CPLEX using C++. I have a heuristic procedure that gives me an initial solution and I want to pass this solution to the optimizer. More specifically, I have an initial 0-1 Matrix that represents a feasible solution to the problem. I tried to use the addMIPstart method but it only accepts Arrays as parameters. I also tried to add each row of the matrix at a time on a loop, but it doesn't seem to work.

Can anybody help me?

Thanks
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

One way to do so will be with the use of the setVectors() method. If var2D is a 2D bool decision variable, var2D_startingVals is the 2D IloNumVar array containing the starting values and i is the 1st dimension of the decision variable, then one such call could look like the following:

for(int c=0;c<i;c++){
 cplex.setVectors(var2D_startingVals[c], 0, var2D[c],0,0,0);
}

Also, do make sure that this code is placed after the 'cplex.extract(model);' line.

Hope this helps.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

This has been asked several times on this Forum. addMIPStart() accepts only a one-dimensional array as argument and you have to convert your matrix to a flat array. Assuming you have a matrix x that stores your variables and a matrix v that stores the initial solution, the code looks something like this:

IloNumVarArray startVars(env);
IloNumArray    startVals(env);
for (int i = 0; i < rows; ++i) {
   for (int j = 0; j < cols; ++j) {
      startVars.add(x[i][j]);
      startVals.add(v[i][j]);
   }
}
cplex.addMIPStart(startVars, startVals);
startVals.end();
startVars.end();

As you can see, all you have to do is to store the variables and values in a temporary one-dimensional array.
#CPLEXOptimizers
#DecisionOptimization