question to understand

[hercule]

Hi !!
I have a question and i want just to understand if that's normal and why

i have to maximize an objectif variable "obj"

IntegerExpressionVariable ie = Choco.constant(0);
//ie is an accumulator where i compute the sum of data
for (int i = 0; i<n ; i++)
    ie = Choco.plus(ie, Choco...)
IntegerVariable obj = Choco.makeIntVar("cout");
m.addConstraint(Choco.eq(obj,ie));
s.maximize(s.getVar(obj), true);

the solver return me a solution in 30 seconds, but when i change the weight of obj the solver takes more time to give the solution.

IntegerVariable obj = Choco.makeIntVar("cout");
m.addConstraint(Choco.eq(obj,Choco.mult(ie,1000)));

in 40 seconds

IntegerVariable obj = Choco.makeIntVar("cout");
m.addConstraint(Choco.eq(obj,Choco.mult(ie,10000)));

in 2 minutes and 39 sec

thank you very much for help

[Arnaud Malapert]

Hello,

The complexity of the filtering algorithm can depend on the domain sizes.
Do you have the same problem using the Choco.sum or Choco.scalar constraint ?
Try also to change the value of model.setDefaultExpressionDecomposition(final Boolean defDecExp)

Best,
Arnaud

[hercule]

hi arnaud,

i can't use Choco.sum or Choco.scalar, i should create a temporary variable ie to compute my objective and then Choco.eq(obj,ie)

but in changing the weight obj=ie , obj=100*ie and obj=10000*ie the duration of resolution change

is that normal in choco??

i tried as you sais to change setDefaultExpressionDecomposition, it work with true (the same result and duration as described) and with false it is much more long i wait for 4 minutes and no results

[Arnaud Malapert]

Hello,

Could you provide a minimal example ?

Best.

[hercule]

this is an example

      int tab [] = new int [1000];      
      IntegerVariable[] IV1 = makeIntVarArray("v", 1000, 0, n);
      IntegerVariable obj1 = Choco.makeIntVar("cout1");
     
      IntegerExpressionVariable tm1 = Choco.constant(0);
      for (int i=0; i<nbre_plan-1; i++) 
          tm1 = Choco.plus(tm1, Choco.mult(Choco.min(IV1[i], 1), Math.min(tab[i+1], 1) ));
      m.addConstraint(Choco.eq(obj1,Choco.mult(tm1, 1000)));
      Solver s = new CPSolver();
      s.read(m);
      s.maximize(s.getVar(obj1), true);

in changing the weight 1000 in this example with 1 or 100000, the duration remarkably change

thanks

[Arnaud Malapert]

Thanks, I will have a look this evening.

[hercule]

a small modification in _

_

      int tab [] = new int [1000];      
      IntegerVariable[] IV1 = makeIntVarArray("v", 1000, 0, n);
      IntegerVariable obj1 = Choco.makeIntVar("cout1");
     
      IntegerExpressionVariable tm1 = Choco.constant(0);
      for (int i=0; i<nbre_plan-1; i++) 
          tm1 = Choco.plus(tm1, Choco.mult(Choco.min(IV1[i], 1), Math.min(tab[i+1], 1) ));
      m.addConstraint(Choco.eq(obj1,Choco.mult(tm1, 1000)));
      Solver s = new CPSolver();
      s.read(m);
      s.maximize(s.getVar(obj1), true);

_

thank you very much_

[hercule]

There is no explanation?

Normally, it is expected that the response time will be the same, Maximizing (X) or Maximizing (1000 X) it is the same.

Thans