sci.op-research

On Mar 4, 3:53 pm, Paul Rubin wrote:
> Amy wrote:
>
> > Just provide you additional information. I tried another example, and
> > CPLEX does use my heuristic after adding that set of inequalities. But
> > I'm still investigating why CPLEX doesn't use my heuristic for my
> > first example.
>
> If CPLEX does accept the heuristic solution most of the time, then I
> suspect the code is generally correct. For the one example where the
> heuristic solution is not accepted, are you sure it is a feasible
> solution to the IP?
>
> You might try adjusting CPLEX's tolerance parameters (both constraint
> tolerance and integrality tolerance) to be slightly looser, and see if
> that causes the one problematic instance to work. Perhaps it is just a
> rounding problem.
>
> /Paul

Hi, Paul,

Yes, the heuristic solution is a feasible IP solution.

How can change the constraint tolerance and integrality tolerance? I'm
using Java.

Thanks.

Amy