[Symphony] Unexpected behaviour of gap_limit in symphony 5.1

[Ted Ralphs]

We have now made a new release 5.1.11 that fixes this bug if you want
to continue to use 5.1. It's actually still present in 5.2, though it
didn't exhibit itself on the example you sent. We fixed it in the
stable branch of 5.2, but have not made a new release yet.

Cheers,

Ted

On Sat, Sep 12, 2009 at 9:31 AM,  <Pierre.Lebodic at lri.fr> wrote:
> Hello,
>
> I mentioned earlier a "bug" involving gap_limit in SYMPHONY 5.1.10. It
> should be possible to reproduce it with the following test instance:
> http://www.lri.fr/~lebodicp/wp-content/uploads/gap_instance.MPS
>
> In the following, cuts are disabled to prevent possible numerical surprises.
> In SYMPHONY 5.1.10, with gap_limit set to 5 (for instance), the following
> happens:
> ****************************************************
> * Optimal Solution Found                           *
> * Now displaying stats and best solution found...  *
> ****************************************************
> ...
> Current Upper Bound:         2.304
> Current Lower Bound:         1.300
> Gap Percentage:              43.58
>
> Solution Found: Node 0, Level 0
> Solution Cost: 2.304
>
> and a feasible solution is displayed. Although the gap percentage is
> 43.58, the solver indicates that the optimal has been found.
>
> In SYMPHONY 5.1.10, with gap_limit set to -1, a better solution is provided:
> ****************************************************
> * Optimal Solution Found                           *
> * Now displaying stats and best solution found...  *
> ****************************************************
> ...
> Solution Found: Node 5, Level 3
> Solution Cost: 2.062
>
> and the optimal solution is displayed.
>
>
> SYMPHONY 5.2 seems to work: with gap_limit lower than 43.58, the optimal
> solution is found. with gap_limit set higher than 43.58, it provides the
> same feasible solution as SYMPHONY 5.1.10.
> Thanks for your help.
>
> Cheers,
>
> Pierre
>
> _______________________________________________
> Symphony mailing list
> Symphony at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/symphony
>



-- 
Dr. Ted Ralphs
Associate Professor, Lehigh University
(610) 628-1280
ted 'at' lehigh 'dot' edu
coral.ie.lehigh.edu/~ted