[Coin-discuss] Inconsistency between Osi Solves

[Pedro Teixeira]

The situation is as follows (I'm running the latest stable of all projects,
should I be running from the trunk?)  :

Both Cbc and Sym give the same straight objetive value of 10.  Where for
this instance the known optimal is 85. The strange thing is that with Clp,
the first result is 110 but then I add an extra cut (subtour) and it becomes
infeasible...

It's really probably something in my code, but it seems Clp was closer to
what I wanted.. I posted the issue here because I was surprised to get
different solutions. I'm also trying to run in cplex, but as you suggested
I'll stick with Cbc and Sym, disregard Clp and will review the code again :/

Ashutosh, I copied the model here:
http://www.pedro.not.br/tmp/model.mps


thanks for the replies.
Pedro

On Feb 7, 2008 1:20 PM, Ted Ralphs <ted at lehigh.edu> wrote:

> As far as I remember, the branchAndBound() method of Clp is a fairly
> basic implementation that is not really intended for practical use. You
> probably want to compare with OsiCbc for solving integer programs. It
> should give better results.
>
> Cheers,
>
> Ted
>
> Pedro Teixeira wrote:
> > Hi,
> >
> > I am new to coin-or and not sure if this is the appropriate list to
> > post. I'm getting very different solutions when using the same model
> > but switching from OsiClpSolverInterface and OsiSymSolverInterface.
> >
> > my current usage is simple:
> >
> > si->loadFromCoinModel(model , true);
> > si->initialSolve();
> > si->branchAndBound();
> >
> > With ClpSolver, it takes much more time and I get (for illustration):
> > Search took 43600 iterations and 7302 nodes
> > Clp0000I Optimal - objective value 115
> >
> > However, with SymSolver I get the result instantaneously:
> > Solution Found: Node 1, Level 1
> > Solution Cost: 10.000
> >
> > Shouldn't I get the same results? Or is there any known issues with Osi?
> > Can anyone shed in any light?
> >
> > cheers,
> > Pedro
> > _______________________________________________
> > Coin-discuss mailing list
> > Coin-discuss at list.coin-or.org
> > http://list.coin-or.org/mailman/listinfo/coin-discuss
>
>
> --
> Dr. Ted Ralphs
> Associate Professor
> Industrial and Systems Engineering
> Lehigh University
> (610)758-4784
> ted at lehigh dot edu
> coral.ie.lehigh.edu/~ted <http://coral.ie.lehigh.edu/%7Eted>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/coin-discuss/attachments/20080207/01942ecb/attachment.html>