[Symphony] Generating multiple solutions sequentially

[Ted Ralphs]

Sorry it took so long for me to answer this. Along with others, it got
stuck in the administrative queue for a while. At the moment, I don't
think there is much you can do other than what you are already doing,
but I am curious to know more about your application. We plan to
support the addition of rows for warm starting in the future, but
there may be a more efficient way of going about the whole algorithm
than what you're currently doing.  Please e-mail me off-line if you're
still interested in discussing this further.

Cheers,

Ted

On Mon, Feb 2, 2009 at 8:11 PM,  <lebodicp at insa-rouen.fr> wrote:
> Hello,
>
> I have implemented a customized solver thanks to the callback functions.
> I'm interested in finding multiple close-to-optimal solutions. My program
> does the following: (1)read the problem data, (2)create and (3)solve the
> problem, (4)cut the current solution by adding at least a row, repeat (3)
> and (4).
>
> I would be interested in optimizing run times. I have tried the warm start
> feature, but as predicted by the manual, problem modifications such as row
> addition are not supported. I also tried saving at root node and loading
> after cutting the current solution, it didn't work. I also did the
> following: use the previous solution found to user-specify a dual bound
> for the next resolution. However, the solver did not seem to benefit from
> this at all. I also tried adding an inequality, with the objective
> function in the lhs, and the dual bound on the rhs. Weirdly, it made the
> B&B stuck at this very value for many iterations, yielding much longer
> execution times. It is an unexpected behaviour, since the polyhedron of
> the LP has been cut. Could it be a bug? Or, could it be that the simplex
> method (I use CLP) tends to slow down with this kind of inequality?
> All in all, what can I save from the previous resolutions to help
> following resolutions?
> Thanks in advance,
>
> 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 (permanent)
Visiting Professor, Monash University (current)
(610) 628-1280
ted 'at' lehigh 'dot' edu
coral.ie.lehigh.edu/~ted

Sent from: Melbourne Vic Australia.