|
From: | Alessandra De Luca |
Subject: | Re: [Help-glpk] Execution time |
Date: | Fri, 14 Sep 2007 11:53:07 +0200 (CEST) |
Alessandra,
it is not an unreasonable long time for such a big problem. You can find
problems with much fewer rows and columns that are practically impossible
to solve using branch-and-bound.
The MIP-solver got a big upgrade starting from version 4.9, so if you are
using 4.8 you will most likely not use it (unless you are infact using a
later GLPK but a 4.8-interface, however even then you will not be able to
access lpx_intopt, which is the new solver).
So, start by upgrading the GLPK-version and try the advanced MIP-solver as
a first step.
Good luck!
/Oscar Gustafsson
On Fri, 14 Sep 2007, Alessandra De Luca wrote:
> I'm trying to solve a MIP problem with 17575 rows, 14960 column and 1018925 non-zeros but Glpk spend a lot of time to solve it (about 40 minutes)....it's normal or there is a way to solve the problem quickly by setting some parameters (for example I've realized that if I set the parameter lpx_k_branch to 0 the execution time for the same problem become larger than 12 hours)?
> In this case what are the parameters I have to set to improve glpk solver and so to reduce the execution time?
> (I'm using glpk from my java code by using Glpk 4.8 java interface).
>
> Thanks
>
>
> ---------------------------------
>
> ---------------------------------
> L'email della prossima generazione? Puoi averla con la nuova Yahoo! Mail
[Prev in Thread] | Current Thread | [Next in Thread] |