Re: [Help-glpk] Conflict graph is too big

[Robbie Morrison]

------------------------------------------------------------
Subject:        Re: [Help-glpk] Conflict graph is too big
From:           Andrew Makhorin
Date:           Mon, 09 May 2011 17:46:52 +0400
------------------------------------------------------------

[snip]

> Your problem is hard for glpk to be solved to
> optimality. In such cases you may either wait for a
> time or, if you are not interested in exact optimum,
> specify a time limit (say, --tmlin 600) or a mip gap
> (say, --mipgap 0.10).
>
> I managed to solve your mip to optimality for 2 minutes
> using --gomory (gomory cuts) and --pcost (pseudo-cost
> branching):

Hello Lizzy, everybody

It would not be difficult to write a script that
submitted different sets of parameters to "GLPSOL
--tmlim 3600" say and produced a report.  The kind of
job that could be run overnight.  Or the --mipgap could
be set wide and increasingly tightened.  Similar
functionality could even be built into GLPSOL ..
although I don't see any real merit in that.

I rather suspect that this comment prompts the notion
of solver tuning and adaptive behavior.  And that there
has been a lot of work done on classifying problems and
matching solvers (that I am totally unaware of).

best wishes, Robbie
---
Robbie Morrison
PhD student -- policy-oriented energy system simulation
Technical University of Berlin (TU-Berlin), Germany
University email (redirected) : address@hidden
Webmail (preferred)           : address@hidden
[from Webmail client]