[Cbc] Another problem "long to be solved"
John Forrest
john.forrest at fastercoin.com
Tue Nov 18 07:03:16 EST 2014
Yves,
In trunk there are some new/modified options.
There was a parameter "experiment". I have left some old meanings, but
introduced some new ones. For difficult problems I would suggest
-experiment 3000 !!. This actually switches on multiple root solvers
(4), does more aggressive preprocessing, adds some extra root cut
generators AND switches on most diving heuristics.
When I tried, the code had no problems with long.lp, but I would suggest
you look at a new parameter divesolves. This can increase the number of
solves a diving heuristic can do. The bottom digit is ignored for
number of solves but can switch on something which could help you a
lot. Normally the diving heuristics fix some integer variables if they
are at their bound, however they do not fix satisfied general integer
variables - not at bound - so if 0 <= x <= 20 and x is at 12 then it
does not get fixed. I have allowed this behaviour to change so
-divesolves 301 will allow up to 300 solves and will fix some satisfied
general integers.
You may also want to look at diveoptions and even set -DDIVE_PRINT=1 in
configure. This switches on some printout which may help you understand
what is happening in which heuristic.
John
On 17/11/14 11:43, Yves Touchard wrote:
> Hello,
>
> I split an huge problem in several little problems which sizes are
> (IMO) equivalent.
> Solving each of them lasts several seconds (no more than 2-3 seconds)
> except one (several minutes).
> I tried different combination of options (w/wo diving, w/wo zero
> root...) without any success. (threads 104 ratio 0.05 cutlength 40
> multiple 4 divingv on divingl on divingf on divingg on divingc on
> divingp on zero root)
>
> Have you an idea to help this problem to be fastly solved ? Or my
> splitting and the cost function are not well balanced ?
> The constraint file may be downloaded here:
> https://www.dropbox.com/s/lbd7o2bdn5392m1/long.lp?dl=0
> One of the sol file may be downloaded here:
> https://www.dropbox.com/s/a7foc860ap6lbmn/long.txt?dl=0
>
> I use the the trunk version, revision #2092, compiled with gcc 4.9.0
>
> Thanks and regards
>
> Yves
>
> ------------------------------------------------------------------------------
> This message and any attachments (the "message") are confidential and
> intended solely for the addressee(s). Any unauthorized use or
> dissemination is prohibited. E-mails are susceptible to alteration.
> Neither DxO Labs nor any of its subsidiaries or affiliates shall be
> liable for the message if altered, changed or falsified.
> Ce message et toutes les pièces jointes (ci-après le "message") sont
> confidentiels et établis à l'intention exclusive de ses destinataires.
> Toute utilisation ou diffusion non autorisé est interdite. Tout
> message électronique est susceptible d'altération. DxO Labs et ses
> filiales déclinent toute responsabilité au titre de ce message s'il a
> été altéré, modifié ou falsifié.
>
>
> _______________________________________________
> Cbc mailing list
> Cbc at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/cbc
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/cbc/attachments/20141118/a917c99c/attachment.html>
More information about the Cbc
mailing list