[Clp] Performance optimization (Hauke Pribnow)
Gleb Belov
gleb.belov at tu-dresden.de
Wed Sep 10 12:19:30 EDT 2008
Hello,
I agree with Tuomo. Keeping basic solution can even be sped up (about
2x) by saving the corresponding factorization. This can be copied 1:1
from the function strongBranching; you only have to extend the status
array by inserting the initial status, e.g., "atLower", for the new var.
Cheers Gleb
> Hi Hauke.
>
> Why not keep an optimal basic solution for the problem without the last column, and use it as a start solution / start basis for LP runs with the last column added. Then you just need to make sure that you reoptimize with the proper LP algorithm, that is, the primal Simplex if the basic solution is primal feasible with respect to the full problem, and the dual Simplex if the basis is dual feasible with respect to the full problem. That should do.
>
> Another option for speedup is to cache solutions for columns appearing often.
>
> Cheers Tuomo
>
>
>
>
> ------------------------------
>
> Message: 2
> Date: Tue, 09 Sep 2008 19:10:38 +0200
> From: Hauke Pribnow <hauke.pribnow at gmx.net>
> Subject: Re: [Clp] Performance optimization
> To: clp at list.coin-or.org
> Message-ID: <48C6AE0E.2040703 at gmx.net>
> Content-Type: text/plain; charset=ISO-8859-1; format=flowed
>
> Thanks a lot for your replies. I'll take a look at your suggestions in
> the next few days.
>
> Hauke
>
>
> ------------------------------
>
> _______________________________________________
> Clp mailing list
> Clp at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/clp
>
>
> End of Clp Digest, Vol 46, Issue 4
> **********************************
>
>
--
Mit freundlichen Gruessen
G. Belov
----------------------------------------------------------------------
Dr. Gleb Belov Gleb.Belov at tu-dresden.de
Wiss. Assistent http://www.math.tu-dresden.de/~belov
Technische Universitaet Dresden Tel: +49 351 463 34186
Institut fuer Numerische Mathematik Fax: +49 351 463 34268
Mommsenstr. 13 Tel. priv.:+49 351 3743754
01062 Dresden
More information about the Clp
mailing list