[Scip] Using a greedy heuristic to improve convergence of branch and price

[Gerald Gamrath]

Dear Abde Ali,

yes, this can be done very easily in SCIP. Please check out "How to add 
primal heuristics" in the SCIP documentation:
http://scip.zib.de/doc/html/HEUR.php
Since you want to implement an improvement heuristic, I would recommend 
to also have a look at the oneopt heuristic:
http://scip.zib.de/doc/html_devel/heur__oneopt_8c_source.php
In particular the tvalues for the heuristic properties and the way it 
decides whether to run or not (seach for "bestsol") might be useful for you.

Best,
Gerald

Am 08.03.2014 08:26, schrieb Abde Ali Kagalwalla:
> Hello SCIP Developers,
>
> I am trying to solve a very large problem with more than 10^6 binary 
> variables. I am currently struggling to get the upper and lower bounds 
> to converge using my implementation of branch and price on SCIP. I 
> have also developed a simple heuristic that can take as input a 
> feasible solution and then try to improve it with some small greedy 
> optimization.
>
> I am wondering if it is possible to embed this greedy heuristic in the 
> branch and bound tree. For example, in my branch and bound every time 
> I find an integral solution, I could call my greedy heuristic and try 
> to improve the solution. This could potentially prune out many nodes 
> in the branch and bound tree.
>
> Any ideas on how I can go about adding such a heuristic within the 
> SCIP framework ? I would really appreciate any suggestions on this.
>
> Sincere Regards,
>
> Abde Ali
>
>
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> http://listserv.zib.de/mailman/listinfo/scip

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.zib.de/pipermail/scip/attachments/20140308/a636c088/attachment.html>