[Scip] how to implement Lagrangian relaxation based branch-and-bound?

[Gerald Gamrath]

Dear Xiangyong,

no, there is no specific documentation about the implementation of a 
Lagrangian relaxation in SCIP, but the general documentation about 
relaxation handlers should be a good point to start with.

Best,
Gerald

Am 19.08.2014 um 16:00 schrieb Xiangyong Li:
>  Dear Dr. Gamrath,
>
> Thanks. Is there any reference regarding implementation of Lagrangian 
> relaxation based branch and bound in the SCIP?
> Best regards,
>
>
> Xiangyong
>
>     *From:* Gerald Gamrath <mailto:gamrath at zib.de>
>     *Date:* 2014-08-19 18:26
>     *To:* lixiangyong at 163.com <mailto:lixiangyong at 163.com>; scip
>     <mailto:scip at zib.de>
>     *Subject:* Re: [Scip] how to implement Lagrangian relaxation based
>     branch-and-bound?
>     Dear Xiangyong,
>
>     yes, you can use SCIP for that. You need to implement a relaxation
>     handler, see
>     http://scip.zib.de/doc/html/RELAX.php
>
>     Additionally, you can turn off LP solving by setting the parameter
>     "lp/solvefreq" to -1.
>
>     Best,
>     Gerald
>
>     Am 19.08.2014 um 03:44 schrieb lixiangyong at 163.com:
>>
>>     Dear all,
>>
>>
>>     I want to implement a Lagrangian relaxation based on branch-and
>>     bound algorithm.  That is,  at each node of the enumeration tree,
>>       I do not solve LP relaxation,
>>     but Lagrangian dual  problem to determine a good lower bound and also
>>     try to get an upper bound. Also I have my own branching strategy.
>>
>>     Can I use SCIP
>>     as a framework to implement my algorithm? In other words, SCIP
>>      takes charge of
>>     most of the job and I only control the solving of Lagrangian Dual and branching.
>>
>>
>>
>>     Is it possible to use solvecallback to solve a Lagrangian dual at each node and turn off solving the LP relaxation?  Is there any way to pass this new lower bound to SCIP
>>     and ask SCIP to control the search tree?
>>
>>     Is there any reference in the SCIP manual?
>>
>>     Would you please give me some suggestions? Thanks.
>>
>>
>>
>>     Xiangyong Li
>>
>>
>>
>>     _______________________________________________
>>     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/20140820/15b9e160/attachment.html>