[Scip] help : How to write dual problem
Abdelkader Ouali
oualiaek at hotmail.fr
Wed Jul 9 22:08:02 CEST 2014
Dear Gregor,
Yes, i confirm very nice as an exercice ;)It's much clear, now
Thank's and regards
---Abdelkader Ouali
Date: Wed, 9 Jul 2014 17:19:06 +0200
From: hendel at zib.de
To: scip at zib.de
Subject: Re: [Scip] help : How to write dual problem
Dear Abdelkader Ouali,
I am afraid there is no ad-hoc method available for this in SCIP.
The methods available through the linear
constraint
handler might be sufficient here:
1. Create a secondary SCIP-instance dualscip and a problem inside
dualscip.
2. Create variables for every linear constraint in dualscip
3. Create a new linear constraint in dualscip for every variable
4. Iterate over the constraints once in order to have their
variables, vals (=coefficients), and number, and add
the coefficients for the corresponding dual variable of this
constraint to every dual constraint it should appear in.
5. Now you can use the LP-writer for the dualscip-Problem.
This is quite a nice SCIP-exercise, isn't it? :)
The reason why SCIP does not come with such a functionality (yet)
is its constraint-based nature. It does not
always make sense to speak of the dual of a SCIP problem. The dual
of the LP that SCIP builds internally might
be different from what you exspect, because the problem has
already undergone a transformation through
various presolving steps.
Kind regards
Gregor
Am 09.07.2014 10:01, schrieb Abdelkader Ouali:
Hello,
How to write dual problem of the original problem in lp
format file
using SCIP routines
Thanks
---
Abdelkader Ouali
_______________________________________________
Scip mailing list
Scip at zib.de
http://listserv.zib.de/mailman/listinfo/scip
_______________________________________________
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/20140709/12a4f797/attachment.html>
More information about the Scip
mailing list