[Coin-lpsolver] getPrimalRay()

[John J Forrest]

Anureet,

I am busy at present, but it might be useful to do what you suggest.  At 
present solvers just return whatever ray they have, but it would be 
possible to change code so that if more than one ray was asked for then 
the original + better rays would be returned.  It would be more useful if 
this facility was done by the base OsiSolverInterface class rather than 
OsiClp or OsiXpr etc.

John Forrest



Anureet Saxena <anureet at yahoo.com> 
Sent by: coin-lpsolver-bounces at list.coin-or.org
10/25/2006 07:34 PM

To
John J Forrest/Watson/IBM at IBMUS
cc
coin-lpsolver at list.coin-or.org, anureet at yahoo.com
Subject
Re: [Coin-lpsolver] getPrimalRay()






hi John,

Thanks for your reply.
Setting OsiDoPresolveInInitial to false (using setHintParam()) does switch 
off
the presolve, but it does not resolve the problem. The extreme ray 
returned
still has the same large numbers.

Scaling is indeed a solution but it is not very good and affects other 
parts of
the code, as you pointed out.

I was wondering if there is a way to extract extreme rays from the tableau
using the textbook method of using columns with no pivotable row. The good
thing about that approach is that if the problem has numerically stable 
basis,
then the extreme ray will also have reasonably sized coefficients. Please
excuse me if this is too much to ask frm the current implementation in 
OsiClp,
as I am not at all familiar with the details.

If anyone has worked with extracting extreme rays using OsiClp and has an
approach to avoid very large numbers, please let me know.

Thanks for your replies,
Regards,
Anureet Saxena



Anureet Saxena
 anureet at cmu.edu
 anureet at yahoo.com
 
--This mail was delivered to you on 100% recyclable electrons.

__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 
_______________________________________________
Coin-lpsolver mailing list
Coin-lpsolver at list.coin-or.org
http://list.coin-or.org/mailman/listinfo/coin-lpsolver

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/clp/attachments/20061026/1e4897ec/attachment.html>