[Coin-discuss] extreme rays and OsiClp

Anureet Saxena anureet at yahoo.com
Fri Mar 16 22:05:34 EDT 2007


hello,
   
  I am using OsiClp to extract extreme points/rays of a certain polyhedron by solving a linear program. When the linear program is unbounded, the extreme rays returned by OsiClp have very large coefficients which creates numerical difficulties. Scaling these extreme rays is one option, but it doesn't seem to work well too. Sample code to reproduce the problem is available at,
   
  http://www.andrew.cmu.edu/user/anureets/tmp/primalRay.tgz

  I was wondering if anyone has used OsiClp to extract extreme rays and knows some simple trick to get around this problem. I have tried putting off the presolve options but with little help.
   
  Interestingly, I had raised this issue a few months back (email attached). Since being able to extract extreme rays/points is one of the fundamental advantages of using a simplex procedure, it will be very nice if we could fix this deficiency in OsiClp.
   
  regards & thanks,
  Anureet Saxena
   
  Earlier email (oct 2006): 
  hi,

I am trying to use OsiClpSolverInterface::getPrimalRay(const int) to extract
extreme rays of a certain unbounded linear program. I have two questions and
would  really appreciate if someone could help me with these: 

1. The primal rays returned by getPrimalRay() are double arrays with very large
numbers. Even for a very simple linear program with 2 variables and one
constraint the extreme ray seems to have numbers of the order of 10^8. A very
simple example and code to illustrate this is available at:

http://www.andrew.cmu.edu/user/anureets/tmp/primalRay.tgz

As I understand, OsiClp is performing some kind of persolve analysis which
leads to this behaviour. 

a. Is it possibe to instruct the presolver to return "neat" extreme rays?
b. Is it possible to turn off this presolve analysis? 

2. The argument to getPrimalRays(const int) seems redundant since the code
anyway returns only one extreme ray (This is also the case with OsiCpx).

regards,
anureet






Anureet Saxena
 anureet at cmu.edu
 anureet at yahoo.com
 
--This mail was delivered to you on 100% recyclable electrons.
 
---------------------------------
Be a PS3 game guru.
Get your game face on with the latest PS3 news and previews at Yahoo! Games.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/coin-discuss/attachments/20070316/7d313293/attachment.html>


More information about the Coin-discuss mailing list