[Coin-discuss] [Fwd: RE: need updated files]

[Matthew Galati]

Peter,

I am fwd'ing this to coin-discuss. Folks, this is a BCP question, but 
bcp-discuss is fairly inactive, so I am using coin-discuss - it relates 
to the use of my example of doing branch and price with BCP found here: 
http://coral.ie.lehigh.edu/~coin/COIN_EXAMPLES/AAP_BP/Doc/aap/index.html

Peter, I would say that in general, "no" CLP is not as "efficient" as 
CPX - but it is competitive. What you are seeing has nothing to do with 
the efficiency of the LP solve. It has to do with the way that BCP works 
with different LP solvers - perhaps the issue is in the OSI layer - 
sometimes, the OSI layer can get out of sync as solver implementations 
change. I have not used BCP in a long time - when I did, I mainly worked 
with CLP - so I can't say too much about whether or not CPX will work 
with it. Hopefully someone on the list can share their experience using 
BCP + CPX.

If no one gets around to helping you, I will have a look sometime later. 
Too busy at the moment.

Thanks,
Matt

-- 
Matthew Galati - Optimization Developer
SAS Institute - Analytical Solutions
Phone 919-531-0332, R5327 
Fax   919-677-4444
http://coral.ie.lehigh.edu/~magh
http://ordlnx2.na.sas.com/projects/OptWiki
http://www.sas.com/technologies/analytics/optimization/



-------- Original Message --------
Subject: 	RE: need updated files
Date: 	Sat, 23 Apr 2005 16:34:20 -0500
From: 	Heesu Hwang <hxh9528 at exchange.uta.edu>
To: 	Matthew Galati <magh at lehigh.edu>



Sorry for bugging you with short knowledge on Unix, compilations, and so on.
However, I have interesting results...
 
I tested CLP and CPX with your small/6.1.aap
and got following results.
 
Can I say CLP is more efficient than CPX for your problem instance?
I just want to make it sure that both will give same solutions.
 
Thanks,
 
Peter.
 
##############CLP results######################

Clp0006I 4  Obj 48.6667
Clp0000I Optimal - objective value 48.6667
LP:   Matrix size: 53 vars x 7 cuts
LP:   Solution value: 48.6667 / 2 , 4
 LP: Default test_feasibility() executed.
 LP: Default test_full() executed.
Clp0006I 0  Obj -100000 Primal inf 8 (8)
Clp0006I 12  Obj -99954
Clp0000I Optimal - objective value -99954
LP:   Terminating and fathoming due to proven high cost.
TM: Running time: 0.550
TM: search tree size: 15   max depth: 5
LP statistics:
   time in cut generation  :        0.000 sec
   time in var generation  :        0.000 sec
   time in heuristics      :        0.000 sec
   time in solving LPs     :        0.270 sec
   time in strong branching:        0.060 sec
 
TM: The best solution found has value 46.000000
 TM: Default BCP_tm_user::display_feasible_solution() executed.
  Algo  var (internal index:   -135                    ) at 1.0000
 
#####################CPX results#########################
LP: *** Starting iteration 29926 ***
LP node written in file sampleAAP.mps
LP:   Matrix size: 29966 vars x 7 cuts
LP:   Solution value: 125.6000 / 4 , 0
 LP: Default test_feasibility() executed.
 LP: Default test_full() executed.
LP:   Primal feasibility lost.
LP:   Generating columns before fathoming/resolving
TM: Time has ran out.
TM: Best lower bound in this phase: 39.333333
TM: Running time: 4509.420
TM: search tree size: 9   max depth: 4
LP statistics:
   time in cut generation  :        0.000 sec
   time in var generation  :        0.000 sec
   time in heuristics      :        0.000 sec
   time in solving LPs     :     3915.640 sec
   time in strong branching:        0.030 sec
 
TM: The best solution found has value 52.000000
 TM: Default BCP_tm_user::display_feasible_solution() executed.
  Algo  var (internal index:    -73                    ) at 1.0000
 

     


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/coin-discuss/attachments/20050423/9b26ea33/attachment.html>