[Coin-lpsolver] Presolve removed variable's values?

John J Forrest jjforre at us.ibm.com
Thu Feb 15 11:51:06 EST 2007


Sathi,

Here are some answers.

coin-lpsolver-bounces at list.coin-or.org wrote on 02/15/2007 11:02:30 AM:

> Hi,
> 
> I have an LP problem in which presolve removes a lot of variables. I 
have
> attached the presolve statistics for a model I have.
> 
> While solving the presolved model, I would like to use some 
relationships
> between the original variables in guiding the pivoting heuristic.  So I 
am
> interested in two issues:
> 
> 1. Which variables are removed? How do the variables in the presolved
> model map to the original variables?

The ClpPresolve object has methods - originalColumns()and originalRows() 
which you can use to find out this information.

> 
> 2. What values are the removed variables fixed at, if any?  Or, are the
> values of some of the removed variables are only fixed during 
postsolving
> process?

For some types of variables, the removed variables will be fixed at their 
bounds - but there is no easy way to find out which bound.  For presolve 
operations which get rid of constraints like x + 5y = 7 then these values 
are only known during postsolve.  In your case most of the operations are 
probably the result of constraints like x + 2y +3z <= 0 in which case they 
would be fixed to lower bound.

> 
> I am at present looking at presolve literature and also the COIN 
presolve
> codes to figure out these issues. Any help would be useful to me.
> 
> Thanks,
> Sathi
> 
> ___________________________________________________________________
> [attachment "presolve_stats.txt" deleted by John J Forrest/Watson/IBM] 
> _______________________________________________
> 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/20070215/445e8c40/attachment.html>


More information about the Clp mailing list