[Coin-discuss] LP and TM solutions (Heesu Hwang)

Laszlo Ladanyi ladanyi at us.ibm.com
Tue Sep 6 15:49:59 EDT 2005


Actually, negative indices may appear in the output, but only in the output of
the LP process. They indicate that this is a variable that is known only to
that LP process (hence it must be an algorithmic variable, i.e., one generated
via column generation). The index is flipped to positive once the variable is
sent to the TM. This happens when the processing of the search tree node is
finished (the node is fathomed or braqnched on) and the variable is still in
the formulation. 

--Laci

On Sun, 4 Sep 2005, Francois Margot wrote:

> 
> Heesu:
> 
> Try to run valgrind on your code. My guess is that memory access errors
> occur somewhere (likely in your own code). Generating heuristic solutions
> properly with Bcp is a little bit tricky. There is an example in 
> the BAC code available in COIN/Examples/BAC.
> 
> Negative variable numbers may exist internally in Bcp, but they should
> not appear in the output. Variable 1481 in a problem with 80 variables
> is impossible if no column generation is done.
> 
> Francois
> 
> On Sun, 4 Sep 2005 coin-discuss-request at list.coin-or.org wrote:
> 
> > Date: Sat, 3 Sep 2005 12:59:23 -0500
> > From: "Heesu Hwang" <hxh9528 at omega.uta.edu>
> > Subject: [Coin-discuss] LP and TM solutions
> > To: "'Discussions about open source software for Operations Research'"
> > 	<coin-discuss at list.coin-or.org>
> > Message-ID: <000001c5b0b1$37dc4690$020ba8c0 at Peter>
> > Content-Type: text/plain;	charset="us-ascii"
> >
> > Hello,
> >
> > Following output is the result of a LP relaxed problem.
> > Notice that it is a binary solution and that I am using Default
> > test_binary() to check the feasibility.
> > LP packs the feasible solution, and TM unpacks it.
> > However the solution that TM displays is different from what LP packed.
> >
> > 1) Anybody has an idea what is going on in here?
> >
> > 2) Do negative internal indices of Algo var make any trouble?
> >
> >
> > Thanks,
> >
> > Heesu Hwang.
> >
> > +++++++++++++++++++++++++++++++++++++++++++++++++++++++
> >
> > LP: *** Starting iteration 11 ***
> >
> > LP:   Matrix size: 80 vars x 31 cuts
> > LP:   Solution value: -1002851.9760 / 2 , 4
> > LP: Default display_lp_solution() executed.
> >  LP : Displaying LP solution (RelaxedSolution) :
> > LP: Default display_lp_solution() executed.
> >   LP : Displaying solution :
> >  Algo  var (internal index:      5                    ) at 1.0000
> >  Algo  var (internal index:     16                    ) at 1.0000
> >  Algo  var (internal index:     33                    ) at 1.0000
> >  Algo  var (internal index:     45                    ) at 1.0000
> >  Algo  var (internal index:     48                    ) at 1.0000
> >  Algo  var (internal index:     49                    ) at 1.0000
> >  Algo  var (internal index:     50                    ) at 1.0000
> >  Algo  var (internal index:     51                    ) at 1.0000
> >  Algo  var (internal index:     57                    ) at 1.0000
> >  Algo  var (internal index:    152                    ) at 1.0000
> >  Algo  var (internal index:   1431                    ) at 1.0000
> >  Algo  var (internal index:  -7060                    ) at 1.0000
> >
> > LP: Default test_binary() executed.
> > LP: Default pack_feasible_solution() executed.
> > TM: Default unpack_feasible_solution() executed.
> > TM: Solution found at 516.580 sec.
> > TM: Solution value: -959109.456000 (best solution value so far:
> > -1061466.200000)
> >
> _______________________________________________
> Coin-discuss mailing list
> Coin-discuss at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/coin-discuss
> 




More information about the Coin-discuss mailing list