Different status for the same problem using different solvers (GLPK, CBC)

[Rémy Roy]

Hello,

I've got a problem with has an infeasible solution. Using the builtin cbc solver, I get an Infeasible status with a partial solution for the problem when I try to solve it. When I use GLPK 4.52.1, I get an Undefined status for the same problem and I do not get the partial solution.

Shouldn't I get the same status for the same problem even if I use different solvers? Would it be possible there is a problem with the parsing of GLPK output for infeasible solutions?

I'm using pulp 1.5.4. You can find the problem LP at http://pastebin.com/3tih0fqg . The output from GLPK can be found at http://pastebin.com/ysQiktir .

Regards,

[Stuart Mitchell]

hmm interesting, 

from the solution file out.sol I get

Problem:    

Rows:       47

Columns:    1345 (1344 integer, 1344 binary)

Non-zeros:  1633

Status:     INTEGER UNDEFINED

Objective:  OBJ = 0 (MINimum)

which pulp reads and assumes it is undefined. I cannot find information on glpk status codes to know whether this is the expected behaviour or not. Perhaps you can ask on the glpk group?

Stu

--
You received this message because you are subscribed to the Google Groups "pulp-or-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pulp-or-discu...@googlegroups.com.
To post to this group, send email to pulp-or...@googlegroups.com.
Visit this group at http://groups.google.com/group/pulp-or-discuss.
For more options, visit https://groups.google.com/groups/opt_out.

--

Stuart Mitchell

PhD Engineering Science

Extraordinary Freelance Programmer and Optimisation Guru

www.stuartmitchell.com

[Rémy Roy]

I sent a mail on the help mailing list for GLPK. Check out https://lists.gnu.org/archive/html/help-glpk/2013-09/msg00011.html if you want to follow the discussion.

Regards,

[Rémy Roy]

This seems to be a bug with GLPK which should be fixed in a future version/release (https://lists.gnu.org/archive/html/bug-glpk/2013-09/msg00001.html).

[Stuart Mitchell]

Thanks for that.

Stu

[Rémy Roy]

A quick follow up on this one. This bug seems to be still present with the latest version of GLPK 4.53 . I'll be following up with the GLPK group.

[Rémy Roy]

Alright, this should have now been fixed in GLPK latest dev source and should be available in a future release. I personally tested the GLPK source change and the status is now INTEGER EMPTY in the solution file and this should map to the LpStatusInfeasible pulp status which should be the same as CBC now.

Regards,