[Cbc] Cbc CLI Infeasible Error

[John Forrest]

Matthew,


Fixed in trunk - do you want it fixed in stable?

OsiPresolve did dual tests before tightening bounds.  As z has positive 
cost and useless coefficient dual tests said put it to lower bound i.e. 
0.4. Tightening then said integer variable is fixed to 0.4 so 
infeasible.  I said it was a stupid bug.

Just moved tightening before dual tests.

John
On 30/07/14 18:17, Matthew Gidden wrote:
> Hi John,
>
>
> On Wed, Jul 30, 2014 at 12:11 PM, John Forrest 
> <john.forrest at fastercoin.com <mailto:john.forrest at fastercoin.com>> wrote:
>
>     Matthew,
>
>     Should be easy to fix - looks like a really stupid bug.
>
>     If I change the lower bounds on y,z to 2,1 it works! As rounding
>     up should be first thing that happens, you can see why I say stupid.
>
> Ok, thanks John! I can change my bounds on my unit test for now, but 
> I'm glad I was able to bring this to your attention. Thanks again for 
> looking into it.
>
>
>     John Forrest
>
>
>     On 30/07/14 16:20, Matthew Gidden wrote:
>>     Hi all,
>>
>>     I've been posting a lot recently.. first of all, thank you for
>>     replying!
>>
>>     I've unit tested my application pretty vigorously, and a Cbc API
>>     unit test broke when I moved to the CLI implementation.
>>     Specifically, I set up the following toy problem:
>>
>>     min 2x + 0.5y + 1.8z
>>     st     x  +     y            > 4.4
>>                        y +     z  < 3.1
>>     x in [1.3, 5], y in [1.7, 5], z in [0.4, 5]
>>     y, z integer
>>
>>     The unit test originally passed using the OsiCbcSolverInterface's
>>     initial solve and branch and bound methods. I updated it to use
>>     an OsiClpSolverInterface that is fed a CbcModel and call
>>     CbcMain0/1 on said model. The unit test now fails with the
>>     following relevant output:
>>
>>         Welcome to the CBC MILP Solver
>>         Version: 2.8.7
>>         Build Date: Dec 28 2013
>>         command line - exchng -solve -quit (default strategy 1)
>>         Continuous objective value is 5.47 - 0.00 seconds
>>         Cgl0000I Cut generators found to be infeasible! (or unbounded)
>>         Pre-processing says infeasible or unbounded
>>
>>
>>     To double check that indeed a feasible solution existed, I solved
>>     the maximization form of the same instance. Attached are MPS
>>     files for the max and min forms, and I have verified the same
>>     behavior using the cbc CLI (i.e., min is returned as infeasible
>>     and max solves).
>>
>>     I'm curious if anyone can replicate this and/or has any advice.
>>
>>     Cheers,
>>
>>     -- 
>>     Matthew Gidden
>>     Ph.D. Candidate, Nuclear Engineering
>>     The University of Wisconsin -- Madison
>>     Ph. 225.892.3192 <tel:225.892.3192>
>>
>>
>>     _______________________________________________
>>     Cbc mailing list
>>     Cbc at list.coin-or.org  <mailto:Cbc at list.coin-or.org>
>>     http://list.coin-or.org/mailman/listinfo/cbc
>
>
>     _______________________________________________
>     Cbc mailing list
>     Cbc at list.coin-or.org <mailto:Cbc at list.coin-or.org>
>     http://list.coin-or.org/mailman/listinfo/cbc
>
>
>
>
> -- 
> Matthew Gidden
> Ph.D. Candidate, Nuclear Engineering
> The University of Wisconsin -- Madison
> Ph. 225.892.3192

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/cbc/attachments/20140730/15a25cce/attachment-0001.html>