[Cbc] [EXTERNAL] Re: Question about "**" prefix in CBC solution file lines

[Watson, Jean-paul]

Hi John,

The infeasibility is minimal – in the 1e-6 range. Your explanation makes sense; I'll just modify our CBC solver plugin to check+suppress the "**" prefix.

As to why it doesn't happen that frequently, I haven't a clue…

Jean-Paul

From: John Forrest <john.forrest at fastercoin.com<mailto:john.forrest at fastercoin.com>>
Date: Wed, 4 Apr 2012 08:56:00 +0100
To: <cbc at list.coin-or.org<mailto:cbc at list.coin-or.org>>
Subject: [EXTERNAL] Re: [Cbc] Question about "**" prefix in CBC solution file lines

Jean-Paul,

How infeasible?

The most likely explanation is that when scaled the problem was feasible within tolerance, but when the solution was unscaled it violated a tolerance.  Clp would have reported that as optimal but with a secondary status flag set.  OsiSolverInterface would drop that information.  Why this happens so rarely is odder.

John

On 03/04/12 04:40, Watson, Jean-paul wrote:
I just encountered our first Coopr user that has generated models such that the resulting CBC solution files contain rows for variable lines that are prefixed by "**".

When I look at the code (CBCGenSolution.cpp), it is obvious that the "**" indicate – in the case of variables – that the bounds have been violated for this particular variable.

This makes sense for a brief moment, until I realize that CBC is reporting the solution is optimal. Solutions can of course violate strict bounds when tolerances are considered, but I would assume I would see this behavior on any of the numerous other models we have solved with CBC.

Can anyone shed some light on this condition, and how the resulting solution should be interpreted?

Thanks,

Jean-Paul



_______________________________________________
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
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/cbc/attachments/20120408/3f8a38b7/attachment.html>