[Cbc] Fwd: CBC

[Petros Karasakalidis]

Hi Giacomo,

Thanks for looking at this. The Matlab interface is using the C++ version
of the cbc as is, so I m not sure if it really matters.

What is interesting is that when I know whether a problem is infeasible the
output of the function is correctly as follows:

------------------------------------------------------------------
 This is CBC v2.8.8
 Author: John J. Forrest

 Problem Properties:
 # Decision Variables:   2730 [0 Integer, 945 Binary]
 # Linear Constraints:   2830 [13013 nz]
------------------------------------------------------------------

 *** TERMINATION: LINEAR RELAXATION INFEASIBLE ***
 Objective Value: 1.79769e+308
 Gap:             1.79769e+308 (100.00%)
 Searched:        0 nodes [0 LP iterations]
------------------------------------------------------------------

What makes it a little further tricky and makes me think there is a bug
here is that in my test run (for which I have the inputs in the dropbox
link above) I get an fval of sth like 1e+50 whereas the f'*x is sth like
0.5011 which is more close to reality. However, I know that the resulting x
is not right because even though satisfies all my constraints, it does
converts my binary variables to reals and gives me this following
(repeating the previous output here).

------------------------------------------------------------------
 This is CBC v2.8.8
 Author: John J. Forrest

 Problem Properties:
 # Decision Variables:   2730 [0 Integer, 945 Binary]
 # Linear Constraints:   2830 [13013 nz]
------------------------------------------------------------------
Pass   1: suminf.    0.81917 (8) obj. -0.0499381 iterations 198

 *** SUCCESSFUL TERMINATION ***
 Objective Value: 1e+50
 Gap:             1e+50 (100.00%)
 Searched:        0 nodes [223 LP iterations]
------------------------------------------------------------------

Any further feedback/clue would be greatly appreciated.

Thanks


On Mon, Feb 17, 2014 at 6:47 PM, Giacomo Nannicini <giacomo.n at gmail.com>wrote:

> Hi,
> I never used the Matlab interface, but looking at the output, it seems
> that the LP relaxation of the problem is infeasible (look at the sum
> of infeasibilities).
> "Successful termination" probably just means that there was no error,
> but the problem has no feasible solution anyway. If you think that
> there must be a feasible solution, I'd suggest checking your data.
>
> Giacomo
>
> On Tue, Feb 18, 2014 at 9:56 AM, Petros Karasakalidis
> <petkaras at gmail.com> wrote:
> > Hi all,
> >
> > I am forwarding this email here as I m not sure my previous email went
> > through since I cannot see it on the active tickets. Any help would be
> > greatly appreciated.
> >
> > Thanks
> >
> > ---------- Forwarded message ----------
> > From: Petros Karasakalidis <petkaras at gmail.com>
> > Date: Sat, Feb 15, 2014 at 8:21 PM
> > Subject: CBC
> > To: cbc-tickets at list.coin-or.org
> >
> >
> > I use CBC through the opti toolbox (having Matlab as an interface) in
> order
> > to solve a MILP problem. Everything was working until the point I used as
> > inputs the data I have here
> >
> > https://www.dropbox.com/s/qwg4x50bcrhq066/cbc_inputs.xlsx
> >
> >
> > The file is about 20MB and wouldnt fit into this email. For some reason,
> the
> > objective results into a very large number (like 1e+50) while the status
> > output shows success as follows:
> >
> > ------------------------------------------------------------------
> >  This is CBC v2.8.8
> >  Author: John J. Forrest
> >
> >  Problem Properties:
> >  # Decision Variables:   2730 [0 Integer, 945 Binary]
> >  # Linear Constraints:   2830 [13013 nz]
> > ------------------------------------------------------------------
> > Pass   1: suminf.    0.81917 (8) obj. -0.0499381 iterations 198
> >
> >  *** SUCCESSFUL TERMINATION ***
> >  Objective Value: 1e+50
> >  Gap:             1e+50 (100.00%)
> >  Searched:        0 nodes [223 LP iterations]
> > ------------------------------------------------------------------
> >
> > Would appreciate any feedback or direction towards the solution. Feel
> free
> > to ask any questions that might be helpful.
> >
> > Thanks
> >
> >
> > _______________________________________________
> > Cbc mailing list
> > 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/20140217/fe97c221/attachment.html>