[Couenne] Infeasibility problems with couenne

[Pietro Belotti]

Hi Jan,

setting nlp_log_level to 0 might provide some more meaningful output as it
cuts all Clp log lines.

The problems seems to be here that Couenne finds the continuous LP
relaxation infeasible, after which the BB is obviously not even started. It
has happened in the past that Couenne returns an infeasible LP for very
large problems, and this seems to be one with 5000s reformulation time. Is
there any small version of your problem that you can try?

Regards,
Pietro



On Sun, Jan 21, 2018 at 10:36 PM, Jan Stampfli <jan.stampfli at outlook.com>
wrote:

> Dear all,
>
>
>
> I have some troubles getting feasible solutions for my MINLP formulation
> using couenne. My model is basically a NLP model but I have to use some
> binary variables for selecting values. Thereby, the binary variables are
> declared as integer with bounds 0 and 1 instead binary. I wrote my model in
> pyomo 5.3 (python 2.7) and using couenne 0.5.  To get feasible solutions, I
> changed the options as follows:
>
>
>
>    - lp_log_level 0
>    - bb_log_level 1
>    - nlp_log_level 1
>    - redcost_bt no
>    - optimality_bt no
>    - aggressive_fbbt no
>    - feas_tolerance 1e-4
>
>
>
> I do not really understand the output, but I think the cbc does not really
> proceed to a lower  depth of the BB. In the appendix I have the output of
> the solver and my couenne.opt. When anyone needs my model please don’t
> hesitate to ask.
>
>
>
> Thanks
>
> Jan
>
>
>
> _______________________________________________
> Couenne mailing list
> Couenne at list.coin-or.org
> https://list.coin-or.org/mailman/listinfo/couenne
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/couenne/attachments/20180122/1cb614a1/attachment.html>