[Cbc] Never-ending

[John Forrest]

Mattia,

Really for flexibility you need to write an event handler (see examples 
in Cbc/examples) - but one parameter is maxnodes.  Better yet is to also 
add  ratiogap or allowablegap to stop when reasonable solution.

Your problem has a lot of symmetry - see if you can break that symmetry.

Optimal answer is 23652.0 - which code gets quite soon - so doing

cbc payg_final.mod%inp-2395.dat -ratiogap 0.01 -solve

gives

Result - Optimal solution found (within gap tolerance)

Objective value:                23652.00000000
Lower bound:                    23587.000
Time 299.32 seconds

John Forrest

On 27/09/14 07:38, Mattia Buccarella wrote:
> Hello everyone.
>
> I am using the standalone Cbc 2.8.9, compiled with the GNU MathProg 
> support.
> I have the problem that it gets stuck forever on certain MILP instances.
>
> One of these instance can be found at the following url:
> INSTANCE: https://dl.dropboxusercontent.com/u/44504233/inp-2395.dat
> MODEL: https://dl.dropboxusercontent.com/u/44504233/payg_final.mod
>
> I tried with the
>
> -cuts on
>
> option to see if something changed but the result was the same, and 
> now I don't know what to try. I tried the options -nodelim and -nodlim 
> to see if I could stop it after a while, but it did not look working.
>
> At the moment, I really need it to stop after spending some time on a 
> certain solution. Do you know how could I do it? The only required 
> thing is that, when I stop it, the found solution (partial) should 
> still be feasible. So optimality is not fundamental at the moment. 
> Feasibility it is instead.
>
> Thank you very much for your time.
>
> --
> Mattia Buccarella
>
>
> _______________________________________________
> 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/20140928/ed6efb02/attachment.html>