[Symphony] solving block-angular structured MILPs

[Rudan János]

Hello,

thanks for the answer.

I'd like to use Symphony because it can solve successfully the
problem. I've generated several problems with the same structure but
with perturbed parameters, and tried to solve them with both DIP and
Symphony.
In case of DIP, usually the ALPS searching seems to stuck in or
progressing very-very slowly towards the solution. I couldn't get
solution after more than 1000 seconds of running.
In case of Symphony, it solves the problem optimally in about 120-130 seconds.

I've also tried out the CbC, which solves each of the same problems in
10-15 seconds, which is quite impressive. Maybe CbC can use the
available structural information?

Best,
János

2012/3/12 Ted Ralphs <ted at lehigh.edu>:
> Hi János,
>
> SYMPHONY does not have the ability to exploit this block structure. Is there
> a reason you don't just use DIP?
>
> Cheers,
>
> Ted
>
> On Mon, Mar 12, 2012 at 11:29 AM, Rudan János <rudanj at gmail.com> wrote:
>>
>> Hello,
>>
>> in my current work I have to solve huge MILPs (having around 25.000
>> constraints and 5000 variables from which 3800 are binary variables).
>> Each of the problems have a really clear block-angular structure, I
>> can specify the border of the blocks in the constraint matrix exactly.
>> I guess that this should significally decrease the solving time,
>> because with the help of them independent subproblems could be
>> formulated and solved.
>>
>> Is there any way to provide this additional structural information to
>> Symphony? I didn't found any way to pass this kind of structural info
>> towards the solver. (For example in case of DIP, there's the MILPBlock
>> application where I can specify explicity the blocks in the constraint
>> matrix - I'm looking for a similar solution for Symphony).
>>
>> Thanks for the help.
>>
>> Best regards,
>> János Rudan
>>
>> ---
>> PhD student
>> PPCU, Budapest, Hungary
>>
>> _______________________________________________
>> Symphony mailing list
>> Symphony at list.coin-or.org
>> http://list.coin-or.org/mailman/listinfo/symphony
>
>
>
>
> --
> Dr. Ted Ralphs
> Associate Professor, Lehigh University
> (610) 628-1280
> ted 'at' lehigh 'dot' edu
> coral.ie.lehigh.edu/~ted
>