[Couenne] Could Couenne accept a known feasible solution

[Pietro Belotti]

Hi Suyun,

if the initial point is specified in the .nl file, then Couenne does read
and check that solution for MINLP feasibility, and if all is good it uses
its objective function as a cutoff.

Regards,
Pietro

On Mon, Apr 2, 2018 at 6:55 PM, Suyun Liu <sul217 at lehigh.edu> wrote:

> Hi Pietro,
>
> Appreciate for your response very much. As far as I know, when
> constructing a model through Pyomo and writing it to the .nl file,  the
> so-called* primal initial guess* is then indicated in that file. Does
> that mean Couenne will ignore the primal initial guess (a possibly
> feasible solution) in the .nl file?
>
> Best Regards,
> Suyun
>
> On Mon, Apr 2, 2018 at 1:18 PM, Pietro Belotti <petr.7b6 at gmail.com> wrote:
>
>> Hi Suyun,
>>
>> Couenne doesn't have a way to add a solution for use in pruning, so I
>> think only art_cutoff (see my other response) can be used.
>>
>> Regards,
>> Pietro
>>
>> On Sat, Mar 31, 2018 at 4:48 PM, Suyun Liu <sul217 at lehigh.edu> wrote:
>>
>>> Hi all,
>>>
>>> I have a question about initialization. Could Couenne accept a known
>>> feasible solution (hence upper bound) to accelerate pruning nodes on the
>>> tree? If so, which options could be used to impose the feasible objective
>>> value to the solver?  It seems like artificial cutoff doesn't work as we
>>> want. I am using Pyomo as the interface and the warmstart is not supported
>>> actually.
>>>
>>> Thanks!
>>>
>>> Best Regards,
>>> Suyun
>>>
>>> _______________________________________________
>>> 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/20180402/6bf8a70e/attachment.html>