[SCIP] Stopping at the next best solution than the one passed

Stefan Vigerske stefan at math.hu-berlin.de
Wed Jun 6 17:27:33 CEST 2018


Hi,

Gerald already explained why limits/objective does not work.
Would it work if you add a new variable z and a new constraint f(x) <= z 
and then minimize z instead of f(x)? Then you could set a variable bound 
on z.

Stefan

On 06/06/2018 05:07 PM, Meenarli Sharma wrote:
> Hello everyone,
> 
> Some more information about my problem.
> 
> I am using SCIP (scipampl) with ampl. I iteratively call scip to solve a
> convex MINLP. I want to pass objective value of (i)th iteration to (i+1)th
> iteration and want scip (at (i+1)th iteration) to stop when it finds a
> solution with better objective value than the one passed.
> 
> I tried modifying (automatically from ampl) the parameter limits/objective
> in the setting file (scip.set) before every iteration. But I get the
> warning, WARNING: unknown parameter <limits/objective>.
> 
> I have cross checked that I am not making any syntax error while writing
> the parameters. Also, I am not getting any such warning when I am not
> having this parameter in the settings file.
> 
> Kindly let me know if there is any suggestion/comment.
> 
> Regards,
> ~Meenarli
> 
> 
> 
> 
>> Hello Ambros,
>>
>> Thank you for the reply. I have already tried this parameter but I got the
>> error: WARNING: unknown parameter <limits/objective>. Any clue why is it
>> happening?
>>
>> Regards,
>> ~Meenarli
>>
>>
>>   already > Hi Meenarli,
>>>
>>> "limits/objective" is the parameter you are looking for.
>>>
>>> You easily inspect a list of all parameters with documentation yourself
>>> in a file by typing "set save scipparams.set" in the interactive shell.
>>>
>>> Best,
>>> Ambros
>>>
>>>
>>> Am 05.06.2018 um 19:12 schrieb Meenarli Sharma:
>>>> Hello Gerald,
>>>>
>>>> Thank you very much for the prompt reply. Is there any option (to be
>>>> written in scip.set file) for passing an upper bound on the objective
>>>> value and then stop at the first solution with objective value smaller
>>>> than the one passed?
>>>>
>>>> Thank you again.
>>>>
>>>> Regards,
>>>> ~Meenarli
>>>>
>>>>    > Dear Meenarli,
>>>>>
>>>>> you want to set limits/bestsol to 1.
>>>>>
>>>>> Best,
>>>>> Gerald
>>>>>
>>>>> On 05.06.2018 17:48, Meenarli Sharma wrote:
>>>>>> Hello everyone,
>>>>>>
>>>>>> I am using SCIP as a solver with AMPL. I pass a feasible solution and
>>>>>> want
>>>>>> SCIP to stop when it finds a solution better than the one that is
>>>>>> being
>>>>>> passed. How can this be achieved?
>>>>>>
>>>>>> I tried limiting the number of feasible solutions but it didn't help
>>>>>> as
>>>>>> heuristics found that many solutions - none of them is better than
>>>>>> what
>>>>>> is
>>>>>> being passed in this case- and it stopped. I can try changing the
>>>>>> limit
>>>>>> on
>>>>>> solutions until I find a better one but I am forced to believe there
>>>>>> must
>>>>>> be a better way to do it.
>>>>>>
>>>>>> Any help will be highly appreciated.
>>>>>>
>>>>>> Thanks and regards,
>>>>>> ~Meenarli
>>>>>>
>>>>>> _______________________________________________
>>>>>> Scip mailing list
>>>>>> Scip at zib.de
>>>>>> https://listserv.zib.de/mailman/listinfo/scip
>>>>>
>>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> Scip mailing list
>>>> Scip at zib.de
>>>> https://listserv.zib.de/mailman/listinfo/scip
>>>>
>>>
>>> --
>>> Ambros Gleixner, Research Group Mathematical Optimization Methods at
>>> Zuse Institute Berlin, http://www.zib.de/gleixner
>>> _______________________________________________
>>> Scip mailing list
>>> Scip at zib.de
>>> https://listserv.zib.de/mailman/listinfo/scip
>>>
>>
>>
>> _______________________________________________
>> Scip mailing list
>> Scip at zib.de
>> https://listserv.zib.de/mailman/listinfo/scip
>>
> 
> 
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> https://listserv.zib.de/mailman/listinfo/scip
> 



More information about the Scip mailing list