[Scip] providing SCIP with a new current solution during BCP

[Hélène Toussaint]

Dear Ambros,

Indeed continue the solving with the new basis would be perfect! Since 
it isn't possible I will try to change my algorithm...
I think your patch could be very usefull, could you send it to me?

Thank you very much for your time.
Best,
Hélène


Le 23/05/2013 15:41, Ambros Gleixner a écrit :
> Dear Héléne,
>
> something as you suggest is not fully possible.  What works is to 
> change the bounds and objective function of the LP temporarily in 
> order to pivot to a different basis and retrieve the primal and dual 
> solution. For this, you could use the diving mode.
>
> However, after ending the diving mode, SCIP will continue from the 
> basis before diving.  It is currently not possible to change this.
>
> Also, diving mode does currently not allow to change the bounds 
> (sides) of the LP rows, which is necessary if you want to restrict 
> yourself to the optimal hyperplane.  We have a patch available that 
> implements this, however, and we could send it to you if you want to 
> play around with this.
>
> Best regards
>
> Ambros
>
>
> Am 23.05.2013 08:56, schrieb Hélène Toussaint:
>> Dear Ambros,
>>
>> I misunderstood your last question in your previous email: I have the
>> new solution of the LP relaxation and I would like to give it to SCIP in
>> such a way SCIP update its internal variables (including the basis) to
>> move to this new solution.
>>
>> Is it possible to initialize SCIP with a solution in such a way the
>> simplex algorithm begin its iterations from this solution?
>>
>> Is there a way to perform degenerate iteration of the Simplex algorithm
>> in order to switch from an optimal solution (of the LP relaxation) to
>> one of the same optimal cost during solving?
>>
>> Best,
>> Hélène
>>
>>
>> Le 22/05/2013 17:22, Ambros Gleixner a écrit :
>>> Dear Hélène,
>>>
>>> I am not sure, yet, how this would be possible.  From which
>>> plugin/callback would you want to provide the new basis?
>>>
>>> Ambros
>>>
>>>
>>> Am 22.05.2013 08:37, schrieb Hélène Toussaint:
>>>> Hi Ambros,
>>>>
>>>>
>>>>> Dear Hélène,
>>>>>
>>>>> sorry for the long wait.  Let me ask back to make sure I have
>>>>> understood
>>>>> you correctly: Are the solutions S and S' you talk about solutions of
>>>>> the LP relaxation?
>>>> Yes.
>>>>>   I.e., do you want to move to another LP solution on
>>>>> the optimal hyperplane?
>>>> Yes, exactly!
>>>>
>>>>> Would this solution be basic and would you know
>>>>> the basis?
>>>> I think yes: I need to know the dual variables associated to this
>>>> solution for my pricing.
>>>>
>>>> Thank you for your answer,
>>>>
>>>> Best,
>>>> Hélène
>>>>>
>>>>> Best regards
>>>>>
>>>>> Ambros
>>>>>
>>>>>
>>>>> Am 21.05.2013 09:18, schrieb Hélène Toussaint:
>>>>>> Hi,
>>>>>>
>>>>>> Can no one help me with this issue? I guess it means the answer 
>>>>>> is no
>>>>>> :(.
>>>>>>
>>>>>> Perhaps is there a way to change the current lpi using the 
>>>>>> methods in
>>>>>> lpi.h?
>>>>>>
>>>>>> Best,
>>>>>> Hélène
>>>>>>
>>>>>> Le 06/05/2013 16:27, Hélène Toussaint a écrit :
>>>>>>> Dear all,
>>>>>>>
>>>>>>> I solve a MIP problem by BCP.
>>>>>>>
>>>>>>> Before the cutting phase I'd like to change the current solution:
>>>>>>> indeed from the current solution S, I compute several other 
>>>>>>> solutions
>>>>>>> with the same objective value and among all these solutions I keep
>>>>>>> the
>>>>>>> solution S' which meets a certain condition. S' is valid for the
>>>>>>> current constraints. I'd like S' to become the current solution and
>>>>>>> update the dual variables so that I can continue the cutting, 
>>>>>>> pricing
>>>>>>> and branching with S'.
>>>>>>>
>>>>>>> Is there a way to do this with SCIP?
>>>>>>>
>>>>>>> Best regards,
>>>>>>>
>>>>>>> Hélène
>>>>>>>
>>>>>>>
>>>>>>> _______________________________________________
>>>>>>> Scip mailing list
>>>>>>> Scip at zib.de
>>>>>>> http://listserv.zib.de/mailman/listinfo/scip
>>>>>>
>>>>>>
>>>>>> _______________________________________________
>>>>>> Scip mailing list
>>>>>> Scip at zib.de
>>>>>> http://listserv.zib.de/mailman/listinfo/scip
>>>>>>
>>>>> _______________________________________________
>>>>> Scip mailing list
>>>>> Scip at zib.de
>>>>> http://listserv.zib.de/mailman/listinfo/scip
>>>>>
>>>>
>>>
>>
>