[Cbc] Incumbent solution

[Gabrielle A. Grün]

 Hi ikhail
there can be multiple solutions for the same objective function value. If you are only after pruning, you can set the cutoff in CbcModel,  and be careful of the small increment. Thanks.       


Gabrielle


 
  ----- Original Message ----- 
  From: Mikhail Nediak 
  To: Stefan Vigerske 
  Cc: cbc at list.coin-or.org 
  Sent: Monday, July 14, 2008 6:30 AM
  Subject: Re: [Cbc] Incumbent solution



  Hi Stefan, 

  Thanks much for the reply. 

  There are two functions by that name. One of them requires "CBC_Message how" as a first parameter. Do you know which one should be used? 
  I actually tried to use them in the following manner: 

  OsiCbcSolverInterface test; 
  //.... 
  test.getModelPtr()->setBestSolution(...); 

  with very strange outcomes. 

  Would setBestObjectiveValue() achieve similar effects in terms of B&B node pruning behavior? 

  Thanks in advance. 
  Best, 
  Mikhail



        Stefan Vigerske <stefan at math.hu-berlin.de> 
        14/07/2008 05:17 AM 
       To cbc at list.coin-or.org  
              cc Mikhail Nediak <MNediak at business.queensu.ca>  
              Subject Re: [Cbc] Incumbent solution 

              

       



  Hi,

  > Right now I use CBC via OSI and would like to pass an incumbent solution to the 
  > B&B (a single solution vector at the start of the run).
  > Does anyone know of an easy and tested way to do it?  Do I need to go through 
  > CbcModel & setBestSolution() or definition of a new heuristic, etc?

  I think going through CbcModel is the easiest solution.
  John recently added another function CbcModel::setBestSolution(...) 
  which you can give a primal solution and then it checks whether it is 
  feasible and keeps it if so.

  Stefan

  -- 
  Stefan Vigerske
  Humboldt University Berlin, Numerical Mathematics
  http://www.math.hu-berlin.de/~stefan




------------------------------------------------------------------------------


  _______________________________________________
  Cbc mailing list
  Cbc at list.coin-or.org
  http://list.coin-or.org/mailman/listinfo/cbc

  ----- Original Message ----- 
  From: Mikhail Nediak 
  To: Stefan Vigerske 
  Cc: cbc at list.coin-or.org 
  Sent: Monday, July 14, 2008 6:30 AM
  Subject: Re: [Cbc] Incumbent solution



  Hi Stefan, 

  Thanks much for the reply. 

  There are two functions by that name. One of them requires "CBC_Message how" as a first parameter. Do you know which one should be used? 
  I actually tried to use them in the following manner: 

  OsiCbcSolverInterface test; 
  //.... 
  test.getModelPtr()->setBestSolution(...); 

  with very strange outcomes. 

  Would setBestObjectiveValue() achieve similar effects in terms of B&B node pruning behavior? 

  Thanks in advance. 
  Best, 
  Mikhail



        Stefan Vigerske <stefan at math.hu-berlin.de> 
        14/07/2008 05:17 AM 
       To cbc at list.coin-or.org  
              cc Mikhail Nediak <MNediak at business.queensu.ca>  
              Subject Re: [Cbc] Incumbent solution 

              

       



  Hi,

  > Right now I use CBC via OSI and would like to pass an incumbent solution to the 
  > B&B (a single solution vector at the start of the run).
  > Does anyone know of an easy and tested way to do it?  Do I need to go through 
  > CbcModel & setBestSolution() or definition of a new heuristic, etc?

  I think going through CbcModel is the easiest solution.
  John recently added another function CbcModel::setBestSolution(...) 
  which you can give a primal solution and then it checks whether it is 
  feasible and keeps it if so.

  Stefan

  -- 
  Stefan Vigerske
  Humboldt University Berlin, Numerical Mathematics
  http://www.math.hu-berlin.de/~stefan




------------------------------------------------------------------------------


  _______________________________________________
  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/20080714/9046b25d/attachment-0002.html