Decision Optimization

Decision Optimization

Delivers prescriptive analytics capabilities and decision intelligence to improve decision-making.

 View Only
  • 1.  Getting candidate variables of Strong branching

    Posted Fri May 22, 2020 05:59 AM
    Hi all,
    I'm using Cplex (12.10) python API to solve VRP related problem and need some information of B&B to do some optimization. I have set strong branching as my variable selection strategy. The default length of MIP strong branching candidate list is 10. How can I get these candidate variables in each node? I didn't find helpful api in cplex.callback package.
    Thanks.

    ------------------------------
    Jianan Zhou
    ------------------------------

    #DecisionOptimization


  • 2.  RE: Getting candidate variables of Strong branching

    Posted Fri May 22, 2020 06:07 AM
    Hello,

    The branching callback is intended to let you decide what the branching decision should be.  It also allows you to check what CPLEX will do if you don't decide anything. But there is no way for the user to get access to the internal information that CPLEX uses to decide which variable to branch on. Also, what makes you say that "the default length of MIP strong branching candidate list is 10"?  Is this something you found in the documentation?

    ------------------------------
    Xavier
    ------------------------------



  • 3.  RE: Getting candidate variables of Strong branching

    Posted Fri May 22, 2020 07:58 AM
    Hi,
    Thanks for your detailed reply. I'm not intend to get the internal info, but to mimic the behaviors of variable selection mentioned in this paper. I don't know how the author get the candidate var in data collection section. I also found this thread. Is it possible to get the info in this way?     
    W.r.t. the SB candidate list, yes, I found the default value here.
    Thanks!

    ------------------------------
    Jianan Zhou
    ------------------------------



  • 4.  RE: Getting candidate variables of Strong branching

    Posted Tue May 26, 2020 04:43 AM
    Hi,

    We can't know exactly how the authors of the paper collected the information. The only way to know would be if you contact the authors.

    My best guess is that they did their own strong branching to collect the labels. There is no API in CPLEX that would allow you to collect the strong branching scores of candidate at each node of the tree.
    Best regards,
    Pierre Bonami

    ------------------------------
    Pierre BONAMI
    ------------------------------



  • 5.  RE: Getting candidate variables of Strong branching

    Posted Tue May 26, 2020 05:35 AM
    Hi,

    Thanks for your clarify!

    ------------------------------
    Jianan Zhou
    ------------------------------