Decision Optimization

Decision Optimization

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

 View Only
Expand all | Collapse all

Criteria in Best Estimate search

  • 1.  Criteria in Best Estimate search

    Posted Tue December 30, 2014 07:20 AM

    Originally posted by: gtoptimizer


    Hey all,

    What is the formula used to compute the "estimated integer objective *" for a node in a branch-and-cut tree (assuming it is common knowledge)?

    Is this the same value based on which the "Best Estimate" node selection strategy chooses the next node to process?

    Thanks in advance!

    * as returned by CPXgetcallbacknodeinfo with whichinfo equal to CPX_CALLBACK_INFO_NODE_ESTIMATE


    #CPLEXOptimizers
    #DecisionOptimization


  • 2.  Re: Criteria in Best Estimate search

    Posted Mon January 05, 2015 04:28 PM

    Originally posted by: EdKlotz


    The specifics of the node estimate calculation for the estimated integer objective CPLEX uses is not publicized.   Here's the fundamental idea.  Best estimate looks at the node LP objective, then adjusts it to account
    for the number and sum of integer infeasibilities to derive an adjusted objective that estimates the likely objective
    of an integer feasible solution found at the node.   Thus, in a minimization problem, a node LP with an objective of 150
    and one integer infeasibility of value 10 is likely to have a more favorable estimated objective than another node LP with
    objective 130 but with 100 integer infeasibilities with sum of infeasibilities of 1000.   By contrast, best estimate node
    selection just considers the node LP objective values of 150 and 130 without considering the amount of integer
    infeasibility repair needed to find an integer feasible solution.
     

    Regarding publicly known strategies for calculating node estimates, try

     

    Land, A. and Powell, S., Computer codes for problems of integer programming, in P.L. Hammer, E.L. Johnson, and B.H. Korte, editors, Discrete Optimization II, Annals of Discrete Mathematics Volume 5, North Holland, Amsterdam, 1979, pages 221-269.

     

    More generally, for a list of papers that contain information about some of the foundational ideas on which CPLEX is based, take a look at https://www-304.ibm.com/support/docview.wss?uid=swg21400019

     

     


    #CPLEXOptimizers
    #DecisionOptimization


  • 3.  Re: Criteria in Best Estimate search

    Posted Mon January 05, 2015 04:51 PM

    Originally posted by: gtoptimizer


    Thanks! Very useful references.


    #CPLEXOptimizers
    #DecisionOptimization