Move the BestBound

Answered

Comments

2 comments

  • Official comment
    Michael Winkler

    Dear Stefan,

     

    what bounds do you get when you manually branch on a variable (do you have special knowledge which variable will have a big impact on moving the dual bound or does it not really matter)?

    Could you make your instance available for us, best in MPS format (compressed with 7zip, bzip2, gzip ...) via a dropbox (or some other data storage website) link?

     

    Some guesses what you can try:

    1. set DUALREDUCTIONS=0

    2. or set the branching priorities on (some of) your 38 binary variables, see http://www.gurobi.com/documentation/8.1/refman/branchpriority.html

     

    Best,

    Michael

    Comment actions Permalink
  • Stefan Bruche

    Thanks for your response and your helpful hints Michael!

    There is one binary variable (X) that has a larger influence on the BestBound (Root=-176510; X_0=-177207; X_1=-177359). So branching this variable would definitely help. But of cause, the identification of this variable is the tough part and largely depends on experience or luck. So I cannot blame Gurobi. I did some experiments trying to use the reduced costs of the relaxed binary variables to push the solver into the right direction, but my success was rather moderate.

    0
    Comment actions Permalink

Please sign in to leave a comment.

Powered by Zendesk