メインコンテンツへスキップ

Termination Criteria when Incumbent and MIPGap are both " - "

回答済み

コメント

4件のコメント

  • 正式なコメント
    Simranjit Kaur
    • Gurobi Staff Gurobi Staff
    This post is more than three years old. Some information may not be up to date. For current information, please check the Gurobi Documentation or Knowledge Base. If you need more help, please create a new post in the community forum. Or why not try our AI Gurobot?.
  • Silke Horn
    • Gurobi Staff Gurobi Staff

    The MIP gap is defined as the relative difference of the current incumbent value and the best bound. If there is no incumbent, then there is no MIP gap.

    In your run, Gurobi did indeed stop because it had exhaustively explored all nodes in the tree. However, I am not aware of any way to know/guess the number of nodes before running the solver. In fact, the number of nodes can depend on a lot of things: With a different solver path (e.g. with different settings, different random seeds, or just a different order of variables or constraints), the number of nodes may differ significantly.

    0
  • Daniel Palmarin
    • Gurobi-versary
    • First Question
    • First Comment

    Hi Silke,

    Thanks for the response. So is the only way to speed this computation up by setting a max node limit or a max time limit (or both)? Based on your experience, would you have a suggested max node limit? Also, as Gurobi continues to search more and more nodes, does the probability of finding a solution decrease or is it static throughout the whole process?

    Dan

    0
  • Silke Horn
    • Gurobi Staff Gurobi Staff

    Hi Dan,

    I guess I would go for a time limit and set it to the maximum duration you are willing to wait.

    You could also experiment with different branching strategies (BranchDir or VarBranch) or other parameter settings to see whether they lead to faster results.

    I don't know the probability distribution of when the first solution is found. Since this really depends on the model and the solver path, I doubt that a general answer would help you much with your model.

    Silke

    0

投稿コメントは受け付けていません。