メインコンテンツへスキップ

Solver takes to long too find Incumbent solution

回答済み

コメント

5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?.
  • Eli Towle
    • Gurobi Staff Gurobi Staff

    It could be difficult to improve the dual bound much more. The first solution Gurobi found was within 0.03% of optimal, so the bound is already pretty good.

    It might be better to focus on trying to find heuristic solutions faster. You can do this with MIPFocus=1 or by increasing the value of the Heuristics parameter. Unfortunately, there's no way of knowing if either of these will help until you try them out on your model.

    1
  • David Mejía
    • Gurobi-versary
    • Conversationalist
    • Curious

    Thank you Eli for your fast response.

    I´ve tried your hint. When I use either MIPFocus=1 or Heuristics=0.1 the solver finds a solution really fast, but no nodes where explored during the optimization process. How can I asses the quality of this solution? How can I know if this is a reliable solution?

    Here is the log with Heuristics = 0.1. The MIPFocus=1 log is pretty much the same.

    Gurobi Optimizer version 9.0.2 build v9.0.2rc0 (win64)
    Warning: Q constraint 0 doesn't have a name
    Warning: default Q constraint names used to write mps file
    Optimize a model with 87241 rows, 81720 columns and 166231 nonzeros
    Model fingerprint: 0xe5ee240c
    Model has 17179 quadratic constraints
    Variable types: 52200 continuous, 29520 integer (17640 binary)
    Coefficient statistics:
    Matrix range [1e-01, 2e+01]
    QMatrix range [5e-04, 1e+01]
    QLMatrix range [2e-01, 2e+02]
    Objective range [3e-03, 4e+02]
    Bounds range [1e+00, 5e+01]
    RHS range [3e-02, 5e+01]
    Presolve removed 44516 rows and 37641 columns
    Presolve time: 0.53s
    Presolved: 76179 rows, 49895 columns, 190513 nonzeros
    Presolved model has 5086 bilinear constraint(s)
    Variable types: 27055 continuous, 22840 integer (11619 binary)

    Deterministic concurrent LP optimizer: primal and dual simplex
    Showing first log only...

    Concurrent spin time: 0.01s

    Solved with dual simplex

    Root relaxation: objective 1.790556e+06, 14311 iterations, 0.75 seconds

    Nodes | Current Node | Objective Bounds | Work
    Expl Unexpl | Obj Depth IntInf | Incumbent BestBd Gap | It/Node Time

    0 0 1790556.08 0 2291 - 1790556.08 - - 3s
    0 0 1790433.35 0 877 - 1790433.35 - - 5s
    0 0 1790431.87 0 738 - 1790431.87 - - 5s
    0 0 1790422.74 0 734 - 1790422.74 - - 6s
    0 0 1790422.72 0 731 - 1790422.72 - - 7s
    0 0 1790421.86 0 675 - 1790421.86 - - 8s
    0 0 1790421.83 0 671 - 1790421.83 - - 9s
    0 0 1790421.78 0 656 - 1790421.78 - - 10s
    0 0 1790421.77 0 648 - 1790421.77 - - 11s
    0 0 1790421.77 0 645 - 1790421.77 - - 11s
    0 0 1790421.77 0 645 - 1790421.77 - - 13s
    H 0 0 -2918.858500 1790421.77 - - 16s
    H 0 0 1790120.9535 1790421.77 0.02% - 16s

    Cutting planes:
    Gomory: 308
    Implied bound: 28
    MIR: 3463
    StrongCG: 217
    Flow cover: 50
    RLT: 7
    Relax-and-lift: 21

    Explored 1 nodes (19991 simplex iterations) in 16.99 seconds
    Thread count was 4 (of 4 available processors)

    Solution count 2: 1.79012e+06 -2918.86

    Optimal solution found (tolerance 1.00e+00)
    Best objective 1.790120953500e+06, best bound 1.790421765679e+06, gap 0.0168%


    0
  • Matthias Miltenberger
    • Gurobi Staff Gurobi Staff

    Hi David,

    You should consider yourself lucky that Gurobi does not have to explore any other nodes except for the root node. This means that no branching was necessary and the problem could be solved using cutting planes and heuristics.

    You can assess the solution quality just as usual by inspecting the reported gap, in this case 0.016%.

    Cheers,
    Matthias

     

    1
  • David Mejía
    • Gurobi-versary
    • Conversationalist
    • Curious

    Thank you for that explanation Matthias. It just seemed too good to be true. 

    Best Regards,

    David

    0

投稿コメントは受け付けていません。