Will Gurobi give sub-optimla solutions even the Gap is very small (<0.01%) for MINLP ?

回答済み

Junhong Liu

2024年03月04日 14:28

• 
• 
• 

I am running a model, MINLP, using Gurobi, and the Gurobi converges very fast and gives the solution.

However, I found a better solution using the distributed optimization algorithm (with sub-problem still solved by Gurobi).

Therefore I wonder whether Gurobi will give a solution for large-scale MINLP that is sub-optimal even when the Gap is very small  (<0.01%) .

Many thanks for your help and time.

---------------------------------------------------------------------------------------------

Academic license - for non-commercial use only - expires 2024-12-05
Gurobi Optimizer version 11.0.0 build v11.0.0rc2 (win64 - Windows 10.0 (19045.2))

CPU model: Intel(R) Core(TM) i5-10500 CPU @ 3.10GHz, instruction set [SSE2|AVX|AVX2]
Thread count: 6 physical cores, 12 logical processors, using up to 12 threads

Optimize a model with 11350 rows, 6312 columns and 23801 nonzeros
Model fingerprint: 0x339931c0
Model has 336 quadratic constraints
Variable types: 5808 continuous, 504 integer (504 binary)
Coefficient statistics:
  Matrix range     [9e-05, 1e+06]
  QMatrix range    [1e+00, 1e+00]
  Objective range  [4e-02, 2e+03]
  Bounds range     [1e+00, 1e+00]
  RHS range        [1e-05, 3e+05]
Presolve removed 7809 rows and 2261 columns
Presolve time: 0.11s
Presolved: 3541 rows, 4051 columns, 11231 nonzeros
Presolved model has 336 quadratic constraint(s)
Variable types: 3575 continuous, 476 integer (476 binary)

Root relaxation: objective 6.639598e+03, 967 iterations, 0.01 seconds (0.01 work units)

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

     0     0 6639.59788    0   21          - 6639.59788      -     -    0s
     0     0 6648.74269    0   25          - 6648.74269      -     -    0s
     0     0 6648.74269    0   23          - 6648.74269      -     -    0s
     0     0 6654.09330    0   29          - 6654.09330      -     -    0s
     0     0 6654.13535    0   26          - 6654.13535      -     -    0s
     0     0 6658.33225    0   41          - 6658.33225      -     -    0s
     0     0 6658.96974    0   32          - 6658.96974      -     -    0s
     0     0 6661.29690    0   32          - 6661.29690      -     -    0s
     0     0 6661.42445    0   32          - 6661.42445      -     -    0s
     0     0 6662.50694    0   26          - 6662.50694      -     -    0s
     0     0 6662.59565    0   30          - 6662.59565      -     -    0s
     0     0 6663.13816    0   29          - 6663.13816      -     -    0s
     0     0 6663.15665    0   29          - 6663.15665      -     -    0s
     0     0 6663.41262    0   29          - 6663.41262      -     -    0s
     0     0 6663.41697    0   29          - 6663.41697      -     -    0s
     0     0 6663.48337    0   29          - 6663.48337      -     -    0s
     0     0 6663.48337    0    7          - 6663.48337      -     -    0s
H    0     0                    6664.0417149 6663.48337  0.01%     -    0s

関連

• Bilinear
• MIP Gap

0

• 正式なコメント

  

  Simranjit Kaur

  • Gurobi Staff

  2026年11月13日 01:29

  This post is more than two 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 try Gurobot, our chatbot interface offering instant, expert-level support.
• 

  Marika Karbstein

  • Gurobi Staff

  2024年03月06日 08:19

  The default MIPGap in Gurobi is 0.01%. As soon as this gap is reached the optimization run finishes. That means that two runs that for example differ in some parameter seetings could result in two slightly different objective values. But for both runs the MIP gap is below 0.01%. If you need more precision, you should reduce the MIPGap parameter.

  0

投稿コメントは受け付けていません。