0
$\begingroup$

Good afternoon. I have a very difficult to a MIP model. I have tried several different strategies to reduce the gap. I am using Gurobi and in this case, I already have an incumbent solution. I've set "Cuts" to 3 (aggressive cut generation), "heuristics" to 0.5, (50% of the time for heuristics) and "MIPfocus" to focus on optimality. Do you have any suggestions for other parameters I can change to alter this performance? Below is gurobi's report for this problem (here I set a maximum time of 180 seconds, but this behaviour would be identical if I left it for 2,3,4 hours).

Academic license - for non-commercial use only - expires 2024-12-10
Gurobi Optimizer version 9.1.1 build v9.1.1rc0 (win64)
Thread count: 6 physical cores, 12 logical processors, using up to 12 threads
Optimize a model with 16566 rows, 109106 columns and 635490 nonzeros
Model fingerprint: 0xea69df7b
Variable types: 108002 continuous, 1104 integer (1104 binary)
Coefficient statistics:
  Matrix range     [7e-02, 2e+05]
  Objective range  [1e+00, 1e+00]
  Bounds range     [1e+00, 1e+00]
  RHS range        [1e+00, 3e+04]

User MIP start produced solution with objective 937.846 (0.09s)
Loaded user MIP start with objective 937.846

Presolve removed 6649 rows and 6530 columns
Presolve time: 0.60s
Presolved: 9917 rows, 102576 columns, 386928 nonzeros
Variable types: 102000 continuous, 576 integer (576 binary)

Root relaxation: objective 5.272822e+02, 16932 iterations, 0.65 seconds
Total elapsed time = 5.30s

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

     0     0  527.28222    0   95  937.84556  527.28222  43.8%     -    6s
     0     0  527.28222    0  277  937.84556  527.28222  43.8%     -   17s
     0     0  527.28222    0  310  937.84556  527.28222  43.8%     -   32s
     0     0  527.28222    0  339  937.84556  527.28222  43.8%     -   87s
     0     0  527.28222    0  339  937.84556  527.28222  43.8%     -  102s
     0     2  527.28222    0  339  937.84556  527.28222  43.8%     -  159s
     1     4  527.28222    1  335  937.84556  527.28222  43.8% 33460  165s

Cutting planes:
  Implied bound: 1713
  MIR: 644
  Flow cover: 1300
  Relax-and-lift: 43

Explored 3 nodes (174099 simplex iterations) in 180.04 seconds
Thread count was 12 (of 12 available processors)

Solution count 1: 937.846

Time limit reached
Best objective 9.378455555555e+02, best bound 5.272822222221e+02, gap 43.7773%

User-callback calls 4853, time in user-callback 0.00 sec

The model is presented below: We need to localize supliers (z1) and (z2), the facilities (yr and yu) and the flux between the supliers and facilities (u's) and from facilities to clients (x's), allowing delays, over the scenario s. enter image description here

$\endgroup$
7
  • $\begingroup$ I would try MIPFocus = 3 (and drop the other options). $\endgroup$ Commented Mar 14 at 18:37
  • $\begingroup$ I have tried, but the result worsed. $\endgroup$ Commented Mar 14 at 18:39
  • $\begingroup$ Another parameter in Gurobi to improve this performance? $\endgroup$ Commented Mar 14 at 18:39
  • $\begingroup$ It appears that you set a time limit of three minutes and spent almost all of that at the root node. You might try reducing the time spent on heuristics and/or cut back on the aggressiveness of cut generation. $\endgroup$
    – prubin
    Commented Mar 14 at 20:36
  • 2
    $\begingroup$ Another option is to improve the model. If you can share the formulation, the community might have some reformulation suggestions. $\endgroup$
    – RobPratt
    Commented Mar 14 at 20:44

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.