我正在实现B&C,并使用添加每个惰性约束后加1的计数器。 解决之后,我的计算结果与Gurobi作为惰性约束检索的结果之间存在很大差异。造成这种差异的原因是什么?
谢谢。
Changed value of parameter LazyConstraints to 1
Prev: 0 Min: 0 Max: 1 Default: 0
Optimize a model with 67 rows, 442 columns and 1154 nonzeros
Variable types: 22 continuous, 420 integer (420 binary)
Coefficient statistics:
Matrix range [1e+00, 1e+00]
Objective range [1e-01, 5e+00]
Bounds range [1e+00, 1e+00]
RHS range [1e+00, 1e+01]
Presolve removed 8 rows and 42 columns
Presolve time: 0.00s
Presolved: 59 rows, 400 columns, 990 nonzeros
Variable types: 1 continuous, 399 integer (399 binary)
Root relaxation: objective 2.746441e+00, 37 iterations, 0.00 seconds
Nodes | Current Node | Objective Bounds | Work
Expl Unexpl | Obj Depth IntInf | Incumbent BestBd Gap | It/Node Time
0 0 4.18093 0 20 - 4.18093 - - 0s
H 0 0 21.2155889 4.18093 80.3% - 0s
0 0 5.91551 0 31 21.21559 5.91551 72.1% - 0s
H 0 0 18.8660609 5.91551 68.6% - 0s
0 0 6.35067 0 38 18.86606 6.35067 66.3% - 0s
H 0 0 17.9145774 6.35067 64.6% - 0s
0 0 6.85254 0 32 17.91458 6.85254 61.7% - 0s
H 0 0 17.7591641 6.85254 61.4% - 0s
0 0 7.20280 0 50 17.75916 7.20280 59.4% - 0s
H 0 0 17.7516768 7.20280 59.4% - 0s
0 2 7.91616 0 51 17.75168 7.91616 55.4% - 0s
* 80 62 30 17.6301180 8.69940 50.7% 10.7 0s
* 169 138 35 16.3820478 9.10423 44.4% 9.9 1s
* 765 486 22 14.6853796 9.65509 34.3% 9.2 2s
* 1315 762 27 14.6428113 9.97011 31.9% 9.4 3s
* 1324 415 14 12.0742408 9.97011 17.4% 9.4 3s
H 1451 459 11.8261154 10.02607 15.2% 9.7 4s
1458 463 11.78416 15 58 11.82612 10.02607 15.2% 9.6 5s
* 1567 461 33 11.6541357 10.02607 14.0% 10.6 6s
4055 906 11.15860 31 36 11.65414 10.69095 8.26% 12.4 10s
Cutting planes:
Gomory: 4
Flow cover: 1
Lazy constraints: 228
Explored 7974 nodes (98957 simplex iterations) in 14.78 seconds
Thread count was 4 (of 4 available processors)
Solution count 10: 11.6541 11.8261 12.0742 ... 17.9146
Optimal solution found (tolerance 1.00e-04)
Best objective 1.165413573861e+01, best bound 1.165413573861e+01, gap 0.0000%
My Lazy constraints counter: 654
答案 0 :(得分:1)
优化完成(或停止)后显示的切割平面统计信息仅显示在解决的最终LP松弛中处于活动状态的切割平面的数量。特别是,最后一个节点处于活动状态的惰性约束的数量可能少于在回调中添加的惰性约束的总数。例如,Gurobi可能会在优化过程中添加控制原始延迟约束的内部切割平面,或者使用回调中的延迟约束来导出其他切割,而不是添加原始切割。