Mixed-Integer Linear Programming solving on Capacitated Facility Location 200x100 (Medium)
67Nodes ExploredDeepBound
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| DeepBoundTraining protocol=trained on easy instances only, Hardware=Intel Xeon Gold 6238R CPU + NVIDIA A100 GPU2026.01 | 67 | 52.24 | 157.56 | — | |
| SCIP-BESTraining protocol=trained on easy instances only, Hardware=Intel Xeon Gold 6238R CPU2026.01 | 76 | 59.3 | 168.02 | — | |
| XGBoostTraining protocol=trained on easy instances only, Hardware=Intel Xeon Gold 6238R CPU2026.01 | 81 | 55.93 | 164.97 | — | |
| He et al.Training protocol=trained on easy instances only, Hardware=Intel Xeon Gold 6238R CPU2026.01 | 83 | 57.19 | 169.33 | — | |
| Heuristic BaselineBaseline Comparison=Subgraph + Rank2025.12 | 117 | — | 126 | — | |
| SCIP2026.03 | 138 | — | 189 | 100 | |
| Heuristic BaselineBaseline Comparison=MPNN + Rank2025.12 | 142 | — | 136 | — | |
| Heuristic BaselineBaseline Comparison=MPNN + Scores2025.12 | 150 | — | 136 | — | |
| Heuristic BaselineBaseline Comparison=Subgraph + Scores2025.12 | 157 | — | 144 | — | |
| Heuristic BaselineBaseline Comparison=MPNN + Pairwise2025.12 | 158 | — | 148 | — | |
| Heuristic BaselineBaseline Comparison=Subgraph + Pairwise2025.12 | 202 | — | 151 | — | |
| VFS2026.03 | 257 | — | 430 | 95 | |
| QL2026.03 | 315 | — | 144 | 100 | |
| QS2026.03 | 322 | — | 147 | 100 | |
| GNN-C2026.03 | 336 | — | 182 | 100 | |
| GNN-G2026.03 | 336 | — | 158 | 100 | |
| MPNNLoss Function=Rank2025.12 | 362 | — | 115 | — | |
| MPNNLoss Function=Pairwise2025.12 | 377 | — | 125 | — | |
| MPNNLoss Function=Scores (Regression)2025.12 | 389 | — | 121 | — | |
| Subgraph GNNLoss Function=Rank2025.12 | 392 | — | 129 | — | |
| Subgraph GNNLoss Function=Scores (Regression)2025.12 | 492 | — | 157 | — | |
| Subgraph GNNLoss Function=Pairwise2025.12 | 501 | — | 146 | — |