Long-Range Interaction Modeling on LRIM-16 hard
-4.345logMSEGPS-RWSE
Evaluation results
| Method | Links | |
|---|---|---|
| GPS-RWSEPreprocessing=O(k · N^2), Computation=O(L · N^2)2026.05 | -4.345 | |
| GPS-RWSETraining Grid Size=16x162026.05 | -4.345 | |
| GPS-BasePreprocessing=-, Computation=O(L · N^2)2026.05 | -4.34 | |
| GPS-BaseTraining Grid Size=16x162026.05 | -4.34 | |
| GPS-LapPE2026.05 | -4.334 | |
| LGSMPreprocessing=O(L · E)† or O(1)‡, Computation=O(D · L · E)2026.05 | -4.284 | |
| GPS-LapPEPreprocessing=O(k^2 · E), Computation=O(L · N^2)2026.05 | -4.248 | |
| GPS-LapPETraining Grid Size=16x162026.05 | -4.248 | |
| GPS-Base2026.05 | -4.211 | |
| GHR-GatedGCN2026.05 | -4.195 | |
| GHR-GatedGCNTraining Grid Size=16x162026.05 | -4.195 | |
| GatedGCNPreprocessing=-, Computation=O(L · E)2026.05 | -3.919 | |
| GatedGCNTraining Grid Size=16x162026.05 | -3.919 | |
| GatedGCN-VNGPreprocessing=O(N), Computation=O(L · E + L · N)2026.05 | -3.756 | |
| GatedGCN-VNGTraining Grid Size=16x162026.05 | -3.756 | |
| GINPreprocessing=-, Computation=O(L · E)2026.05 | -2.406 | |
| GINTraining Grid Size=16x162026.05 | -2.406 | |
| GIN2026.05 | 2.533 | |
| GatedGCN2026.05 | 3.844 | |
| GPS-RWSE2026.05 | 4.011 | |
| GatedGCN-VNG2026.05 | 4.068 |