Edge flow prediction on Traffic (test)
0.057RMSEFlowSymm
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| FlowSymm2026.01 | 0.057 | 0.028 | 0.85 | |
| Bil-GCNformulation=bilevel, encoder=GCN2026.01 | 0.062 | 0.034 | 0.82 | |
| GINformulation=data-driven, layers=22026.01 | 0.065 | 0.038 | — | |
| EGNNformulation=data-driven, level=edge-level via line-graph2026.01 | 0.065 | 0.036 | — | |
| GCNformulation=data-driven, layers=22026.01 | 0.066 | 0.04 | — | |
| MLP-Divformulation=hybrid, prior=MLP2026.01 | 0.066 | 0.041 | 0.81 | |
| Bil-MLPformulation=bilevel, encoder=MLP2026.01 | 0.069 | 0.038 | 0.79 | |
| Divformulation=physics-based, regularization=node-wise divergence2026.01 | 0.071 | 0.041 | 0.76 | |
| GCN-Divformulation=hybrid, prior=GCN2026.01 | 0.071 | 0.048 | 0.81 | |
| MLPformulation=data-driven, architecture=two-layer feedforward2026.01 | 0.083 | 0.055 | — |