Node-level Regression Uncertainty Quantification on Chicago
100PICPER-GNN
Evaluation results
| Method | Links | ||
|---|---|---|---|
| ER-GNN2026.05 | 100 | 3.06 | |
| BayesianNN2026.05 | 100 | 2.99 | |
| QpiGNNWidth penalty (lambda)=0.12026.05 | 99 | 0.6 | |
| QpiGNNWidth penalty (lambda)=opt.2026.05 | 98 | 0.36 | |
| QpiGNNWidth penalty (lambda)=0.52026.05 | 97 | 0.36 | |
| CF-GNNoptimized=true2026.05 | 91 | 2.26 | |
| CF-GNN2026.05 | 90 | 3.12 | |
| RQRadj.-GNN2026.05 | 88 | 0.3 | |
| SQR-GNN2026.05 | 87 | 0.21 | |
| MC Dropout2026.05 | 34 | 0.07 |