Node Classification on Citeseer (test) (BCD & Certifiable Robustness)
72Accuracy (Clean)GCN
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| GCNAdversarial Training=false2023.06 | 72 | 54.7 | 51.7 | 45.3 | 38 | 38.3 | 1.7 | 4.8 | |
| Robust Diffusion (ChebNetII)Adversarial Training=true, Training Attack=LR-BCD2023.06 | 3.7 | 11.5 | 11.5 | 19.8 | 16.4 | 62.3 | 59.8 | 65.4 | |
| Robust Diffusion (ChebNetII)Adversarial Training=true, Training Attack=PR-BCD2023.06 | 3.4 | 13.2 | 14.3 | 11.1 | 11.1 | 59.8 | 65.4 | 62.9 | |
| GPRGNNAdversarial Training=false2023.06 | 2.2 | 3.6 | 7.9 | 4.2 | 9.1 | 24.8 | 22.9 | 42.4 | |
| Robust Diffusion (GPRGNN)Adversarial Training=true, Training Attack=LR-BCD2023.06 | 1.7 | 15.7 | 13.6 | 19.5 | 24.8 | 67.4 | 69 | 65.9 | |
| ChebNetIIAdversarial Training=false2023.06 | 1.1 | 5.8 | 5 | 10.4 | 17.9 | 55.6 | 54 | 60.3 | |
| GATAdversarial Training=true, Training Attack=PR-BCD2023.06 | 1.1 | 3.7 | 8.3 | 14.6 | 13.2 | 19.3 | 41 | 41.9 | |
| SoftMedianAdversarial Training=false2023.06 | 0.9 | 9.5 | 9.3 | 16.2 | 25.2 | 1.7 | 4.2 | 10.7 | |
| GATAdversarial Training=true, Training Attack=LR-BCD2023.06 | 0.8 | 5.9 | 9 | 8.4 | 13.1 | 21.7 | 41.3 | 44.1 | |
| Robust Diffusion (GPRGNN)Adversarial Training=true, Training Attack=PR-BCD2023.06 | 0.6 | 15 | 15.9 | 16 | 23.2 | 65 | 65.9 | 62.3 | |
| APPNPAdversarial Training=false2023.06 | 0.2 | 3.7 | 4.2 | 5.5 | 8.9 | 31.2 | 31.6 | 24.6 | |
| GCNAdversarial Training=true, Training Attack=PR-BCD2023.06 | 0 | 6.9 | 5.3 | 8.9 | 5.8 | -2 | 1.4 | 1.2 | |
| GCNAdversarial Training=true, Training Attack=LR-BCD2023.06 | -0.2 | 7.8 | 5.9 | 10.9 | 8.1 | -10.1 | 13.1 | 11.4 | |
| APPNPAdversarial Training=true, Training Attack=PR-BCD2023.06 | -0.2 | 6.2 | 5.5 | 9 | 10.3 | 6.5 | 32.9 | 31.5 | |
| APPNPAdversarial Training=true, Training Attack=LR-BCD2023.06 | -0.8 | 7.6 | 5.6 | 8.6 | 11.1 | 21.8 | 22.9 | 16.4 | |
| GATAdversarial Training=false2023.06 | -3.6 | -3.9 | -15.9 | -2.3 | -3 | -1.7 | -4.8 | -14 |