Table 3 Comparison of node classification accuracy with other GNN methods (highest accuracy highlighted in bold).
From: Co-embedding of edges and nodes with deep graph convolutional neural networks
Method | Cora | Citeseer | Pubmed | Disease | Airport |
|---|---|---|---|---|---|
GCN2 | 81.5 ± 0.5 | 70.4 ± 0.4 | 78.1 ± 0.4 | 69.8 ± 0.5 | 81.4 ± 0.6 |
GAT5 | 83.0 ± 0.5 | 71.6 ± 0.8 | 78.2 ± 0.4 | 70.4 ± 0.5 | 81.6 ± 0.4 |
AMC-GCN47 | 84.8 ± 0.4 | 72.8 ± 0.5 | 78.9 ± 0.3 | 70.8 ± 0.2 | 80.5 ± 0.5 |
NIGCN48 | 82.1 ± 1.1 | 71.4 ± 0.8 | 80.9 ± 2.0 | 68.5 ± 1.5 | 82.1 ± 1.1 |
DropEdge31 (64 layers) | 78.9 ± 0.3 | 65.1 ± 0.5 | 76.9 ± 0.6 | 69.7 ± 1.6 | 82.8 ± 1.5 |
NodeNorm8 (64 layers) | 83.4 ± 0.6 | 73.8 ± 0.8 | 80.4 ± 1.2 | 69.6 ± 0.8 | 83.9 ± 1.2 |
GCNII25 (64 layers) | 85.5 ± 0.4 | 73.4 ± 0.2 | 79.7 ± 0.3 | 71.3 ± 0.4 | 84.5 ± 0.5 |
GDC49 | 83.8 ± 0.2 | 73.3 ± 0.3 | 79.9 ± 0.3 | 70.2 ± 0.1 | 83.6 ± 0.2 |
DeepGWC50 (64 layers) | 86.4 ± 0.2 | 74.9 ± 0.5 | 80.7 ± 0.2 | 70.8 ± 0.7 | 84.7 ± 1.3 |
CensNet36 | 79.4 ± 1.0 | 62.5 ± 1.5 | 69.9 ± 2.1 | 64.4 ± 2.1 | 78.6 ± 1.8 |
NENN37 | 82.6 ± 0.1 | 68.2 ± 0.1 | 77.7 ± 0.1 | 67.7 ± 0.1 | 79.8 ± 0.1 |
EGAT51 | 82.1 ± 0.7 | 70.3 ± 0.5 | 78.1 ± 0.4 | 69.1 ± 0.6 | 80.4 ± 0.5 |
CEN-DGCNN (Ours) | 87.1 ± 0.5 | 75.0 ± 0.8 | 81.8 ± 0.4 | 73.5 ± 0.6 | 85.8 ± 0.6 |
