Table 2 Accuracy(%) comparison with different models on nine datasets.
From: GTAT: empowering graph neural networks with cross attention
Model | Core | Citeseer | Pubmed | Computers | Photo | Physics | CS | WikiCS | Arxiv |
|---|---|---|---|---|---|---|---|---|---|
GCN | 80.67±1.10 | 68.10±1.47 | 78.22±1.49 | 82.43±1.47 | 90.84±0.72 | 92.98±0.82 | 91.45±0.31 | 75.04±0.82 | 70.33 ± 0.32 |
FAGCN | 80.64±1.49 | 68.13±1.34 | 78.85±1.80 | 83.63±1.37 | 91.71±1.06 | 93.47±0.75 | 92.52±0.34 | 74.43±1.05 | 67.12 ± 0.85 |
SGC | 79.51±1.49 | 67.75±1.39 | 76.24±2.35 | 83.27±1.21 | 91.01±1.02 | 92.40±0.29 | 91.95±0.48 | 74.32±1.54 | 70.23 ± 0.26 |
SAGE | 80.50±1.54 | 68.60±1.20 | 78.65±1.83 | 82.77±1.23 | 91.52±0.76 | 92.92±0.55 | 90.98±0.50 | 73.99±1.40 | 69.98 ± 0.20 |
GAT | 81.06±1.03 | 68.61±1.22 | 78.51±1.63 | 83.21±1.42 | 91.33±0.80 | 93.09±0.77 | 91.34±0.40 | 75.21±0.98 | 70.41 ± 0.14 |
GAT2 | 81.16±1.34 | 68.40±1.17 | 78.51±1.76 | 83.62±1.51 | 91.47±0.98 | 92.89±0.74 | 91.32±0.29 | 75.39±1.14 | 70.98 ± 0.27 |
mGCMN | 81.21±0.91 | 68.69±1.31 | 78.87±1.60 | 82.88±1.22 | 91.34±0.77 | 93.38±0.54 | 91.12±0.36 | 73.94±0.11 | 70.12 ± 0.10 |
Hyper-Conv. | 80.11±1.02 | 67.41±1.47 | 78.16±1.04 | 79.47±1.85 | 88.59±0.68 | 92.65±0.49 | 89.14±0.35 | 73.26±2.08 | 70.08 ± 0.26 |
Dir-GNN | 77.89±1.51 | 67.44±1.12 | 75.46±2.29 | 80.19±1.67 | 90.42±1.22 | 93.00±0.65 | 91.92±0.23 | 73.61±1.05 | 69.11 ± 0.26 |
GTAT | 81.50±1.27 | 68.91±1.51 | 79.34±0.80 | 83.93±1.09 | 91.70±0.39 | 93.41±0.35 | 91.68±0.21 | 76.01±1.15 | 71.02 ± 0.17 |
GTAT2 | 81.65±1.49 | 68.78±1.58 | 79.05±1.33 | 84.37±1.46 | 91.79±0.41 | 93.39±0.53 | 91.75±0.37 | 75.93±1.15 | 71.37 ±0.14 |
