notesum.ai
Published at October 22Fast Graph Sharpness-Aware Minimization for Enhancing and Accelerating Few-Shot Node Classification
cs.CL
cs.AI
Released Date: October 22, 2024
Authors: Yihong Luo1, Yuhan Chen2, Siya Qiu1, Yiwei Wang3, Chen Zhang4, Yan Zhou4, Xiaochun Cao5, Jing Tang1
Aff.: 1The Hong Kong University of Science and Technology; 2School of Computer Science and Engineering, Sun Yat-sen University; 3University of California, Merced; 4Createlink Technology; 5School of Cyber Science and Technology, Shenzhen Campus of Sun Yat-sen University

| Setting | Corafull | Avg | DBLP | Avg | ogbn-arXiv | Avg | ||||||||||||
| 5N3K | 5N5K | 10N3K | 10N5K | acc | time | 5N3K | 5N5K | 10N3K | 10N5K | acc | time | 5N3K | 5N5K | 10N3K | 10N5K | acc | time | |
| MAML models | ||||||||||||||||||
| Meta-GCN | 70.25 | 77.00 | 51.19 | 58.85 | 64.32 | 9.48 | 82.60 | 85.20 | 65.96 | 70.85 | 76.15 | 17.57 | 49.32 | 54.37 | 30.68 | 28.20 | 40.64 | 40.99 |
| w/ SAM | 70.23 | 75.82 | 54.77 | 58.18 | 64.75 | 19.03 | 82.50 | 85.04 | 68.31 | 71.22 | 76.77 | 35.30 | 54.80 | 55.19 | 25.10 | 31.79 | 41.72 | 82.54 |
| w/ FGSAM | 70.97 | 77.64 | 55.53 | 59.30 | 65.86 | 10.83 | 82.66 | 85.26 | 69.22 | 71.80 | 77.24 | 19.15 | 52.45 | 57.05 | 28.92 | 31.03 | 42.36 | 42.48 |
| w/ FGSAM+ | 71.54 | 78.97 | 58.73 | 61.61 | 67.71 | 6.51 | 82.40 | 84.24 | 68.97 | 72.18 | 76.95 | 10.62 | 52.98 | 58.08 | 31.09 | 33.38 | 43.88 | 22.11 |
| AMM-GNN | 72.92 | 80.44 | 57.58 | 57.29 | 67.06 | 15.00 | 81.02 | 83.48 | 66.40 | 71.31 | 75.55 | 26.73 | 51.95 | 57.79 | 28.71 | 26.74 | 41.30 | 42.33 |
| w/ SAM | 68.47 | 74.10 | 52.43 | 57.94 | 63.24 | 30.83 | 80.54 | 83.45 | 66.29 | 71.50 | 75.45 | 54.76 | 49.42 | 50.75 | 30.57 | 32.42 | 40.79 | 84.93 |
| w/ FGSAM | 71.67 | 77.72 | 60.15 | 62.11 | 67.91 | 17.60 | 84.01 | 85.32 | 67.12 | 71.70 | 77.04 | 30.16 | 48.69 | 55.89 | 35.59 | 32.57 | 43.19 | 44.41 |
| w/ FGSAM+ | 72.79 | 79.18 | 59.59 | 62.61 | 68.54 | 10.00 | 81.24 | 85.07 | 70.37 | 71.32 | 77.00 | 16.26 | 51.02 | 50.49 | 33.60 | 34.05 | 42.29 | 23.19 |
| non-MAML models | ||||||||||||||||||
| GPN | 65.23 | 65.67 | 50.48 | 51.23 | 58.15 | 1.89 | 76.05 | 75.02 | 65.41 | 64.52 | 70.25 | 3.28 | 55.35 | 57.50 | 42.72 | 41.54 | 49.28 | 7.70 |
| w/ SAM | 67.28 | 65.02 | 55.06 | 52.30 | 59.92 | 3.62 | 79.44 | 77.66 | 67.88 | 67.78 | 73.19 | 6.78 | 56.18 | 58.65 | 39.91 | 39.92 | 48.67 | 15.98 |
| w/ FGSAM | 69.54 | 69.37 | 57.85 | 56.49 | 63.31 | 2.33 | 80.10 | 79.61 | 68.50 | 69.44 | 74.41 | 4.00 | 57.58 | 58.23 | 47.67 | 48.20 | 52.92 | 8.57 |
| w/ FGSAM+ | 69.40 | 69.96 | 57.74 | 56.10 | 63.30 | 1.83 | 80.02 | 79.69 | 68.94 | 69.51 | 74.54 | 2.56 | 57.39 | 58.04 | 46.59 | 49.49 | 52.88 | 4.66 |
| TENT | 71.24 | 75.49 | 57.29 | 60.35 | 66.09 | 10.88 | 80.67 | 82.74 | 69.04 | 71.79 | 76.06 | 11.36 | 60.44 | 67.34 | 47.14 | 54.88 | 57.45 | 12.90 |
| w/ SAM | 71.38 | 75.29 | 56.86 | 61.85 | 66.35 | 22.03 | 82.13 | 85.10 | 68.96 | 73.62 | 77.45 | 22.86 | 63.58 | 69.30 | 50.79 | 55.21 | 59.72 | 26.43 |
| w/ FGSAM | 71.10 | 76.72 | 57.86 | 63.71 | 67.35 | 20.28 | 82.99 | 86.13 | 70.31 | 73.41 | 78.21 | 20.95 | 63.88 | 71.15 | 53.32 | 57.08 | 61.36 | 23.40 |
| w/ FGSAM+ | 72.85 | 77.77 | 58.37 | 63.04 | 68.01 | 15.10 | 83.64 | 85.97 | 71.15 | 73.72 | 78.62 | 15.58 | 66.20 | 69.14 | 50.66 | 53.56 | 59.89 | 16.86 |