Table 3 Comparison results with different regularization terms. Data in bold indicates optimal values.
From: Dynamic graph structure evolution for node classification with missing attributes
Dataset | Case | 5% | 15% | 35% | 55% | 75% |
|---|---|---|---|---|---|---|
Cora | EGS (w/o all) | 81.91 | 81.03 | 80.58 | 79.58 | 78.06 |
EGS (w/o \(\lambda _{2}\)) | 83.47 | 83.01 | 81.66 | 81.02 | 79.81 | |
EGS (w/o \(\lambda _{1}\)) | 82.50 | 81.64 | 81.52 | 80.42 | 79.03 | |
EGS (ours) | 84.77 | 84.57 | 84.48 | 83.95 | 82.13 | |
Pubmed | EGS (w/o all) | 75.95 | 75.83 | 75.29 | 74.33 | 73.21 |
EGS (w/o \(\lambda _{2}\)) | 78.17 | 77.91 | 76.56 | 74.86 | 73.68 | |
EGS (w/o \(\lambda _{1}\)) | 76.75 | 76.02 | 75.72 | 74.78 | 73.58 | |
EGS (ours) | 80.37 | 79.79 | 79.34 | 78.95 | 77.01 |
