Link Property Prediction On Ogbl Collab
评估指标
Ext. data
Number of params
Test Hits@50
Validation Hits@50
评测结果
各个模型在此基准测试上的表现结果
| Paper Title | Repository | |||||
|---|---|---|---|---|---|---|
| E2N | No | 526851 | 0.9515 ± 0.1410 | 0.9546 ± 0.1270 | Edge2Node: Reducing Edge Prediction to Node Classification | - |
| HyperFusion | No | 1064446212 | 0.7129 ± 0.0018 | 0.7385 ± 0.0099 | - | - |
| GIDN@YITU | No | 60449025 | 0.7096 ± 0.0055 | 0.9620 ± 0.0040 | GIDN: A Lightweight Graph Inception Diffusion Network for High-efficient Link Prediction | - |
| PLNLP + SIGN | No | 34980864 | 0.7087 ± 0.0033 | 1.0000 ± 0.0000 | - | - |
| PLNLP (random walk aug.) | No | 34980864 | 0.7059 ± 0.0029 | 1.0000 ± 0.0000 | Pairwise Learning for Neural Link Prediction | |
| HOP-REC | No | 30191104 | 0.7012 ± 0.0016 | 1.0000 ± 0.0000 | - | - |
| PLNLP+ LRGA | No | 35200656 | 0.6909 ± 0.0055 | 1.0000 ± 0.0000 | Global Attention Improves Graph Networks Generalization | |
| PLNLP (val as input) | No | 35112192 | 0.6872 ± 0.0052 | 1.0000 ± 0.0000 | Pairwise Learning for Neural Link Prediction | |
| Refined-GAE | No | 126669825 | 0.6816 ± 0.0041 | 1.0000 ± 0.0000 | Reconsidering the Performance of GAE in Link Prediction | |
| TopoLink | No | 483363845 | 0.6792 ± 0.0074 | 0.6771 ± 0.0083 | - | - |
| S3GRL (PoS Plus) | No | 5913025 | 0.6683 ± 0.0030 | 0.9861 ± 0.0006 | Simplifying Subgraph Representation Learning for Scalable Link Prediction | |
| ELPH | No | 3284065 | 0.6636 ± 0.5876 | 0.6631 ± 0.0021 | - | - |
| BUDDY | No | 1184867 | 0.6572 ± 0.0053 | 0.6621 ± 0.0016 | - | - |
| Adamic Adar+Edge Proposal Set | No | 0 | 0.6548 ± 0.0000 | 0.9735 ± 0.0000 | Edge Proposal Sets for Link Prediction | |
| SEAL-nofeat (val as input) | No | 501570 | 0.6474 ± 0.0043 | 0.6495 ± 0.0043 | Labeling Trick: A Theory of Using Graph Neural Networks for Multi-Node Representation Learning | |
| Adamic Adar | No | 0 | 0.6417 ± 0.0000 | 0.6349 ± 0.0000 | - | - |
| Common Neighbor | No | 0 | 0.6137 ± 0.0000 | 0.6036 ± 0.0000 | - | - |
| SEAL-nofeat | No | 501570 | 0.5471 ± 0.0049 | 0.6495 ± 0.0043 | Labeling Trick: A Theory of Using Graph Neural Networks for Multi-Node Representation Learning | |
| GraphSAGE (val as input) | No | 460289 | 0.5463 ± 0.0112 | 0.5688 ± 0.0077 | Inductive Representation Learning on Large Graphs | |
| NGNN + GraphSAGE | No | 591873 | 0.5359 ± 0.0056 | 0.6281 ± 0.0046 | Network In Graph Neural Network | - |
0 of 34 row(s) selected.