Link Property Prediction on ogbl-collab
Test Hits@50
- 2014-03-26 — DeepWalk: Test Hits@50 0.5037
- 2017-06-07 — GraphSAGE (val as input): Test Hits@50 0.5463
- 2020-06-14 — PLNLP+ LRGA: Test Hits@50 0.6909
- 2021-12-06 — PLNLP (random walk aug.): Test Hits@50 0.7059
- 2022-10-04 — GIDN@YITU: Test Hits@50 0.7096
- 2023-11-06 — E2N: Test Hits@50 0.9515
| Rank | Model | Test Hits@50 | Ext. data | Validation Hits@50 | Number of params | Paper | Code | Year |
|---|---|---|---|---|---|---|---|---|
| 1 | E2N | 0.9515 ± 0.1410 | No | 0.9546 ± 0.1270 | 526851 | Edge2Node: Reducing Edge Prediction to Node Classification | 2023 | |
| 2 | HyperFusion | 0.7129 ± 0.0018 | No | 0.7385 ± 0.0099 | 1064446212 | |||
| 3 | GIDN@YITU | 0.7096 ± 0.0055 | No | 0.9620 ± 0.0040 | 60449025 | GIDN: A Lightweight Graph Inception Diffusion Network for High-efficient Link Prediction | 2022 | |
| 4 | PLNLP + SIGN | 0.7087 ± 0.0033 | No | 1.0000 ± 0.0000 | 34980864 | |||
| 5 | PLNLP (random walk aug.) | 0.7059 ± 0.0029 | No | 1.0000 ± 0.0000 | 34980864 | Pairwise Learning for Neural Link Prediction | zhitao-wang/PLNLP · zhitao-wang/plnlp | 2021 |
| 6 | HOP-REC | 0.7012 ± 0.0016 | No | 1.0000 ± 0.0000 | 30191104 | |||
| 7 | PLNLP+ LRGA | 0.6909 ± 0.0055 | No | 1.0000 ± 0.0000 | 35200656 | Global Attention Improves Graph Networks Generalization | omri1348/LRGA · omri1348/LRGA · chuanqichen/cs224w | 2020 |
| 8 | PLNLP (val as input) | 0.6872 ± 0.0052 | No | 1.0000 ± 0.0000 | 35112192 | Pairwise Learning for Neural Link Prediction | zhitao-wang/PLNLP · zhitao-wang/plnlp | 2021 |
| 9 | Refined-GAE | 0.6816 ± 0.0041 | No | 1.0000 ± 0.0000 | 126669825 | Reconsidering the Performance of GAE in Link Prediction | GraphPKU/Refined-GAE · graphpku/refined-gae | 2024 |
| 10 | TopoLink | 0.6792 ± 0.0074 | No | 0.6771 ± 0.0083 | 483363845 | |||
| 11 | S3GRL (PoS Plus) | 0.6683 ± 0.0030 | No | 0.9861 ± 0.0006 | 5913025 | Simplifying Subgraph Representation Learning for Scalable Link Prediction | venomouscyanide/s3grl · venomouscyanide/s3grl_ogb · venomouscyanide/S3GRL_OGB | 2023 |
| 12 | ELPH | 0.6636 ± 0.5876 | No | 0.6631 ± 0.0021 | 3284065 | |||
| 13 | BUDDY | 0.6572 ± 0.0053 | No | 0.6621 ± 0.0016 | 1184867 | |||
| 14 | Adamic Adar+Edge Proposal Set | 0.6548 ± 0.0000 | No | 0.9735 ± 0.0000 | 0 | Edge Proposal Sets for Link Prediction | sangyx/gtrick · CUAI/Edge-Proposal-Sets | 2021 |
| 15 | SEAL-nofeat (val as input) | 0.6474 ± 0.0043 | No | 0.6495 ± 0.0043 | 501570 | Labeling Trick: A Theory of Using Graph Neural Networks for Multi-Node Representation Learning | dmlc/dgl · facebookresearch/SEAL_OGB | 2020 |
| 16 | Adamic Adar | 0.6417 ± 0.0000 | No | 0.6349 ± 0.0000 | 0 | |||
| 17 | Common Neighbor | 0.6137 ± 0.0000 | No | 0.6036 ± 0.0000 | 0 | |||
| 18 | SEAL-nofeat | 0.5471 ± 0.0049 | No | 0.6495 ± 0.0043 | 501570 | Labeling Trick: A Theory of Using Graph Neural Networks for Multi-Node Representation Learning | dmlc/dgl · facebookresearch/SEAL_OGB | 2020 |
| 19 | GraphSAGE (val as input) | 0.5463 ± 0.0112 | No | 0.5688 ± 0.0077 | 460289 | Inductive Representation Learning on Large Graphs | pyg-team/pytorch_geometric · dmlc/dgl · dmlc/dgl · +17 | 2017 |
| 20 | NGNN + GraphSAGE | 0.5359 ± 0.0056 | No | 0.6281 ± 0.0046 | 591873 | Network In Graph Neural Network | 2021 |