NS - CUK Seminar: V.T.Hoang, Review on "Two-Dimensional Weisfeiler-Lehman Graph Neural Networks for Link Prediction", arXiv 2022
- 1. Van Thuy Hoang Dept. of Artificial Intelligence, The Catholic University of Korea [email protected] Two-Dimensional Weisfeiler-Lehman Graph Neural Networks for Link Prediction Hu, et. al, 2022
- 2. 2 Link Prediction (LP) Problem Given an incomplete network, predict whether two nodes are likely to have a link. Applications: • Friend recommendation in social networks • Product recommendation in ecommerce • Reaction prediction in metabolic networks • Knowledge graph completion • ...
- 3. 3 Heuristic Methods for LP Calculate a proximity score for each pair of nodes.
- 4. 4 Heuristic Methods for LP Notations: 𝛤 𝑥 is the neighbor set of x in the graph • The common neighbors (CN) heuristic: |𝛤 𝑥 ∩ 𝛤 𝑦 | x and y are likely to have a link if they have many common neighbors. y x
- 5. 5 Heuristic Methods for LP The preferential attachment (PA) heuristic: |𝛤(𝑥)|∙|𝛤(𝑦)| First-order heuristic, only involves 1-hop neighbors. y x x prefers to connect to y if y is popular.
- 6. 6 Problems of Heuristic Methods There are over 50 different heuristics for LP. Many new heuristics come out every year. Problems: • No universally good heuristic. A heuristic may work well on a kind of networ ks, but poorly on others. • There are networks where no existing heuristics work well.
- 7. 7 Problems of Heuristic Methods Why not directly learn such features from the network? Weisfeiler-Lehman Neural Machine (WLNM) WLNM is a neural network model for link prediction which automatically learns graph features from links’ enclosing subgraphs.
- 8. 8 GNNs cannot distinguish links
- 9. 9 Framework • a particular 2-WL test, works for link prediction. It takes links as mess age passing units. • Given the centering directed link to predict, local 2-FWL aggregates i nformation from its neighboring links in each iteration.
- 10. 10 Representations for k-tuples of nodes • k-WL defines the jth neighborhood of k-tuple s as • 2-WL is updated respectively by
- 11. 11 Experiments • eleven benchmark datasets.
- 12. 12 Conclusions • two-dimensional Weisfeiler-Lehman graph neural networks for link pr ediction. • have stronger power than 1-WL. • Show the effectiveness of 2-WL-GNNs for link prediction.