Poster
A Topological Perspective on Demystifying GNN-Based Link Prediction Performance
Yu Wang · Tong Zhao · Yuying Zhao · Yunchao Liu · Xueqi Cheng · Neil Shah · Tyler Derr
Halle B #230
Graph Neural Networks (GNNs) have shown great promise in learning node embeddings for link prediction (LP). While numerous studies improve the overall GNNs' LP performance, none have explored their varying performance across different nodes and the underlying reasons. To this end, we demystify which nodes perform better from the perspective of their local topology. Despite the widespread belief that low-degree nodes exhibit worse LP performance, we surprisingly observe an inconsistent performance trend. This prompts us to propose a node-level metric, Topological Concentration (TC), based on the intersection of the local subgraph of each node with the ones of its neighbors. We empirically demonstrate that TC correlates with LP performance more than other node-level topological metrics, better identifying low-performing nodes than using degree. With TC, we discover a novel topological distribution shift issue in which nodes' newly joined neighbors tend to become less interactive with their existing neighbors, compromising the generalizability of node embeddings for LP at testing time. To make the computation of TC scalable, We further propose Approximated Topological Concentration (ATC) and justify its efficacy in approximating TC with reduced computation complexity. Given the positive correlation between node TC and its LP performance, we explore the potential of boosting LP performance via enhancing TC by re-weighting edges in the message-passing and discuss its effectiveness with limitations. Our code is publicly available at https://github.com/YuWVandy/TopoLPGNN.