A Topological Perspective on Demystifying GNN-Based Link Prediction Performance

TL;DR

This paper introduces Topological Concentration (TC), a new metric based on local topology, to better understand and improve GNN-based link prediction performance, especially for low-degree nodes, and proposes scalable approximations and enhancement strategies.

Contribution

It proposes Topological Concentration (TC) as a novel metric for node performance prediction in GNNs, introduces an efficient approximation (ATC), and explores methods to boost link prediction by enhancing TC.

Findings

TC correlates better with LP performance than degree or density.

A topological distribution shift affects newly joined neighbors.

Enhancing TC can improve LP performance, with some limitations.

Abstract

Graph Neural Networks (GNNs) have shown great promise in learning node embeddings for link prediction (LP). While numerous studies aim to improve the overall LP performance of GNNs, none have explored its varying performance across different nodes and its underlying reasons. To this end, we aim to demystify which nodes will perform better from the perspective of their local topology. Despite the widespread belief that low-degree nodes exhibit poorer LP performance, our empirical findings provide nuances to this viewpoint and prompt us to propose a better 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 has a higher correlation with LP performance than other node-level topological metrics like degree and subgraph density, offering a better way to identify low-performing nodes than using cold-start. With TC, we discover a novel topological distribution shift issue in which newly joined neighbors of a node tend to become less interactive with that node's 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 theoretically/empirically justify its efficacy in approximating TC and reducing the 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/Topo_LP_GNN.