Long-range Meta-path Search on Large-scale Heterogeneous Graphs
This work addresses a critical bottleneck in graph neural networks for heterogeneous graphs, offering a scalable solution for researchers and practitioners dealing with large, complex graph data.
The paper tackles the problem of utilizing long-range dependencies in large-scale heterogeneous graphs by introducing an automatic framework called LMSPS, which dynamically searches for effective meta-paths to reduce computational costs and mitigate over-smoothing, achieving state-of-the-art performance across diverse datasets.
Utilizing long-range dependency, a concept extensively studied in homogeneous graphs, remains underexplored in heterogeneous graphs, especially on large ones, posing two significant challenges: Reducing computational costs while maximizing effective information utilization in the presence of heterogeneity, and overcoming the over-smoothing issue in graph neural networks. To address this gap, we investigate the importance of different meta-paths and introduce an automatic framework for utilizing long-range dependency on heterogeneous graphs, denoted as Long-range Meta-path Search through Progressive Sampling (LMSPS). Specifically, we develop a search space with all meta-paths related to the target node type. By employing a progressive sampling algorithm, LMSPS dynamically shrinks the search space with hop-independent time complexity. Through a sampling evaluation strategy, LMSPS conducts a specialized and effective meta-path selection, leading to retraining with only effective meta-paths, thus mitigating costs and over-smoothing. Extensive experiments across diverse heterogeneous datasets validate LMSPS's capability in discovering effective long-range meta-paths, surpassing state-of-the-art methods. Our code is available at https://github.com/JHL-HUST/LMSPS.