HeMI: Multi-view Embedding in Heterogeneous Graphs
This work addresses the problem of learning effective embeddings for heterogeneous graphs, which is crucial for data mining tasks in domains like social networks or bioinformatics, but it appears incremental as it builds on existing meta-path and self-supervised approaches.
The paper tackled representation learning for heterogeneous graphs by proposing a self-supervised method that maximizes mutual information among meta-path representations to encode shared semantics, resulting in performance improvements of 1% to 10% over competing methods on tasks like node classification, clustering, and link prediction.
Many real-world graphs involve different types of nodes and relations between nodes, being heterogeneous by nature. The representation learning of heterogeneous graphs (HGs) embeds the rich structure and semantics of such graphs into a low-dimensional space and facilitates various data mining tasks, such as node classification, node clustering, and link prediction. In this paper, we propose a self-supervised method that learns HG representations by relying on knowledge exchange and discovery among different HG structural semantics (meta-paths). Specifically, by maximizing the mutual information of meta-path representations, we promote meta-path information fusion and consensus, and ensure that globally shared semantics are encoded. By extensive experiments on node classification, node clustering, and link prediction tasks, we show that the proposed self-supervision both outperforms and improves competing methods by 1% and up to 10% for all tasks.