[Paper Review] A Survey on Heterogeneous Graph Embedding: Methods, Techniques, Applications and Sources

Xiao Wang, Deyu Bo+4 more|arXiv (Cornell University)|Nov 30, 2020

Advanced Graph Neural NetworksComputer Science42 citations

TL;DR

This survey comprehensively reviews heterogeneous graph embedding methods, taxonomy, techniques, real-world deployments, datasets, and open-source tools, highlighting challenges and future directions.

ABSTRACT

Heterogeneous graphs (HGs) also known as heterogeneous information networks have become ubiquitous in real-world scenarios; therefore, HG embedding, which aims to learn representations in a lower-dimension space while preserving the heterogeneous structures and semantics for downstream tasks (e.g., node/graph classification, node clustering, link prediction), has drawn considerable attentions in recent years. In this survey, we perform a comprehensive review of the recent development on HG embedding methods and techniques. We first introduce the basic concepts of HG and discuss the unique challenges brought by the heterogeneity for HG embedding in comparison with homogeneous graph representation learning; and then we systemically survey and categorize the state-of-the-art HG embedding methods based on the information they used in the learning process to address the challenges posed by the HG heterogeneity. In particular, for each representative HG embedding method, we provide detailed introduction and further analyze its pros and cons; meanwhile, we also explore the transformativeness and applicability of different types of HG embedding methods in the real-world industrial environments for the first time. In addition, we further present several widely deployed systems that have demonstrated the success of HG embedding techniques in resolving real-world application problems with broader impacts. To facilitate future research and applications in this area, we also summarize the open-source code, existing graph learning platforms and benchmark datasets. Finally, we explore the additional issues and challenges of HG embedding and forecast the future research directions in this field.

Motivation & Objective

• Introduce the basic concepts of heterogeneous graphs and identify the unique challenges for HG embedding
• Systematically categorize state-of-the-art HG embedding methods by information used in learning
• Analyze the pros and cons and transformativeness of HG embedding methods in real-world settings
• Summarize open-source code, platforms, and benchmark datasets to facilitate future work
• Discuss open issues and forecast future directions in heterogeneous graph embedding

Proposed method

• Categorize HG embedding methods into structure-preserved, attribute-assisted, application-oriented, and dynamic approaches
• Review structure-preserved methods based on links, meta-paths, and subgraphs such as meta-paths, meta-graphs and hyperedges
• Discuss path- and subgraph-based methods including metapath2vec, metagraph2vec, mg2vec, and DHNE
• Explain how heterogeneous attributes require fusion and influence model design
• Highlight the role of meta-paths and meta-graphs in capturing higher-order relationships
• Present examples of relation-specific metrics and learning objectives for preserving heterogeneity

Experimental results

Research questions

• RQ1 What are the fundamental challenges posed by heterogeneity in HG embedding compared with homogeneous graphs?
• RQ2 How can HG embedding methods leverage different information types (structure, attributes, domain knowledge) to address these challenges?
• RQ3 What are the relative strengths and limitations of link-based, path-based, and subgraph-based HG embedding methods?
• RQ4 How do HG embedding techniques translate to real-world deployments and applications?
• RQ5 What open-source resources and benchmarks exist to support future HG embedding research?

Key findings

• HG embedding methods must address complex HG structures and heterogeneous attributes, which differ from homogeneous graphs
• There is a taxonomy of HG embedding methods based on the information used: structure, attributes, application cues, and dynamics
• Link-based, path-based, and subgraph-based approaches offer complementary perspectives on preserving heterogeneity, with trade-offs in complexity and scalability
• Meta-path and meta-graph concepts enable high-order semantic capture, while random-walk and hybrid-relations methods provide diverse learning strategies
• Several methods use relation-specific projections, translated distance functions, or joint multi-task objectives to model heterogeneous relations
• The survey maps open-source code, platforms, and benchmark datasets to facilitate future HG embedding research