This is a page that curates AI-related papers published worldwide. All content here is summarized using Google Gemini and operated on a non-profit basis. Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.
Leveraging Multi-facet Paths for Heterogeneous Graph Representation Learning
Created by
Haebom
Author
Jongwoo Kim, Seongyeub Chu, Hyeongmin Park, Bryan Wong, Keejun Han, Mun Yong Yi
Outline
This paper highlights the limitation of existing heterogeneous graph neural networks (HGNNs) in capturing complex interactions due to their reliance on domain-specific, predefined metapaths. To address this, we propose a novel model, MF2Vec. Instead of predefined metapaths, MF2Vec uses fine-grained paths to extract paths through random walks and generates multifaceted vectors. It learns various aspects of nodes and relationships and constructs a uniform network to generate node embeddings, which are then applied to classification, link prediction, and clustering tasks. Experimental results demonstrate that MF2Vec outperforms existing methods and provides a more flexible and comprehensive framework for complex network analysis. The source code can be found at https://anonymous.4open.science/r/MF2Vec-6ABC .
Takeaways, Limitations
•
Takeaways:
◦
It provides a more flexible and comprehensive network analysis framework by removing the dependency on predefined meta-paths.
◦
Effectively learn different aspects of nodes and relationships using multifaceted paths.
◦
It outperforms existing methods in various tasks such as classification, link prediction, and clustering.
•
Limitations:
◦
Further research may be needed to explore the efficiency and scalability of random walk-based path extraction methods.
◦
An analysis of the complexity and computational cost of the multifaceted vector generation process is required.
◦
Performance evaluations for certain types of graph structures may be lacking.
