HMSG: Heterogeneous Graph Neural Network based on Metapath Subgraph Learning

• DOI: 10.48550/arxiv.2109.02868
• Publication Date: 2021-01-01
• Authors: Xinjun Cai; Jiaxing Shang; Fei Hao; Dajiang Liu; Linjiang Zheng
• Keywords: Social and Information Networks (cs.SI); FOS: Computer and information sciences; Computer Science - Machine Learning; 68T07; Artificial Intelligence (cs.AI); Computer Science - Artificial Intelligence; I.2.6; 0202 electrical engineering, electronic engineering, information engineering; Computer Science - Social and Information Networks; 02 engineering and technology; Machine Learning (cs.LG)

Many real-world data can be represented as heterogeneous graphs with different types of nodes and connections. Heterogeneous graph neural network model aims to embed or subgraphs into low-dimensional vector space for various downstream tasks such node classification, link prediction, etc. Although several models were proposed recently, they either only aggregate information from the same type neighbors, just indiscriminately treat homogeneous neighbors in way. Based on these observations, we propose a new named HMSG comprehensively capture structural, semantic attribute both neighbors. Specifically, first decompose multiple metapath-based subgraphs, each subgraph associates specific structural information. Then message aggregation methods are applied independently, so that learned more targeted efficient manner. Through type-specific transformation, attributes also transferred among nodes. Finally, fuse together get complete representation. Extensive experiments datasets clustering prediction show achieves best performance all evaluation metrics than state-of-the-art baselines.