GripNet: Graph information propagation on supergraph for heterogeneous graphs. (January 2023)

• Record Type: Journal Article
• Authors: Xu, Hao; Sang, Shengqi; Bai, Peizhen; Li, Ruike; Yang, Laurence; Lu, Haiping
• Is Part Of: Pattern recognition. Volume 133(2023)
• Publication Date: 2023-01
• Subjects: Graph representation learning -- Heterogeneous graph -- Data integration -- Multi-relational link prediction -- Node classification; Pattern perception -- Periodicals; Perception des structures -- Périodiques; Patroonherkenning; 006.4
• DOI: 10.1016/j.patcog.2022.108973
• Languages: English
• ISSNs: 0031-3203

Abstract

Highlights: A novel supergraph data structure to segregate a heterogeneous graph into interconnected, semantically-coherent subgraphs for efficient learning. A new graph representation learning framework based on supergraph for heterogeneous graphs and graph-based data integration. Extensive experiments on seven large-scale datasets for link prediction and node classification tasks. Abstract: Heterogeneous graph representation learning aims to learn low-dimensional vector representations of different types of entities and relations to empower downstream tasks. Existing popular methods either capture semantic relationships but indirectly leverage node/edge attributes in a complex way, or leverage node/edge attributes directly without taking semantic relationships into account. When involving multiple convolution operations, they also have poor scalability. To overcome these limitations, this paper proposes a flexible and efficient Gr aph i nformation p ropagation Net work (GripNet) framework. Specifically, we introduce a new supergraph data structure consisting of supervertices and superedges. A supervertex is a semantically-coherent subgraph. A superedge defines an information propagation path between two supervertices. GripNet learns new representations for the supervertex of interest by propagating information along the defined path using multiple layers. We construct multiple large-scale graphs and evaluate GripNet against competing methods to show its superiority in link prediction, node classification, and data integration. The code and data are available at https://github.com/nyxflower/GripNet .