ProtoHG: Prototype-Enhanced Hypergraph Learning for Heterogeneous Information Networks

Abstract

The variety and complexity of relations in real-world data lead to Heterogeneous Information Networks (HINs). Capturing the semantics from such networks requires approaches capable of utilizing the full richness of the HINs. Existing methods for modeling HINs employ techniques originally designed for graph neural networks in combination with HIN decomposition analysis, especially using manually predefined metapaths. In this paper, we introduce a novel hypergraph learning approach for node classification in HINs. Using hypergraphs instead of graphs, our method captures higher-order relationships among heterogeneous nodes and extracts semantic information without relying on metapaths. Our method leverages the power of prototypes to improve the robustness of the hypergraph learning process, and we further discuss the advantages that our method can bring to scalability, which due to their expressiveness is an important issue for hypergraphs. Extensive experiments on three real-world HINs demonstrate the effectiveness of our method.