Principled Hyperedge Prediction with Structural Spectral Features and Neural Networks

• DOI: 10.48550/arxiv.2106.04292
• Publication Date: 2021-01-01
• Authors: Changlin Wan; Muhan Zhang; Hao Wei; Sha Cao; Pan Li; Chi Zhang
• Keywords: Social and Information Networks (cs.SI); FOS: Computer and information sciences; Computer Science - Machine Learning; 0303 health sciences; 03 medical and health sciences; Computer Science - Social and Information Networks; Machine Learning (cs.LG)

Hypergraph offers a framework to depict the multilateral relationships in real-world complex data. Predicting higher-order relationships, i.e hyperedge, becomes fundamental problem for full understanding of complicated interactions. The development graph neural network (GNN) has greatly advanced analysis ordinary graphs with pair-wise relations. However, these methods could not be easily extended case hypergraph. In this paper, we generalize challenges GNN representing data principle, which are edge- and node-level ambiguities. To overcome challenges, present SNALS that utilizes bipartite structural features collectively tackle two ambiguity issues. captures joint interactions hyperedge by its local environment, is retrieved collecting spectrum information their connections. As result, achieves nearly 30% performance increase compared most recent GNN-based models. addition, applied predict genetic on 3D genome organization showed consistently high prediction accuracy across different chromosomes, generated novel findings 4-way gene interaction, further validated existing literature.