Interpretable Hypergraph Learning via Neural Additive Models

2026-10-05

AI Summary

This study addresses the limitation that existing hypergraph learning methods are predominantly black-box models, making it difficult to disentangle the independent contributions of node attributes and higher-order structural information. To tackle this issue, we propose HGNAN, a framework that extends neural additive models to hypergraph data for the first time. By leveraging a synergistic mechanism combining feature-level nonlinear decomposition with hypergraph-aware structural aggregation, our method renders the prediction process transparent. Extensive experiments demonstrate that HGNAN achieves state-of-the-art performance across multiple benchmark datasets. Furthermore, it significantly enhances intrinsic interpretability while maintaining high predictive accuracy, thereby establishing a new paradigm for interpretable hypergraph learning.

Abstract

Hypergraphs offer a natural framework for modeling networked data, where dependencies among entities are governed by higher-order interactions. While hypergraph learning methods such as hypergraph neural networks have demonstrated remarkable predictive performance, most existing approaches rely on black-box message-passing architectures, making it difficult to disentangle the contributions of node attributes and higher-order structural information. To address this challenge, we introduce the hypergraph neural additive network (HGNAN), an inherently interpretable framework for learning on hypergraph-structured data. HGNAN extends classical neural additive models to higher-order relational data by integrating feature-wise nonlinear decomposition with hypergraph-aware structural aggregation, enabling transparent prediction for both node- and hyperedge-level tasks. Extensive experiments on benchmark datasets demonstrate that HGNAN achieves performance comparable with state-of-the-art hypergraph learning methods while providing intrinsic and meaningful interpretability.

Problem

• Hypergraph Learning
• Interpretability
• Black-box Models
• Higher-order Interactions
• Neural Additive Models

Innovation

• Hypergraph Learning
• Neural Additive Models
• Interpretability
• Higher-order Interactions
• HGNAN