HeHLP-HHAN: Heterogeneous Hyperlink Prediction method based on Hyperbolic Hypergraph Attention Network

doi:10.1016/j.physa.2026.131675

HeHLP-HHAN: Heterogeneous Hyperlink Prediction method based on Hyperbolic Hypergraph Attention Network

2026-05-20

0

PRE

AI

DOI：10.1016/j.physa.2026.131675

Original

Original request for help

Save

Abstract

En 中文

• Propose a novel heterogeneous hyperlink prediction method based on hyperbolic hypergraph attention networks, which can learn heterogeneous hypergraph embedding representations in hyperbolic space and reduce embedding distortion compared to Euclidean space. • Design two types of attention mechanism: type-level attention and node-level attention, which can better capture the complex relationships of heterogeneous hypergraphs and improve the expressive power of the model. • We proposed a hyperlink scoring function based on Fermi-Dirac distribution in hyperbolic space, which is more suitable for measuring the co-occurrence probability of nodes in hyperbolic space than traditional Euclidean distance-based scoring functions, and fully utilizes the geometric characteristics of hyperbolic space. • Conduct experiments on five types of real world datasets. The results show that our method outperforms the state-of-the-art baselines in both AP and AUC metrics.

Keywords:

Hyperbolic space

Heterogeneous hypergraph

Attention mechanism

Hyperlink prediction

Embedding representation

HeHLP-HHAN: Heterogeneous Hyperlink Prediction method based on Hyperbolic Hypergraph Attention Network

Abstract

Journal

Organization