Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Hypergraph Transformer for Semi-Supervised Classification

Zexi Liu, Bohan Tang, Ziyuan Ye, Xiaowen Dong, Siheng Chen, Yanfeng Wang

TL;DR

This work addresses the challenge of leveraging global information in hypergraphs for semi-supervised node classification. It introduces HyperGraph Transformer (HyperGT), a Transformer-based framework that fuses global node–hyperedge interactions with local structural cues through incidence-matrix based positional encoding and a star-expansion–driven regularization term. Empirical results on four real-world hypergraph benchmarks show state-of-the-art performance, with notable gains on Walmart, and ablations confirm the contribution of each design component. The approach highlights the value of integrating global attention with hypergraph locality, offering practical improvements for hypergraph-structured data analysis.

Abstract

Hypergraphs play a pivotal role in the modelling of data featuring higher-order relations involving more than two entities. Hypergraph neural networks emerge as a powerful tool for processing hypergraph-structured data, delivering remarkable performance across various tasks, e.g., hypergraph node classification. However, these models struggle to capture global structural information due to their reliance on local message passing. To address this challenge, we propose a novel hypergraph learning framework, HyperGraph Transformer (HyperGT). HyperGT uses a Transformer-based neural network architecture to effectively consider global correlations among all nodes and hyperedges. To incorporate local structural information, HyperGT has two distinct designs: i) a positional encoding based on the hypergraph incidence matrix, offering valuable insights into node-node and hyperedge-hyperedge interactions; and ii) a hypergraph structure regularization in the loss function, capturing connectivities between nodes and hyperedges. Through these designs, HyperGT achieves comprehensive hypergraph representation learning by effectively incorporating global interactions while preserving local connectivity patterns. Extensive experiments conducted on real-world hypergraph node classification tasks showcase that HyperGT consistently outperforms existing methods, establishing new state-of-the-art benchmarks. Ablation studies affirm the effectiveness of the individual designs of our model.

Hypergraph Transformer for Semi-Supervised Classification

TL;DR

This work addresses the challenge of leveraging global information in hypergraphs for semi-supervised node classification. It introduces HyperGraph Transformer (HyperGT), a Transformer-based framework that fuses global node–hyperedge interactions with local structural cues through incidence-matrix based positional encoding and a star-expansion–driven regularization term. Empirical results on four real-world hypergraph benchmarks show state-of-the-art performance, with notable gains on Walmart, and ablations confirm the contribution of each design component. The approach highlights the value of integrating global attention with hypergraph locality, offering practical improvements for hypergraph-structured data analysis.

Abstract

Hypergraphs play a pivotal role in the modelling of data featuring higher-order relations involving more than two entities. Hypergraph neural networks emerge as a powerful tool for processing hypergraph-structured data, delivering remarkable performance across various tasks, e.g., hypergraph node classification. However, these models struggle to capture global structural information due to their reliance on local message passing. To address this challenge, we propose a novel hypergraph learning framework, HyperGraph Transformer (HyperGT). HyperGT uses a Transformer-based neural network architecture to effectively consider global correlations among all nodes and hyperedges. To incorporate local structural information, HyperGT has two distinct designs: i) a positional encoding based on the hypergraph incidence matrix, offering valuable insights into node-node and hyperedge-hyperedge interactions; and ii) a hypergraph structure regularization in the loss function, capturing connectivities between nodes and hyperedges. Through these designs, HyperGT achieves comprehensive hypergraph representation learning by effectively incorporating global interactions while preserving local connectivity patterns. Extensive experiments conducted on real-world hypergraph node classification tasks showcase that HyperGT consistently outperforms existing methods, establishing new state-of-the-art benchmarks. Ablation studies affirm the effectiveness of the individual designs of our model.
Paper Structure (12 sections, 11 equations, 2 figures, 4 tables)

This paper contains 12 sections, 11 equations, 2 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Preliminary
  4. Hypergraph Transformer
  5. Hypergraph incidence matrix based positional encoding
  6. Hypergraph attention
  7. Hypergraph structure regularization
  8. Training strategy
  9. Experiments
  10. Experimental settings
  11. Results and analysis
  12. Conclusion

Figures (2)

  • Figure 1: Left: A hypergraph and the associated hypergraph star-expansion. Hypergraph star-expansion fully preserves the structural information of the hypergraph. Right: Illustration of the previous message-passing method and our HyperGT framework for the hypergraph depicted on the left. We build interactions among all nodes and hyperedges by hypergraph attention.
  • Figure 2: System architecture of HyperGT, taking node features $\mathbf{X}_\mathcal{V}$, hyperedge features $\mathbf{X}_\mathcal{E}$, and the hypergraph incidence matrix $\mathbf{H}$ as inputs. There are three key components: 1) Positional encoding using the hypergraph incidence matrix for node-node and hyperedge-hyperedge interactions; 2) Hypergraph attention module creating connections from one node/hyperedge to any other nodes/hyperedges; and 3) Hypergraph structure regularization for node-hyperedge connectivities.