Read the original document by opening this link in a new tab.
Table of Contents
Abstract
Introduction
Related Work
Hypergraph Neural Networks
Implementation
Experiments
Dataset
Citation network classification
Visual object recognition
Conclusion
Summary
In this paper, the authors present a hypergraph neural networks (HGNN) framework for data representation learning. They propose to incorporate a hypergraph structure to encode high-order data correlation, especially beneficial for dealing with complex data. The HGNN method outperforms traditional methods in experiments on citation network classification and visual object recognition tasks. The paper introduces the concept of hypergraph learning and its application in neural networks, highlighting the efficiency and effectiveness of the proposed HGNN model.