A Graph Convolutional Network with Dynamic Updating of Graph Topology Structure
Abstract
Graph Convolutional Networks (GCNs) have demonstrated notable strengths in handling this type of data. However, most existing methods assume that graph-structured data is homophilous, meaning nodes of the same type are more likely to form edges. In contrast, relatively little attention has been paid to heterophilous graph data. To address this issue, we propose a novel Feature and Label-based Graph Convolutional Network (FL-GCN) framework, which incorporates a graph structure learning strategy based on features and labels. Experimental results demonstrate that our proposed model outperforms existing methods across multiple real-world datasets.
Fields of Study & Subjects
Advanced Graph Neural Networks
Graph Theory and Algorithms
Complex Network Analysis Techniques
Publication details
- Journal / source
- IEICE Transactions on Fundamentals of Electronics Communications and Computer Sciences
- Volume
- E109.A
- Issue
- 7
- Pages
- 1309–1313
- Publication date
- 13 January 2026
- Language
- EN