Geometric Deep Learning: A Bibliometric Analysis of Research Evolution and Emerging Directions
IEEE Access, vol.14, pp.124642-124662, 2026 (SCI-Expanded, Scopus)
- Publication Type: Article / Article
- Volume: 14
- Publication Date: 2026
- Doi Number: 10.1109/access.2026.3723132
- Journal Name: IEEE Access
- Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Compendex, INSPEC, Directory of Open Access Journals
- Page Numbers: pp.124642-124662
- Keywords: Bibliometric analysis, geometric deep learning (GDL), graph neural networks (GNNs), non-Euclidean data, science mapping
- Middle East Technical University Affiliated: Yes
Abstract
Geometric Deep Learning (GDL) has emerged as a powerful framework for learning from graph-structured and non-Euclidean data, addressing fundamental limitations of conventional deep learning approaches. This study presents a structured bibliometric analysis of GDL research based on 733 publications retrieved from the Web of Science Core Collection (2017-2025), examining the field's evolution, intellectual structure, and emerging research directions. The findings reveal rapid and sustained growth driven by increasing interdisciplinary adoption across computer science, biomedical engineering, and data-intensive scientific domains. GDL is characterized by a stable methodological core rooted in graph-based and geometric principles, alongside a growing diversification of applications including computer vision, computational biology, and network analysis. Collaboration network analysis reveals a research ecosystem shaped by concentrated institutional leadership and globally distributed partnerships, facilitating the diffusion of key methodologies across disciplinary boundaries. Despite these advances, challenges remain in bridging theoretical developments with real-world applications and in broadening global research participation. The bibliometric evidence suggests that GDL is consolidating into a significant and rapidly growing methodological framework within modern AI, with future progress likely to emerge from the integration of theoretical innovation, interdisciplinary collaboration, and application-driven research.