The rapid advancement of Graph Neural Networks (GNNs) has revolutionized how machine learning addresses structured, relational, and topological data. GNNs are now foundational tools for modeling interconnected information, powering breakthroughs across domains from social networks to biomedicine. Yet, their growing impact brings forth a host of new challenges—including deepening our theoretical understanding, enhancing expressiveness and generalization, tackling the scalability of massive and evolving graphs, and fostering seamless integration with emerging paradigms like Large Language Models (LLMs).

As GNNs are increasingly deployed in real-world systems featuring heterogeneous, multimodal, temporal, and dynamic graph structures, there is pressing need for continued innovation. This encompasses development of novel architectures, robust interpretability frameworks, and domain-specific deployments to unlock their full potential.

This Research Topic aims to feature research driving progress in the theoretical foundations, architectural advances, and impactful applications of GNNs across a spectrum of scientific, engineering, and socio-technical landscapes.

Areas of interest include but are not limited to:
• Theoretical Advancements in Graph Learning
• Expressive power and mathematical limitations of GNNs
• Sample and label efficiency in training GNNs
• Innovations in GNN Architectures and Learning Paradigms
• Scalable and efficient GNN models for large-scale graphs
• Dynamic, temporal, and streaming graph learning techniques
• Graph-based retrieval-augmented generation (RAG) methods
• Transportation, climate, and urban systems modeling

We invite original research articles, reviews, perspectives, and case studies that propel the field of GNNs forward—bridging gaps, tackling open problems, and demonstrating innovations across theory, methodology, and practice.

Authors

Frontiers in Complex Systems

Multi- and Cross-Disciplinary Complexity