A survey of dynamic graph neural networks
Yanping Zheng, Lu Yi, Zhewei Wei
TL;DR
This survey addresses the challenge of learning on graphs whose structure and attributes evolve over time by categorizing dynamic GNN approaches into discrete-time and continuous-time paradigms and outlining architectures that couple spatial GNNs with temporal models. It systematically reviews scalable processing and training for large-scale dynamic graphs, surveys datasets and benchmarks, and discusses advances in transfer learning, pretraining, and future directions such as interpretability and the integration of LLMs. The work highlights a landscape of methods ranging from snapshot-based to fully integrated temporal networks, and from Hawkes-process–driven CTDG models to attention-based and memory-augmented approaches, emphasizing practical considerations for scalability and evaluation. Overall, the paper provides a comprehensive foundation for researchers to compare methods, identify gaps, and pursue scalable, interpretable, and data-efficient dynamic graph learning with broad real-world impact.
Abstract
Graph neural networks (GNNs) have emerged as a powerful tool for effectively mining and learning from graph-structured data, with applications spanning numerous domains. However, most research focuses on static graphs, neglecting the dynamic nature of real-world networks where topologies and attributes evolve over time. By integrating sequence modeling modules into traditional GNN architectures, dynamic GNNs aim to bridge this gap, capturing the inherent temporal dependencies of dynamic graphs for a more authentic depiction of complex networks. This paper provides a comprehensive review of the fundamental concepts, key techniques, and state-of-the-art dynamic GNN models. We present the mainstream dynamic GNN models in detail and categorize models based on how temporal information is incorporated. We also discuss large-scale dynamic GNNs and pre-training techniques. Although dynamic GNNs have shown superior performance, challenges remain in scalability, handling heterogeneous information, and lack of diverse graph datasets. The paper also discusses possible future directions, such as adaptive and memory-enhanced models, inductive learning, and theoretical analysis.