A Comprehensive Survey of Dynamic Graph Neural Networks: Models, Frameworks, Benchmarks, Experiments and Challenges

TL;DR

The paper delivers a comprehensive survey of dynamic graph neural networks (DGNNs), introducing a novel taxonomy that covers 81 DGNN models and 12 training frameworks. It provides a structured evaluation across six standard datasets, comparing nine representative CTDG models and three DTDG models, using metrics for convergence accuracy, efficiency, and GPU memory. The work also details 12 benchmarks and discusses open challenges, such as the need for unified frameworks and scalable training on large-scale dynamic graphs. Overall, it offers practical guidance for model and framework design, emphasizing reproducibility and scalability in dynamic graph learning tasks.

Abstract

Dynamic Graph Neural Networks (GNNs) combine temporal information with GNNs to capture structural, temporal, and contextual relationships in dynamic graphs simultaneously, leading to enhanced performance in various applications. As the demand for dynamic GNNs continues to grow, numerous models and frameworks have emerged to cater to different application needs. There is a pressing need for a comprehensive survey that evaluates the performance, strengths, and limitations of various approaches in this domain. This paper aims to fill this gap by offering a thorough comparative analysis and experimental evaluation of dynamic GNNs. It covers 81 dynamic GNN models with a novel taxonomy, 12 dynamic GNN training frameworks, and commonly used benchmarks. We also conduct experimental results from testing representative nine dynamic GNN models and three frameworks on six standard graph datasets. Evaluation metrics focus on convergence accuracy, training efficiency, and GPU memory usage, enabling a thorough comparison of performance across various models and frameworks. From the analysis and evaluation results, we identify key challenges and offer principles for future research to enhance the design of models and frameworks in the dynamic GNNs field.

Figures (19)

• Figure 1: A toy example of dynamic graph representation and learning.
• Figure 2: The taxonomy for DTDG in our survey.
• Figure 3: Node update methods based on event streams.
• Figure 4: The taxonomy for CTDG in our survey.
• Figure 5: Val.ap after each training epoch.