Key Papers and Contributions in Dynamic Graph Neural Networks (GNNs)

Give explain with real citation for dynamic GNN papers with truth referneces that exist

Dynamic Graph Neural Networks (GNNs) are an extension of traditional GNNs designed to handle graphs that evolve over time. These models are particularly useful for applications where the underlying graph structure changes, such as social networks, communication networks, and recommendation systems. Below are some key papers in the field of dynamic GNNs, along with citations and brief explanations of their contributions.

1. EvolveGCN: Evolving Graph Convolutional Networks for Dynamic Graphs

   • Citation: Pareja, A., Domeniconi, G., Chen, J., Ma, T., Suzumura, T., Kanezashi, H., ... & Leiserson, C. (2020). EvolveGCN: Evolving Graph Convolutional Networks for Dynamic Graphs. Proceedings of the AAAI Conference on Artificial Intelligence, 34(04), 5363-5370.
   • Explanation: EvolveGCN introduces a method to adapt Graph Convolutional Networks (GCNs) to dynamic graphs by evolving the GCN parameters over time. The model uses a recurrent neural network (RNN) to update the GCN parameters, allowing it to capture temporal dependencies in the graph data.

2. Dynamic Graph Representation Learning via Self-Attention Networks

   • Citation: Xu, D., Ruan, C., Korpeoglu, E., Kumar, S., & Achan, K. (2020). Inductive representation learning on temporal graphs. International Conference on Learning Representations (ICLR).
   • Explanation: This paper presents a self-attention mechanism for dynamic graph representation learning. The model, known as Temporal Graph Networks (TGNs), leverages self-attention to capture the temporal dynamics of the graph, enabling it to make predictions based on the evolving structure.

3. DySAT: Deep Neural Representation Learning on Dynamic Graphs via Self-Attention Networks

   • Citation: Sankar, A., Wu, Y., Gou, L., Zhang, W., & Yang, H. (2020). DySAT: Deep Neural Representation Learning on Dynamic Graphs via Self-Attention Networks. Proceedings of the 13th International Conference on Web Search and Data Mining (WSDM), 519-527.
   • Explanation: DySAT employs self-attention mechanisms to learn node representations in dynamic graphs. The model captures both structural and temporal dependencies by applying self-attention layers over the graph's temporal snapshots, making it effective for tasks like link prediction and node classification.

4. JODIE: Predicting Dynamic Embedding Trajectory in Temporal Interaction Networks

   • Citation: Kumar, S., Zhang, X., & Leskovec, J. (2019). Predicting Dynamic Embedding Trajectory in Temporal Interaction Networks. Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining (KDD), 1269-1278.
   • Explanation: JODIE (Joint Dynamic User-Item Embeddings) focuses on temporal interaction networks, such as user-item interactions in recommendation systems. The model predicts the future trajectory of embeddings by jointly learning the dynamic representations of users and items, capturing the temporal evolution of interactions.

5. Temporal Graph Networks for Deep Learning on Dynamic Graphs

   • Citation: Rossi, E., Chamberlain, B., Frasca, F., Eynard, D., Monti, F., & Bronstein, M. (2020). Temporal Graph Networks for Deep Learning on Dynamic Graphs. arXiv preprint arXiv:2006.10637.
   • Explanation: Temporal Graph Networks (TGNs) are designed to handle continuous-time dynamic graphs. The model uses a combination of message passing and memory modules to update node embeddings over time, allowing it to efficiently process and learn from dynamic graph data.

These papers represent significant advancements in the field of dynamic GNNs, each introducing novel methods to handle the complexities of evolving graph structures. The citations provided are from reputable conferences and journals, ensuring the credibility and impact of the research.