Dynamic Graph
Dynamic graphs represent evolving relationships between entities, focusing on modeling changes in both network structure and node/edge attributes over time. Current research emphasizes developing efficient and expressive model architectures, such as dynamic graph neural networks (DyGNNs) and graph transformers, often incorporating techniques like graph coarsening, attention mechanisms, and temporal convolutional layers to capture complex spatiotemporal patterns. This field is significant for its applications in diverse areas including urban planning, social network analysis, and brain imaging, enabling improved prediction, anomaly detection, and a deeper understanding of complex systems. Furthermore, research is actively addressing challenges related to scalability, robustness to distribution shifts, and explainability of model predictions.
Papers
A Deep Autoregressive Model for Dynamic Combinatorial Complexes
Ata TunaImperial College LondonQuality Measures for Dynamic Graph Generative Models
Ryien Hosseini, Filippo Simini, Venkatram Vishwanath, Rebecca Willett, Henry HoffmannUniversity of Chicago●Argonne National Laboratory●NSF-Simons National Institute for Theory and Mathematics in Biology
Spectral Theory for Edge Pruning in Asynchronous Recurrent Graph Neural Networks
Nicolas BessoneIT University of CopenhagenUniDyG: A Unified and Effective Representation Learning Approach for Large Dynamic Graphs
Yuanyuan Xu, Wenjie Zhang, Xuemin Lin, Ying ZhangThe University of New South Wales●Shanghai Jiao Tong University●Zhejiang Gongshang University
Efficient Dynamic Attributed Graph Generation
Fan Li, Xiaoyang Wang, Dawei Cheng, Cong Chen, Ying Zhang, Xuemin LinDG-Mamba: Robust and Efficient Dynamic Graph Structure Learning with Selective State Space Models
Haonan Yuan, Qingyun Sun, Zhaonan Wang, Xingcheng Fu, Cheng Ji, Yongjian Wang, Bo Jin, Jianxin Li