The School of Computer Science at the University of Windsor is pleased to present …
Benchmarking GNN and Graph Transformer Models for Dynamic Link Prediction
PhD. Seminar by: Nahid Abdolrahmanpour Holagh
Date: Monday, November 3rd, 2025
Time: 3:00 pm
Location: MS Teams
Networks in our world—such as social, communication, and citation systems—are not static; they evolve continuously as new connections form and existing ones disappear. Understanding these evolving patterns is crucial for anticipating relationships, detecting emerging communities, and improving predictive intelligence in dynamic environments.
Dynamic Link Prediction (DLP) aims to forecast such temporal interactions by learning both the structure and evolution of graphs. While Graph Neural Networks (GNNs) have achieved notable progress in this area, they are inherently limited by local neighbourhood aggregation. In contrast, Graph Transformers introduce global attention mechanisms that can model long-range dependencies and capture complex temporal-structural relationships across evolving networks.
This research investigates how Graph Transformers reshape dynamic graph learning by bridging local structure awareness with global contextual reasoning. The seminar will discuss their theoretical advantages and potential to unify spatial, temporal, and semantic dimensions of graph evolution. Ultimately, this work highlights the emerging role of Graph Transformers as a foundational paradigm for scalable dynamic link prediction in real-world systems.
Internal Reader: Dr Dan Wu
Internal Reader: Dr Jianguo Lu
External Reader: Dr Narayan Kar
Advisor (s): Dr Ziad Kobti
