Presentation at Uppsala University on Learning-on-Graph (LoG) Research Topics

This week, the local events of Learning on Graphs (LoG) Conference took place at Uppsala University, gathering researchers from academia and industry to discuss cutting-edge advancements in machine learning on graphs and geometry. The two-day workshop, affiliated with the LoG Conference, featured keynotes, contributed talks, and poster sessions, fostering discussions on the latest research in graph learning.

I had the opportunity to present our work on Continuous-Time Dynamic Graphs (CTDGs), where I highlighted how temporal graphs can model unstructured event streams and facilitate downstream tasks such as node classification and link prediction. My talk covered representation learning techniques designed to capture evolving structures in graphs, contributing to advancements in dynamic graph modeling.

Additionally, Sofiane Ennadir from our team presented his research on robust learning in graph-based models. His talk, If You Want to Be Robust, Be Wary of Initialization, emphasized the impact of initialization strategies on the robustness of graph learning algorithms and how adversarial attacks can exploit weak points in graph models.

The meetup provided a fantastic platform to exchange ideas with researchers tackling real-world problems using graph-based learning. With keynotes from leading experts, discussions on equivariant neural networks, adversarial robustness, and relational inductive biases, and opportunities for collaboration, the event reinforced the importance of graph learning in AI research and applications.

Special thanks to the organizers of LoG Meetup Sweden for a well-structured and engaging event! Looking forward to continued collaborations in this exciting space.

A few snapshots from our presentations at the event:

Stay tuned for more updates on our research and upcoming events!