Unveiling Graph Datasets

Bastian Rieck, a tenured professor of machine learning at the University of Fribourg and head of the AIDOS lab, brings a unique perspective combining pure mathematics, machine learning, and graph theory. In this episode, listeners will learn about the challenges of evaluating graph learning methods and datasets through the lens of his work on the "No Metric to Rule Them All" paper, introducing the RINGS framework for assessing how much a graph's structure and node features truly contribute to learning tasks. Key insights include how many popular graph datasets might be misleading researchers by not requiring genuine graph reasoning, the importance of evaluating datasets under structural perturbations, and why researchers should rethink how datasets are curated and validated and what are the "right" questions we should ask them. Follow our guest Github Bastian's Webpage Bluesky Papers in focus No Metric to Rule Them All: Toward Principled Evaluations of Graph-Learning Datasets. (2025)

Guest

Bastian Grossenbacher Rieck: Bastian Rieck, M.Sc., Ph.D., is a Full Professor of Machine Learning at the University of Fribourg in Switzerland and the Principal Investigator of the AIDOS Lab, which focuses on developing novel machine learning methods driven by geometry and topology. Bastian’s research has received multiple awards, including an ERC Starting Grant. He is also a member of ELLIS, the European Laboratory for Learning and Intelligent Systems, and AI-LIFE, the International Network of Artificial Intelligence for the Life Sciences. Wearing yet another hat, he serves as the co-director of the Applied Algebraic Topology Research Network. Bastian received his M.Sc. degree in mathematics, as well as his Ph.D. in computer science, from Heidelberg University in Germany, graduating both times with distinction. He is a big proponent of scientific outreach and enjoys blogging about his research, academia in general, and software development.