Csaba Szepesvári

Csaba Szepesvári is a Professor and Canada CIFAR AI Chair in the Department of Computing Science at the University of Alberta, affiliated with Amii. Since 2017, he has been on partial leave, leading the Foundations team at DeepMind in Edmonton. His work bridges theoretical machine learning and practical reinforcement learning systems.

His research interests lie at the intersection of reinforcement learning, sequential decision making, bandit algorithms, and theoretical machine learning. He develops and analyzes algorithms for efficient learning in complex environments, focusing on sample efficiency, function approximation, and optimization. His recent work explores policy gradient methods, offline and online RL, uncertainty estimation, and foundational limits of learning algorithms.

The trends in his recent publications (2023–2024) emphasize theoretical advances in RL and bandits, particularly in q π-realizability, natural policy gradient, ensemble sampling, and lower bounds. The work spans NeurIPS, ICML, COLT, and AISTATS, reflecting deep theoretical engagement with practical implications.

He is recognized as a Canada CIFAR AI Chair, a prestigious award supporting leading researchers in artificial intelligence.

Csaba Szepesvári mentors students and collaborates extensively, often with researchers like András György, Tor Lattimore, Gellért Weisz, and Dale Schuurmans. He has co-organized RL theory seminars and workshops and co-authored the influential book Bandit Algorithms (2020). His leadership in both academic and industrial research (DeepMind) underscores his impact on the field.

He is actively involved in the machine learning community through research, mentorship, and event organization, contributing to both foundational theory and real-world applications.