Wouter M. Koolen-Wijkstra

Wouter M. Koolen-Wijkstra is a Professor of Mathematical Machine Learning at the University of Twente (Statistics group) and a Scientific Staff Member at Centrum Wiskunde & Informatica (CWI), Amsterdam, in the Machine Learning department. His research bridges theoretical machine learning, game theory, and statistics, with active projects on multi-armed bandits, online learning, and safe inference methodologies. He co-leads INRIA-CWI associate teams (6PAC and 4TUNE) and is an ELLIS Scholar.

His work emphasizes provable guarantees in learning algorithms, including:

• Regret minimization under risk-averse scenarios
• Multi-scale adaptation in online decision-making
• Game-theoretic equilibria computation
• Anytime-valid statistical inference via e-processes

Recent publications demonstrate a focus on robust learning frameworks, particularly in bandit problems, hypothesis testing, and Nash equilibrium characterization, often leveraging information-theoretic and optimization principles.

Awards include:

• Veni Grant (2015) for 'Learning at the Intrinsic Task Pace'
• QUT Vice-Chancellor's Fellowship (2013) for multitask learning
• Rubicon Grant (2010) for game-theoretic online learning
• ELLIS Scholar recognition

He teaches graduate courses on Machine Learning Theory and Graphical Models at CWI. Current grants include collaborations with INRIA (4TUNE and 6PAC teams) and industry partnerships (e.g., PPS Booking.COM).