Zak Mhammedi

Zak Mhammedi is a Postdoctoral Associate at the Massachusetts Institute of Technology working with Sasha Rakhlin on theoretical problems in machine learning. His research spans Reinforcement Learning, Control Theory, and Optimization with strong theoretical foundations.

His educational background includes:

• PhD in Computer Science, 2021, Australian National University
• MPhil in Computer Science, 2016-2017, The University of Melbourne

Mhammedi's research focuses on bridging theoretical guarantees with practical algorithms, particularly in online learning, projection-free optimization, and reinforcement learning. His work often addresses fundamental limitations in statistical learning while developing computationally efficient methods. He has made significant contributions to understanding risk-monotonicity, developing parameter-free online learning algorithms, and creating efficient exploration strategies for reinforcement learning.

His publication record shows a consistent trajectory in theoretical machine learning, with numerous papers in top-tier venues including NeurIPS, ICML, and COLT. His work demonstrates expertise across multiple subfields including PAC-Bayesian theory, online convex optimization, and control theory applications to machine learning.

His scientific achievements include:

• Oral Presentation at NeurIPS 2021 (awarded to approximately 1% of submissions)
• Spotlight Presentation at NeurIPS 2020
• Spotlight Presentation at NeurIPS 2018 (awarded to approximately 3% of submissions)

Mhammedi actively contributes to the academic community through program committee service for COLT 2021 and 2022, and as a reviewer for major conferences including COLT, NeurIPS, and ICML across multiple years. His technical expertise spans adaptive methods, computational efficiency, constrained optimization, and theoretical aspects of deep learning and reinforcement learning.