Quan M. Nguyen
Postdoctoral Fellow, University of British Columbia (UBC)
UBC Email: manhquan.nguyen[at]ubc.ca
(On-leave) Lecturer at University of Information Technology (UIT)
Vietnam National University, Ho Chi Minh City
UIT Email: quannm[at]uit.edu.vn
Personal: manhquan233@gmail.com
CV (updated September 2026): Link
I am a postdoctoral fellow at the University of British Columbia, working with Mark Schmidt and Christos Thrampoulidis. I am also affiliated with University of Information Technology, VNU-HCM.
My main interest is in developing theory for algorithms behind modern machine learning: how optimizers like Adam should be tuned, why training at large learning rates is unstable yet effective, and how feature geometry impacts the convergence properties of post-training algorithms for language models. My research leverages techniques developed from my earlier works on adversarial bandits, distributionally robust optimization and online reinforcement learning.
I received my PhD in Computer Science from University of Victoria (Canada) in 2026, advised by Nishant Mehta. During my PhD, I spent Fall 2024 working with Shinji Ito and Junpei Komiyama at RIKEN AIP and University of Tokyo. Before that, I worked as a research scientist at University of Hamburg (Germany) after having obtained a MSc. in Intelligent Adaptive Systems there, advised by Mikko Lauri and Simone Frintrop. Prior to Hamburg, I did a Bachelor of Engineering from FPT University (Vietnam), supervised by Phan Duy Hung and Lan V. Truong.
How to pronounce my name: I go by "Quan", which can be pronounced like |Kwan|.
Oct 15, 2026: I am affiliated with University of Information Technology (UIT) of VNU-HCM as an on-leave limited-term lecturer.
Sep 29, 2026: Two workshop papers accepted to NeurIPS OPT as posters.
May 1, 2026: I joined UBC as a postdoc, working with Mark Schmidt and Christos Thrampoulidis.
Apr 13, 2026: I left A5 Labs.
Apr 6, 2026: I will give an invited talk on my ALT 2026 paper at the Dartmouth/Berkeley/Rice Reading Group on Foundation Models/AI.
Feb 22-28, 2026: I attended ALT 2026 at the Fields Institute in Toronto. At ALT, I gave two talks: one filling-in talk for Hamish on sparse non-parametric bandits, and one talk on my own paper on tuning Adam's momentum factors.
Jan 15, 2026: I successfully defended my PhD thesis.
Dec 2025 (2): One single-author paper accepted to ALT.
Dec 2025 (1): I joined A5Labs / AceGuardian (a California startup) as an AI Researcher (remote).
Quan Nguyen. How to Set β1, β2 in Adam: An Online Learning Perspective. International Conference on Algorithmic Learning Theory (ALT) 2026.
Quan Nguyen, Shinji Ito, Junpei Komiyama, Nishant A. Mehta. Data-dependent Bounds with T-Optimal Best-of-Both-Worlds Guarantees in Multi-Armed Bandits using Stability-Penalty Matching. Conference on Learning Theory (COLT) 2025.
Quan Nguyen, Nishant A. Mehta, Cristóbal Guzmán. Beyond Minimax Rates in Group Distributionally Robust Optimization via a Novel Notion of Sparsity. International Conference on Machine Learning (ICML) 2025.
Quan Nguyen, Nishant A. Mehta. Near-optimal Per-Action Regret Bounds for Sleeping Bandits. International Conference on Artificial Intelligence and Statistics (AISTATS) 2024. Arxiv.
Quan Nguyen, Nishant A. Mehta. Adversarial Online Multi-Task Reinforcement Learning. International Conference on Algorithmic Learning Theory (ALT) 2023. Long version.
Quan Nguyen, Julius Richter, Mikko Lauri, Timo Gerkmann, Simone Frintrop. Improving mix-and-separate training in audio-visual sound source separation with an object prior. International Conference on Pattern Recognition (ICPR) 2020. (Oral Presentation, top 4%). PDF.
Quan Nguyen, Mikko Lauri, Simone Frintrop. Distance dependent Maximum margin Dirichlet Process Mixture. The 16th Pacific Rim International Conferences on Artificial Intelligence (PRICAI), 2019
Quan Nguyen*, Sharan Vaswani*, Christos Thrampoulidis. Sample Complexity of On-Policy RLVR Algorithms: A Generic Analysis using Non-Uniform Smoothness and Aligned Features. NeurIPS Optimization for Machine Learning (OPT) 2026. (*) Co-first authors.
Quan Nguyen*, Robin Yadav*, Param Mody, Mark Schmidt. Shallow ReLU Classifiers on Non-Linearly Separable Data: The Loss Landscape, Edge of Stability, and Beyond. NeurIPS Optimization for Machine Learning (OPT) 2026. (*) Co-first authors. Long version is available here: Arxiv.
Quan Nguyen, Nishant A. Mehta, Cristóbal Guzmán. Beyond Minimax Rates in Group Distributionally Robust Optimization via a Novel Notion of Sparsity. AAAI Workshop on Biased and Scarce Data, 2024.
Quan Nguyen, Thanh Nguyen-Tang. One-Layer Transformers are Provably Optimal for In-context Reasoning and Distributional Association Learning in Next-Token Prediction Tasks. 2025. Under review.
N. Churamani, P. Anton, M. Bruegger, E. Fliesswasser, T. Hummel, J. Mayer, W. Mustafa, H. G. Ng, T. Nguyen, Q. Nguyen, M. Soll, S. Springenberg, S. Griffiths, S. Heinrich, N. Navarro-Guerrero, E. Strahl, J. Twiefel, C. Weber, S. Wermter. (2017). The Impact of Personalisation on Human-Robot Interaction in Learning Scenarios. Proceedings of the Fifth International Conference on Human Agent Interaction (HAI), pages 171–180. PDF.
H. G. Ng, P. Anton, M.Bruegger, N. Churamani, E. Fliesswasser, T. Hummel, J. Mayer, W. Mustafa, T. Nguyen, Q. Nguyen, M. Soll, S. Springenberg, S. Griffiths, S. Heinrich, N. Navarro-Guerrero, E. Strahl, J. Twiefel, C. Weber, S. Wermter. (2017). Hey Robot, Why Don’t You Talk To Me?. Proceedings of the IEEE International Symposium on Robot and Human Interactive Communication (RO-MAN), pages 728–731.
How to Set β1, β2 in Adam: An Online Learning Perspective.
International Conference on Algorithmic Learning Theory (ALT), Toronto, February 2026.
Dartmouth/Berkeley/Rice Reading Group on Foundation Models/AI, April 2026.
One-layer transformers for In-context recall tasks: Bayes Optimality and Out-of-distribution Generalization. Invited Talk. University of British Columbia, Vancouver, September 2025.
Beyond Minimax Rates in Group Distributionally Robust Optimization via a Novel Notion of Sparsity.
AAAI Workshop on Artificial Intelligence with Biased or Scarce Data, Vancouver, March 2024.
ICML Poster, July 2025.
Data-dependent Bounds with Best-of-both-world Guarantees for Multi-armed Bandits using Stability-Penalty Matching.
RIKEN Sequential Decision Making Seminar, Tokyo, November 2024.
Conference on Learning Theory (COLT), Lyon, July 2025
Adversarial Online Multi-Task Reinforcement Learning. International Conference on Algorithmic Learning Theory (ALT), Singapore, February 2023.
Improving mix-and-separate training in audio-visual sound source separation with an object prior. International Conference on Pattern Recognition (ICPR), Milan, Italy, 2020.
Distance dependent Maximum margin Dirichlet Process Mixture. The 16th Pacific Rim International Conferences on Artificial Intelligence (PRICAI), Fiji, 2019.
Non-parametric Bayesian clustering with applications in object candidate detection. Machine Learning Oberseminar, University of Hamburg, Germany, 2017.
PC member for ECAI (2025)
Reviewer for NeurIPS (2023 - 2025), ICML (2025-), ICLR (2024, 2025), AISTATS (2025), ICRA (2018), CVPR (2018), TMLR (2025), Artificial Intelligence Journal (2026)
Teaching Assistant for CSC 531/ CSC 431 Machine Learning Theory, University of Victoria. Fall 2024.
Teaching Assistant for CSC 503 / SENG 474 Data Mining. University of Victoria. Spring 2022, Spring 2023, Fall 2024, Summer 2025.
Teaching Assistant for Cognitive Computer Vision. University of Hamburg. 2018 and 2019.
Mentoring:
Irene Hang Duong (BSc CS at UVic. Now a software engineer at Google).
Robin Yadav (MSc CS at UBC)
Param Mody (BSc Math at UBC)
NeurIPS's Top Reviewer, 2024.
Charles S. Humphrey Graduate Student Awards, University of Victoria, 2023.
University of Victoria Conference Travel Awards, 2023.
University of Victoria Graduate Awards for Top-Performing PhD Students. 2021 - 2023.
University of Victoria Graduate Fellowship Awards. 2021-2024.
Instagram Machine Learning Competition. Second Prize. 2017.
Vietnam ACM ICPC National Algorithm Programming Competition. Bronze prize. 2013.
Badminton is one of my biggest hobbies. I am a big fan of Taufik Hidayat and Tai Tzu Ying and often practice their signature shots (in particular, backhand smash and reverse slices). Also, I have been a member of UVic's Badminton Club since 2021.
I also watch Xiangqi (Chinese chess) videos from time to time. I particularly enjoy the matches played by Hu Rong Hua and Li Laiqun. Sometimes I watch Western chess videos as well.