Vanilla Policy Optimization Is Both Optimal and Differentially Private for Stochastic Contextual Bandits
Preprint,
I am a first-year PhD student at the University of Toronto, grateful to be advised by Murat Erdogdu, Wenlong Mou, and Stanislav Volgushev. My research interests are broadly in the mathematical foundations of machine learning, especially learning and decision-making with limited feedback, memory, or compute.
Before starting my PhD, I studied mathematics and computer science at the University of Toronto and did research at the Vector Institute, where I had the pleasure of working with Daniel Roy and Idan Attias.
Recent
-
New publications: Three papers to appear at NeurIPS 2026.
paper -
New preprint: Vanilla Policy Optimization Is Both Optimal and Differentially Private for Stochastic Contextual Bandits.
paper -
Started my PhD in Statistical Sciences at the University of Toronto.
life -
Presented Reinforcement Learning with Action-Triggered Observations at ICML 2026.
paper -
Featured in The Logic’s Top Prospects 2025.
life -
Presented my first paper, Capacity-Constrained Online Learning with Delays, at COLT 2025.
paper -
Graduated from the University of Toronto and received the Governor General’s Silver Medal.
life
Preprints
Publications
-
Adaptively Incorporating Directional Hints into Zeroth-Order Optimization
Conference on Neural Information Processing Systems (NeurIPS), (to appear)
-
Capacity-Constrained Online Convex Optimization with Delayed Feedback
Conference on Neural Information Processing Systems (NeurIPS), (to appear)
-
A Reduction from Delayed to Immediate Feedback for Online Convex Optimization
Conference on Neural Information Processing Systems (NeurIPS), (to appear)
-
Reinforcement Learning with Action-Triggered Observations
International Conference on Machine Learning (ICML),
-
Capacity-Constrained Online Learning with Delays: Scheduling Frameworks and Regret Trade-offs
Conference on Learning Theory (COLT),