Statistics Seminar
Lan Wang: New Results for Distributional Reinforcement Learning
Event Details
Date
Thursday, September 3, 2026
Time
1-2 p.m.
Location
7560 Morgridge Hall
Description
Distributional reinforcement learning (RL) models the entire distribution of returns and is particularly useful for risk-sensitive decision-making. Quantile temporal difference (QTD) learning is a widely used model-free distributional RL method with strong empirical performance, yet its theoretical guarantees remain less developed. We provide nonasymptotic performance guarantees for QTD.
Cost
Free
Accessibility
We value inclusion and access for all participants and are pleased to provide reasonable accommodations for this event. Please email thomas.jilk@wisc.edu to make a disability-related accommodation request. Reasonable effort will be made to support your request.