The 69th Seminar
Date and Time: May 29th 3:00 pm – 4:00 pm(JST)
Speaker: Yihan Du (Tsinghua University)
Title: Risk-aware Online Decision Making
Venue: Zoom webinar
Classic online decision-making (e.g., bandit and reinforcement learning) algorithms focus mainly on the risk-neutral criterion, i.e., maximizing the expected cumulative reward, and can fail to avoid rare but disastrous situations. As a result, existing online decision-making algorithms cannot be directly applied to tackle real-world risk-sensitive tasks, such as autonomous driving and clinical treatment planning, where policies that ensure low risk of getting into catastrophic situations are strongly preferred. Motivated by the above facts, we study risk-aware online decision making. Specifically, we investigate bandit and reinforcement learning (RL) problems equipped with risk-sensitive criteria, e.g., mean-covariance and conditional value-at-risk (CVaR), which aim to control the risk during decision making and prevent the agent from getting into catastrophic states. We design efficient risk-aware bandit and RL algorithms with rigorous theoretical guarantees.
Yihan Du is currently a fifth-year Ph.D. student at the Institute for Interdisciplinary Information Sciences (headed by Prof. Andrew Chi-Chih Yao) of Tsinghua University, advised by Prof. Longbo Huang. She was a research intern at Microsoft Research Asia supervised by Dr. Wei Chen (IEEE Fellow, Director of Microsoft Research Asia Theory Center), and a visiting Ph.D. student at Cornell University supervised by Prof. Wen Sun. She is broadly interested in the area of machine learning, with emphases on online learning, reinforcement learning and representation learning.