|
Name of Speaker
|
Seungki Min |
|
Schedule
|
30 Jan 2026, 10am – 11.30am (60 min talk + 30 min Q&A) |
|
Venue
|
BIZ1 0302
|
| Link to register
(via Zoom) |
|
|
Title
|
An Information-Theoretic Analysis of Nonstationary Bandit Learning
|
|
Abstract
|
In many real-world bandit learning problems, the underlying environment evolves over time, requiring decision-makers to continually acquire information and adapt their action selection accordingly. In this talk, I study Bayesian formulations of nonstationary bandit problems, where environmental dynamics are modeled as stochastic processes, and develop an information-theoretic framework for analyzing attainable performance.
Our analysis yields generic regret upper bounds that extend classical results from stationary Bayesian bandits to nonstationary settings. A key insight is that the entropy rate of the optimal action process naturally quantifies the intrinsic difficulty introduced by nonstationarity. I further connect our results to existing frequentist analyses of nonstationary bandits, showing that several well-known regret bounds in the literature can be recovered as special cases within our unified framework.
|
|
About the Speaker
|
Seungki Min is an Assistant Professor of Operations Management at Seoul National University Business School. His research focuses on bandit optimization and reinforcement learning, with an emphasis on principled frameworks for dynamic decision problems arising in business and engineering applications, including online platforms, pricing, and finance. His research has appeared in Operations Research, Management Science, and leading AI/ML conferences such as ICML and NeurIPS. He earned his Ph.D. from Columbia Business School. Prior to academia, he worked in high-frequency trading domain.
|
