Technical - Partner Event
AI Seminar – Bingshan Hu
When
Jan. 21, 2022 - Jan. 21, 2022
12:00 PM – 1:00 PM MST
Where
Online
Online
Presenter: Bingshan Hu
Title: (Near)-optimal Regret Bound for Differentially Private Thompson Sampling
Abstract: A Multi-armed bandit problem is a classical sequential decision-making problem in which the goal is to accumulate as much reward as possible. In this learning model, only a limited amount of information is revealed in each round. The imperfect feedback model results in the learning algorithm being in a dilemma between exploration (gaining information) and exploitation (accumulating reward). Thompson Sampling is one of the classical learning algorithms that can make a good balance between exploration and exploitation and it always has a very competitive empirical performance.
In the standard non-private learning, the learning algorithm can always get access to the true revealed information to make future decisions. However, if the revealed information is about individuals, to preserve privacy, the decisions made by the learning algorithm should not rely on the true revealed information. In this talk, I will present a Thompson Sampling-based algorithm, DP-TS, for private stochastic bandits. The regret upper bound for DP-TS matches the discovered regret lower bound up to an extra loglogT factor.
Bio: Bingshan Hu is an Amii Postdoctoral Fellow co-hosted by Prof. Nidhi Hegde from University of Alberta and Prof. Mark Schmidt from University of British Columbia. She completed her PhD from University of Victoria under the supervision of Prof. Nishant Mehta in 2021. Her research lies in the theoretical side of machine learning, aiming at devising efficient and private online learning algorithms. She serves as reviewers for conferences such as NeurlPS, ICML, and AISTATS. She was recognized as one of the top 10% of high-scoring reviewers at NeurIPS 2020.
Prior to pursuing her PhD studies, she worked in industry research labs as a wireless technology specialist. She invented/co-invented around 20 patents with more than half of them having been granted by either the European patent office or US patent office. Besides the foundations of online learning, she is also interested in the usage of online learning in novel applications in wireless networks.
The University of Alberta Artificial Intelligence (AI) Seminar is a weekly meeting where researchers (including students, developers, and professors) interested in AI can share their current research. Presenters include local speakers from the University of Alberta and industry as well as other institutions. The seminars discuss a wide range of topics related in any way to Artificial Intelligence, from foundational theoretical work to innovative applications of AI techniques to new fields and problems are of interest. Learn more at the AI Seminar website and by subscribing to the mailing list!
Looking to build AI capacity? Need a speaker at your event?
