- This event has passed.
AMS Weekly Seminar | Hamsa Bastani
February 27 @ 1:30 pm - 2:30 pm
Location: Gilman 50
When: February 27th at 1:30 p.m.
Title: Multitask Learning and Bandits via Robust Statistics
Abstract: We study the “many bandits” setting, where a decision-maker must simultaneously learn across multiple distinct but related linear contextual bandit instances. Such problems arise in a number of applications, including personalized medicine, targeted COVID-19 screening, dynamic pricing, and adaptive auditing. We study the setting where the unknown parameter in each bandit instance can be decomposed into a global parameter plus a sparse instance-specific term. Then, we propose a novel two-stage estimator that exploits this structure in a sample-efficient way by using a combination of robust statistics (to learn across similar instances) and LASSO regression (to debias the results). We embed this estimator within a bandit algorithm, and prove that it improves asymptotic regret bounds in the context dimension; this improvement is exponential for data-poor instances. We further demonstrate how our results depend on the underlying network structure of bandit instances. Joint work with Kan Xu.
Zoom link: https://wse.zoom.us/j/93287142219?pwd=z9fqWnRMzmzS0SGijRiie5yN3kHRSZ.1