BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Thompson Sampling for Stochastic Bandits with Noisy Contexts: An I
 nformation-Theoretic Analysis - Dr Sharu Jose\, University of Birmingham
DTSTART:20231025T130000Z
DTEND:20231025T140000Z
UID:TALK204664@talks.cam.ac.uk
CONTACT:Dr Varun Jog
DESCRIPTION:Decision-making in the face of uncertainty is a practical chal
 lenge found across various areas such as control and robotics\, clinical t
 rials\, communication\, and ecology. An extensively studied decision-makin
 g framework is that of stochastic contextual bandits (CBs) which uses side
  information\, termed context\, for sequential decision making.  Prior res
 earch on CBs has mostly focussed on models where the contexts are well-def
 ined. This\, however\, is not true in real-world applications where the co
 ntexts are either noisy or are indicative of predictive measurements. In t
 his talk\, we focus on noisy CBs where the learner observes only a noisy\,
  corrupted\, version of the true context through an unknown noise channel.
   We introduce a Thompson Sampling algorithm for Gaussian bandits with Gau
 ssian context noise that can ‘approximate’ the action policy of an ora
 cle which has access to the predictive distribution of the true context fr
 om the observed noisy context. Using information-theoretic tools\, we stud
 y the Bayesian regret of the proposed algorithm.
LOCATION:MR5\, CMS Pavilion A
END:VEVENT
END:VCALENDAR