BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Statistical Inference for Bandit Data
DTSTART:20211203T151500
DTEND:20211203T170000
DTSTAMP:20260407T094037Z
UID:5eb96e31cb23271152c000d3142ebac173615c243ddecbeeae8b4040
CATEGORIES:Conferences - Seminars
DESCRIPTION:Kelly Zhang\, Harvard University\nBandit algorithms are increa
 singly used in real-world sequential decision-making problems. Associated 
 with this is an increased desire to be able to use the resulting datasets 
 to answer scientific questions like: Did one type of ad lead to more purch
 ases? In which contexts is a mobile health intervention effective? However
 \, classical statistical approaches fail to provide valid confidence inter
 vals when used with data collected with bandit algorithms. Alternative met
 hods have recently been developed for simple models (e.g.\, comparison of 
 means).\nYet there is a lack of general methods for conducting statistical
  inference using more complex models on data collected with (contextual) b
 andit algorithms\; for example\, current methods cannot be used for valid 
 inference on parameters in a logistic regression model for a binary reward
 . In this work\, we develop theory justifying the use of M-estimators -- w
 hich includes estimators based on empirical risk minimization as well as m
 aximum likelihood -- on data collected with adaptive algorithms\, includin
 g (contextual) bandit algorithms.\nSpecifically\, we show that M-estimator
 s\, modified with particular adaptive weights\, can be used to construct a
 symptotically valid confidence regions for a variety of inferential target
 s.\n 
LOCATION:https://epfl.zoom.us/j/67359428841
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR