BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Risk Minimization from Adaptively Collected Data: Guarantees for P
 olicy Learning
DTSTART:20211217T151500
DTEND:20211217T170000
DTSTAMP:20260407T152508Z
UID:7846f0abd93231ae67427975d24ccd8dad3ac617468f36dd81c1436a
CATEGORIES:Conferences - Seminars
DESCRIPTION:Antoine Chambaz\, Université de Paris        \nEmpiric
 al risk minimization (ERM) is the workhorse of machine learning but its mo
 del-agnostic guarantees can fail when using data collected in an adaptive 
 fashion\, like in the setting of a contextual bandit algorithm for instanc
 e.  In this setting\, and focusing on policy learning\, I will present a 
 generic importance sampling weighted ERM algorithm and its regret guarante
 es\, which close an open gap in the existing literature whenever explorati
 on decays to zero.  An empirical investigation validates the theory.\n\nT
 his is a joint work with Aurélien Bibaut (Netflix)\, Nathan Kallus (Corne
 ll University and Netflix)\, Maria Dimakopoulou (Netflix) and Mark J. van 
 der laan (UC Berkeley)\n\n 
LOCATION:https://epfl.zoom.us/j/61899097082
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR