BEGIN:VCALENDAR
PRODID:-//eluceo/ical//2.0/EN
VERSION:2.0
CALSCALE:GREGORIAN
BEGIN:VEVENT
UID:www.tcs.tifr.res.in/event/1468
DTSTAMP:20240829T095226Z
SUMMARY:Performativity in Reinforcement Learning
DESCRIPTION:Speaker: Debmalya Mandal (University of Warwick)\n\nAbstract: \
 nHow should we design machine learning systems when the underlying environ
 ment (e.g. data distribution) changes in response to the deployed model? I
 n the context of supervised learning\, the framework of performative predi
 ction provides game-theoretic solution concepts that a learner can optimiz
 e in the presence of decision-dependent distributions. In this talk\, I wi
 ll provide an overview of our work to model such “performativity” in t
 he context of reinforcement learning. In particular\, I will describe how 
 to reach a stable policy in a setting where the underlying MDP reacts to t
 he deployed policy. I will end with some open questions\, and if time perm
 its\, some of our recent works on reinforcement learning with human feedb
 ack.\nShort Bio:\nDebmalya Mandal is an assistant professor at the Univers
 ity of Warwick\, UK. He completed his PhD from Harvard University and was
  subsequently a postdoc at Columbia University and Max Planck Institute.
  He is broadly interested in problems at the interface of machine learni
 ng and multi-agent systems.\n
URL:https://www.tcs.tifr.res.in/web/events/1468
DTSTART;TZID=Asia/Kolkata:20240903T100000
DTEND;TZID=Asia/Kolkata:20240903T110000
LOCATION:A-201 (STCS Seminar Room)
END:VEVENT
END:VCALENDAR