BEGIN:VCALENDAR
VERSION:2.0
PRODID:icalendar-ruby
CALSCALE:GREGORIAN
X-WR-CALNAME:Seminar @ Cornell Tech: Aviral Kumar
X-WR-TIMEZONE:Eastern Time (US & Canada)
BEGIN:VEVENT
DTSTAMP:20260519T041606Z
UID:tag:localist.com\,2008:EventInstance_42792114953811
DTSTART:20230420T143000Z
DTEND:20230420T153000Z
DESCRIPTION:Reinforcement Learning from Static Datasets: Algorithms\, Analy
 sis and Applications\n\nTypically\, reinforcement learning (RL) methods re
 ly on trial-and-error interaction with the environment from scratch to dis
 cover effective behaviors. While this sort of paradigm has the potential t
 o discover good strategies\, this paradigm also inhibits RL methods from c
 ollecting enough experience or training data in real-world problems where 
 active interaction is expensive (e.g.\, in drug design) or dangerous (e.g.
 \, for robots operating around humans). My work develops approaches to all
 eviate this limitation: how can we learn policies to effectively make deci
 sions entirely from previously-collected\, static datasets in an offline m
 anner? In this talk\, I will discuss challenges that appear in this kind o
 f offline reinforcement learning (offline RL) and develop algorithms and t
 echniques to address these challenges. I will then discuss how my approach
 es for offline RL and decision-making have enabled us to make progress in 
 real-world problems such as hardware accelerator design\, robotic manipula
 tion\, and computational chemistry. Finally\, I will discuss how we can st
 art to enable offline RL methods to benefit from generalization capabiliti
 es offered by large and expressive models by understanding how these metho
 ds behave.\n\n \n\nSpeaker Bio\n\nAviral Kumar is a final year Ph.D. stude
 nt at UC Berkeley. His research focuses on developing effective and reliab
 le approaches for (sequential) decision-making. Towards this goal\, he foc
 uses on designing reinforcement learning techniques to static datasets and
  on understanding and applying these methods in practice. Before his Ph.D.
 \, Aviral obtained his B.Tech. in Computer Science from IIT Bombay in Indi
 a. He is a recipient of the C.V. & Daulat Ramamoorthy Distinguished Resear
 ch Award\, awarded to 1 PhD student in Berkeley EECS for outstanding contr
 ibutions to a new area of research in computer science\, Facebook Ph.D. Fe
 llowship in Machine Learning and Apple Scholars in AI/ML Ph.D. Fellowship.
LOCATION:Cornell Tech\, Bloomberg Center\, Room 401
SUMMARY:Seminar @ Cornell Tech: Aviral Kumar
URL;VALUE=URI:https://events.cornell.edu/event/seminar_cornell_tech_aviral_
 kumar
CATEGORIES:Seminar
END:VEVENT
END:VCALENDAR