BEGIN:VCALENDAR
VERSION:2.0
PRODID:icalendar-ruby
CALSCALE:GREGORIAN
X-WR-CALNAME:Representation-based Reinforcement Learning: Bridging World Mo
 del and Policy via Spectral Representations
X-WR-TIMEZONE:Eastern Time (US & Canada)
BEGIN:VEVENT
DTSTAMP:20260817T050841Z
UID:tag:localist.com\,2008:EventInstance_51152181520207
DTSTART:20251107T160000Z
DTEND:20251107T170000Z
DESCRIPTION:Electrical Engineering Seminar\n\nFriday\, November 7 at 11am\n
 \nSEC LL2.224\n\n \n\n"Representation-based Reinforcement Learning: Bridgi
 ng World Model and Policy via Spectral Representations"\n\nSpeaker: Bo Dai
 \, Assistant Professor\, Georgia Tech\, Staff Research Scientist\, Google 
 DeepMind\n\nReinforcement learning often faces a trade-off between world m
 odel flexibility and computational tractability. Linear dynamics induces t
 ractable planning and exploration but with huge approximation gap\, while 
 flexible models can capture complex dynamics in world model and policy but
  often introduce nonlinearity\, making planning and exploration challengin
 g. In this talk\, we explore how representation learning can help overcome
  this dilemma and bridge world model to policy. We present algorithms that
  extract flexible representations\, which enabling practical and provable 
 planning and exploration from flexible world model. We provide theoretical
  guarantees our algorithm for RL in MDP and POMDP settings\, and empirical
  results demonstrating the superiority of our approach on various benchmar
 ks.\n\n \n\n\nBio: Bo Dai is an assistant professor in Georgia Tech and a 
 staff research scientist in Google DeepMind. He obtained his Ph.D. from Ge
 orgia Tech. His research interest lies in developing principled and practi
 cal algorithms for reinforcement learning and generative models. He regula
 rly serves as area chair or senior program committee member at major AI/ML
  conferences such as ICML\, NeurIPS\, AISTATS\, and ICLR.
GEO:42.363197;-71.127278
LOCATION:Science and Engineering Complex (SEC)\, LL2.224
SUMMARY:Representation-based Reinforcement Learning: Bridging World Model a
 nd Policy via Spectral Representations
URL;VALUE=URI:https://events.seas.harvard.edu/event/representation-based-re
 inforcement-learning-bridging-world-model-and-policy-via-spectral-represen
 tations
CATEGORIES:Colloquia / Seminar / Lecture
END:VEVENT
END:VCALENDAR