BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Latent Action Space for Offline Reinforcement Learning - Wenxuan Z
 hou\, Carnegie Mellon University (USA)
DTSTART:20210126T131500Z
DTEND:20210126T141500Z
UID:TALK156091@talks.cam.ac.uk
CONTACT:Mateja Jamnik
DESCRIPTION:"Join us on Zoom":https://zoom.us/j/99166955895?pwd=SzI0M3pMVE
 kvNmw3Q0dqNDVRalZvdz09\n\nThe goal of offline reinforcement learning is to
  learn a policy from a fixed dataset\, without further interactions with t
 he environment. This setting will be an increasingly more important paradi
 gm for real-world applications of reinforcement learning such as robotics\
 , in which data collection is slow and potentially dangerous. In this talk
 \, we will discuss the challenges of applying existing off-policy algorith
 ms on static datasets and the reasonings behind the objectives of offline 
 RL. We will then introduce our approach Policy in the Latent Action Space 
 (PLAS) which naturally satisfies the objectives. Our method is evaluated o
 n continuous control benchmarks in simulation and the cloth-sliding task w
 ith a physical robot. We demonstrate that our method provides competitive 
 performance consistently across various continuous control tasks and diffe
 rent types of datasets\, outperforming previous offline reinforcement lear
 ning methods with explicit constraints.\n\nBio:\nWenxuan Zhou is a Ph.D. s
 tudent at the Robotics Institute at Carnegie Mellon University\, advised b
 y Prof. David Held. Her research interests lie at the intersection of robo
 tics and reinforcement learning. Previously\, she received her Master's de
 gree in Robotics at CMU advised by Prof. Abhinav Gupta. Prior to that\, sh
 e obtained her dual B.S. degrees in Electrical and Computer Engineering fr
 om Shanghai Jiao Tong University and Mechanical Engineering from the Unive
 rsity of Michigan. She will be joining DeepMind as an intern in Summer 202
 1.
LOCATION:Zoom
END:VEVENT
END:VCALENDAR