BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Reinforcement learning with a corrupted reward function - Tom McGr
 ath\, Imperial College London
DTSTART:20171129T170000Z
DTEND:20171129T183000Z
UID:TALK96529@talks.cam.ac.uk
CONTACT:Adrià Garriga Alonso
DESCRIPTION:No real-world reward function is perfect. Sensory errors and s
 oftware bugs may result in RL agents observing higher (or lower) rewards t
 han they should. For example\, a reinforcement learning agent may prefer s
 tates where a sensory error gives it the maximum reward\, but where the tr
 ue reward is actually small. Two ways around the problem are investigated.
LOCATION: Cambridge University Engineering Department\, CBL Seminar room B
 E4-38.  For directions see http://learning.eng.cam.ac.uk/Public/Directions
END:VEVENT
END:VCALENDAR