BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Denver
X-LIC-LOCATION:America/Denver
BEGIN:DAYLIGHT
TZOFFSETFROM:-0700
TZOFFSETTO:-0600
TZNAME:MDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0600
TZOFFSETTO:-0700
TZNAME:MST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20200129T163559Z
LOCATION:502-503-504
DTSTART;TZID=America/Denver:20191118T113000
DTEND;TZID=America/Denver:20191118T120000
UID:submissions.supercomputing.org_SC19_sess115_ws_mlhpce101@linklings.com
SUMMARY:Metaoptimization on a Distributed System for Deep Reinforcement Le
 arning
DESCRIPTION:Workshop\n\nMetaoptimization on a Distributed System for Deep 
 Reinforcement Learning\n\nHeinrich, Frosio\n\nTraining intelligent agents 
 through reinforcement learning (RL) is a notoriously unstable procedure.  
 Massive parallelization on GPUs and distributed systems has been exploited
  to generate a large amount of training experiences and consequently reduc
 e instabilities, but the success of training remains strongly influenced b
 y the choice of the hyperparameters.  To overcome this issue, we introduce
  HyperTrick, a new metaoptimization algorithm, and show its effective appl
 ication to tune hyperparameters in the case of deep RL, while learning to 
 play different Atari games on a distributed system.  Our analysis provides
  evidence of the interaction between the identification of the optimal hyp
 erparameters and the learned policy, that is peculiar of the case of metao
 ptimization for deep RL.  When compared with state-of-the-art metaoptimiza
 tion algorithms, HyperTrick is characterized by a simpler implementation a
 nd it allows learning similar policies, while making a more effective use 
 of the computational resources in a distributed system.\n\nTag: Workshop R
 eg Pass, Machine Learning\n\nRegistration Category: Workshop Reg Pass, Mac
 hine Learning
URL:https://sc19.supercomputing.org/presentation/?id=ws_mlhpce101&sess=ses
 s115
END:VEVENT
END:VCALENDAR

