BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//132.216.98.100//NONSGML kigkonsult.se iCalcreator 2.20.4//
BEGIN:VEVENT
UID:20260612T021441EDT-1575o1GTUE@132.216.98.100
DTSTAMP:20260612T061441Z
DESCRIPTION:NCRN Seminar\n\nSpeaker: Marc G. Bellemare\, Google Brain in Mo
 ntreal\n\nZoom: Registration Link\n\nAfter registering\, you will receive 
 a confirmation email containing information about joining the meeting.ABST
 RACT:\n	Efficiently navigating a superpressure balloon in the stratosphere 
 requires the integration of a multitude of cues\, such as wind speed and s
 olar elevation\, and the process is complicated by forecast errors and spa
 rse wind measurements. Coupled with the need to make decisions in real tim
 e\, these factors rule out the use of conventional control techniques. Thi
 s talk describes the use of reinforcement learning to create a high-perfor
 ming flight controller for Loon superpressure balloons. Our algorithm uses
  data augmentation and a self-correcting design to overcome the key techni
 cal challenge of reinforcement learning from imperfect data\, which has pr
 oved to be a major obstacle to its application to physical systems. We dep
 loyed our controller to station Loon balloons at multiple locations across
  the globe\, including a 39-day controlled experiment over the Pacific Oce
 an. Analyses show that the controller outperforms Loon's previous algorith
 m and is robust to the natural diversity in stratospheric winds. These res
 ults demonstrate that reinforcement learning is an effective solution to r
 eal-world autonomous control problems in which neither conventional method
 s nor human intervention suffice\, offering clues about what may be needed
  to create artificially intelligent agents that continuously interact with
  real\, dynamic environments.BIOGRAPHY:\n	Marc G. Bellemare leads the reinf
 orcement learning efforts at Google Brain in Montreal and holds a Canada C
 IFAR AI Chair at Mila. He received his Ph.D. from the University of Albert
 a\, where he developed the highly-successful Arcade Learning Environment b
 enchmark. From 2013 to 2017 he worked at DeepMind in London\, UK\, where h
 e made major contributions to deep reinforcement learning\, in particular 
 pioneering the distributional method. Marc G. Bellemare is also a CIFAR Le
 arning in Machines & Brains Fellow and an adjunct professor at McGill Univ
 ersity.\n
DTSTART:20210318T180000Z
DTEND:20210318T190000Z
LOCATION:CA\, ZOOM
SUMMARY:Autonomous navigation of stratospheric balloons using reinforcement
  learning
URL:https://www.mcgill.ca/channels/channels/event/autonomous-navigation-str
 atospheric-balloons-using-reinforcement-learning-329332
END:VEVENT
END:VCALENDAR