BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//https://techplay.jp//JP
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-WR-CALDESC:[16th AIP Open Seminar] Talks by Approximate Bayesian Inferenc
 e Team
X-WR-CALNAME:[16th AIP Open Seminar] Talks by Approximate Bayesian Inferenc
 e Team
X-WR-TIMEZONE:Asia/Tokyo
BEGIN:VTIMEZONE
TZID:Asia/Tokyo
BEGIN:STANDARD
DTSTART:19700101T000000
TZOFFSETFROM:+0900
TZOFFSETTO:+0900
TZNAME:JST
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
UID:808207@techplay.jp
SUMMARY:[16th AIP Open Seminar] Talks by Approximate Bayesian Inference Tea
 m
DTSTART;TZID=Asia/Tokyo:20210310T150000
DTEND;TZID=Asia/Tokyo:20210310T170000
DTSTAMP:20260518T082216Z
CREATED:20210212T060018Z
DESCRIPTION:イベント詳細はこちら\nhttps://techplay.jp/event/80820
 7?utm_medium=referral&utm_source=ics&utm_campaign=ics\n\nApproximate Baye
 sian Inference Team at RIKEN AIP (https://aip.riken.jp/labs/generic_tech/
 approx_bayes_infer/) \n\nSpeaker 1: Emtiyaz Khan (45 mins)\nTitle: Bayesi
 an principles for Learning-Machines\nAbstract: \nHumans and animals have 
 a natural ability to autonomously learn and quickly adapt to their surrou
 ndings. How can we design machines that do the same? In this talk\, I wil
 l present Bayesian principles to bridge such gaps between humans and mach
 ines. I will show that a wide-variety of machine-learning algorithms are 
 instances of a single learning-rule derived from Bayesian principles. The
  rule unravels a dual-perspective yielding new mechanisms for knowledge t
 ransfer in learning machines. In the end\, I will summarize the research 
 done by the group in the last 4 years. Overall\, my hope is to convince t
 he audience that Bayesian principles are indispensable for an AI that lea
 rns as efficiently as we do.\n\nSpeaker 2: Dharmesh Tailor (25 mins) \nTi
 tle:  Memorable Experiences of Learning-Machines\nAbstract: \nHumans and 
 other animals have a natural ability to identify useful past experiences.
  How can machines do the same? We present "memorable experiences" to iden
 tify a machine's relevant past experiences and understand its current kno
 wledge. The approach is based on a new notion of duality which is an exte
 nsion of similar ideas used in kernel methods. We demonstrate the applica
 tion of memorable examples as a tool to understand knowledge learned by s
 everal types of machine-learning models.\n\nSpeaker 3: Pierre Alquier  (3
 5 mins) \nTitle:  Meta-Strategy for Hyperparameter Tuning with Guarantees
 \nAbstract: \nOnline gradient methods\, like the online gradient algorith
 m (OGA)\, often depend on tuning parameters that are difficult to set in 
 practice. We consider an online meta-learning scenario\, and we propose a
  meta-strategy to learn these parameters from past tasks. Our strategy is
  based on the minimization of a regret bound. It allows to learn the init
 ialization and the step size in OGA with guarantees. We provide a regret 
 analysis of the strategy in the case of convex losses. It suggests that\,
  when there are parameters θ1\,…\,θT solving well tasks 1\,…\,T res
 pectively and that are close enough one to each other\, our strategy inde
 ed improves on learning each task in isolation. In the context of Approxi
 mate Bayesian Inference\, our method can be interpreted as learning the m
 ean and variance of a Gaussian prior. This opens new perspectives on more
  general methods to learn priors. \n\n\n\nAll participants are required t
 o agree with the AIP Open Seminar Series Code of Conduct.\nPlease see the
  URL below.\nhttps://aip.riken.jp/event-list/termsofparticipation/?lang=e
 n\n\nRIKEN AIP will expect adherence to this code throughout the event. W
 e expect cooperation from all participants to help ensure a safe environm
 ent for everybody.\n\n
LOCATION:オンライン
URL:https://techplay.jp/event/808207?utm_medium=referral&utm_source=ics&utm
 _campaign=ics
END:VEVENT
END:VCALENDAR