BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Europe/Stockholm
X-LIC-LOCATION:Europe/Stockholm
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20190719T085743Z
LOCATION:HG E 3
DTSTART;TZID=Europe/Stockholm:20190612T130000
DTEND;TZID=Europe/Stockholm:20190612T133000
UID:submissions.pasc-conference.org_PASC19_sess126_msa112@linklings.com
SUMMARY:Demystifying Parallel and Distributed Deep Learning
DESCRIPTION:Minisymposium\nComputer Science and Applied Mathematics, Emerg
 ing Application Domains, Physics, Life Sciences\n\nDemystifying Parallel a
 nd Distributed Deep Learning\n\nBen-Nun, Hoefler\n\nDeep Neural Networks (
 DNNs) are becoming an important tool in modern computing applications. Acc
 elerating their training is a major challenge and techniques range from di
 stributed algorithms to low-level circuit design. The talk outlines deep l
 earning from a theoretical perspective, followed by approaches for its par
 allelization. We present trends in DNN architectures and the resulting imp
 lications on parallelization strategies. We then review and model the diff
 erent types of concurrency in DNNs: from the single operator, through para
 llelism in network inference and training, to distributed deep learning. W
 e discuss asynchronous stochastic optimization, distributed system archite
 ctures, communication schemes, and neural architecture search. Based on th
 ose approaches, we extrapolate potential directions for parallelism in dee
 p learning.
END:VEVENT
END:VCALENDAR