BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/New_York
X-LIC-LOCATION:America/New_York
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20210402T160040Z
LOCATION:Track 1
DTSTART;TZID=America/New_York:20201118T104500
DTEND;TZID=America/New_York:20201118T113000
UID:submissions.supercomputing.org_SC20_sess299_inv111@linklings.com
SUMMARY:Can We Build a Virtuous Cycle Between Machine Learning and HPC?
DESCRIPTION:Invited Talk\n\nCan We Build a Virtuous Cycle Between Machine 
 Learning and HPC?\n\nYoung\n\nMachine learning (ML) draws on many HPC comp
 onents; AlexNet was enabled by GPUs, and GPU-based MLPerf submissions conn
 ect data replicas using Infiniband. In other ways, it feels like HPC and M
 L are diverging. Just a handful of the current ML startups are consid
 ering HPC applications. Google's TPUs use a proprietary interconnect and a
  limited set of collective operations. Many HPC codes are written for
  double precision; ML machines reach for peta-ops of low-precision fl
 oating-point. Programming stacks are very different, with rare commonality
  (Horovod integrates MPI into TensorFlow). Can we construct a better, virt
 uous cycle between ML and HPC? In this talk I'll examine the opportunities
  for parallelization, mixed precision and new algorithms (particularly spa
 rsity) in uniting the two fields and solving common problems.\n\nTag:
  Applications, Big Data, Extreme Scale Computing, Machine Learning, Deep L
 earning and Artificial Intelligence\n\nRegistration Category: Tech Program
  Reg Pass, Exhibits Reg Pass
END:VEVENT
END:VCALENDAR