BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Chicago
X-LIC-LOCATION:America/Chicago
BEGIN:DAYLIGHT
TZOFFSETFROM:-0600
TZOFFSETTO:-0500
TZNAME:CDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0500
TZOFFSETTO:-0600
TZNAME:CST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20211207T055404Z
LOCATION:Online
DTSTART;TZID=America/Chicago:20211114T143000
DTEND;TZID=America/Chicago:20211114T150000
UID:submissions.supercomputing.org_SC21_sess430_ws_exampi101@linklings.com
SUMMARY:A FACT-Based Approach: Making ML Collective Autotuning Feasible on
  Exascale Systems
DESCRIPTION:Workshop\n\nA FACT-Based Approach: Making ML Collective Autotu
 ning Feasible on Exascale Systems\n\nWilkins, Guo, Thakur, Hardavellas, Di
 nda...\n\nMachine learning (ML) autotuners use supervised learning to sele
 ct MPI collective algorithms, significantly improving collective performan
 ce. However, a user may find it difficult to understand the benefit of aut
 otuners because we lack a methodology to quantify their performance. Addit
 ionally, to obtain the advertised performance, ML model training requires 
 benchmark data from a vast majority of the feature space.  Collecting such
  data regularly on large scale systems consumes far too much time and reso
 urces.  To address these challenges, we contribute (1) a performance evalu
 ation framework to compare and improve collective autotuner designs and (2
 ) the Feature scaling, Active learning, Converge, Tune hyperparameters (FA
 CT) approach, a three-part methodology to minimize the training data colle
 ction time (and thus maximize practicality at larger scale) without sacrif
 icing accuracy. On a production scale system, our methodology produces a m
 odel of equal accuracy using 6.88x less training data collection time.\n\n
 Tag: Online Only, Extreme Scale Computing, Parallel Programming Languages 
 and Models, Performance\n\nRegistration Category: Workshop Reg Pass
END:VEVENT
END:VCALENDAR