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:20210916T132453Z
LOCATION:Jean Calvin
DTSTART;TZID=Europe/Stockholm:20210708T123000
DTEND;TZID=Europe/Stockholm:20210708T130000
UID:submissions.pasc-conference.org_PASC21_sess114_msa176@linklings.com
SUMMARY:On Sparsity in AI/ML and Earth Science Applications, and its Archi
 tectural Implications
DESCRIPTION:Minisymposium\n\nOn Sparsity in AI/ML and Earth Science Applic
 ations, and its Architectural Implications\n\nSpeiser\n\nThe topic of spar
 sity relates to many aspects of the interplay between Earth Science, AI/ML
 , and computing. Many types of data are intrinsically sparse, and in situa
 tions when data are naturally dense there are many mechanisms which genera
 te sparsity in data collection and aggregation. We discuss some of these d
 ata types and mechanisms, and also draw some connections to fields where s
 parsity is a fundamental consideration, such as Compressed Sensing. In the
  context of Machine Learning, sparsity is also a consideration that is lik
 ely to grow in importance over time, for several reasons, ranging from int
 erpretability of models to performance. We give an overview of sparse aspe
 cts in ML generally and in Deep Learning more specifically, exploring moti
 vations and techniques for sparse models. Finally, we consider the "migrat
 ing threads" computer architecture, which seems to exhibit good strong sca
 ling for typical sparse workloads, such as sparse matrix-vector multiplica
 tion. Likewise, recent results indicate good performance on sparse variant
 s of Machine Learning as well as sparse graph processing.\n\nDomain: CS an
 d Math, Emerging Applications, Climate and Weather, Solid Earth Dynamics
END:VEVENT
END:VCALENDAR