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:20181221T160904Z
LOCATION:C2/3/4 Ballroom
DTSTART;TZID=America/Chicago:20181113T083000
DTEND;TZID=America/Chicago:20181113T170000
UID:submissions.supercomputing.org_SC18_sess322_post153@linklings.com
SUMMARY:AI Matrix – Synthetic Benchmarks for DNN
DESCRIPTION:Poster\nTech Program Reg Pass, Exhibits Reg Pass\n\nAI Matrix 
 – Synthetic Benchmarks for DNN\n\nWei, Xu, Jin, Zhang, Zhang\n\nThe curren
 t AI benchmarks suffer from a number of drawbacks. First, they cannot adap
 t to the emerging changes of deep learning (DL) algorithms and are fixed o
 nce selected. Second, they contain tens to hundreds of applications and ha
 ve very long running time. Third, they are mainly selected from open sourc
 es, which are restricted by copyright and not representable of the proprie
 tary applications. To address these drawbacks, this work firstly proposes 
 a synthetic benchmark framework that generates a small number of benchmark
 s that best represent a broad range of applications using their profiled w
 orkload characteristics. The synthetic benchmarks can adapt to new DL algo
 rithms by re-profiling new applications and updating itself, greatly reduc
 e number of benchmark tests and running time, and strongly represent DL ap
 plications of interests. The framework is validated by using log data prof
 iled from DL models running on Alibaba AI platform, and is representable o
 f real workload characteristics.
URL:https://sc18.supercomputing.org/presentation/?id=post153&sess=sess322
END:VEVENT
END:VCALENDAR