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:20190719T085745Z
LOCATION:HG E 3
DTSTART;TZID=Europe/Stockholm:20190614T143000
DTEND;TZID=Europe/Stockholm:20190614T150000
UID:submissions.pasc-conference.org_PASC19_sess164_msa134@linklings.com
SUMMARY:FPGA-Accelerated Machine Learning Inference as a Solution for Part
 icle Physics Computing Challenges
DESCRIPTION:Minisymposium\nComputer Science and Applied Mathematics, Physi
 cs\n\nFPGA-Accelerated Machine Learning Inference as a Solution for Partic
 le Physics Computing Challenges\n\nLiu\n\nResources required for high-thro
 ughput computing in large scale particle physics experiments face challeng
 ing demands both now and in the future. The growing exploration of machine
  learning algorithms in particle physics offers new solutions to
  simulation, reconstruction,  and analysis. These new machine learnin
 g solutions often lead to increased parallelization and faster reconstruct
 ions times on dedicated hardware, here specifically Field Programmable Gat
 e Arrays.  We explore the possibility that applications of machi
 ne learning simultaneously also solve the increasing computing challe
 nges.  Employing machine learning acceleration as a web service, we d
 emonstrate a heterogeneous compute solution for particle physics
  experiments that requires minimal modification to the current computing m
 odel. First results with Project Brainwave by Microsoft Azure, using the&n
 bsp;Resnet-50 image classification model as an example, demonstrate i
 nference times of approximately 50 (10) milliseconds with our experim
 ental physics software framework using Brainwave as a cloud (edge) se
 rvice. We also adapt the image classifier, for example, physics applicatio
 ns using transfer learning: jet identification in the CMS experiment 
 and event classification in the Nova neutrino experiment at Fermilab.
  Solutions explored here are potentially applicable sooner than may have&n
 bsp;been initially realized.
END:VEVENT
END:VCALENDAR