BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Sabre//Sabre VObject 4.6.1//EN
CALSCALE:GREGORIAN
BEGIN:VTIMEZONE
TZID:Europe/Zurich
X-LIC-LOCATION:Europe/Zurich
TZURL:http://tzurl.org/zoneinfo/Europe/Zurich
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19810329T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19961027T030000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
UID:news221@dmi.unibas.ch
DTSTAMP;TZID=Europe/Zurich:20180716T175831
DTSTART;TZID=Europe/Zurich:20180518T110000
SUMMARY:Seminar in Numerical Analysis: Holger Fröning (Universität Heidel
 berg)
DESCRIPTION:We are observing a continuous increase in concurrency and heter
 ogeneity for computing systems of any scale\, ranging from small mobile de
 vices to huge datacenters\, and driven by a steady demand for more computi
 ng power. One of the prime examples for an application with virtually unli
 mited computational requirements is machine learning\, in particular deep 
 neural networks (DNN). At the level of data-centers\, DNN training has alr
 eady led to a ubiquitous use of graphics processing units (GPUs)\, forming
  a prime example for specialization for computational improvement. Still\,
  this application is strongly hindered by insufficient compute power and b
 y scalability limitations. Contrary\, mobile architectures for DNN inferen
 ce are still nascent\, and a large amount of proposals have been published
  in the recent years. Both applications\, training and inference\, can fur
 thermore benefit a lot from algorithmic optimizations to reduce the comput
 ational requirements. This talk presents a short introduction of the appli
 cation\, a summary of our observations\, and our own research on reduced p
 recision by extreme forms of quantizations. Finally\, this talk will offer
  some opinions on anticipated research problems.
X-ALT-DESC:We are observing a continuous increase in concurrency and hetero
 geneity for computing systems of any scale\, ranging from small mobile dev
 ices to huge datacenters\, and driven by a steady demand for more computin
 g power. One of the prime examples for an application with virtually unlim
 ited computational requirements is machine learning\, in particular deep n
 eural networks (DNN). At the level of data-centers\, DNN training has alre
 ady led to a ubiquitous use of graphics processing units (GPUs)\, forming 
 a prime example for specialization for computational improvement. Still\, 
 this application is strongly hindered by insufficient compute power and by
  scalability limitations. Contrary\, mobile architectures for DNN inferenc
 e are still nascent\, and a large amount of proposals have been published 
 in the recent years. Both applications\, training and inference\, can furt
 hermore benefit a lot from algorithmic optimizations to reduce the computa
 tional requirements. This talk presents a short introduction of the applic
 ation\, a summary of our observations\, and our own research on reduced pr
 ecision by extreme forms of quantizations. Finally\, this talk will offer 
 some opinions on anticipated research problems. 
DTEND;TZID=Europe/Zurich:20180518T120000
END:VEVENT
END:VCALENDAR