BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Date iCal//NONSGML kigkonsult.se iCalcreator 2.20.2//
METHOD:PUBLISH
X-WR-CALNAME;VALUE=TEXT:Eventi DIAG
BEGIN:VTIMEZONE
TZID:Europe/Paris
BEGIN:STANDARD
DTSTART:20211031T030000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
RDATE:20221030T030000
TZNAME:CET
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20220327T020000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:calendar.24878.field_data.0@www.u-gov-ricerca.uniroma1.it
DTSTAMP:20260405T115910Z
CREATED:20220428T132054Z
DESCRIPTION:Marco Canini associate professor at KAUST will visit the diag t
 he 29 of April. During his visit he will give the following seminar lectur
 es in the Aula Magna of DIAGAbstract:Scaling deep learning to a large clus
 ter of workers is challenging due to high communication overheads that dat
 a-parallelism entails. This talk describes our efforts to rein in distribu
 ted deep learning's communication bottlenecks. We describe SwitchML\, the 
 state-of-the-art in-network aggregation system for collective communicatio
 n using programmable network switches. We introduce OmniReduce\, an effici
 ent streaming aggregation system that exploits sparsity to maximize effect
 ive bandwidth use. We touch on our work to develop compressed gradient com
 munication algorithms that perform efficiently and adapt to network condit
 ions. Lastly\, we take a broad look at the challenges to accelerated decen
 tralized training in the federated learning setting where heterogeneity is
  an intrinsic property of the environment.Bio:Marco does not know what the
  next big thing will be. But he's sure that our next-gen computing and net
 working infrastructure must be a viable platform for it. Marco's research 
 spans a number of areas in computer systems\, including distributed system
 s\, large-scale/cloud computing and computer networking with emphasis on p
 rogrammable networks. His current focus is on designing better systems sup
 port for AI/ML and providing practical implementations deployable in the r
 eal-world.Marco is an associate professor in Computer Science at KAUST. Ma
 rco obtained his Ph.D. in computer science and engineering from the Univer
 sity of Genoa in 2009 after spending the last year as a visiting student a
 t the University of Cambridge. He was a postdoctoral researcher at EPFL an
 d a senior research scientist at Deutsche Telekom Innovation Labs & TU Ber
 lin. Before joining KAUST\, he was an assistant professor at UCLouvain. He
  also held positions at Intel\, Microsoft and Google.
DTSTART;TZID=Europe/Paris:20220429T090000
DTEND;TZID=Europe/Paris:20220429T090000
LAST-MODIFIED:20220428T153913Z
LOCATION:Aula Magna
SUMMARY:Accelerated Deep Learning via Efficient\, Compressed and Managed Co
 mmunication  - Marco Canini
URL;TYPE=URI:http://www.u-gov-ricerca.uniroma1.it/node/24878
END:VEVENT
END:VCALENDAR