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:20211207T054807Z
LOCATION:230-231-232
DTSTART;TZID=America/Chicago:20211117T143000
DTEND;TZID=America/Chicago:20211117T150000
UID:submissions.supercomputing.org_SC21_sess174_pap488@linklings.com
SUMMARY:FedAT: A High-Performance and Communication-Efficient Federated Le
 arning System with Asynchronous Tiers
DESCRIPTION:Paper\n\nFedAT: A High-Performance and Communication-Efficient
  Federated Learning System with Asynchronous Tiers\n\nChai, Chen, Anwar, Z
 hao, Cheng...\n\nFederated learning (FL) involves training a model over ma
 ssive distributed devices, while keeping the training data localized and p
 rivate. This form of collaborative learning exposes new tradeoffs among mo
 del convergence speed, model accuracy, balance across clients and communic
 ation cost, with new challenges including the straggler problem and commun
 ication bottleneck. To address these issues, we present FedAT, a novel fed
 erated learning system with asynchronous tiers. FedAT synergistically comb
 ines synchronous, intra-tier training and asynchronous, cross-tier trainin
 g. By bridging the synchronous and asynchronous training through tiering, 
 FedAT minimizes the straggler with improved test accuracy. FedAT uses a we
 ighted aggregation heuristic to balance the training across clients for fu
 rther accuracy improvement.  FedAT compresses uplink and downlink communic
 ations using an efficient compression algorithm, which minimizes the commu
 nication cost. Results show that FedAT improves the prediction performance
  by up to 21.09% and reduces the communication cost by up to 8.5x, compare
 d to state-of-the-art FL methods.\n\nTag: Machine Learning and Artificial 
 Intelligence\n\nRegistration Category: Tech Program Reg Pass
END:VEVENT
END:VCALENDAR