BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Los_Angeles
X-LIC-LOCATION:America/Los_Angeles
BEGIN:DAYLIGHT
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
TZNAME:PDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20240626T180034Z
LOCATION:3008\, 3rd Floor
DTSTART;TZID=America/Los_Angeles:20240626T110000
DTEND;TZID=America/Los_Angeles:20240626T111500
UID:dac_DAC 2024_sess126_RESEARCH501@linklings.com
SUMMARY:AdaptiveFL: Adaptive Heterogeneous Federated Learning for Resource
 -Constrained AIoT Systems
DESCRIPTION:Research Manuscript\n\nChentao Jia (East China Normal Universi
 ty), Ming Hu (Nanyang Technological University), Zekai Chen and Yanxin Yan
 g (East China Normal University), Xiaofei Xie (Singapore Management Univer
 isity), Yang Liu (Nanyang Technological University), and Mingsong Chen (Ea
 st China Normal University)\n\nAlthough Federated Learning (FL) is promisi
 ng to enable collaborative learning among Artificial Intelligence of Thing
 s (AIoT) devices, it suffers from the problem of low classification perfor
 mance due\nto various heterogeneity factors (e.g., computing capacity, mem
 ory size) of devices and uncertain operating environments. To address thes
 e issues, this paper introduces an effective FL approach named AdaptiveFL 
 based on a novel fine-grained width-wise model pruning strategy, which can
  generate various heterogeneous local models for heterogeneous AIoT device
 s. By using our proposed reinforcement learning-based device selection mec
 hanism, AdaptiveFL can adaptively dispatch suitable heterogeneous models t
 o corresponding AIoT devices on the fly based on their available resources
  for \nlocal training. Experimental results show that, compared to state-o
 f-the-art methods, AdaptiveFL can achieve up to 16.83% inference improveme
 nts for both IID and non-IID scenarios.\n\nTopic: AI, Design\n\nKeyword: A
 I/ML System and Platform Design\n\nSession Chair: Hsien-Hsin Sean Lee (Int
 el Corporation)
END:VEVENT
END:VCALENDAR