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:20250625T183020Z
LOCATION:3006\, Level 3
DTSTART;TZID=America/Los_Angeles:20250624T103000
DTEND;TZID=America/Los_Angeles:20250624T104500
UID:dac_DAC 2025_sess148_RESEARCH2469@linklings.com
SUMMARY:Resilient Federated Learning on Embedded Devices with Constrained 
 Network Connectivity
DESCRIPTION:Zihan Li, Han Liu, Ao Li, Ching-hsiang Chan, Yevgeniy Vorobeyc
 hik, and William Yeoh (Washington University, St. Louis); Wenjing Lou (Vir
 ginia Polytechnic Institute and State University); and Ning Zhang (Washing
 ton University, St. Louis)\n\nFederated learning enables decentralized mod
 el training while preserving data privacy. However, since the learning pro
 cess overlays the physical network infrastructure, the efficiency of learn
 ing can be impacted by network connectivity. In this work, we conducted ex
 tensive experiments to empirically characterize the impacts and leverage t
 he insights to propose an adaptive federation protocol, where clients with
  limited bandwidth are only prompted to transmit adaptively compressed gra
 dient updates when the gradient similarity score is similar between the lo
 cal and global models. Our evaluation in simulated environments and on rea
 l hardware devices shows bandwidth savings of 60% to 78% compared to state
 -of-the-art methods.\n\nTopics: Security\n\nTracks: SEC1: AI/ML Security/P
 rivacy\n\nSession Chairs: Adnan Siraj Rakin (Binghamton University) and Ay
 esha Siddique (University of Maine)\n\n
END:VEVENT
END:VCALENDAR