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:20260402T024533Z
LOCATION:3000\, Level 3
DTSTART;TZID=America/Los_Angeles:20250624T143000
DTEND;TZID=America/Los_Angeles:20250624T144500
UID:dac_DAC 2025_sess101_RESEARCH2348@linklings.com
SUMMARY:FF-INT8: Efficient Forward-Forward DNN Training on Edge Devices wi
 th INT8 Precision
DESCRIPTION:Jingxiao Ma (Brown University), Priyadarshini Panda (Yale Univ
 ersity), and Sherief Reda (Brown University)\n\nBackpropagation, while fou
 ndational to neural network training, is inefficient for resource-constrai
 ned edge devices due to its high time and energy consumption. While low-pr
 ecision quantization has been explored for inference speed-up, its use in 
 training remains underexplored. The Forward-Forward (FF) algorithm offers 
 an alternative by replacing the backward pass with an additional forward p
 ass, reducing memory and computation. This paper introduces an INT8 quanti
 zed training approach using FF's layer-by-layer strategy to stabilize grad
 ient quantization and proposes a "look-afterward" scheme to improve accura
 cy. Experiments a edge device show 4.6% faster training, 8.3% energy savin
 gs, and 27.0% reduced memory usage, with competitive accuracy.\n\nTopics: 
 AI\n\nTracks: AI1: AI/ML Algorithms\n\nSession Chairs: Fan Yang (Fudan Uni
 versity) and Li Shang (Fudan University)\n\n
END:VEVENT
END:VCALENDAR