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:20240626T180033Z
LOCATION:3001\, 3rd Floor
DTSTART;TZID=America/Los_Angeles:20240625T104500
DTEND;TZID=America/Los_Angeles:20240625T110000
UID:dac_DAC 2024_sess103_RESEARCH865@linklings.com
SUMMARY:LEAF: An Adaptation Framework against Noisy Data on Edge through U
 ltra Low-Cost Training
DESCRIPTION:Research Manuscript\n\nZihan Xia (University of California, Sa
 n Diego); Jinwook Kim (SK hynix); and Mingu Kang (University of California
 , San Diego)\n\nIn real-world neural network deployments, incoming data of
 ten contains noise and imperfections. Retraining on resource-constrained e
 dge devices becomes essential to maintain performance. To tackle this chal
 lenge, we introduce LEAF, a hardware-efficient framework designed for adap
 ting to degraded images. By analyzing neural network behavior on degraded 
 images, we propose two techniques: 1) Selective Experience Replay for skip
 ping unimportant images, reducing computation, and 2) Pseudo Noise Ditheri
 ng for extremely low precision (3 or 4-bit) gradient quantization, enablin
 g nearly full-integer training. Extensive experiments on CIFAR10 and Tiny 
 ImageNet datasets, with various image degradations, demonstrate LEAF's ult
 ra-low cost with minimal accuracy loss.\n\nTopic: AI\n\nKeyword: AI/ML Alg
 orithms\n\nSession Chairs: Hongyang Jia (Tsinghua University) and Grace Li
  Zhang (Technische Universität Darmstadt)
END:VEVENT
END:VCALENDAR