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:Level 2 Lobby
DTSTART;TZID=America/Los_Angeles:20240625T180000
DTEND;TZID=America/Los_Angeles:20240625T190000
UID:dac_DAC 2024_sess236_RESEARCH814@linklings.com
SUMMARY:ODILO: On-Device Incremental Learning Via Lightweight Operations
DESCRIPTION:Work-in-Progress Poster\n\nQing Wang (Yunnan University), Di L
 iu (Norwegian University of Science and Technology), and Shengfa Miao and 
 Mingxiong Zhao (Yunnan University)\n\nIn this paper, we propose ODILO, a n
 ew on-device incremental\nlearning framework for edge systems. The key par
 t of ODILO is a new module, namely Efficient Incremental Module (EIM). EIM
  is composed of normal convolutions and lightweight operations. During inc
 remental learning, EIM exploits some lightweight operations, called adapte
 rs, to effectively and efficiently learn features for new classes such tha
 t it can improve the accuracy of incremental learning while reducing model
  complexity as well as training overhead. The efficiency of ODILO is furth
 er bolstered by adapter fusion, prototypes, and efficient data augmentatio
 n. We conduct extensive experiments on the CIFAR-100 and Tiny-ImageNet dat
 asets. Experimental results show that ODILO improves the accuracy by up to
  4.21% over existing methods while reducing around 50% of model complexity
 . In addition, evaluations on real edge systems demonstrate its applicabil
 ity for on-device machine learning. The code will be available upon accept
 ance.\n\nTopic: AI, Autonomous Systems, Cloud, Design, EDA, Embedded Syste
 ms, IP, Security
END:VEVENT
END:VCALENDAR