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:3001\, 3rd Floor
DTSTART;TZID=America/Los_Angeles:20240626T110000
DTEND;TZID=America/Los_Angeles:20240626T111500
UID:dac_DAC 2024_sess102_RESEARCH005@linklings.com
SUMMARY:SMORE: Similarity-Based Hyperdimensional Domain Adaptation for Mul
 ti-Sensor Time Series Classification
DESCRIPTION:Research Manuscript\n\nJunyao Wang and Mohammad Al Faruque (Un
 iversity of California, Irvine)\n\nMany real-world applications of the Int
 ernet of Things (IoT) employ machine learning (ML) algorithms to analyze t
 ime series information collected by interconnected sensors. However, distr
 ibution shift, a fundamental challenge in data-driven ML, arises when a mo
 del is deployed on a data distribution different from the training data an
 d can substantially degrade model performance. Additionally, increasingly 
 sophisticated deep neural networks (DNNs) are proposed to capture intricat
 e spatial and temporal dependencies in multi-sensor time series data, ofte
 n exceeding the capabilities of today's edge devices. In this paper, we pr
 opose SMORE, a novel resource-efficient domain adaptation (DA) algorithm f
 or multi-sensor time series classification, leveraging the ultra-efficient
  operations of hyperdimensional computing. SMORE dynamically customizes te
 st-time models with explicit consideration of the domain context of each s
 ample to provide accurate predictions when confronted with domain shifts. 
 Our evaluation on a variety of multi-sensor time series classification tas
 ks shows that SMORE achieves on average 1.98% higher accuracy than state-o
 f-the-art (SOTA) DNN-based DA algorithms with 18.81× faster training and 4
 .85× faster inference.\n\nTopic: AI\n\nKeyword: AI/ML Algorithms\n\nSessio
 n Chairs: Parivesh Choudhary (Synopsys) and Anca Molnos (CEA-List)
END:VEVENT
END:VCALENDAR