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:3004\, 3rd Floor
DTSTART;TZID=America/Los_Angeles:20240625T154500
DTEND;TZID=America/Los_Angeles:20240625T160000
UID:dac_DAC 2024_sess129_RESEARCH1762@linklings.com
SUMMARY:VAE-HDC: Efficient and Secure Hyper-dimensional Encoder Leveraging
  Variation Analog Entropy
DESCRIPTION:Research Manuscript\n\nBoyang Cheng, Jianbo Liu, Steven Davis,
  Zephan Enciso, Yiyang Zhang, and Ningyuan Cao (University of Notre Dame)\
 n\nHyperdimensional computing (HDC) is a bio-inspired machine learning par
 adigm utilizing hyperdimensional spaces for data representation. HDC signi
 ficantly improves the ability to learn from sparse data and enhances noise
  robustness, and also enables parallel computation. Despite these advantag
 es, HDC's reliance on high dimensionality and operational simplicity can l
 ead to increased hardware costs and potential security vulnerabilities. Th
 is paper introduces a novel HDC encoding strategy using variation-based an
 alog entropy (VAE), aiming to reduce memory footprint, lower power/energy 
 consumption, and enhance security with physically-unclonable entropy gener
 ation. The VAE cell, with high entropy robustness 30.23-57.76 dB SNR and a
  small footprint 10 transistors, allows HDC to achieve a 14.3x reduction i
 n vector dimensions, a 4.4x decrease in unit entropy cell area, and a 2% i
 ncrease in accuracy compared to binary/multi-bit HDC. These benefits lead 
 to a 1.3-4.4x area and a 327x leakage power reduction when compared to an 
 SRAM baseline. We have designed custom low-power circuits that enable end-
 to-end analog entropy storage, distribution management, binding, permutati
 on, and bundling. This analog implementation prevents data conversion duri
 ng feature vector encoding, thereby significantly enhancing energy efficie
 ncy 48.5 nJ per query. Furthermore, with hardware-secured basis vectors, d
 ata security is significantly improved, as evidenced by the markedly degra
 ded visual distinguish-ability of retrieved image data and maximum of 11dB
  lower PSNR.\n\nTopic: AI, Design\n\nKeyword: AI/ML, Digital, and Analog C
 ircuits\n\nSession Chairs: Yu Cao (University of Minnesota) and Xin Zhang 
 (IBM Research)
END:VEVENT
END:VCALENDAR