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:20260402T024532Z
LOCATION:3001\, Level 3
DTSTART;TZID=America/Los_Angeles:20250623T164500
DTEND;TZID=America/Los_Angeles:20250623T170000
UID:dac_DAC 2025_sess124_RESEARCH1340@linklings.com
SUMMARY:DANN: Diffractive Acoustic Neural Network for in-sensor computing 
 system target at multi-biomarker diagnosis
DESCRIPTION:Lewei He, Ning Lin, Binbin Cui, Xinran Zhang, and Shiming Zhan
 g (University of Hong Kong) and Zhongrui Wang (Southern University of Scie
 nce and Technology)\n\nAnalog machine learning hardware platforms, such as
  those using wave physics, present potential for edge artificial intellige
 nce (AI) applications due to in-sensor computing architecture, offering su
 perior energy efficiency compared to digital circuits. While the diffracti
 ve neural network has been implemented in optical systems, its deployment 
 on integrated acoustic systems has not been achieved due to the challenges
  associated with hardware optimization. In this paper, we propose the Diff
 ractive Acoustic Neural Network (DANN), a novel approach that applies diff
 ractive neural network algorithms to surface acoustic wave (SAW) systems f
 or in-sensor multi-biomarker diagnosis. To address the optimization challe
 nges, we introduce a novel training methodology that combines Finite Eleme
 nt Analysis (FEA) with gradient descent. We validate our method on Major D
 epressive Disorder (MDD) and prostate cancer, achieving accuracies of 74.0
 7% and 86.0%, respectively, which nearly reaches the accuracy levels of cl
 inical diagnoses. By comparing the co-training method with the traditional
  gradient descent training method and direct training on the FEA model, th
 e co-training method demonstrates its advantages in balancing training eff
 iciency and accuracy. Furthermore, a comparison of power consumption betwe
 en the traditional method and the in-sensor computing system is conducted,
  indicating 66% energy savings attributed to its high level of integration
 .\n\nTopics: Design\n\nTracks: DES5: Emerging Device and Interconnect Tech
 nologies\n\nSession Chairs: Doo Seok Jeong (Hanyang University) and Xunzha
 o Yin (Zhejiang University)\n\n
END:VEVENT
END:VCALENDAR