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:3003\, 3rd Floor
DTSTART;TZID=America/Los_Angeles:20240627T110000
DTEND;TZID=America/Los_Angeles:20240627T111500
UID:dac_DAC 2024_sess121_RESEARCH331@linklings.com
SUMMARY:MERSIT: A Hardware-Efficient 8-bit Data Format with Enhanced Post-
 Training Quantization DNN Accuracy
DESCRIPTION:Research Manuscript\n\nNguyen-Dong Ho, Gyujun Jeong, Cheol-Min
  Kang, Seungkyu Choi, and Ik Joon Chang (Kyung Hee University)\n\nPost-tra
 ining quantization (PTQ) models utilizing conventional 8-bit Integer or fl
 oating-point formats still exhibit significant accuracy drops in modern de
 ep neural networks (DNNs), rendering them unreliable. This paper presents 
 MERSIT, a novel 8-bit PTQ data format designed for various DNNs. While lev
 eraging the dynamic configuration of exponent and fraction bits derived fr
 om Posit data format, MERSIT demonstrates enhanced hardware efficiency thr
 ough the proposed merged decoding scheme. Our evaluation indicates that ME
 RSIT yields more reliable 8-bit PTQ models, exhibiting superior accuracy a
 cross various DNNs compared to conventional floating-point formats.\n\nTop
 ic: AI, Design\n\nKeyword: AI/ML Architecture Design\n\nSession Chairs: Se
 rcan Aygun (University of Louisiana) and Charbel Sakr (NVIDIA)
END:VEVENT
END:VCALENDAR