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:Level 2 Lobby
DTSTART;TZID=America/Los_Angeles:20240626T180000
DTEND;TZID=America/Los_Angeles:20240626T190000
UID:dac_DAC 2024_sess237_RESEARCH1768@linklings.com
SUMMARY:Distribution-Guided Fairness Calibration in Learning
DESCRIPTION:Work-in-Progress Poster\n\nYi Sheng, Junhuan Yang, Zhepeng Wan
 g, and Weiwen Jiang (George Mason University); Qian Lou (University of Cen
 tral Florida); and Lei Yang (George Mason University)\n\nIn recent years, 
 the emphasis on model fairness in AI applications for edge devices has gro
 wn. However, the traditional focus on AI model optimization has revolved a
 round accuracy and efficiency,\nformulating it as a two-objective problem.
  Consequently, the fairness of the model often goes unaddressed, potential
 ly leading to unjust treatment of minorities. To rectify this oversight, i
 t is imminent\nand essential to include fairness in the model optimization
 , making it a three-objective optimization problem in terms of accuracy, e
 fficiency, and fairness. By examining the existing methods, we found\nthat
  the weight distribution will affect efficiency and fairness, but these tw
 o metrics are always considered separately. Confronting the obstacle, we p
 ropose a novel optimization framework namelyFAIST, calibrating a fair mode
 l by controlling weight distribution to optimize fairness, efficiency, and
  accuracy simultaneously. We first devise an optimization algorithm that c
 an guide the training\nto generate model weights following the desired dis
 tribution. Then, we integrate the optimizer into a reinforcement learning 
 process to identify hyperparameters of distribution to yield high performa
 nce. Evaluation of dermatology and face attribute datasets demonstrates FA
 IST's simultaneous improvements, with a notable 27.24% fairness improvemen
 t on the ISIC2019 dataset\n\nTopic: AI, Autonomous Systems, Cloud, Design,
  EDA, Embedded Systems, IP, Security
END:VEVENT
END:VCALENDAR