BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Denver
X-LIC-LOCATION:America/Denver
BEGIN:DAYLIGHT
TZOFFSETFROM:-0700
TZOFFSETTO:-0600
TZNAME:MDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0600
TZOFFSETTO:-0700
TZNAME:MST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260422T000712Z
LOCATION:607
DTSTART;TZID=America/Denver:20231112T105000
DTEND;TZID=America/Denver:20231112T111000
UID:submissions.supercomputing.org_SC23_sess415_ws_esp105@linklings.com
SUMMARY:Optimized Uncertainty Estimation for Vision Transformers: Enhancin
 g Adversarial Robustness and Performance Using Selective Classification
DESCRIPTION:Erik Pautsch (Loyola University, Chicago); John LI (University
  of California San Diego, Argonne National Laboratory (ANL)); Silvio Rizzi
  (Argonne National Laboratory (ANL)); George Thiruvathukal (Loyola Univers
 ity, Chicago); and Maria Pantoja (California Polytechnic State University,
  San Luis Obispo)\n\nDeep Learning models frequently produce high-confiden
 ce softmax outputs for out-of-distribution (OOD) inputs, which would ideal
 ly be classified as "I don't know". To enhance our model's trustworthiness
 , we incorporate selective classification, which entails abstaining from p
 redictions in situations of doubt. This approach requires initial uncertai
 nty estimation. Subsequently, instead of offering a singular prediction, w
 e provide a distribution over predictions, enabling users to discern if th
 e model is trustworthy or if consultation with a human expert is necessary
 . In this paper, we assess uncertainty in two baseline models: a Convoluti
 onal Neural Network (CNN) and a Vision Transformer (ViT). Leveraging these
  uncertainty values, we minimize errors by refraining from predictions dur
 ing high uncertainty. Additionally, we evaluate these models across variou
 s distributed architectures.\n\nTag: Performance Optimization\n\nRegistrat
 ion Category: Workshop Reg Pass\n\nSession Chairs: Tiernan Casey (Sandia N
 ational Laboratories) and Antigoni Georgiadou (Oak Ridge National Laborato
 ry (ORNL))\n\n
END:VEVENT
END:VCALENDAR