BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Chicago
X-LIC-LOCATION:America/Chicago
BEGIN:DAYLIGHT
TZOFFSETFROM:-0600
TZOFFSETTO:-0500
TZNAME:CDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0500
TZOFFSETTO:-0600
TZNAME:CST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260522T150116Z
LOCATION:C2/3/4 Ballroom
DTSTART;TZID=America/Chicago:20181113T083000
DTEND;TZID=America/Chicago:20181113T170000
UID:submissions.supercomputing.org_SC18_sess322_post201@linklings.com
SUMMARY:Fast and Accurate Training of an AI Radiologist
DESCRIPTION:Lucas A. Wilson, Vineet Gundecha, Srinivas Varadharajan, Alex 
 Filby, Pei Yang, and Quy Ta (Dell EMC); Valeriu Codreanu and Damian Podare
 anu (SURFsara); and Vikram Saletore (Intel Corporation)\n\nThe health care
  industry is expected to be an early adopter of AI and deep learning to im
 prove patient outcomes, reduce costs, and speed up diagnosis. We have deve
 loped models for using AI to diagnose pneumonia, emphysema, and other thor
 acic pathologies from chest x-rays. Using the Stanford University CheXNet 
 model as inspiration, we explore ways of developing accurate models for th
 is problem with fast parallel training on Zenith, the Intel Xeon-based sup
 ercomputer at Dell EMC's HPC and AI Innovation Lab. We explore various net
 work topologies to gain insight into what types of neural networks scale w
 ell in parallel and improve training time from days to hours. We then expl
 ore transferring this learned knowledge to other radiology subdomains, suc
 h as mammography, and whether this leads to better models than developing 
 subdomain models independently.\n\nRegistration Category: Tech Program Reg
  Pass, Exhibits Reg Pass\n\n
END:VEVENT
END:VCALENDAR