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:20200129T163603Z
LOCATION:502-503-504
DTSTART;TZID=America/Denver:20191117T160000
DTEND;TZID=America/Denver:20191117T163000
UID:submissions.supercomputing.org_SC19_sess101_ws_dls121@linklings.com
SUMMARY:Evolving Larger Convolutional Layer Kernel Sizes for a Settlement 
 Detection Deep-Learner on Summit
DESCRIPTION:Workshop\n\nEvolving Larger Convolutional Layer Kernel Sizes f
 or a Settlement Detection Deep-Learner on Summit\n\nColetti, Lunga, Basset
 t, Rose\n\nDeep-learner  hyper-parameters,  such  as kernel sizes, batch s
 izes, and learning rates, can significantly influence the quality of train
 ed models. The state of the art for finding optimal hyper-parameters gener
 ally uses a brute force, grid search approach, random search,  or  Bayesia
 n-based  optimization  among  other techniques.  We  applied  an  evolutio
 nary  algorithm  to optimize kernel sizes for a convolutional neural netwo
 rk used to detect settlements in satellite imagery. Usually convolutional 
 layer kernel sizes are small – typically one, three, or five – but we foun
 d that the system converged at, or near, kernel sizes of nine for the last
  convolutional layer,  and  that  this  occurred  for  multiple  runs  usi
 ng two different datasets. Moreover, the larger kernel sizes had fewer fal
 se positives than the 3x3 kernel sizes found as optimal via a brute force 
 uniform grid search. This suggests that this large kernel size may be leve
 raging patterns  found  in  larger  areal  features  in  the  source image
 ry, and that this may be generalized as possible guidance for similar remo
 te sensing deep-learning tasks.\n\nTag: Workshop Reg Pass, Deep Learning, 
 Scientific Computing\n\nRegistration Category: Workshop Reg Pass, Deep Lea
 rning, Scientific Computing
URL:https://sc19.supercomputing.org/presentation/?id=ws_dls121&sess=sess10
 1
END:VEVENT
END:VCALENDAR