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:20181221T160726Z
LOCATION:D167/174
DTSTART;TZID=America/Chicago:20181111T153000
DTEND;TZID=America/Chicago:20181111T160000
UID:submissions.supercomputing.org_SC18_sess221_ws_mlhpce118@linklings.com
SUMMARY:Ramifications of Evolving Misbehaving Convolutional Neural Network
  Kernel and Batch Sizes
DESCRIPTION:Workshop\nApplications, Deep Learning, Machine Learning, Works
 hop Reg Pass\n\nRamifications of Evolving Misbehaving Convolutional Neural
  Network Kernel and Batch Sizes\n\nColetti, Lunga, Berres, Sanyal, Rose\n\
 nDeep-learners have many hyper-parameters including learning rate, batch s
 ize, kernel size --- all playing a significant role toward estimating high
  quality models.  Discovering useful hyper-parameter guidelines is an acti
 ve area of research, though the state of the art generally uses a brute fo
 rce, uniform grid approach or random search for finding ideal settings.  W
 e share the preliminary results of using an alternative approach to deep l
 earner hyper-parameter tuning that uses an evolutionary algorithm to impro
 ve the accuracy of a deep-learner models used in satellite imagery buildin
 g footprint detection. We found that the kernel and batch size hyper-param
 eters surprisingly differed from sizes arrived at via a brute force unifor
 m grid approach.  These differences suggest a novel role for evolutionary 
 algorithms in determining the number of convolution layers, as well as sma
 ller batch sizes in improving deep-learner models.
URL:https://sc18.supercomputing.org/presentation/?id=ws_mlhpce118&sess=ses
 s221
END:VEVENT
END:VCALENDAR