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:D165
DTSTART;TZID=America/Chicago:20181111T153000
DTEND;TZID=America/Chicago:20181111T160000
UID:submissions.supercomputing.org_SC18_sess147_ws_cafcw107@linklings.com
SUMMARY:HPC-Based Hyperparameter Search of MT-CNN for Information Extracti
 on from Cancer Pathology Reports
DESCRIPTION:Workshop\nApplications, Deep Learning, Exascale, Workshop Reg 
 Pass\n\nHPC-Based Hyperparameter Search of MT-CNN for Information Extracti
 on from Cancer Pathology Reports\n\nYoon, Alawad, Christian, Hinkle, Raman
 athan...\n\nFinding optimal hyperparameters is necessary to identify the b
 est performing deep learning models, but the process is costly. In this pa
 per, we applied model-based optimization, also known as Bayesian optimizat
 ion, using the CANDLE framework implemented on a High-Performance Computin
 g environment. As a use case, we selected information extraction from canc
 er pathology reports using a multi-task convolutional neural network, a mo
 del with 10 hyperparameters to be optimized. We utilized a synthesized tex
 t corpus of 8,000 training cases and 2,000 validation cases with four type
 s of clinical task labels including primary cancer site, laterality, behav
 ior, and histological grade. We conducted the experiments on the Titan sup
 ercomputer at the Oak Ridge Leadership Computing Facility (OLCF), reported
  the optimal hyperparameters found, and demonstrated that hyperparameter o
 ptimization using the CANDLE framework is a feasible approach with respect
  to both scalability and clinical task performance.
URL:https://sc18.supercomputing.org/presentation/?id=ws_cafcw107&sess=sess
 147
END:VEVENT
END:VCALENDAR