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:20260522T150120Z
LOCATION:D167/174
DTSTART;TZID=America/Chicago:20181112T113000
DTEND;TZID=America/Chicago:20181112T120000
UID:submissions.supercomputing.org_SC18_sess151_ws_mlhpce105@linklings.com
SUMMARY:Automated Labeling of Electron Microscopy Images Using Deep Learni
 ng
DESCRIPTION:Gunther Weber (Lawrence Berkeley National Laboratory; Universi
 ty of California, Davis) and Colin Ophus and Lavanya Ramakrishnan (Lawrenc
 e Berkeley National Laboratory)\n\nSearching for scientific data requires 
 metadata providing a relevant context. Today, generating metadata is a tim
 e and labor intensive manual process that is often neglected, and importan
 t datasets are not accessible through search. We investigate the use of ma
 chine learning to generalize metadata from a subset of labeled data, thus 
 increasing the availability of meaningful metadata for search. Specificall
 y, we consider electron microscopy images collected at the National Center
  for Electron Microscopy at the Lawrence Berkeley National Laboratory and 
 use of deep learning to discern characteristics from a small subset of lab
 eled images and transfer labels to the entire image corpus.\n\nRelatively 
 small training set sizes and a minimum resolution of 512x512 pixels requir
 ed by the application domain pose unique challenges. We overcome these cha
 llenges by using a simple yet powerful convolutional network architecture 
 that limits the number of free parameters to lower the required amount of 
 computational power and reduce the risk of overfitting. We achieve a class
 ification accuracy of approximately 80% in discerning between images recor
 ded in two operating modes of the electron microscope---transmission elect
 ron microscopy (TEM) and scanning transmission electron microscopy (STEM).
  We use transfer learning–i.e., re-using the pre-trained convolution layer
 s from the TEM vs. STEM classification problem–to generalize labels and ac
 hieve an accuracy of approximately 70% despite current experiments being l
 imited to small training data sets. We present these predictions as sugges
 tions to domain scientists to accelerate the labeling process with the goa
 l of further validating our approach and improving the accuracy of automat
 ically created labels.\n\nTag: Deep Learning, Machine Learning\n\nRegistra
 tion Category: Workshop Reg Pass\n\n
END:VEVENT
END:VCALENDAR