BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/New_York
X-LIC-LOCATION:America/New_York
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20210402T160557Z
LOCATION:Track 7
DTSTART;TZID=America/New_York:20201111T173000
DTEND;TZID=America/New_York:20201111T180000
UID:submissions.supercomputing.org_SC20_sess199_ws_dls109@linklings.com
SUMMARY:DeepGalaxy: Deducing the Properties of Galaxy Mergers from Images 
 Using Deep Neural Networks
DESCRIPTION:Workshop\n\nDeepGalaxy: Deducing the Properties of Galaxy Merg
 ers from Images Using Deep Neural Networks\n\nCai, Bédorf, Saletore, Codre
 anu, Podareanu...\n\nGalaxy mergers, the dynamical process during which tw
 o galaxies collide, are among the most spectacular phenomena in the Univer
 se. During this process, the two colliding galaxies are tidally disrupted,
  producing significant visual features that evolve as a function of time. 
 These visual features contain valuable clues for deducing the physical pro
 perties of the galaxy mergers. In this work, we propose DeepGalaxy, a visu
 al analysis framework trained to predict the physical properties of galaxy
  mergers based on their morphology. Based on an encoder-decoder architectu
 re, DeepGalaxy encodes the input images to a compressed latent space z, an
 d determines the similarity of images according to the latent-space distan
 ce. DeepGalaxy consists of a fully convolutional autoencoder (FCAE) which 
 generates activation maps at its 3D latent-space, and a variational autoen
 coder (VAE) which compresses the activation maps into a 1D vector, and a c
 lassifier that generates labels from the activation maps. The backbone of 
 the FCAE can be fully customized according to the complexity of the images
 . DeepGalaxy demonstrates excellent scaling performance on parallel machin
 es. On the Endeavour supercomptuer, the scaling efficiency exceeds 0.93 wh
 en trained on 128 workers, and it maintains above 0.73 when trained with 5
 12 workers. Without having to carry out expensive numerical simulations, D
 eepGalaxy makes inference of the physical properties of galaxy mergers dir
 ectly from images, and thereby achieves a speed up factor of about 10^5.\n
 \nRegistration Category: Workshop Reg Pass
END:VEVENT
END:VCALENDAR