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:20260422T000711Z
LOCATION:401-402
DTSTART;TZID=America/Denver:20231115T143000
DTEND;TZID=America/Denver:20231115T150000
UID:submissions.supercomputing.org_SC23_sess168_pap413@linklings.com
SUMMARY:High Throughput Training of Deep Surrogates from Large Ensemble Ru
 ns
DESCRIPTION:Lucas Meyer (National Institute for Research in Digital Scienc
 e and Technology, Electricité de France); Marc Schouler and Robert Alexand
 er Caulk (National Institute for Research in Digital Science and Technolog
 y); Alejandro Ribes (Electricité de France); and Bruno Raffin (National In
 stitute for Research in Digital Science and Technology)\n\nRecent years ha
 ve seen a surge in deep learning approaches to accelerate numerical solver
 s, which provide faithful but computationally intensive simulations of the
  physical world. These deep surrogates are generally trained in a supervis
 ed manner from limited amounts of data slowly generated by the same solver
  they intend to accelerate. We propose an open-source framework that enabl
 es the online training of these models from a large ensemble run of simula
 tions. It leverages multiple levels of parallelism to generate rich datase
 ts. The framework avoids I/O bottlenecks and storage issues by directly st
 reaming the generated data. A training reservoir mitigates the inherent bi
 as of streaming while maximizing GPU throughput. Experiment on training a 
 deep surrogate for the heat equation shows the proposed approach enables t
 raining on 8TB of data in 2 hours with an accuracy improved by 47% and a b
 atch throughput multiplied by 13 compared to a traditional offline procedu
 re.\n\nTag: Artificial Intelligence/Machine Learning\n\nRegistration Categ
 ory: Tech Program Reg Pass\n\nReproducibility Badges: Artifact Available, 
 Artifact Functional, Results Reproduced\n\nSession Chair: Shuaiwen Leon So
 ng (Microsoft Corporation)\n\n
END:VEVENT
END:VCALENDAR