BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Europe/Stockholm
X-LIC-LOCATION:Europe/Stockholm
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20230831T095745Z
LOCATION:Dischma
DTSTART;TZID=Europe/Stockholm:20230626T153000
DTEND;TZID=Europe/Stockholm:20230626T160000
UID:submissions.pasc-conference.org_PASC23_sess162_msa186@linklings.com
SUMMARY:Online Learning of Sub-Grid Stress Models for Large Eddy Simulatio
 n of Wall Bounded Turbulent Flows
DESCRIPTION:Minisymposium\n\nRiccardo Balin (Argonne National Laboratory),
  Aviral Prakash and Basu Parmar (University of Colorado Boulder), Filippo 
 Simini (Argonne National Laboratory), and John A. Evans and Kenneth E. Jan
 sen (University of Colorado Boulder)\n\nData-driven approaches for the dev
 elopment of sub-grid stress (SGS) closure models for large eddy simulation
  (LES) have been gaining popularity as they offer encouraging results for 
 improved predictive capacity over traditional models. Due to the time and 
 scale-resolving nature of LES, these models must be trained on instantaneo
 us high-fidelity turbulent data. As we enter the age of exascale simulatio
 ns, generating new databases of training data for traditional offline lear
 ning creates an I/O and storage bottleneck. This limitation is resolved by
  performing online (in situ) learning, wherein the ML model is trained con
 currently with the flow simulation producing the training data. Moreover, 
 online learning allows fine tuning of existing models to new problem class
 es on the fly. This talk will cover the software infrastructure developed 
 to perform online learning at scale. We leveraged an open-source tool call
 ed SmartSim to deploy a database to store training data in-memory and enab
 le two-way asynchronous communication between the simulation and model tra
 ining applications running concurrently on an HPC cluster. In addition, we
  demonstrate online training of a SGS model on a number of flow problems a
 nd compare the model’s predictive capacity against other classical and off
 line trained SGS models.\n\nDomain: Computer Science, Machine Learning, an
 d Applied Mathematics 
\n\nSession Chairs: Timothy C Germann (Los Alamos N
 ational Laboratory) and Ramesh Balakrishnan (Argonne National Laboratory)
END:VEVENT
END:VCALENDAR