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:20211207T055407Z
LOCATION:Online
DTSTART;TZID=America/Chicago:20211115T155500
DTEND;TZID=America/Chicago:20211115T162500
UID:submissions.supercomputing.org_SC21_sess423_ws_mlhpce103@linklings.com
SUMMARY:Colmena: Scalable Machine-Learning-Based Steering of Ensemble Simu
 lations for High Performance Computing
DESCRIPTION:Workshop\n\nColmena: Scalable Machine-Learning-Based Steering 
 of Ensemble Simulations for High Performance Computing\n\nWard, Sivaraman,
  Pauloski, Babuji, Chard...\n\nScientific applications that involve simula
 tion ensembles can be accelerated greatly by using experiment design metho
 ds to select the best simulations to perform. Methods that use machine lea
 rning (ML) to create proxy models of simulations show particular promise f
 or guiding ensembles but are challenging to deploy because of the need to 
 coordinate dynamic mixes of simulation and learning tasks. We present Colm
 ena, an open-source Python framework that allows users to steer campaigns 
 by providing just the implementations of individual tasks plus the logic u
 sed to choose which tasks to execute when. Colmena handles task dispatch, 
 results collation, ML model invocation, and ML model (re)training, using P
 arsl to execute tasks on HPC systems. We describe the design of Colmena an
 d illustrate its capabilities by applying it to electrolyte design, where 
 it both scales to \num{65536} CPUs and accelerates the discovery rate for 
 high-performance molecules by a factor of 100 over unguided searches.\n\nT
 ag: Online Only, Machine Learning and Artificial Intelligence\n\nRegistrat
 ion Category: Workshop Reg Pass
END:VEVENT
END:VCALENDAR