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:20190719T085745Z
LOCATION:HG D 3.2
DTSTART;TZID=Europe/Stockholm:20190614T113000
DTEND;TZID=Europe/Stockholm:20190614T120000
UID:submissions.pasc-conference.org_PASC19_sess112_msa265@linklings.com
SUMMARY:Calibration of Stochastic Models with SPUX, a Flexible and Portabl
 e Framework with Multilevel Task-Based Parallelization
DESCRIPTION:Minisymposium\nComputer Science and Applied Mathematics, Emerg
 ing Application Domains, Engineering\n\nCalibration of Stochastic Models w
 ith SPUX, a Flexible and Portable Framework with Multilevel Task-Based Par
 allelization\n\nBacci, Šukys\n\nCalibration of models and propagation of u
 ncertainties are ubiquitous problems in computational science. The SPUX so
 ftware package is a recent effort devoted to two main goals. The first one
  is to offer a problem-agnostic solution for stochastic model calibration 
 and uncertainty quantification. The second goal is to allow calibration ca
 mpaigns to profit from high-performance computing infrastructures, indepen
 dently of the type of software that represents the user’s model. In 
 SPUX, model parameters are estimated by resorting to particle filtering, a
  communication-intensive but efficient scheme for the estimation of the ma
 rginal likelihood of stochastic dynamics. Portability to different applica
 tions is ensured by the high-level of abstraction characteristic of task-b
 ased parallelization schemes, while the expressive programming language Py
 thon facilitates integration with the user’s code. Here, we will loo
 k at SPUX with a focus on design, (inter-)operability, and usage on comput
 ing clusters. We will show the software architecture, and real-world appli
 cations in the fields of ecology and environmental science, which will als
 o serve as specific examples of integration. Our contribution will end wit
 h considerations regarding future directions as suggested by the gained ex
 perience on scalability, performance, portability, and users’ needs 
 and expectations.
END:VEVENT
END:VCALENDAR

