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:20250822T115804Z
LOCATION:Room 5.0B56
DTSTART;TZID=Europe/Stockholm:20250616T170000
DTEND;TZID=Europe/Stockholm:20250616T173000
UID:submissions.pasc-conference.org_PASC25_sess168_pap121@linklings.com
SUMMARY:Scalable Bayesian Inference of Large Simulations via Asynchronous 
 Prefetching Multilevel Delayed Acceptance
DESCRIPTION:Maximilian Kruse (Karlsruhe Institute of Technology); Zihua Ni
 u (Ludwig Maximilian University of Munich); Sebastian Wolf (Technical Univ
 ersity of Munich); Mikkel Lykkegaard (Danish Technological Institute); Mic
 hael Bader (Technical University of Munich); Alice-Agnes Gabriel (Universi
 ty of California San Diego, Ludwig Maximilian University of Munich); and L
 inus Seelinger (Karlsruhe Institute of Technology)\n\nBayesian inference e
 nables greater scientific insight into simulation models, determining mode
 l parameters and meaningful confidence regions from observed data. With hi
 erarchical methods like Multilevel Delayed Acceptance (MLDA) drastically r
 educing compute cost, sampling Bayesian posteriors for computationally int
 ensive models becomes increasingly feasible. Pushing MLDA towards the stro
 ng scaling regime (i.e. high compute resources, short time-to-solution) re
 mains a challenge: Even though MLDA only requires a moderate number of hig
 h-accuracy simulation runs, it inherits the sequential chain structure and
  need for chain burn-in from Markov chain Monte Carlo (MCMC). We present f
 ully asynchronous parallel prefetching for MLDA, adding an axis of scalabi
 lity complementary to forward model parallelization and parallel chains. A
  thorough scaling analysis demonstrates that prefetching is advantageous i
 n strong scaling scenarios. We investigate the behavior of prefetching MLD
 A in small-scale test problems. A large-scale geophysics application, name
 ly parameter identification for non-linear earthquake modelling, highlight
 s interaction with coarse-level quality and model scalability.\n\nDomain: 
 Computational Methods and Applied Mathematics\n\nSession Chair: Tobias Rib
 izel (Technical University of Munich)\n\n
END:VEVENT
END:VCALENDAR