KIT | KIT-Bibliothek | Impressum | Datenschutz

Scalable Bayesian Inference of Large Simulations via Asynchronous Prefetching Multilevel Delayed Acceptance

Kruse, Maximilian ORCID iD icon 1; Niu, Zihua; Wolf, Sebastian; Lykkegaard, Mikkel B.; Bader, Michael; Gabriel, Alice-Agnes; Seelinger, Linus ORCID iD icon 1
1 Scientific Computing Center (SCC), Karlsruher Institut für Technologie (KIT)

Abstract:

Bayesian inference enables greater scientific insight into simulation models, determining model parameters and meaningful confidence regions from observed data. With hierarchical methods like Multilevel Delayed Acceptance (MLDA) drastically reducing compute cost, sampling Bayesian posteriors for computationally intensive models becomes increasingly feasible. Pushing MLDA towards the strong scaling regime (i.e. high compute resources, short time-to-solution) remains a challenge: Even though MLDA only requires a moderate number of high-accuracy simulation runs, it inherits the sequential chain structure and need for chain burn-in from Markov chain Monte Carlo (MCMC).We present fully asynchronous parallel prefetching for MLDA, adding an axis of scalability complementary to forward model parallelization and parallel chains. A thorough scaling analysis demonstrates that prefetching is advantageous in strong scaling scenarios. We investigate the behavior of prefetching MLDA in small-scale test problems. A large-scale geophysics application, namely parameter identification for non-linear earthquake modelling, highlights interaction with coarse-level quality and model scalability.


Verlagsausgabe §
DOI: 10.5445/IR/1000184542
Veröffentlicht am 05.09.2025
Cover der Publikation
Zugehörige Institution(en) am KIT Scientific Computing Center (SCC)
Publikationstyp Proceedingsbeitrag
Publikationsdatum 16.06.2025
Sprache Englisch
Identifikator ISBN: 979-84-00-71886-1
KITopen-ID: 1000184542
Erschienen in Proceedings of the Platform for Advanced Scientific Computing Conference
Veranstaltung Platform for Advanced Scientific Computing Conference (PASC 2025), Brugg, Schweiz, 16.06.2025 – 18.06.2025
Verlag Association for Computing Machinery (ACM)
Seiten 1–13
Nachgewiesen in Dimensions
OpenAlex
Scopus
Relationen in KITopen
KIT – Die Universität in der Helmholtz-Gemeinschaft
KITopen Landing Page