EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Neural Importance Sampling for Rapid and Reliable Gravitational-Wave Inference

Maximilian Dax, Stephen R Green, Jonathan Gair, Michael Pürrer, Jonas Wildberger, Jakob H Macke, Alessandra Buonanno, Bernhard Schölkopf

Physical Review Letters · 26.04.2023 · 130 (17)

DOI: 10.1103/physrevlett.130.171403 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Physical Review Letters, online erschienen am 26.04.2023. Verfasst von Maximilian Dax und 7 weiteren Autor:innen. Beteiligte Einrichtung: Max Planck Institute for Intelligent Systems, Max-Planck-Ring 4, 72076 Tübingen, Germany. Die Arbeit wurde bisher 111-mal zitiert (Crossref) und nennt 55 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Science Foundation, Science and Technology Facilities Council, Max-Planck-Gesellschaft.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

We combine amortized neural posterior estimation with importance sampling for fast and accurate gravitational-wave inference. We first generate a rapid proposal for the Bayesian posterior using neural networks, and then attach importance weights based on the underlying likelihood and prior. This provides (1) a corrected posterior free from network inaccuracies, (2) a performance diagnostic (the sample efficiency) for assessing the proposal and identifying failure cases, and (3) an unbiased estimate of the Bayesian evidence. By establishing this independent verification and correction mechanism we address some of the most frequent criticisms against deep learning for scientific inference. We carry out a large study analyzing 42 binary black hole mergers observed by LIGO and Virgo with the SEOBNRv4PHM and IMRPhenomXPHM waveform models. This shows a median sample efficiency of ≈10% (2 orders of magnitude better than standard samplers) as well as a tenfold reduction in the statistical uncertainty in the log evidence. Given these advantages, we expect a significant impact on gravitational-wave inference, and for this approach to serve as a paradigm for harnessing deep learning methods in scientific applications.

Autor:innen und Institutionen

  • Max Planck Institute for Intelligent Systems, Max-Planck-Ring 4, 72076 Tübingen, Germany.
  • Max Planck Institute for Gravitational Physics (Albert Einstein Institute), Am Mühlenberg 1, 14476 Potsdam, Germany.
  • School of Mathematical Sciences, University of Nottingham, University Park, Nottingham, NG7 2RD, United Kingdom.
  • Department of Physics, East Hall, University of Rhode Island, Kingston, Rhode Island 02881, USA.
  • URI Research Computing, Tyler Hall, University of Rhode Island, Kingston, Rhode Island 02881, USA.
  • Machine Learning in Science, University of Tübingen, 72076 Tübingen, Germany.

Förderung

  • National Science Foundation
  • Science and Technology Facilities Council
  • Max-Planck-Gesellschaft
  • Australian Research Council
  • Centre National de la Recherche Scientifique
  • Instituto Nazionale di Fisica Nucleare
  • Ministry of Education, Culture, Sports, Science and Technology
  • Japan Society for the Promotion of Science

Zitieren

APA

Dax, M., Green, S. R., Gair, J., Pürrer, M., Wildberger, J., Macke, J. H., Buonanno, A., & Schölkopf, B. (2023). Neural Importance Sampling for Rapid and Reliable Gravitational-Wave Inference. Physical Review Letters, 130(17). https://doi.org/10.1103/physrevlett.130.171403

Vancouver

Dax M, Green SR, Gair J, Pürrer M, Wildberger J, Macke JH, et al. Neural Importance Sampling for Rapid and Reliable Gravitational-Wave Inference. Physical Review Letters. 2023;130(17). doi:10.1103/physrevlett.130.171403

BibTeX
@article{Dax2023,
  author = {Dax, Maximilian and Green, Stephen R and Gair, Jonathan and Pürrer, Michael and Wildberger, Jonas and Macke, Jakob H and Buonanno, Alessandra and Schölkopf, Bernhard},
  title = {Neural Importance Sampling for Rapid and Reliable Gravitational-Wave Inference},
  journal = {Physical Review Letters},
  year = {2023},
  volume = {130},
  number = {17},
  doi = {10.1103/physrevlett.130.171403}
}
Kostenlos starten →