Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

  • Perspective
  • Published:

Hardware-accelerated inference for real-time gravitational-wave astronomy

Nature Astronomy volume 6, pages 529–536 (2022)Cite this article

Subjects

Abstract

Computational demands in gravitational-wave astronomy are expected to at least double over the next five years. As kilometre-scale interferometers are brought to design sensitivity, real-time delivery of gravitational-wave alerts will become increasingly important to enable multimessenger follow-up. Here we discuss a novel implementation and deployment of deep learning inference for real-time data denoising and astrophysical source identification. This objective is accomplished using a generic inference-as-a-service model capable of adapting to the future needs of gravitational-wave data analysis. The implementation allows seamless incorporation of hardware accelerators and also enables the use of commercial or private as-a-service computing. Low-latency and offline computing in gravitational-wave astronomy addresses key challenges in scalability and reliability and provides a data analysis platform particularly optimized for deep learning applications.

This is a preview of subscription content, access via your institution

Access options

Buy this article

  • Purchase on Springer Link
  • Instant access to full article PDF
Buy now

Prices may be subject to local taxes which are calculated during checkout

Fig. 1: Example IaaS deployment scenarios.
Fig. 2: Distributions of time and cost required to process 1 s of data.
Fig. 3: Flow of data through ensemble of individual modules on inference service in end-to-end pipeline.
Fig. 4: End-to-end server-side latency for multiple GPU counts as a function of r.
Fig. 5: The number of seconds of gravitational-wave data processed per second as a function of time for a sustained test using HEPCloud.
Fig. 6: Performance of the online-deployed DeepClean noise regression as a function of the aggregation latency—length of the time-series data excluded from each cleaned segment, to avoid noisy edges.

Similar content being viewed by others

Data availability

The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.

Code availability

All code used to generate the data for both the online and offline experiments in this study is available at https://github.com/fastmachinelearning/gw-iaas.

References

  1. Dietrich, T. et al. New constraints on the supranuclear equation of state and the Hubble constant from nuclear physics—multi-messenger astronomy. Science 370, 1450–1453 (2020).

    Article  ADS  MathSciNet  Google Scholar 

  2. Aartsen, M. G. et al. Multimessenger observations of a flaring blazar coincident with high-energy neutrino IceCube-170922A. Science 361, eaat1378 (2018).

  3. Abbott, B. P. et al. Observation of gravitational waves from a binary black hole merger. Phys. Rev. Lett. 116, 061102 (2016).

    Article  ADS  MathSciNet  Google Scholar 

  4. Abbott, B. et al. GW170817: observation of gravitational waves from a binary neutron star inspiral. Phys. Rev. Lett. 119, 161101 (2017).

    Article  ADS  Google Scholar 

  5. Harry, G. M. et al. Advanced LIGO: the next generation of gravitational wave detectors. Class. Quantum Gravity 27, 084006 (2010).

    Article  ADS  MathSciNet  Google Scholar 

  6. Aasi, J. et al. Advanced LIGO. Class. Quantum Gravity 32, 074001 (2015).

    Article  ADS  Google Scholar 

  7. Acernese, F. et al. Advanced Virgo. Class. Quantum Gravity 32, 024001 (2015).

    Article  ADS  Google Scholar 

  8. Abbott, B. P. et al. Prospects for observing and localizing gravitational-wave transients with Advanced LIGO, Advanced Virgo and KAGRA. Living Rev. Relativ. 23, 3 (2020).

  9. Usman, S. A. et al. The PyCBC search for gravitational waves from compact binary coalescence. Class. Quantum Gravity 33, 215004 (2016).

    Article  ADS  Google Scholar 

  10. Cannon, K. et al. GstLAL: a software framework for gravitational wave discovery. Preprint at https://arxiv.org/abs/2010.05082 (2020).

  11. Abbott, B. et al. Search for subsolar mass ultracompact binaries in Advanced LIGO’s second observing run. Phys. Rev. Lett. 123, 161102 (2019).

  12. Aso, Y. et al. Interferometer design of the KAGRA gravitational wave detector. Phys. Rev. D 88, 043007 (2013).

    Article  ADS  Google Scholar 

  13. Somiya, K. Detector configuration of KAGRA—the Japanese cryogenic gravitational-wave detector. Class. Quantum Gravity 29, 124007 (2012).

    Article  ADS  Google Scholar 

  14. Zevin, M. et al. Gravity Spy: integrating advanced LIGO detector characterization, machine learning, and citizen science. Class. Quantum Gravity 34, 064003 (2017).

    Article  ADS  Google Scholar 

  15. Davis, D. et al. LIGO detector characterization in the second and third observing runs. Class. Quantum Gravity 38, 135014 (2021).

    Article  ADS  Google Scholar 

  16. Cuoco, E. et al. Enhancing gravitational-wave science with machine learning. Mach. Learn. Sci. Technol. 2, 011002 (2021).

    Article  Google Scholar 

  17. Vajente, G. et al. Machine-learning nonstationary noise out of gravitational-wave detectors. Phys. Rev. D 101, 042003 (2020).

  18. Ormiston, R., Nguyen, T., Coughlin, M., Adhikari, R. X. & Katsavounidis, E. Noise reduction in gravitational-wave data via deep learning. Phys. Rev. Res. 2, 033066 (2020).

  19. George, D. & Huerta, E. A. Deep learning for real-time gravitational wave detection and parameter estimation: results with Advanced LIGO data. Phys. Lett. B 778, 64–70 (2018).

    Article  ADS  Google Scholar 

  20. Schäfer, M. B. & Nitz, A. H. From one to many: A deep learning coincident gravitational-wave search. Phys. Rev. D 105, 043003 (2022).

    Article  ADS  MathSciNet  Google Scholar 

  21. Green, S. R., Simpson, C. & Gair, J. Gravitational-wave parameter estimation with autoregressive neural network flows. Phys. Rev. D 102, 104057 (2020).

    Article  ADS  MathSciNet  Google Scholar 

  22. Gabbard, H., Messenger, C., Heng, I.S. et al. Bayesian parameter estimation using conditional variational autoencoders for gravitational-wave astronomy. Nat. Phys. 18, 112–117 (2022).

    Article  Google Scholar 

  23. Mukund, N., Abraham, S., Kandhasamy, S., Mitra, S. & Philip, N. S. Transient classification in LIGO data using difference boosting neural network. Phys. Rev. D 95, 104059 (2017).

    Article  ADS  Google Scholar 

  24. Mittal, S. & Vetter, J. S. A survey of CPU–GPU heterogeneous computing techniques. ACM Comput. Surv. 47, 69 (2015).

  25. Duarte, J. et al. FPGA-accelerated machine learning inference as a service for particle physics computing. Comput. Softw. Big Sci. 3, 13 (2019).

    Article  Google Scholar 

  26. Caulfield, A. et al. A cloud-scale acceleration architecture. In Proc. 49th Annual IEEE/ACM International Symposium on Microarchitecture (IEEE, 2016).

  27. Sfiligoi, I. et al. Expanding IceCube GPU computing into the Clouds. In 2021 IEEE 17th International Conference on eScience (eScience) 227–228 (IEEE, 2021).

  28. Wang, M. et al. GPU-Accelerated Machine Learning Inference as a Service for Computing in Neutrino Experiments. Front. Big Data 3 (2021).

  29. Krupa, J. et al. GPU coprocessors as a service for deep learning inference in high energy physics. Mach. Learn. Sci. Technol. 2, 035005 (2021).

    Article  Google Scholar 

  30. Rankin, D. et al. FPGAs-as-a-Service Toolkit (FaaST). In 2020 IEEE/ACM International Workshop on Heterogeneous High-performance Reconfigurable Computing (H2RC) 38–47 (IEEE, 2020).

  31. Abbott, B. P. et al. Characterization of transient noise in Advanced LIGO relevant to gravitational wave signal GW150914. Class. Quantum Gravity 33, 134001 (2016).

    Article  ADS  Google Scholar 

  32. Triton Inference Server v.2.12.0 https://developer.nvidia.com/nvidia-triton-inference-server (NVIDIA, 2021).

  33. Abadie, J. et al. First low-latency LIGO+Virgo search for binary inspirals and their electromagnetic counterparts. Astron. Astrophys. 541, A155 (2012).

    Article  Google Scholar 

  34. Nitz, A. H., Dal Canton, T., Davis, D. & Reyes, S. Rapid detection of gravitational waves from compact binary mergers with PyCBC Live. Phys. Rev. D 98, 024050 (2018).

  35. Abbott, R. et al. Search for gravitational waves associated with gamma-ray bursts detected by Fermi and Swift during the LIGO–Virgo run O3a. Astrophys. J. 915, 86 (2021).

    Article  ADS  Google Scholar 

  36. Usman, S. A. et al. The PyCBC search for gravitational waves from compact binary coalescence. Class. Quantum Gravity 33, 215004 (2016).

    Article  ADS  Google Scholar 

  37. Abbott, R. et al. GWTC-2: compact binary coalescences observed by LIGO and Virgo during the first half of the third observing run. Phys. Rev. X 11, 021053 (2021).

  38. Holzman, B. et al. HEPCloud, a new paradigm for HEP facilities: CMS Amazon Web Services investigation. Comput. Softw. Big Sci. 1, 1 (2017).

  39. Reitze, D. et al. Cosmic Explorer: The U.S. Contribution to Gravitational-Wave Astronomy beyond LIGO. Bull. Am. Astron. Soc. 51 (2019).

  40. Kurtzer, G. M. et al. hpcng/singularity: singularity 3.7.3. https://zenodo.org/record/1310023 (2021).

Download references

Acknowledgements

We are grateful for computational resources provided by the LIGO Laboratory at Caltech, Livingston, LA, and Hanford, WA. The LIGO Laboratory has been supported under National Science Foundation (NSF) grants PHY-0757058 and PHY-0823459. A.G., D.R., T.N., P.H. and E.K. are supported by NSF grants 1934700 and 1931469, and D.R. additionally by the IRIS-HEP grant 1836650. J.K. is supported by NSF grant 190444. M.S. and M.C. are supported by NSF grant PHY-2010970. Work supported by the Fermi National Accelerator Laboratory, managed and operated by Fermi Research Alliance, LLC under contract DE-AC02-07CH11359 with the US Department of Energy. The US Government retains and the publisher, by accepting the article for publication, acknowledges that the US Government retains a non-exclusive, paid-up, irrevocable, world-wide license to publish or reproduce the published form of this manuscript, or allow others to do so, for US Government purposes. Cloud credits for this study were provided by the Internet2-managed Exploring Clouds for Acceleration of Science (NSF grant PHY-190444). Additionally we would like to thank the NSF Institute for AI and Fundamental Interactions (cooperative agreement PHY-2019786). We are also grateful for the support provided by S. Anderson in the realization and testing of our workflow within the LDG. Finally, we thank A. Pace for providing useful comments on the manuscript.

Author information

Authors and Affiliations

  1. Department of Physics, MIT, Cambridge, MA, USA

    Alec Gunny, Dylan Rankin, Jeffrey Krupa, Tri Nguyen, Philip Harris & Erik Katsavounidis

  2. LIGO Laboratory, MIT, Cambridge, MA, USA

    Alec Gunny, Tri Nguyen & Erik Katsavounidis

  3. The NSF Institute for Artificial Intelligence and Fundamental Interactions, Cambridge, MA, USA

    Alec Gunny, Dylan Rankin, Jeffrey Krupa, Tri Nguyen & Philip Harris

  4. School of Physics and Astronomy, University of Minnesota, Minneapolis, MN, USA

    Muhammed Saleem & Michael Coughlin

  5. Fermi National Accelerator Laboratory, Batavia, IL, USA

    Steven Timm & Burt Holzman

Authors
  1. Alec Gunny
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Dylan Rankin
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Jeffrey Krupa
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Muhammed Saleem
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Tri Nguyen
    View author publications

    You can also search for this author in PubMed Google Scholar

  6. Michael Coughlin
    View author publications

    You can also search for this author in PubMed Google Scholar

  7. Philip Harris
    View author publications

    You can also search for this author in PubMed Google Scholar

  8. Erik Katsavounidis
    View author publications

    You can also search for this author in PubMed Google Scholar

  9. Steven Timm
    View author publications

    You can also search for this author in PubMed Google Scholar

  10. Burt Holzman
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

A.G. and D.R. are the primary authors of the manuscript. J.K. integrated applications in HEPCloud. S.T. and B.H. support and operate HEPCloud. M.S., M.C., E.K. and T.N. support development of DeepClean. All authors contributed to editing of the manuscript.

Corresponding authors

Correspondence to Alec Gunny or Dylan Rankin.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Astronomy thanks Alexander Nitz and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Supplementary Discussion including Supplementary Fig. 1.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Gunny, A., Rankin, D., Krupa, J. et al. Hardware-accelerated inference for real-time gravitational-wave astronomy. Nat Astron 6, 529–536 (2022). https://doi.org/10.1038/s41550-022-01651-w

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1038/s41550-022-01651-w

This article is cited by

Search

Advanced search

Quick links

Nature Briefing AI and Robotics

Sign up for the Nature Briefing: AI and Robotics newsletter — what matters in AI and robotics research, free to your inbox weekly.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing: AI and Robotics