First Machine Learning Gravitational-Wave Search Mock Data Challenge - Research Results

Aug 2, 2026

Upload by Noah

First Machine Learning Gravitational-Wave Search Mock Data Challenge results evaluate six entered algorithms for recovering binary black hole merger signals embedded in progressively more realistic detector noise. The final dataset uses real noise from the O3a observing run and includes signals up to 20 s with precession and higher-order modes. Average sensitivity distance and runtime are measured using one month of blinded test data. Best machine learning methods reach up to 95% of matched-filtering sensitivity for simulated Gaussian noise at FAR of one per month, while real-noise performance leads to 70% and improves again at higher FAR.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages24

Reading Time1 h 0 min

AccessFree to download

Introduction

Gravitational-wave observations and observing runs

Matched filtering and search methods

First machine learning gravitational-wave search mock data challenge Marlin B. Schäfer, 1,2 Ondrˇej Zelenka,3,4 Alexander H. Nitz, 1,2 He Wang,5 Shichao Wu, 1,2 Zong-Kuan Guo,5 Zhoujian Cao,6 Zhixiang Ren,7 Paraskevi Nousi,8 Nikolaos Stergioulas,9 Panagiotis Iosif, 10,9 Alexandra E. Koloniari,9 Anastasios Tefas,8 Nikolaos Passalis,8 Francesco Salemi, 11,12 Gabriele Vedovato, 13 Sergey Klimenko,14 Tanmaya Mishra, 14 Bernd Brügmann,3,4 Elena Cuoco, 15,16,17 E. A. Huerta, 18,19 Chris Messenger,20 and Frank Ohme1,2 1Max-Planck-Institut für Gravitationsphysik, Albert-Einstein-Institut, D-30167 Hannover, Germany 2Leibniz Universität Hannover, D-30167 Hannover, Germany 3Friedrich-Schiller-Universität Jena, D-07743 Jena, Germany 4Michael Stifel Center Jena, D-07743 Jena, Germany 5CAS Key Laboratory of Theoretical Physics, Institute of Theoretical Physics, Chinese Academy of Sciences, Beijing 100190, China 6Department of Astronomy, Beijing Normal University, Beijing 100875, China 7Peng Cheng Laboratory, Shenzhen, 518055, China 8Department of Informatics, Aristotle University of Thessaloniki, GR-54124 Thessaloniki, Greece 9Department of Physics, Aristotle University of Thessaloniki, GR-54124 Thessaloniki, Greece 10GSI Helmholtz Center for Heavy Ion Research, Planckstraße 1, 64291 Darmstadt, Germany 11 Universita` di Trento, Dipartimento di Fisica, I-38123 Povo, Trento, Italy 12INFN, Trento Institute for Fundamental Physics and Applications, I-38123 Povo, Trento, Italy 13INFN, Sezione di Padova, I-35131 Padova, Italy 14Department of Physics, University of Florida, PO Box 118440, Gainesville, Florida 32611-8440, USA 15European Gravitational Observatory (EGO), I-56021 Cascina, Pisa, Italy 16Scuola Normale Superiore, Piazza dei Cavalieri 7, I-56126 Pisa, Italy 17INFN, Sezione di Pisa, Largo Bruno Pontecorvo, 3, I-56127 Pisa, Italy 18Data Science and Learning Division, Argonne National Laboratory, Lemont, Illinois 60439, USA 19Department of Computer Science, University of Chicago, Chicago, Illinois 60637, USA 20SUPA, School of Physics and Astronomy, University of Glasgow, Glasgow G12 8QQ, United Kingdom (Received 23 September 2022; accepted 28 November 2022; published 27 January 2023) We present the results of the first Machine Learning Gravitational-Wave Search Mock Data Challenge. For this challenge, participating groups had to identify gravitational-wave signals from binary black hole mergers of increasing complexity and duration embedded in progressively more realistic noise. The final of the 4 provided datasets contained real noise from the O3a observing run and signals up to a duration of 20 s with the inclusion of precession effects and higher order modes. We present the average sensitivity distance and run-time for the 6 entered algorithms derived from 1 month of test data unknown to the participants prior to submission. Of these, 4 are machine learning algorithms. We find that the best machine learning based algorithms are able to achieve up to 95% of the sensitive distance of matched-filtering based production analyses for simulated Gaussian noise at a false-alarm rate (FAR) of one per month. In contrast, for real noise, the leading machine learning search achieved 70% . For higher FARs the differences insensitive distance shrink to the point where select machine learning submissions outperform traditional search algorithms at FARs ≥ 200 per month on some datasets. Our results show that current machine learning search algorithms may already be sensitive enough in limited parameter regions to be useful for some production settings. To improve the state-of-the-art, machine learning algorithms need to reduce the false-alarm rates at which they are capable of detecting signals and extend their validity to regions of parameter space where modeled searches are computationally expensive to run. Based on our findings we compile a list of research areas that we believe are the most important to elevate

FAQs

What was the goal of the First Machine Learning Gravitational-Wave Search Mock Data Challenge?

Participating groups had to identify gravitational-wave signals from binary black hole mergers with increasing complexity and duration, embedded in progressively more realistic noise.

What does the final dataset in the challenge contain?

It uses real noise from the O3a observing run and includes signals up to 20 s, incorporating precession effects and higher order modes.

How did machine learning algorithms perform compared with matched filtering?

For simulated Gaussian noise at a false-alarm rate of one per month, the best machine learning algorithms reached up to 95% of the sensitive distance of matched-filtering production analyses. For real noise, the leading machine learning search achieved about 70%.

Aug 2, 2026

Upload by Noah