Ongoing research
paperAuthors
Credibility Rating
Good quality. Reputable source with community review or editorial standards, but less rigorous than peer-reviewed venues.
Rating inherited from publication venue: arXiv
This paper presents PyMerger, a deep learning tool for gravitational wave detection using ResNets, demonstrating machine learning applications in scientific data analysis that relate to AI safety through robust neural network design and real-world safety-critical applications.
Paper Details
Metadata
Abstract
We present PyMerger, a Python tool for detecting binary black hole (BBH) mergers from the Einstein Telescope (ET), based on a Deep Residual Neural Network model (ResNet). ResNet was trained on data combined from all three proposed sub-detectors of ET (TSDCD) to detect BBH mergers. Five different lower frequency cutoffs ($F_{\text{low}}$): 5 Hz, 10 Hz, 15 Hz, 20 Hz, and 30 Hz, with match-filter Signal-to-Noise Ratio ($MSNR$) ranges: 4-5, 5-6, 6-7, 7-8, and >8, were employed in the data simulation. Compared to previous work that utilized data from single sub-detector data (SSDD), the detection accuracy from TSDCD has shown substantial improvements, increasing from $60\%$, $60.5\%$, $84.5\%$, $94.5\%$ to $78.5\%$, $84\%$, $99.5\%$, $100\%$, and $100\%$ for sources with $MSNR$ of 4-5, 5-6, 6-7, 7-8, and >8, respectively. The ResNet model was evaluated on the first Einstein Telescope mock Data Challenge (ET-MDC1) dataset, where the model demonstrated strong performance in detecting BBH mergers, identifying 5,566 out of 6,578 BBH events, with optimal SNR starting from 1.2, and a minimum and maximum $D_{L}$ of 0.5 Gpc and 148.95 Gpc, respectively. Despite being trained only on BBH mergers without overlapping sources, the model achieved high BBH detection rates. Notably, even though the model was not trained on BNS and BHNS mergers, it successfully detected 11,477 BNS and 323 BHNS mergers in ET-MDC1, with optimal SNR starting from 0.2 and 1, respectively, indicating its potential for broader applicability.
Summary
PyMerger is a Python tool that uses a Deep Residual Neural Network (ResNet) to detect binary black hole (BBH) mergers from the Einstein Telescope gravitational wave detector. The model was trained on combined data from all three proposed ET sub-detectors (TSDCD), achieving substantially improved detection accuracy compared to single sub-detector approaches—reaching 78.5-100% accuracy across different signal-to-noise ratio ranges. When evaluated on the Einstein Telescope mock Data Challenge dataset, the model identified 5,566 out of 6,578 BBH events and unexpectedly demonstrated strong generalization by detecting BNS and BHNS mergers despite not being trained on them.
Cited by 1 page
| Page | Type | Quality |
|---|---|---|
| AI Proliferation | Risk | 60.0 |
Cached Content Preview
[2310.10409] Einstein Telescope: binary black holes gravitational wave signals detection from three detectors combined data using deep learning
1 1 institutetext: Particle Astrophysics Science and Technology Centre, Nicolaus Copernicus Astronomical Center,
Rektorska 4, 00-614 Warsaw, Poland
2 2 institutetext: Astronomical Observatory, University of Warsaw, Aleje Ujazdowskie 4, 00-478 Warsaw, Poland
Einstein Telescope: binary black holes gravitational wave signals detection from three detectors combined data using deep learning
Wathela Alhassan
,
E-mail: wathelahamed@gmail.com11
T. Bulik
1122
M. Suchenek
11
(Received October 15, 2023; accepted March 16, 1997)
Abstract
Context. Continuing from our prior work (Alhassan et al. 2022 ) , where a single detector data of the Einstein Telescope (ET) was evaluated for the detection of binary black hole (BBHs) using deep learning (DL).
Aims. In this work we explored the detection efficiency of BBHs using data combined from all the three proposed detectors of ET, with five different lower frequency cutoff ( F l o w subscript 𝐹 𝑙 𝑜 𝑤 F_{low} ): 5 Hz, 10 Hz, 15 Hz, 20 Hz and 30 Hz, and the same previously used SNR ranges of: 4-5, 5-6, 6-7, 7-8 and ¿8.
Methods. Using ResNet model (which had the best overall performance on single detector data), the detection accuracy has improved from 60 % percent 60 60\% , 60.5 % percent 60.5 60.5\% , 84.5 % percent 84.5 84.5\% , 94.5 % percent 94.5 94.5\% and 98.5 % percent 98.5 98.5\% to 78.5 % percent 78.5 78.5\% , 84 % percent 84 84\% , 99.5 % percent 99.5 99.5\% , 100 % percent 100 100\% and 100 % percent 100 100\% for sources with SNR of 4-5, 5-6, 6-7, 7-8 and ¿8 respectively.
Results. The results show a great improvement in accuracy for lower SNR ranges: 4-5, 5-6 and 6-7 by 18.5 % percent 18.5 18.5\% , 24.5 % percent 24.5 24.5\% , 13 % percent 13 13\% respectively, and by 5.5 % percent 5.5 5.5\% and 1.5 % percent 1.5 1.5\% for higher SNR ranges: 7-8 and ¿8 respectively. In a qualitative evaluation, ResNet model was able to detect sources at 86.601 Gpc, with 3.9 averaged SNR (averaged SNR from the three detectors) and 13.632 chirp mass at 5 Hz. It was also shown that the use of the three detectors combined data is appropriate for near-real-time detection, and can be significantly improved using more powerful setup.
Key Words.:
gravitational waves –
instrumentation: detectors –
methods: data analysis
1 Introduction
This is a continuation of our previous work (Alhassan et al. 2022 ) (WTM1 hereafter) on the detection of binary black holes (BBHs) gravitational wave (GWs) signals from the Einstein Telescope (ET) (Punturo et al. 2010 ; Abernathy et al. 2011 ; Maggiore et al. 2020 ) using deep learning (DL). ET is designed 1 1 1 See https://www.et-gw.eu/index.php/relevant-et-documents for relevant documents on the ET design. to be an underground GWs detector,
... (truncated, 28 KB total)4303919b455b8d6d | Stable ID: sid_CKUts1gXJV