Gravitational wave denoising of binary black hole mergers with deep learning

Abstract

Gravitational wave detection requires an in-depth understanding of the physical properties of gravitational wave signals, and the noise from which they are extracted. Understanding the statistical properties of noise is a complex endeavor, particularly in realistic detection scenarios. In this article we demonstrate that deep learning can handle the non-Gaussian and non-stationary nature of gravitational wave data, and showcase its application to denoise the gravitational wave signals generated by the binary black hole mergers GW150914, GW170104, GW170608 and GW170814 from advanced LIGO noise. To exhibit the accuracy of this methodology, we compute the overlap between the time-series signals produced by our denoising algorithm, and the numerical relativity templates that are expected to describe these gravitational wave sources, finding overlaps O ∼ > 0.99. We also show that our deep learning algorithm is capable of removing noise anomalies from numerical relativity signals that we inject in real advanced LIGO data. We discuss the implications of these results for the characterization of gravitational wave signals.

Application of dictionary learning to denoise LIGO’s blip noise transients

“…The approaches are as diverse as Bayesian inference, machine learning, deep learning, and citizen science [14][15][16][17][18][19][20][21][22]. Recent examples of glitch mitigation are reported in [23][24][25][26]. Reference [23] describes various deglitching methods to extract the strong glitch present in the LIGO-Livingston detector about 1s before the merger of the binary neutron star that produced the signal GW170817 [4].…”

“…Glitch reduction, together with other techniques, was shown to improve the statistical significance of a GW trigger. In addition, deep learning approaches have also proven very effective to recover the true GW signal even in the presence of glitches [26].…”

“…For each glitch, the data correspond to a one-second window centered at the GPS time of the glitch as provided by Gravity Spy [16]. As in [23,24,26] the goal of our work is to mitigate the impact of glitches in GW data in order to increase the statistical significance of astrophysical triggers. Our results show that dictionary-learning techniques are able to model blip glitches and to subtract them from the data without significantly disturbing the background.…”

Data-driven Detection of Multimessenger Transients

“…When used for source detection, the inferred rate of false positives of these methods strongly depends on the completeness of the training data. For instance, one must verify that training includes all possible sources of systematics, including glitches (non-Gaussian noise; Wei & Huerta (2020); George et al (2018); Zevin et al (2017)). One must also take care that such systematics are distributed in realistic proportion to each other and to true signal events.…”

Gravitational-Wave Burst Signals Denoising Based on the Adaptive Modification of the Intersection of Confidence Intervals Rule

“…In [40], machine learning methods based on the dictionaries built from numerical-relativity templates of gravitational-wave signals were applied for data denoising, with satisfactory results for signals embedded in simulated Gaussian noise and some promising results for application on real gravitational-wave signals. Deep learning was applied in [41,42] for the noise reduction in the gravitational-wave detector data. The authors in [43] proposed deep filtering, which utilized deep learning with convolutional neural networks (CNNs) for detection and parameter estimation of gravitational-waves from BBH mergers, with signals being embedded in actual LIGO noise.…”