Detecting and Denoising Gravitational Wave Signals from Binary Black Holes using Deep Learning

Abstract

We present a convolutional neural network, designed in the auto-encoder configuration that can detect and denoise astrophysical gravitational waves from merging black hole binaries, orders of magnitude faster than the conventional matched-filtering based detection that is currently employed at advanced LIGO (aLIGO). The Neural-Net architecture is such that it learns from the sparse representation of data in the time-frequency domain and constructs a non-linear mapping function that maps this representation into two separate masks for signal and noise, facilitating the separation of the two, from raw data. This approach is the first of its kind to apply machine learning based gravitational wave detection/denoising in the 2D representation of gravitational wave data. We applied our formalism to the first gravitational wave event detected, GW150914, successfully recovering the signal at all three phases of coalescence at both detectors. This method is further tested on the gravitational wave data from the second observing run (O2) of aLIGO, reproducing all binary black hole mergers detected in O2 at both the aLIGO detectors. The Neural-Net seems to have uncovered a pattern of 'ringing' after the ringdown phase of the coalescence, which is not a feature that is present in the conventional binary merger templates. This method can also interpolate and extrapolate between modeled templates and explore gravitational waves that are unmodeled and hence not present in the template bank of signals used in the matched-filtering detection pipelines. Faster and efficient detection schemes, such as this method, will be instrumental as ground based detectors reach their design sensitivity, likely to result in several hundreds of potential detections in a few months of observing runs.

Intelligent noise suppression for gravitational wave observational data

“…In case of the GW170823 (Fig. 5c), a BBH event with high chirp mass 29.2M ⊙ , both Bacon et al [25] and Murali et al [26] could recover the phase of original GW signal with certain cycles but failed to recover the complete evaluation in amplitude scale. In the contrast, we observe a clear match in the amplitude of peaks of the extracted GW170823 waveform, with an overlap of 96.95% and 99.00% for Handfold and Livingston, respectively.…”

“…Overlap and matched-filtering signal-to-noise (MFSNR) [37] are calculated to represent phase and amplitude recovery performance. We calculate the overlap over the same signal duration [24] for phase recovery and obtain the similar overlaps with [25,26]. With respect to the overlap distribution among the validation dataset for Handford O3b (more samiliar results in other observating runs are provided in Supplementary Materials), overlap is higher than 0.9 for most waveforms (Fig.…”

“…The majority of state-of-the-art machine learning algorithms, however, have trouble in dealing with real-world noise that tainted by non-stationary and non-Gaussian noise artifacts [21]. Alternatively, several studies [22][23][24][25][26][27] have employed the deep learning approach to denoise the GW signals from advanced LIGO noise. Although they can be used to enhance the quality of data in some aspects, none of them can be used to properly recover phase and amplitude information (lost in the normalization process) of GW waveforms while simultaneously reducing the amplitude of loud noise artifacts for overall data quality improvement.…”

“…Although they can be used to enhance the quality of data in some aspects, none of them can be used to properly recover phase and amplitude information (lost in the normalization process) of GW waveforms while simultaneously reducing the amplitude of loud noise artifacts for overall data quality improvement. Impressive progress on the large-scale deep learning model with billions of parameters, AlphaFold2 [28] for example, suggests that by scaling up data, model size, and training time in the right way, model performance [29,30] on difficult low signal-to-noise ratio (SNR) denoising task might be better than that of aforementioned million-parameter-scale deep learning [22][23][24][25][26] methods. This work aspires to contribute to these ongoing efforts by developing, for the first time, a billion-parameter-scale transformer-based model with highly-expressiveness background noise suppression of LIGO observations.…”

WaveFormer: transformer-based denoising method for gravitational-wave data

“…• We further investigated WaveFormer's capacity to recover signals from observational data in terms of phase and amplitude recovery. We achieved state-of-the-art accuracy compared with other deep learning methods [29][30][31][32]. On majority of the detected binary black hole (BBH) events, the phase overlaps are higher than 0.99 (1% error).…”