Deep learning network to distinguish binary black hole signals from short-duration noise transients

• DOI: 10.1103/physrevd.107.024030
• Publication Date: 2023-01-24T15:14:37Z
• Authors: Sunil Choudhary; Anupreeta More; Sudhagar Suyampra...; Sukanta Bose
• Keywords: Physics - Data Analysis, Statistics and Probability; 0103 physical sciences; FOS: Physical sciences; General Relativity and Quantum Cosmology (gr-qc); Astrophysics - Instrumentation and Methods for Astrophysics; Instrumentation and Methods for Astrophysics (astro-ph.IM); 01 natural sciences; General Relativity and Quantum Cosmology; Data Analysis, Statistics and Probability (physics.data-an)

Blip glitches, a type of short-duration noise transient in the LIGO--Virgo data, are nuisance for binary black hole (BBH) searches. They affect BBH search sensitivity significantly because their time-domain morphologies very similar, and that creates difficulty vetoing them. In this work, we construct deep-learning neural network to efficiently distinguish signals from blip glitches. We introduce sine-Gaussian projection (SGP) maps, which projections GW frequency-domain data snippets on basis sine-Gaussians defined by quality factor central frequency. feed SGP maps our network, classifies blips. Whereas simulated, blips used taken real throughout analysis. show improves identification comparison results obtained using traditional-$\chi^2$ $\chi^2$. For example, 75% at false-positive rate $10^{-2}$ BBHs with total mass range $[80,140]~M_{\odot}$ SNR $[3,8]$. Also, it correctly identifies 95% events GWTC-3. The computation time classification is few minutes thousands single core. With further optimisation next version algorithm, expect reduction computational cost. Our proposed method can potentially improve veto process analysis conceivably support identifying low-latency pipelines.