Utilizing Machine Learning to Characterize Transient Noise in Advanced LIGO

POSTER

Abstract

LIGO searches for gravitational waves in two detectors at Hanford, WA and Livingston, LA. Glitches are transient noises that contaminate data by elevating background noise levels, and some glitches may even mimic real gravitational wave signals. In this study we investigate the behavior of glitches during the first part of the fourth observation run (O4a) at LIGO Hanford. Glitches are categorized based on their morphologies in a time-frequency spectrogram. We analyze glitches during O4a at LIGO-Hanford (H1) using t-SNE (t-distributed Stochastic Neighbor Embedding), a machine learning technique that reduces the dimensions of a data vector, providing useful visualization by clustering data points according to features present in the data. We look at differences in the clustering over time which helps to identify the origin of transient noises.

*We acknowledge support from the National Science Foundation from its REU Site in Physics and Astronomy (NSF Grant No. 2150445) at Louisiana State University, as well as the support from the LIGO Scientific Collaboration for providing data and tools used in this study.

Publication: [1] Comparison between t-SNE and cosine similarity, Tabata Aira Ferreira and Cesar Augusto Costa 2022 Class. Quantum Grav. 39 165013
[2] Advanced LIGO, The LIGO Scientific Collaboration et al 2015 Class. Quantum Grav. 32 074001
[3] Observation of Gravitational Waves from a Binary Black Hole Merger, The LIGO Scientific Collaboration et al., Phys. Rev. Lett. 116, 061102 (2016)
[4] LIGO detector characterization in the second and third observing runs, D Davis et al 2021 Class. Quantum Grav. 38 135014
[5] A guide to LIGO–Virgo detector noise and extraction of transient gravitational-wave signals, B P Abbott et al 2020 Class. Quantum Grav. 37 055002

Presenters

  • Osvaldo Salas

    • Austin College

Authors

  • Osvaldo Salas

    • Austin College