Classification methods for noise transients in advanced gravitational-wave detectors II: performance tests on Advanced LIGO data. (13th January 2017)

• Record Type: Journal Article
• Authors: Powell, Jade; Torres-Forné, Alejandro; Lynch, Ryan; Trifirò, Daniele; Cuoco, Elena; Cavaglià, Marco; Heng, Ik Siong; Font, José A
• Is Part Of: Classical and quantum gravity. Volume 34:Number 3(2017:Feb.)
• Publication Date: 2017-01-13
• Subjects: gravitational waves -- LIGO -- machine learning; Quantum gravity -- Periodicals; Gravitation -- Periodicals; Relativity (Physics) -- Periodicals; Space and time -- Periodicals; Periodicals; 521.1
• DOI: 10.1088/1361-6382/34/3/034002
• Languages: English
• ISSNs: 0264-9381

Abstract

Abstract: The data taken by the advanced LIGO and Virgo gravitational-wave detectors contains short duration noise transients that limit the significance of astrophysical detections and reduce the duty cycle of the instruments. As the advanced detectors are reaching sensitivity levels that allow for multiple detections of astrophysical gravitational-wave sources it is crucial to achieve a fast and accurate characterization of non-astrophysical transient noise shortly after it occurs in the detectors. Previously we presented three methods for the classification of transient noise sources. They are Principal Component Analysis for Transients (PCAT), Principal Component LALInference Burst (PC-LIB) and Wavelet Detection Filter with Machine Learning (WDF-ML). In this study we carry out the first performance tests of these algorithms on gravitational-wave data from the Advanced LIGO detectors. We use the data taken between the 3rd of June 2015 and the 14th of June 2015 during the 7th engineering run (ER7), and outline the improvements made to increase the performance and lower the latency of the algorithms on real data. This work provides an important test for understanding the performance of these methods on real, non stationary data in preparation for the second advanced gravitational-wave detector observation run, planned for later this year. We show that all methods can classify transients in non stationary data with a high level of accuracy and show the benefits of using multiple classifiers.