Identifying Witnesses to LIGO Glitches Using Auxiliary Channels

Other Versions:

Abstract:: The exquisite sensitivity of the Laser Interferometer Gravitational-Wave Observatory (LIGO) makes it extremely vulnerable to glitches, which are short-lived noise transients that surpass the Gaussian-distributed background noise. Glitches occur in the main channel at a high rate, in comparison to real signals. As a result, these triggers can mask or mimic real gravitational wave signals derived from astrophysical sources resulting in high false positive rate in the search pipelines. Along with the main channel, LIGO maintains a large set of auxiliary channels. These auxiliary channels monitor the state of the detector and can witness these glitches. In this work, we present an exploration of Boolean Matrix Factorization (BMF) as a possible method for clustering loud triggers in a set of auxiliary channels that coincide with glitches and selecting certain channels as glitch witnesses. We test the factorization against the real data and simulated data, to determine how BMF performs in comparison to other factorization models such as the Non-Negative Matrix Factorization (NMF). By doing so, our goal is to examine if some Gravity Spy glitch classes have consistently correlated loud triggers in certain sets of channels which may allow domain experts to localize and fix glitches at their source.

Files in Document:

Other Files:

LIGO Document T2200207-v3