http://arxiv.org/abs/2205.09204
In the field of gravitational-wave (GW) interferometers, the most severe limitation to the detection of GW signals from astrophysical sources comes from random noise, which reduces the instrument sensitivity and impacts the data quality. For transient searches, the most problematic are transient noise artifacts, known as glitches, happening at a rate around $1 \text{min}^{-1}$. As they can mimic GW signals, there is a need for better modeling and inclusion of glitches in large-scale studies, such as stress testing the searches pipelines and increasing confidence of a detection. In this work, we employ Generative Adversarial Networks (GAN) to learn the underlying distribution of blip glitches and to generate artificial populations. Taking inspiration from the field of image processing, we implement Wasserstein GAN with consistency term penalty for the generation of glitches in time domain. Furthermore, we share the trained weights through the \texttt{gengli}, a user-friendly open-source software package for fast glitch generation and provide practical examples about its usage.
M. Lopez, V. Boudart, S. Schmidt, et. al.
Fri, 20 May 22
18/65
Comments: 6 pages, 5 figures. arXiv admin note: text overlap with arXiv:2203.06494
You must be logged in to post a comment.