The low-latency characterization of detector noise is a crucial step in the detection of gravitational waves. In particular, a rapid classification and identification of transient noise sources, usually referred to as glitches, is very important when candidate signals are sent as gravitational alerts to the astronomical community. Machine learning is emerging as a promising alternative to standard methodologies used so far for the data characterization in the gravitational wave community. In particular, deep learning approach looks very promising in tackling the problem of rapid classification of noise transients in the second-generation interferometers like Advanced LIGO and Advanced Virgo. We will then discuss some possible approaches for establishing the quality of data, reducing the noise and classifying transient noise sources. We will also present some results based on simulated and real data, showing the performance of deep learning and its feasibility as a new and efficient approach to data characterization in gravitational wave interferometers. At the same time, we will show how to use machine learning techniques to search for unmodeled or unknown signals.
Machine Learning for the Characterization of Gravitational Wave Data
Massimiliano Razzano;
2021-01-01
Abstract
The low-latency characterization of detector noise is a crucial step in the detection of gravitational waves. In particular, a rapid classification and identification of transient noise sources, usually referred to as glitches, is very important when candidate signals are sent as gravitational alerts to the astronomical community. Machine learning is emerging as a promising alternative to standard methodologies used so far for the data characterization in the gravitational wave community. In particular, deep learning approach looks very promising in tackling the problem of rapid classification of noise transients in the second-generation interferometers like Advanced LIGO and Advanced Virgo. We will then discuss some possible approaches for establishing the quality of data, reducing the noise and classifying transient noise sources. We will also present some results based on simulated and real data, showing the performance of deep learning and its feasibility as a new and efficient approach to data characterization in gravitational wave interferometers. At the same time, we will show how to use machine learning techniques to search for unmodeled or unknown signals.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
