An Insight on Gravity Spy dataset and Machine Learning techniques for Glitch Classification in Gravitational Waves

Abstract

Gravitational waves are distortions in the space-time continuum produced when heavenly bodies interact. In 1916,as per relativity theory, Albert Einstein hypothesized the presence of gravitational waves. These waves are difficult to detect, yet they contain an abundance of data about the things that generated them. LIGO which stands for ”Laser Interferometer Gravitational-Wave Observatory” discovered gravitational waves for the first time in 2015, begetting more research and analysis. By analysing the gravitational waves released by the earlier objects in the universe, cosmologists can gain insight into the conditions of the universe that prevailed after the Big Bang. These Waves can also be used to study how matter is distributed in Universe and to discover new objects such as dark matter, exoplanets and black holes. The existence of glitches, which are noise transients, affects the processing of GW data. If some glitches excite the detector at the same frequency as gravitational waves, they may impede analysis. In order to recover segments of gravitational wave signals that coincide with a glitch, it is necessary to correctly interpret glitches. Classification of these glitches is also very important as it tells the origin of the glitches and hence help in their removal from the gravitational wave. The recent developments in the disciplines of Data Science and Artificial Intelligence have unveiled new and robust detection and analytical tools. This work includes a survey report on the identification and classification of glitches by different researchers, as well as critical insights and analyses on the topic.This work at its first part explores the dataset - Gravity Spy where we have identifies few issues which may be unknowingly have not been identified in any of the work we have studied for this paper.On the other side, it is demonstrated with illustrations, that methodologies studied thus far may fail when applied to glitches belonging to new classes and new unknown glitch images.