Optimisation of Gravitational-Wave Data Analysis: Employing Machine Learning in Searches for Cosmic Strings and Binary Black Holes

Gravity is the phenomenon of mutual attraction between masses, and it is gravity that determines the trajectories of masses. Examples include planets in orbit, or humans walking the Earth. General relativity, the famous theory of gravity by Einstein, posits that nothing can move faster than light. At almost 300 million meters per second, this speed relates distance to time, and therefore so must the model of gravity. This leads to the concept of spacetimes, where trajectories in time and space can be calculated based on gravity. General relativity further states that gravity is modelled through the shape of spacetime, dictated by mass. Since mass can move, this shape is subject to change and can even show waves, aptly called gravitational waves. These waves can for instance be generated by two merging black holes in a binary black hole, or hypothetical objects called cosmic strings. Gravitational-wave signals can then be measured using extremely sensitive detectors, with the unfortunate side effect of also registering large amounts of noise and glitches. In this thesis new methods are developed to optimise the analysis of detector data, refining gravitational-wave searches. Machine learning models are proposed that can accurately distinguish signals from cosmic strings or specific binary black holes from glitches, where their similarity currently confuses pipelines. This mitigation can potentially greatly increase search efficiency. Additionally, machine learning algorithms that can significantly accelerate gravitational-wave signal prediction are studied. Such improvements can mean saving full days in searches, or even weeks for future detectors.