My PhD thesis will focus on designing a machine learning based approach to detect transient noise artefacts called 'glitches' in gravitational wave (GW) data for the LISA mission. Glitches limit the precision of parameter estimates from GW data, reduce the effective observation time of GW detectors, and may be mistaken for GWs from unknown sources. Thus, it is crucial to have reliable methods for detecting and, where possible, removing them from experimental data.
My project will draw upon the methods used so far to classify and model glitches from the LIGO and LISA Pathfinder (LPF) experiments. I aim to build a classification system (illustrated in Figure 1) that can detect glitches in LISA data, sort them into groups that correspond to distinct physical causes, and then mitigate their effect on the parameter estimate uncertainty for GW sources. I will need to incorporate inputs from multiple data channels, and the system will need to be able to work with an unbalanced, unlabelled dataset. The success of my work can be judged based on how well it performs on glitches in the blind datasets produced for the LISA Data Challenge.