Coherence monitoring and broadband denoising in GW detectors

There are multiple sources of noise beyond the fundamental ones that impact the sensitivity of LIGO. This includes instrumental and environmental such sources that may couple to the gravitational-wave measurement in a linear or non-linear way. When such a noise source is witnessed by auxiliary channels (i.e., other than the GW one), it presents an opportunity to establish its presence and variation via coherence measurements as well as potentially regress it out of the GW channel. In this presentation, we present our work on first identifying these couplings with coherence monitoring and then cleaning the broadband noise using a convolutional neural network algorithm we call DeepClean. We discuss how this cleaning pipeline is automated and how it improves the sensitivity of the instruments in various frequency bands where such noise currently exists.