A novel Machine Learning pipeline for vetoing and denoising on GW interferometer data leveraging Generative Neural Networks to map carefully selected auxiliary channels (sensitive to the noise but not to GW signals) into the main channel of the interferometer. The generated output is then subtracted from the interferometer main channel data.
The module allows for visualization of original, generated and cleaned data, as well as metadata relative to veto flags and potentially failed to be cleaned data such as the area and Signal-to-Noise Ratio (SNR) of glitches still visible after cleaning.
Release Notes
The final release of the module is available at https://github.com/interTwin-eu/DT-Virgo-notebooks/tree/main/Final_Release/GlitchFlow. The module is implemented as an itwinai plugin, and is installable as pip install itwinai-virgo
Future Plans
The ultimate goal is to integrate the DT in the low latency pipeline at Virgo and Einstein Telescope.
We also plan to extend the module to work with time domain data instead of time-frequency domain for better integrability with the rest of downstream data analysis tools in the interferometer pipeline.
