Denoising Time-Series Data from Gravitational Wave Detectors with Autoencoders based on Deep Recurrent Neural Networks

Extracting gravitational waves whose amplitude is much smaller than the background noise and inferring accurate parameters of their sources in real-time is crucial in enabling multimessenger astrophysics. We are working on reducing the noise in the data from the LIGO detectors and extracting gravitational wave signals using denoising auto-encoders based on recurrent neural networks. This will be used as a preprocessing step to improve the accuracy of our Deep Filtering algorithm as well as that of existing detection and parameter estimation pipelines in gravitational wave detectors.

Project Members: Hongyu Shen, Daniel George, Eliu Huerta

Collaborators: Zhizhen Zhao (U. Illinois)

Funding: NCSA Faculty Fellow Program 2017-2018