Machine learning algorithms for transient signal classification for gravitational wave astronomy

Scope

Gravitational wave observations are limited by a background of transient signals from instrumental and environmental origin.

This package includes a set of machine learning tools that allow to classify those transient signals, in order to better characterize their large population, give hints about their source, and provide new ways for mitigating this background. The algorithms take a labelled set of transients, extract features from the time series, and learn a classifier (neural network) using standard ML libraries.

While this toolbox is intended for gravitational wave data specifically, it is general enough to be adapted to problems in other fields.

The package will be released later this year (2018).

Contact person

Eric Chassande-Mottin, Michal Bejger APC/CNRS IN2P3