This working group aims at investigating Machine Learning (ML) techniques to classify Gravitational Wave (GW) signals, recognise noise and disturbances from the instrument, as well as identify GW signals from known mechanisms and GW signals from yet unknown mechanisms. More specifically, the goal is to design problem-specific ML pipelines by considering ready-to-use algorithms for solving classical ML tasks (clustering, classification and regression), specific data structures for the development of novel ML algorithms, as well as specific frameworks for the development of GPU-based ML solutions.

Gravitational wave detectors are complex instruments and their data reflect this complexity. The Gravitational Wave Open Science Center ( https://www.gw-openscience.org ) is the web page where you find the data along with documentation, tutorials, and online tools for finding and viewing data.

In particular we suggest to have a look at the tutorial pages:
( https://www.gw-openscience.org/tutorials/ ) where you can find web courses (https://www.gwopenscience.org/static/workshop2/course.html ) containing lecture videos and material for hands-on sessions. If you are interested in glitches you can have a look at the material we prepared for the school in Braga (all the links can be found in the presentation link https://indico.lip.pt/event/557/contributions/1585/attachments/1518/1890/introduction_data_challenge.pdf.)

WG1 IS STRUCTURED IN SEVERAL TASKS AND SUBTASKS IN THE FOLLOWING WAY:

We have confirmed the following 3 tasks and defined their leaders.

1. Access to the detectors data – main and auxiliary data (Task Leaders: Agata Trovato, Luca Rei)
(a) Preparing the exemplary data sets.
(b) Exploration of alternative representation of data.
(c) Spectrograms, Wavelet transforms, Q-transforms, Time series, Fourier
Transform data, Other (quadratic) time-frequency representations of time-series alternative to spectrograms.

2. Detection and classification/characterization of GW data (Task Leaders: Massimiliano Razzano, Eftim Zdravevski)
(a) HPC/large scale computing, Origin of glitches, discovery of new classes of glitches, continuous signals (GW and data characterization).

3. Alternative data denoising approaches (Task Leaders: Alejandro Torres-Forne, Jonathan Lerga, Toni Font)