Dataset generation for gravitational-wave physics

This repo is meant to generate data for ML tasks in gravitational wave physics. It is based on ml4gw and primarily uses its functionality.

To generate data, run

python main.py --config config.yaml --data path/to/data/folder --out path/to/output/directory

with

• config: This is the config file containing all general and signal-specific setups. In the config folder, you can find example configs for BBH and BNS signals.
• data: path to directory containing open (background) data. The data can be downloaded using the load_data.py script.
• out: output directory where to store the dataset.

To update the number of generated signal and background events, change the num_waveforms parameter in the config.yaml file.

The generated output are sig.h5 and bkg.h5 files, which contain the time series for the two detectors (data) to be used for classification. The sig.h5 files additionally contain all parameters defining the signal waveform to be used in regression tasks or for more detailed studies.