The neural network included in this repository predicts the mismatch between the gravitational waves emitted from any two binary black hole systems with symmetric mass ratio,
The network and its uses are described in the paper by Deborah Ferguson at https://arxiv.org/abs/2209.15144.
This repository includes two jupyter notebooks to serve as examples of how to use this network.
predict_mismatch.ipynb: predict the mismatch between two binary black hole systems by setting their initial parameters and feeding them into the network.
predict_many_mismatches: predict the mismatches between many pairs of binary systems by setting their initial parameters and feeding multiple pairs into the network at once.
The network inputs the initial parameters that describe two binary black hole systems.
This includes their symmetric mass ratio (
The mismatch between two gravitational waves is defined as:
where
and
This network was built using tensorflow. It is fully connected, consisting of 15 hidden layers with 56 nodes per layer, as described in the figure below.
It has been trained on the SXS public catalog of numerical relativity waveforms (https://data.black-holes.org/waveforms) using the
If you use this network, please cite Ferguson 2022 (https://arxiv.org/abs/2209.15144).
Using this network will require tensorflow and numpy.
A requirements.txt is included in this repository. You can use this to pip install all the necessary libraries with the command:
pip install -r requirements.txt