pespace: a tool of response generation and likelihood evaluation for space-borne gravitational wave detectors

last commit code style: black docs Publish Python package

Warning

This is an experimental project under active development. The design and APIs are not stable and may change frequently.

This package can be used to generate detector responses and evaluate the likelihood function under the stationary Gaussian noise assumption for space-borne detectors, with a focus on parameter estimation of massive black hole binaries. Core computations are implemented with taichi-lang and jax, enabling automatic differentiation and hardware acceleration across multiple architectures. More details can be found in the paper or the document.

Installation

Install from PyPI:

pip install pespace

Install the latest or specific commit version:

# install the latest development version
pip install git+https://github.com/nnrui/pespace
# install a specific commit
pip install git+https://github.com/nnrui/pespace@<commit-hash>

By default, only the basic dependencies required by the taichi-lang backend for response and likelihood computations are installed. If (backend="jax") is used, install it additionally:

pip install "pespace[jax]"

No samplers are installed by default either, please manually install them according to the needs. To install multiple currently supported samplers in one go:

pip install "pespace[samplers]"

Note

We have not yet successfully obtained reliable posteriors sampling with flowMC and jaxns in our tests. The configurations provided in the example scripts should be regarded as preliminary and have not been fully validated.

To use the sampler interface provided by bilby:

pip install "pespace[bilby]"

When the jax backend is used, the waveform relies on a customized ripple fork (import ripplegw) which is not published on PyPI and need to be installed manually:

pip install "rippleGW @ git+https://github.com/nnrui/ripple"

Usage

The basic functionality of generating detector responses is demonstrated in the tutorial. Example scripts of the full Bayesian parameter estimation for a massive black hole binary merger signal can be found here for a single LISA-like detector, and here for the LISA-Taiji-Tianqin network.

Similar Packages

If pespace cannot meet your needs, you may find other packages for similar functionality (welcome to open issues or pull requests if you have more). But differences in conventions need to be noted.

Contact

The author strive to make this project easy-to-use and maintainable. But the author’s experience and knowledge in software engineering is limited. Any feedback, comments, and suggestions are greatly appreciated. Feel free to open issues or contact.

Citation

If you think this package is useful, please consider cite arxiv: 2602.12011.

The development of this package depends on many previous works. Please cite the original works for the corresponding modules you have used.

The frequency domain response model:

  • Sylvain Marsat, John G. Baker, Fourier-domain modulations and delays of gravitational-wave signals, arxiv: 1806.10734;

The adaptive tempering SMC with NUTS and Hessian: