blackjax-gw

Gravitational wave parameter estimation using BlackJAX nested sampling and Jim gravitational wave analysis tools.

Core Dependencies

This project requires two main repositories:

1. BlackJAX (nested sampling branch): git+https://github.com/handley-lab/blackjax@nested_sampling

   • Provides Bayesian nested sampling algorithms
   • Used for parameter estimation and evidence calculation

2. Jim: git+https://github.com/kazewong/jim

   • Gravitational wave analysis framework
   • Provides detectors (H1, L1), likelihood functions, waveforms, priors, and transforms

Installation

pip install uv
uv pip install "jax[cuda12_pip]"==0.4.31 -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html git+https://github.com/handley-lab/blackjax@nested_sampling git+https://git.ligo.org/lscsoft/ligo-segments.git git+https://github.com/kazewong/jim anesthetic

Usage

The main analysis script is ICML/run_GW150419.py, which performs parameter estimation on gravitational wave data using:

• RippleIMRPhenomD waveform model
• HeterodynedTransientLikelihoodFD likelihood
• 11-parameter prior (masses, spins, sky position, etc.)
• Custom periodic boundary handling for angular parameters

Jim Version Notes

• OLD VERSION: 03842792626c3a14ec8885f8076833324c391db2 (current codebase)
• NEW VERSION: 3d0116844d921ea7a035973137386d5bbb71c0d4 (updated)
• Ming's version: https://github.com/ming-256/jim/commits/main/

Jim provides:

• LIGO detector data loading
• Various likelihood implementations (phase marginalisation, heterodyning)
• Detector models for signal injection