Abstract:We present an accelerated pipeline, based on high-performance computing techniques and normalizing flows, for joint Bayesian parameter estimation and model selection and demonstrate its efficiency in gravitational wave astrophysics. We integrate the Jim inference toolkit, a normalizing flow-enhanced Markov chain Monte Carlo (MCMC) sampler, with the learned harmonic mean estimator. Our Bayesian evidence estimates run on $1$ GPU are consistent with traditional nested sampling techniques run on $16$ CPU cores, while reducing the computation time by factors of $5\times$ and $15\times$ for $4$-dimensional and $11$-dimensional gravitational wave inference problems, respectively. Our code is available in well-tested and thoroughly documented open-source packages, ensuring accessibility and reproducibility for the wider research community.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to accelerate Bayesian parameter estimation and model selection in gravitational - wave astrophysics. Specifically, the authors propose an accelerated pipeline based on high - performance computing techniques and normalizing flows to achieve joint Bayesian parameter estimation and model selection, and demonstrate its efficiency in gravitational - wave astrophysics. ### Main problems 1. **Computational efficiency**: Although the traditional nested sampling method can accurately calculate Bayesian evidence, the computational cost is very high in the case of high - dimensional parameter spaces or multimodal posterior distributions. Therefore, a faster and more scalable method is needed to accelerate Bayesian inference. 2. **Bayesian evidence estimation**: Bayesian evidence is a key quantity for model comparison, but it is very difficult to calculate. Existing methods such as nested sampling are tightly coupled with sampling strategies, which limits the application of acceleration techniques. Therefore, an efficient Bayesian evidence estimation method independent of sampling strategies is required. ### Solutions The authors solve the above problems by combining the following two key components: 1. **JIM inference toolkit**: JIM is an inference toolkit based on JAX. It uses the MCMC sampling method enhanced by normalizing flows and utilizes hardware accelerators such as GPUs, significantly improving the speed of parameter estimation. 2. **Learned Harmonic Mean Estimator**: This estimator is independent of sampling strategies and can be used in combination with JIM, thereby significantly reducing the calculation time of Bayesian evidence while maintaining accuracy. ### Experimental verification The authors carried out experimental verification through simulated gravitational - wave signals, demonstrating the effectiveness and acceleration effect of their pipeline in 4 - dimensional and 11 - dimensional parameter spaces. The results show that, at the same accuracy, the new method is 5.4 times and 14.8 times faster than the traditional method respectively. ### Conclusion This paper successfully constructs an end - to - end accelerated pipeline that can provide accurate Bayesian evidence estimation while significantly reducing the calculation time, thus providing an efficient solution for parameter estimation and model selection in gravitational - wave astrophysics. In addition, this method can also be applied to other scientific fields that rely on Bayesian inference.

Accelerated Bayesian parameter estimation and model selection for gravitational waves with normalizing flows

Accelerated nested sampling with $β$-flows for gravitational waves

Advancing space-based gravitational wave astronomy: Rapid parameter estimation via normalizing flows

Massively parallel Bayesian inference for transient gravitational-wave astronomy

Gravitational-wave parameter estimation with autoregressive neural network flows

Rapid parameter estimation for pulsar-timing-array datasets with variational inference and normalizing flows

Rapid Parameter Estimation for Merging Massive Black Hole Binaries Using Continuous Normalizing Flows

Cosmological Inference using Gravitational Waves and Normalising Flows

Rapid Likelihood Free Inference of Compact Binary Coalescences using Accelerated Hardware

Fast Parameter Inference on Pulsar Timing Arrays with Normalizing Flows

An efficient pipeline for joint gravitational wave searches from individual binaries and a gravitational wave background with Hamiltonian sampling

Efficient pipeline for joint gravitational wave searches from individual binaries and a gravitational wave background with Hamiltonian sampling

Improving the scalability of Gaussian-process error marginalization in gravitational-wave inference

Parallelized Inference for Gravitational-Wave Astronomy

Gravitational wave populations and cosmology with neural posterior estimation

Efficient parameter inference for gravitational wave signals in the presence of transient noises using temporal and time-spectral fusion normalizing flow

Accelerating global parameter estimation of gravitational waves from Galactic binaries using a genetic algorithm and GPUs

Accelerated parameter estimation in Bilby with relative binning