Accelerated Sequential Posterior Inference via Reuse for Gravitational-Wave Analyses

Area of Science:

• Astrophysics
• Computational Science
• Statistical Modeling

Background:

• Gravitational-wave astronomy requires frequent reanalysis of events with new models.
• Existing methods are computationally intensive, demanding significant resources.
• Bayesian inference is crucial for interpreting complex astronomical data.

Purpose of the Study:

• To introduce a novel framework, ASPIRE, for accelerating Bayesian posterior inference.
• To enable efficient updates of existing Bayesian results without rerunning full analyses.
• To reduce computational costs and time in scientific reanalysis.

Main Methods:

• ASPIRE combines normalizing flows with a generalized sequential Monte Carlo (SMC) scheme.
• It reuses existing posterior samples and Bayesian evidence estimates.
• The framework allows for unbiased results under alternative models.

Main Results:

• ASPIRE reduces likelihood evaluations by up to 5.8x and wall times by 5.5x.
• It accurately reproduces full Bayesian results when switching waveform models or adding physical effects.
• Significant gains in efficiency are observed per posterior sample.

Conclusions:

• ASPIRE provides a statistically robust and computationally efficient method for Bayesian reanalysis.
• It facilitates systematic studies, scalable reanalyses of large datasets, and broad applicability across scientific domains.
• This framework accelerates discovery in fields like gravitational-wave astronomy.