Simulation-based Inference for Gravitational-waves from Intermediate-Mass Binary Black Holes in Real Noise

TL;DR

This paper explores using simulation-based inference with real detector noise to improve parameter estimation of intermediate-mass binary black holes in gravitational-wave data, showing promising bias reduction.

Contribution

It introduces a flow-matching posterior estimation method trained on real noise for gravitational-wave parameter inference, demonstrating potential improvements over traditional techniques.

Findings

Significant reduction in measurement bias for black hole parameters.

Effective posterior estimation using real detector noise.

Highlighting the need for further development for practical application.

Abstract

We present an exploratory investigation into using Simulation-based Inference techniques, specifically Flow-Matching Posterior Estimation, to construct a posterior density estimator trained using real gravitational-wave detector noise. Our prototype estimator is trained on a 9-dimensional space, and for training efficiency outputs posterior probability distributions for the binary black holes chirp mass and mass ratio. We use this prototype estimator to investigate possible effects on parameter estimation for Intermediate-Mass Binary Black Holes, and show statistically significant reduction in measurement bias. Although the results show potential for improved measurements, they also highlight the need for further work.