Accelerating parameter estimation of gravitational waves from compact binary coalescence using adaptive frequency resolutions

Abstract

Bayesian parameter estimation of gravitational waves from compact binary coalescence (CBC) typically requires more than millions of evaluations of computationally expensive template waveforms. We propose a technique to reduce the cost of waveform generation by exploiting the chirping behavior of CBC signal. Our technique does not require waveforms at all frequencies in the frequency range used in the analysis, and does not suffer from the fixed cost due to the upsampling of waveforms. Our technique speeds up the parameter estimation of typical binary neutron star signal by a factor of $\mathcal{O}(10)$ for the low-frequency cutoff of $20\,\mathrm{Hz}$, and $\mathcal{O}(10^2)$ for $5\,\mathrm{Hz}$. It does not require any offline preparations or accurate estimates of source parameters provided by detection pipelines.