LIGO Document T2300266-v2

Inferring Gravitational Wave Source Properties from Intermediate Pipeline Output with Machine Learning

Document #:
LIGO-T2300266-v2
Document type:
T - Technical notes
Other Versions:
Abstract:
The LIGO-Virgo-KAGRA collaboration provides low-latency (near-real time) localization using the signal-to-noise ratio measured for a single point in the search parameter space. Parameter estimation pipelines subsequently samples the full parameter space to obtain more accurate estimates of the localization. However, this process is computationally expensive. The multi-messenger detection of the binary neutron star merger GW170817 confirmed the need for accurate and fast data products. Some detection pipelines utilize singular value decomposition to reduce the filtering cost. This project uses machine learning to input signal-to-noise ratios from singular value decomposition time-series into simulation-based inference (SBI), a likelihood-free inference algorithm, which outputs a posterior with an accurate parameter estimation, such as a sky map, to localize compact binary coalescences and infer other source properties.
Files in Document:
Other Files:
Topics:
Keywords:
SURF23

DCC Version 3.4.2, contact DCC Help