Resolving Galactic binaries in LISA data using particle swarm optimization and cross-validation

TL;DR

This paper introduces an iterative method using particle swarm optimization and cross-validation to resolve thousands of Galactic binary signals in LISA data, effectively reducing noise and spurious sources.

Contribution

The novel combination of PSO with cross-validation significantly improves the accuracy of source estimation in LISA data analysis.

Findings

Successfully identified around 10,000 binaries in mock data

Reduced residual noise to instrumental noise level at higher frequencies

Demonstrated effectiveness in a simulated data challenge

Abstract

The space-based gravitational wave (GW) detector LISA is expected to observe signals from a large population of compact object binaries, comprised predominantly of white dwarfs, in the Milky Way. Resolving individual sources from this population against its self-generated confusion noise poses a major data analysis problem. We present an iterative source estimation and subtraction method to address this problem based on the use of particle swarm optimization (PSO). In addition to PSO, a novel feature of the method is the cross-validation of sources estimated from the same data using different signal parameter search ranges. This is found to greatly reduce contamination by spurious sources and may prove to be a useful addition to any multi-source resolution method. Applied to a recent mock data challenge, the method is able to find $O(10^4)$ Galactic binaries across a signal frequency range of $[0.1,15]$ mHz, and, for frequency $\gtrsim 4$ mHz, reduces the residual data after subtracting out estimated signals to the instrumental noise level.