Bayesian ICA-based source separation of Cosmic Microwave Background by a discrete functional approximation

TL;DR

This paper introduces a fast Bayesian source separation method using a functional approximation for analyzing Cosmic Microwave Background data, enabling efficient full-sky map reconstructions.

Contribution

It presents a novel functional approximation technique for Bayesian source separation, significantly reducing computational time compared to traditional MCMC methods.

Findings

Orders of magnitude faster than MCMC approaches

Effective full-sky map reconstructions at WMAP and Planck scales

Modeling spatial smoothness with Gaussian Markov random fields

Abstract

A functional approximation to implement Bayesian source separation analysis is introduced and applied to separation of the Cosmic Microwave Background (CMB) using WMAP data. The approximation allows for tractable full-sky map reconstructions at the scale of both WMAP and Planck data and models the spatial smoothness of sources through a Gaussian Markov random field prior. It is orders of magnitude faster than the usual MCMC approaches. The performance and limitations of the approximation are also discussed.