CRUSH: fast and scalable data reduction for imaging arrays

TL;DR

CRUSH is a scalable, efficient data reduction method for submillimeter imaging arrays that uses iterative statistical estimators to separate source signals from noise, suitable for large datasets.

Contribution

It introduces a novel, parallelizable data analysis approach tailored for large-scale submillimeter imaging arrays, with adjustable filtering properties.

Findings

Linear scaling of computing requirements with data size

Effective separation of source and noise signals

Well-characterized filtering properties

Abstract

CRUSH is an approach to data analysis under noise interference, developed specifically for submillimeter imaging arrays. The method uses an iterated sequence of statistical estimators to separate source and noise signals. Its filtering properties are well-characterized and easily adjusted to preference. Implementations are well-suited for parallel processing and its computing requirements scale linearly with data size -- rendering it an attractive approach for reducing the data volumes from future large arrays.