2019/03/11 by Luke Pratley, Pratley, Luke, Jason D. McEwen +11 · 1 citation
Engineering · Environmental Science · Physics and Astronomy · #Antenna Design and Optimization #FOS: Physical sciences #Instrumentation and Methods for Astrophysics (astro-ph.IM) #Radio Astronomy Observations and Technology #Soil Moisture and Remote Sensing #Sparse and Compressive Sensing Techniques
paper · pdf · doi:10.48550/arxiv.1903.04502
openalex publication_date 2019/03/11 · openalex created_date 2022/07/29 · openalex updated_date 2026/07/28
Next generation radio interferometric telescopes are entering an era of big\ndata with extremely large data sets. While these telescopes can observe the sky\nin higher sensitivity and resolution than before, computational challenges in\nimage reconstruction need to be overcome to realize the potential of\nforthcoming telescopes. New methods in sparse image reconstruction and convex\noptimization techniques (cf. compressive sensing) have shown to produce higher\nfidelity reconstructions of simulations and real observations than traditional\nmethods. This article presents distributed and parallel algorithms and\nimplementations to perform sparse image reconstruction, with significant\npractical considerations that are important for implementing these algorithms\nfor Big Data. We benchmark the algorithms presented, showing that they are\nconsiderably faster than their serial equivalents. We then pre-sample gridding\nkernels to scale the distributed algorithms to larger data sizes, showing\napplication times for 1 Gb to 2.4 Tb data sets over 25 to 100 nodes for up to\n50 billion visibilities, and find that the run-times for the distributed\nalgorithms range from 100 milliseconds to 3 minutes per iteration. This work\npresents an important step in working towards computationally scalable and\nefficient algorithms and implementations that are needed to image observations\nof both extended and compact sources from next generation radio interferometers\nsuch as the SKA. The algorithms are implemented in the latest versions of the\nSOPT (https://github.com/astro-informatics/sopt) and PURIFY\n(https://github.com/astro-informatics/purify) software packages (Versions\n3.1.0), which have been released alongside of this article.\n