http://arxiv.org/abs/1903.04502
Next generation radio interferometric telescopes are entering an era of big data with extremely large data sets. While these telescopes can observe the sky in higher sensitivity and resolution than before, computational challenges in image reconstruction need to be overcome to realize the potential of forthcoming telescopes. New methods in sparse image reconstruction and convex optimisation techniques (cf. compressive sensing) have shown to produce higher fidelity reconstructions of simulations and real observations than traditional methods. This article presents distributed and parallel algorithms and implementations to perform sparse image reconstruction. Both computations and data can be distributed (with MPI) across nodes of a computing cluster, while on each node multi-core parallelization (on GPUs or across CPU cores) can be used for further optimisation. The algorithms are implemented in the latest versions of the SOPT (https://github.com/astro-informatics/sopt) and PURIFY (https://github.com/astro-informatics/purify) software packages (Versions 3.0.1), which have been released alongside of this article. We benchmark the algorithms presented, showing that they are considerably faster than their serial equivalents and can scale to big data sets. This work presents an important step towards the computationally scalable and efficient algorithms and implementations that are needed to image observations from next generation radio interferometers such as the SKA.
L. Pratley, J. McEwen, M. d’Avezac, et. al.
Wed, 13 Mar 19
7/125
Comments: 25 pages, 5 figures