Project

Back to overview

Wide-band imaging in the SKA era

English title Wide-band imaging in the SKA era
Applicant Thiran Jean-Philippe
Number 170863
Funding scheme Bilateral programmes
Research institution Laboratoire de traitement des signaux 5 EPFL - STI - IEL - LTS5
Institution of higher education EPF Lausanne - EPFL
Main discipline Information Technology
Start/End 01.11.2017 - 31.10.2021
Approved amount 349'832.00
Show all

All Disciplines (2)

Discipline
Information Technology
Astronomy, Astrophysics and Space Sciences

Keywords (5)

Astronomy: imaging; Signal processing: sparsity; Astronomy: radio interferometry; Signal processing: convex optimisation; Signal processing: inverse problems

Lay Summary (French)

Lead
La radiointerférometrie est une technique permettant l’imagerie du ciel avec grande précision. Les mesures associées ne fournissent que des informations incomplètes et des algorithmes complexes sont utilisés pour la reconstruction d’image. Les radiotéléscopes du futur comme ‘SKA’ ont pour objectif d’atteindre un nouveau régime de résolution et de sensitivité. Les énormes volumes de données associés représentent un défi gigantesque en termes algorithmiques, en particulier pour l’imagerie à large bande où le ciel est observé simultanément à de multiples radiofréquences. Un besoin aigu de nouveaux algorithmes d’imagerie se fait jour, qui puissent traiter des volumes des données inégalés.
Lay summary

L’objectif du projet est de développer une méthode d’imagerie à large bande se basant sur des structures algorithmiques modernes permettant un traitement parallèle sur de multiples noeuds de calcul. Un nouveau modèle d’image sera développé tirant avantage de l’existence de structures spatiales et des corrélations inter-fréquentielles. Un nouvel algorithme sera défini qui séparera les données en une multitude de blocs pouvant être traités en parallèle. La validation sera réalisée à partir de données acquises par ‘MeerKat’, précurseur sud-africain de SKA.

 Le projet aura un impact pour la recherche et l’éducation scientifique de pointe, en particulier dans des régions défavorisées du globe. Du côté sud-africain le projet offre des opportunités pour des jeunes chercheurs issus de populations défavorisées. Du côté suisse un chercheur sera formé dans un domaine d’une importance cruciale pour le pays, qui envisage une participation officielle au projet international SKA. Les compétences acquises en traitement de données seront hautement transférables dans le monde industriel.

Direct link to Lay Summary Last update: 08.06.2017

Responsible applicant and co-applicants

Employees

Name Institute

Publications

Publication
A Faceted Prior for Scalable Wideband Imaging: Application to Radio Astronomy
Thouvenin Pierre-Antoine, Abdulaziz Abdullah, Jiang Ming, Repetti Audrey, Wiaux Yves (2019), A Faceted Prior for Scalable Wideband Imaging: Application to Radio Astronomy, in 2019 IEEE International Workshop on Computational Advances in Multi-Sensor Adaptive Processing, IEEE, -.
A faceted prior for scalable wideband computational imaging
ThouveninPierre-Antoine, AbdulazizAbdullah, JiangMing, RepettiAudrey, YvesWiaux (2019), A faceted prior for scalable wideband computational imaging, in SPARS 2019, SPARS, Toulouse.
Deep Post-Processing for Sparse Image Deconvolution
TerrisMatthieu, AbdulazizAbdullah, DabbechArwa, JiangMing, RepettiAudrey, PesquetJean-Christophe, WiauxYves (2019), Deep Post-Processing for Sparse Image Deconvolution, in SPARS 2019, SPARS, Toulouse.
Fourier dimensionality reduction for fast radio transients
JiangMing, VijayKartik, ThiranJean-Philippe (2019), Fourier dimensionality reduction for fast radio transients, in BASP Frontiers 2019, BASP Frontiers, Villars-sur-Ollon.
Cygnus A super-resolved via convex optimization from VLA data
Dabbech A, Onose A, Abdulaziz A, Perley R A, Smirnov O M, Wiaux Y (2018), Cygnus A super-resolved via convex optimization from VLA data, in Monthly Notices of the Royal Astronomical Society, 476(3), 2853-2866.

Collaboration

Group / person Country
Types of collaboration
Centre for High Performance Computing, Cape Town: Prof. Catherine Cress South Africa (Africa)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
- Research Infrastructure
Astrophysics Lab, EPFL Lausanne: Prof. JP. Kneib Switzerland (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
Biomedical and Astronomical Signal Processing group, Heriot-Watt Edinburgh: Prof. Y. Wiaux Great Britain and Northern Ireland (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
- Research Infrastructure
- Exchange of personnel
Heriot Wattt Edinburgh: BASP group, Professor Wiaux Great Britain and Northern Ireland (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
- Research Infrastructure
- Exchange of personnel
SKA South Africa, Cape Town: Dr Ludwig Schwardt South Africa (Africa)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
Signal Processing Lab 2, EPFL Lausanne: Prof. P. Vandergheynst Switzerland (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
Department of Computer Science, University of Cape Town: Prof. J. Gain South Africa (Africa)
- in-depth/constructive exchanges on approaches, methods or results
- Publication

Scientific events

Active participation

Title Type of contribution Title of article or contribution Date Place Persons involved
SWISS SKA DAYS 2019 Talk given at a conference Parallel precision wide-band imaging in the SKA era with scalable optimization algorithms 19.06.2019 University of Bern, Switzerland Jiang Ming;
BASP Frontiers 2019 Poster Fourier dimensionality reduction for fast radio transients 03.02.2019 Villars-sur-Ollon, Switzerland Thiran Jean-Philippe; Smirnov Oleg; Jiang Ming;
SWISS SKA DAYS 2018 Talk given at a conference Wide-band imaging in the SKA era – convex optimization for precision scalable imaging 11.06.2018 Institute for Data Science FHNW, Brugg, Switzerland Smirnov Oleg; Jiang Ming;


Self-organised

Title Date Place
Edinburgh Meeting July 30 - August 03 Smirnov and Wiaux groups 30.07.2018 Edinburgh, Great Britain and Northern Ireland
Cape Town Meeting Jan 29 - Feb 03 Smirnov and Wiaux groups 29.01.2018 Cape Town, South Africa

Associated projects

Number Title Start Funding scheme
146594 Next-generation calibration and imaging in radio interferometry 01.08.2013 Project funding (Div. I-III)

Abstract

Radio interferometry (RI) allows the observation of radio emissions of our Universe with great sensitivity and angular resolution. Aperture synthesis in RI consists in correlating electric signals from pairs of antennae to produce the so-called visibilities which, under the simplifying assumptions of non-polarised monochromatic incoherent radiation on a small field of view, provide an incomplete Fourier sampling of the underlying two-dimensional sky brightness image of interest. New radio telescopes, such as the future flagship Square Kilometre Array (SKA) are intended to provide images at a totally new range of resolutions and sensitivities, and on a wide frequency band. Data rate estimates, for the first phase of development of the telescope only, are around few terabytes per second. The massive amounts of data to be acquired will represent a great challenge for the infrastructure and signal processing, and the methods solving the inverse problem associated with the image reconstruction need to be fast and to scale well with the data volumes. The celebrated CLEAN imaging algorithm and its variants, that have driven RI imaging so far, will simply not scale to the SKA Big-Data regime. Clearly, an acute need for a totally new generation of imaging algorithms is emerging.The project aims at tackling this RI imaging challenge in the wide-band setting, i.e. when a third imaging dimension is introduced to account for observation over a whole radio frequency band, by leveraging modern parallelised and distributed algorithmic structures in convex optimisation. We will define a new advanced wide-band signal model relying on low-rankness and average joint-sparsity of a matricisation of the wide-band image cube of interest. A convex optimisation problem will be formulated that encompasses multiple data fidelity terms and regularisation priors accounting for the signal model. A recently proposed primal-dual algorithmic structure will be leveraged to solve the optimisation problem, presenting the extreme advantage of full splitting of the objective function: the algorithm enables to split not only the data but also the image into a large number of blocks that can be processed in parallel at each iteration of the image reconstruction process. It is also shipped with a very powerful randomisation functionality, which enables a random selection of the data and image blocks to be updated at each iteration for lower computational requirement. The same algorithmic structure will be adapted for Faraday depth synthesis when wide-band polarimetric data are considered. The algorithm validation will be performed with real data from the South African SKA precursor MeerKAT, of which first dishes are coming online simultaneously with the start of this project, as well as with simulated SKA data. A high performance computing (HPC) implementation will also be developed.All investigators and collaborators of the project have been contributing to advance RI imaging in the past decade. On the astronomy community side, Prof. Smirnov (PI Rhodes University, South Africa, and SKA South Africa) has led the development of the RI measurement equation formalism and of calibration algorithms. He leads research on new RI algorithmic developments with SKA South Africa. On the signal processing side, Prof. Wiaux (Heriot-Watt Edinburgh, UK, previously at EPFL, Switzerland) has pioneered a novel approach to image reconstruction, leveraging the recent theory of compressed sensing and associated convex optimisation algorithmic structures, providing conceptual advances on both grounds of imaging quality and algorithm scalability. His work, partly in collaboration with Prof. Thiran (PI EPFL, Switzerland), was widely taken up by the RI community. The project will be scientifically led by Prof. Thiran, Prof. Smirnov, and Prof. Wiaux, together with their collaborators Prof. Vandergheynst (EPFL, Switzerland), Dr Schwardt (SKA, South Africa), Prof. Gain (University of Cape Town, South Africa), Prof. Cress (Cape Town HPC Centre, South Africa), and Prof. Kneib (EPFL, Switzerland).The project will have significant impact for research, scientific training, and human capacity development. On the South African side, the SKA project's national prominence has been instrumental both in attracting the best young minds into radio astronomy and technical fields related to radio astronomy, and in increasing the diversity of research groups. Our project will provide research opportunities for four PhD students, in particular young researchers from previously disadvantaged demographics. On the Swiss side, a postdoctoral researcher will be trained in a field of increasing importance to Switzerland. The country is indeed, since very recently, considering its membership to the SKA project in the context of its roadmap for infrastructure 2017-2020. The scientific skills acquired by the PhD students and postdoctoral researcher on data processing will be highly transferable to the industrial world where they are highly demanded.
-