Optimizing sparse RFI prediction using deep learning

Joshua Kerrigan, Paul la Plante, Saul Kohn, Jonathan C. Pober, James Aguirre, Zara Abdurashidova, Paul Alexander, Zaki S. Ali, Yanga Balfour, Adam P. Beardsley, Gianni Bernardi, Judd D. Bowman, Richard F. Bradley, Jacob Burba, Chris L. Carilli, Carina Cheng, David R. DeBoer, Matt Dexter, Eloy de Lera Acedo, Joshua S. DillonJulia Estrada, Aaron Ewall-Wice, Nicolas Fagnoni, Randall Fritz, Steve R. Furlanetto, Brian Glendenning, Bradley Greig, Jasper Grobbelaar, Deepthi Gorthi, Ziyaad Halday, Bryna J. Hazelton, Jack Hickish, Daniel C. Jacobs, Austin Julius, Nicholas S. Kern, Piyanat Kittiwisit, Matthew Kolopanis, Adam Lanman, Telalo Lekalake, Adrian Liu, David MacMahon, Lourence Malan, Cresshim Malgas, Matthys Maree, Zachary E. Martinot, Eunice Matsetela, Andrei Mesinger, Mathakane Molewa, Miguel F. Morales, Tshegofalang Mosiane, Abraham R. Neben, Aaron R. Parsons, Nipanjana Patra, Samantha Pieterse, Nima Razavi-Ghods, Jon Ringuette, James Robnett, Kathryn Rosie, Peter Sims, Craig Smith, Angelo Syce, Nithyanandan Thyagarajan, Peter K.G. Williams, Haoxuan Zheng

Research output: Contribution to journalArticlepeer-review

17 Scopus citations


Radio frequency interference (RFI) is an ever-present limiting factor among radio telescopes even in the most remote observing locations. When looking to retain the maximum amount of sensitivity and reduce contamination for Epoch of Reionization studies, the identification and removal of RFI is especially important. In addition to improved RFI identification, we must also take into account computational efficiency of the RFI-Identification algorithm as radio interferometer arrays such as the Hydrogen Epoch of Reionization Array (HERA) grow larger in number of receivers. To address this, we present a deep fully convolutional neural network (DFCN) that is comprehensive in its use of interferometric data, where both amplitude and phase information are used jointly for identifying RFI. We train the network using simulated HERA visibilities containing mock RFI, yielding a known ‘ground truth’ data set for evaluating the accuracy of various RFI algorithms. Evaluation of the DFCN model is performed on observations from the 67 dish build-out, HERA-67, and achieves a data throughput of 1.6 × 105 HERA time-ordered 1024 channelled visibilities per hour per GPU. We determine that relative to an amplitude only network including visibility phase adds important adjacent time–frequency context which increases discrimination between RFI and non-RFI. The inclusion of phase when predicting achieves a recall of 0.81, precision of 0.58, and F2 score of 0.75 as applied to our HERA-67 observations.

Original languageEnglish (US)
Pages (from-to)2605-2615
Number of pages11
JournalMonthly Notices of the Royal Astronomical Society
Issue number2
StatePublished - Sep 11 2019


  • Methods: data analysis – Techniques: interferometric

ASJC Scopus subject areas

  • Astronomy and Astrophysics
  • Space and Planetary Science


Dive into the research topics of 'Optimizing sparse RFI prediction using deep learning'. Together they form a unique fingerprint.

Cite this