Master project: Performance analysis of radio frequency interference detection in radio astronomy

Michael Mesarcik & Rob van Nieuwpoort, UvA

Radio Frequency Interference (RFI) is a growing concern for modern radio telescopes due to their increasing sensitivity and the proliferation of consumer electronics that depend on electromagnetic emissions. As a result, approaches for RFI detection and mitigation have become a necessity in modern radio observatories. Processing pipelines are employed in observatories that perform RFI detection and mitigation in a post correlation setting, using algorithms such as AO-Flagger [1] or more recently using deep learning architectures based on UNET [2].

Currently, deep learning approaches to RFI detection have been based on supervised architectures [3, 4, 5, 6]. In this case, due to the unavailability of human-labelled datasets, supervised methods are trained and evaluated on simulated interference, or ground truth maps generated by heuristic methods such as the AO-Flagger. In effect, the generalisability of RFI detection algorithms is limited on out-of-distribution RFI samples that the algorithms have not yet been exposed to during training. Therefore we propose the use of the Nearest-Latent-Neighbours (NLN) [7] to overcome the generalisability problems of supervised methods for RFI detection.

In this master’s project you will be required to do the following:

1. Analyse and model the performance of existing supervised RFI detection algorithms.

2. Design a computationally efficient and scalable nearest neighbour search framework for the implementation of the NLN algorithm [7]. This is expected to be the main component of this project.

3. Compare the performance of the two approaches and suggest future approaches for efficient RFI detection.


[1] A. R. Offringa, J. J. Van De Gronde, and J. B. Roerdink, “A morphological algorithm for improving radio-frequency interference detection,” Astronomy and Astrophysics, vol. 539, 2012.

[2] O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 9351, pp. 234–241, 2015.

[3] J. Akeret, C. Chang, A. Lucchi, and A. Refregier, “Radio frequency interference mitigation using deep convolutional neural networks,” Astronomy and Computing, vol. 18, pp. 35–39, 2017.

[4] A. V. Sadr, B. A. Bassett, N. Oozeer, Y. Fantaye, and C. Finlay, “Deep learning improves identification of Radio Frequency Interference,” Monthly Notices of the Royal Astronomical Society, vol. 499, no. 1, pp. 379–390, 2020.

[5] J. Kerrigan, P. L. Plante, S. Kohn, J. C. Pober, J. Aguirre et al., “Optimizing Sparse RFI Prediction using Deep Learning,” Monthly Notices of the Royal Astronomical Society, vol. 11, no. February, pp. 1–11, 7 2019. [Online]. Available: Optimizing sparse RFI prediction using deep learning | Monthly Notices of the Royal Astronomical Society | Oxford Academic

[6] Z. Yang, C. Yu, J. Xiao, and B. Zhang, “Deep residual detection of radio frequency interference for FAST,” Monthly Notices of the Royal Astronomical Society, vol. 492, no. 1, pp. 1421–1431, 2020.

[7] M. Mesarcik, E. Ranguelova, A.-J. Boonstra, and R. V. van Nieuwpoort, “Improving Novelty Detection using the Reconstructions of Nearest Neighbours,” Under review, 11 2021. [Online]. Available: