Deep Learning improves identification of Radio Frequency Interference

Abstract

Flagging of Radio Frequency Interference (RFI) is an increasingly important challenge in radio astronomy. We present R-Net, a deep convolutional ResNet architecture that significantly outperforms existing algorithms -- including the default MeerKAT RFI flagger, and deep U-Net architectures -- across all metrics including AUC, F1-score and MCC. We demonstrate the robustness of this improvement on both single dish and interferometric simulations and, using transfer learning, on real data. Our R-Net model's precision is approximately $90\%$ better than the current MeerKAT flagger at $80\%$ recall and has a 35\% higher F1-score with no additional performance cost. We further highlight the effectiveness of transfer learning from a model initially trained on simulated MeerKAT data and fine-tuned on real, human-flagged, KAT-7 data. Despite the wide differences in the nature of the two telescope arrays, the model achieves an AUC of 0.91, while the best model without transfer learning only reaches an AUC of 0.67. We consider the use of phase information in our models but find that without calibration the phase adds almost no extra information relative to amplitude data only. Our results strongly suggest that deep learning on simulations, boosted by transfer learning on real data, will likely play a key role in the future of RFI flagging of radio astronomy data.

arXiv comment12 pages, 15 figures. Published on MNRAS

References in corpus (14)

• Post-correlation radio frequency interference classification methods
• Applying Deep Learning to Fast Radio Burst Classification
• Radio Frequency Interference Mitigation at the WSRT
• Optimizing Sparse RFI Prediction using Deep Learning
• DeepSource: Point Source Detection using Deep Learning
• Deep residual detection of radio frequency interference for FAST
• Neutral hydrogen and magnetic fields in M83 observed with the SKA Pathfinder KAT-7
• Engineering and Science Highlights of the KAT-7 Radio Telescope
• Optimising Baryon Acoustic Oscillation Surveys - I: Testing the concordance LCDM cosmology
• Eyes Wide Open - Optimising Cosmological Surveys in a Crowded Market
• HIDE & SEEK: End-to-End Packages to Simulate and Process Radio Survey Data
• Detecting Radio Frequency Interference in radio-antenna arrays with the Recurrent Neural Network algorithm
• CTLearn: Deep Learning for Gamma-ray Astronomy
• Polarization-based online interference mitigation in radio interferometry

Cited by in corpus (14)

• Intelligent Computing: The Latest Advances, Challenges and Future
• Learning to detect RFI in radio astronomy without seeing it
• Trajectory Based RFI Subtraction and Calibration for Radio Interferometry
• Using autoencoders and deep transfer learning to determine the stellar parameters of 286 CARMENES M dwarfs
• Bayesian approach to radio frequency interference mitigation
• Radio Frequency Interference Detection Using Efficient Multi-Scale Convolutional Attention UNet
• A Self-Learning Neural Network Approach for RFI Detection and Removal in Radio Astronomy
• RFI Detection with Spiking Neural Networks
• TABASCAL: Removing multi-satellite interference from radio interferometry observations
• Detecting Modeling Bias with Continuous Time Flow Models on Weak Lensing Maps
• Spiking Neural Networks for Radio Frequency Interference Detection in Radio Astronomy
• Advancing RFI-Detection in Radio Astronomy with Liquid State Machines
• The Impact of GSM towers in Radio Astronomy
• Operational Non-identifiability of Single-epoch Low-rank RFI Mitigation: Controlled Failure-mode Analysis and HERA Evidence