RFI Detection with Spiking Neural Networks
Nicholas J. Pritchard, Andreas Wicenec, Mohammed Bennamoun, Richard Dodson
14 January 2024
Author Verified?
true
High School Summary
This study focuses on the detection and mitigation of radio frequency interference (RFI) in radio telescopes using a type of artificial intelligence called Spiking Neural Networks (SNNs). SNNs are inspired by how our brains work and are good at processing data that changes over time and space. The researchers adapted a previous algorithm and architecture to work with SNNs, allowing for simplified RFI detection by sampling the internal spiking neurons. They tested the performance of their SNN approach using simulated telescope data and hand-labeled datasets. While the SNN approach was competitive with other methods for the simulated telescope data, it faced challenges with the other datasets. However, the SNN approach was able to maintain its performance while reducing the computational requirements of the original algorithm. This study shows that SNNs have potential for RFI detection in radio telescopes and is the first to apply SNNs in astronomy.
University Summary
This study focuses on detecting and mitigating Radio Frequency Interference (RFI) in radio telescopes using machine learning methods. The researchers introduce Spiking Neural Networks (SNNs) to the field of astronomy for the first time, specifically for RFI detection. They adapt existing algorithms and architectures to work with SNNs, allowing for simplified downstream RFI detection. The performance of the SNN approach is evaluated using simulated telescope data and a hand-labelled dataset. The results show that the SNN approach is competitive with existing methods for the simulated dataset but faces challenges with other datasets. However, the SNN method eliminates the need for a computationally intensive step found in other approaches. This study demonstrates the potential of SNNs for machine learning-based RFI detection in radio telescopes and is the first to apply SNNs in astronomy.