Applying Deep Learning to Fast Radio Burst Classification

Upcoming fast radio burst (FRB) surveys will search ∼10³ beams on the sky with a very high duty cycle, generating large numbers of single-pulse candidates. The abundance of false positives presents an intractable problem if candidates are to be inspected by eye, making it a good application for artificial intelligence (AI). We apply deep learning to single-pulse classification and develop a hierarchical framework for ranking events by their probability of being astrophysical transients. We construct a treelike deep neural network that takes multiple or individual data products as input (e.g., dynamic spectra and multibeam information) and trains on them simultaneously. We have built training and test sets using false-positive triggers from real telescopes, simulated FRBs, and pulsar single pulses. Training the network was independently done for both the CHIME Pathfinder and Apertif. High accuracy and recall can be achieved with a labeled training set of a few thousand events. Even with high triggering rates, classification can be done very quickly on graphical processing units, which is essential for selective voltage dumps or real-time VOEvents. We investigate whether dedispersion back ends could be replaced by a real-time DNN classifier. It is shown that a single forward propagation through a moderate convolutional network could be faster than brute-force dedispersion, but the low signal-to-noise per pixel makes such a classifier suboptimal for this problem. Real-time automated classification will prove useful for bright, unexpected signals, both now and when searchable parameter spaces outgrow our ability to manually inspect data, such as for the SKA and ngVLA.