Applying Deep Learning to Fast Radio Burst Classification

Abstract

Upcoming Fast Radio Burst (FRB) surveys will search $\sim$10\,${}^3$ beams on sky with 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 true astrophysical transients. We construct a tree-like deep neural network (DNN) that takes multiple or individual data products as input (e.g. dynamic spectra and multi-beam detection information) and trains on them simultaneously. We have built training and test sets using false-positive triggers from real telescopes, along with simulated FRBs, and single pulses from pulsars. Training of the DNN was independently done for two radio telescopes: the CHIME Pathfinder, and Apertif on Westerbork. High accuracy and recall can be achieved with a labelled training set of a few thousand events. Even with high triggering rates, classification can be done very quickly on Graphical Processing Units (GPUs). That speed is essential for selective voltage dumps or issuing real-time VOEvents. Next, we investigate whether dedispersion back-ends could be completely 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 sub-optimal for this problem. Real-time automated classification may prove useful for bright, unexpected signals, both now and in the era of radio astronomy when data volumes and the searchable parameter spaces further outgrow our ability to manually inspect the data, such as for SKA and ngVLA.

References in corpus (39)

• A bright millisecond radio burst of extragalactic origin
• A Population of Fast Radio Bursts at Cosmological Distances
• cuDNN: Efficient Primitives for Deep Learning
• A Repeating Fast Radio Burst
• The Host Galaxy and Redshift of the Repeating Fast Radio Burst FRB 121102
• Fast Radio Burst Discovered in the Arecibo Pulsar ALFA Survey
• An extreme magneto-ionic environment associated with the fast radio burst source FRB 121102
• FRBCAT: The Fast Radio Burst Catalogue
• The Repeating Fast Radio Burst FRB 121102 as Seen on Milliarcsecond Angular Scales
• Dense magnetized plasma associated with a fast radio burst
• Why does deep and cheap learning work so well?
• The repeating Fast Radio Burst FRB 121102: Multi-wavelength observations and additional bursts
• A real-time fast radio burst: polarization detection and multiwavelength follow-up
• Fifty Years of Pulsar Candidate Selection: From simple filters to a new principled real-time classification approach
• The magnetic field and turbulence of the cosmic web measured using a brilliant fast radio burst
• Pulsar searches and timing with the SKA
• The detection of an extremely bright fast radio burst in a phased array feed survey
• Accelerating incoherent dedispersion
• Searching for pulsars using image pattern recognition
• The first interferometric detections of Fast Radio Bursts
• The LOFAR Pilot Surveys for Pulsars and Fast Radio Transients
• Identifying the source of perytons at the Parkes radio telescope
• An accurate and efficient algorithm for detection of radio bursts with an unknown dispersion measure, for single dish telescopes and interferometers
• The Euclidean distribution of Fast Radio Bursts
• Single-pulse classifier for the LOFAR Tied-Array All-sky Survey
• A Machine Learning Classifier for Fast Radio Burst Detection at the VLBA
• Finding pulsars with LOFAR
• Detection of Dispersed Radio Pulses: A machine learning approach to candidate identification and classification
• CHIME FRB: An application of FFT beamforming for a radio telescope
• Real-time, fast radio transient searches with GPU de-dispersion
• The Pulsar Search Collaboratory: Discovery and Timing of Five New Pulsars
• Real-Time Dedispersion for Fast Radio Transient Surveys, using Auto Tuning on Many-Core Accelerators
• Real-time searches for fast transients with Apertif and LOFAR
• Pulsar Searches with the SKA
• Limits on the ultra-bright Fast Radio Burst population from the CHIME Pathfinder
• ALFABURST: A commensal search for Fast Radio Bursts with Arecibo
• In-Datacenter Performance Analysis of a Tensor Processing Unit
• VOEvent Standard for Fast Radio Bursts
• Pulsar Candidate Identification with Artificial Intelligence Techniques

Cited by in corpus (42)

• Fast Radio Bursts: An Extragalactic Enigma
• FETCH: A deep-learning based classifier for fast transient classification
• Surveying the reach and maturity of machine learning and artificial intelligence in astronomy
• Chromatic periodic activity down to 120 MHz in a Fast Radio Burst
• Repeating fast radio bursts with WSRT/Apertif
• Five new real-time detections of Fast Radio Bursts with UTMOST
• A bright, high rotation-measure FRB that skewers the M33 halo
• Comprehensive analysis of a dense sample of FRB 121102 bursts
• On Neural Architectures for Astronomical Time-series Classification with Application to Variable Stars
• Machine learning classification of CHIME fast radio bursts -- I. Supervised methods
• Deep Learning improves identification of Radio Frequency Interference
• Transfer learning for radio galaxy classification
• Machine learning classification of CHIME fast radio bursts: II. Unsupervised Methods
• A fast radio burst with sub-millisecond quasi-periodic structure
• Single Pulse Detection Algorithms for Real-time Fast Radio Burst Searches using GPUs
• Dual-frequency single-pulse study of PSR B0950+08
• Searching for broadband pulsed beacons from 1883 stars using neural networks
• Deep Learning Assisted Data Inspection for Radio Astronomy
• Identification of BASS DR3 Sources as Stars, Galaxies and Quasars by XGBoost
• Searching for hot subdwarf stars from the LAMOST spectra. III. classifying the hot subdwarf stars from LAMOST DR4 using deep learning method
• Estimating fast transient detection pipeline efficiencies at UTMOST via real-time injection of mock FRBs
• Galactic Radio Explorer: an all-sky monitor for bright radio bursts
• Advances in Deep Space Exploration via Simulators & Deep Learning
• The ROAD to discovery: machine learning-driven anomaly detection in radio astronomy spectrograms
• 81 New Candidate Fast Radio Bursts in Parkes Archive
• Improved Type III solar radio burst detection using congruent deep learning models
• The Apertif Radio Transient System (ARTS): Design, Commissioning, Data Release, and Detection of the first 5 Fast Radio Bursts
• Reproducible and Portable Workflows for Scientific Computing and HPC in the Cloud
• Applying saliency-map analysis in searches for pulsars and fast radio bursts
• Verifying and Reporting Fast Radio Bursts
• An activity transition in FRB 20201124A: methodological rigor, detection of frequency-dependent cessation, and a geometric magnetar model
• Machine Learning Classification of Repeating FRBs from FRB121102
• Comprehensive analysis of the Apertif Fast Radio Burst sample: similarities with young, energetic neutron stars
• New upper limits on low-frequency radio emission from isolated neutron stars with LOFAR
• Search for fast radio transients using Arecibo drift-scan observations at 1.4 GHz
• Detecting Pulsars with Neural Networks: A Proof of Concept
• SPARKESX: Single-dish PARKES data sets for finding the uneXpected -- A data challenge
• Datacube segmentation via Deep Spectral Clustering
• Apertif, Phased Array Feeds for the Westerbork Synthesis Radio Telescope
• Neural networks in the search for fast radio bursts with RATAN-600
• Operational Non-identifiability of Single-epoch Low-rank RFI Mitigation: Controlled Failure-mode Analysis and HERA Evidence
• Radio Galaxy Zoo: Giant Radio Galaxy Classification using Multi-Domain Deep Learning