Applying saliency-map analysis in searches for pulsars and fast radio bursts

Abstract

To investigate the use of saliency-map analysis to aid in searches for transient signals, such as fast radio bursts and individual pulses from radio pulsars. We aim to demonstrate that saliency maps provide the means to understand predictions from machine learning algorithms and can be implemented in piplines used to search for transient events. We have implemented a new deep learning methodology to predict whether or not any segment of the data contains a transient event. The algorithm has been trained using real and simulated data sets. We demonstrate that the algorithm is able to identify such events. The output results are visually analysed via the use of saliency maps. We find that saliency maps can produce an enhanced image of any transient feature without the need for de-dispersion or removal of radio frequency interference. Such maps can be used to understand which features in the image were used in making the machine learning decision and to visualise the transient event. Even though the algorithm reported here was developed to demonstrate saliency-map analysis, we have detected, in archival data, a single burst event with dispersion measure of $41$\,cm${}^{-3}$pc that is not associated with any currently known pulsar.

arXiv comment16 pages, 6 figures, accepted for publication in A&A

References in corpus (25)

• Methods for Interpreting and Understanding Deep Neural Networks
• A bright millisecond radio burst of extragalactic origin
• PSRCHIVE and PSRFITS An Open Approach to Radio Pulsar Data Storage and Analysis
• A Repeating Fast Radio Burst
• Transient radio bursts from rotating neutron stars
• Fast Radio Bursts: An Extragalactic Enigma
• A Living Theory Catalogue for Fast Radio Bursts
• CHIME/FRB Detection of Eight New Repeating Fast Radio Burst Sources
• Considerations for a Multi-beam Multi-purpose Survey with FAST
• A Second Source of Repeating Fast Radio Bursts
• Fast Radio Burst 121102 Pulse Detection and Periodicity: A Machine Learning Approach
• Accelerating incoherent dedispersion
• Discovery of 14 radio pulsars in a survey of the Magellanic Clouds
• Identifying the source of perytons at the Parkes radio telescope
• The Parkes Observatory Pulsar Data Archive
• Applying Deep Learning to Fast Radio Burst Classification
• Five new real-time detections of Fast Radio Bursts with UTMOST
• Discovery of Two New Pulsars in 47 Tucanae (NGC 104)
• Single-pulse classifier for the LOFAR Tied-Array All-sky Survey
• Detection of Dispersed Radio Pulses: A machine learning approach to candidate identification and classification
• Piggyback search for fast radio bursts using Nanshan 26m and Kunming 40m radio telescopes -- I. Observing and data analysis systems, discovery of a mysterious peryton
• A new fast radio burst in the datasets containing the Lorimer burst
• A Comparison of Photometric Redshift Techniques for Large Radio Surveys
• Search for a Radio Pulsar in the Remnant of Supernova 1987A
• Discovery of Five New Pulsars in Archival Data

Cited by in corpus (6)

• Interpreting automatic AGN classifiers with saliency maps
• Fast radio burst detection in the presence of coloured noise
• The Importance of Being Interpretable: Toward An Understandable Machine Learning Encoder for Galaxy Cluster Cosmology
• High time resolution search for prompt radio emission from the long GRB 210419A with the Murchison Widefield Array
• Vela Pulsar: Single Pulses Analysis with Machine Learning Techniques
• SPARKESX: Single-dish PARKES data sets for finding the uneXpected -- A data challenge