Abstract
ABSTRACT Machine learning methods are increasingly helping astronomers identify new radio pulsars. However, they require a large amount of labelled data, which is time consuming to produce and biased. Here, we describe a Semi-supervised generative adversarial network, which achieves better classification performance than the standard supervised algorithms using majority unlabelled data sets. We achieved an accuracy and mean F-Score of 94.9 per cent trained on only 100 labelled candidates and 5000 unlabelled candidates compared to our standard supervised baseline which scored at 81.1 per cent and 82.7 per cent, respectively. Our final model trained on a much larger labelled data set achieved an accuracy and mean F-score value of 99.2 per cent and a recall rate of 99.7 per cent. This technique allows for high-quality classification during the early stages of pulsar surveys on new instruments when limited labelled data are available. We open-source our work along with a new pulsar-candidate data set produced from the High Time Resolution Universe – South Low Latitude Survey. This data set has the largest number of pulsar detections of any public data set and we hope it will be a valuable tool for benchmarking future machine learning models.
Highlights
Discovering a new pulsar can often lead to new and exciting science
We present results from training a machine learning algorithm to address the practical scenario where we typically have a small amount of labelled data along with a large amount of unlabelled data
We find that the ensemble Supervised Generative Adversarial Network (SGAN) outperforms the standard supervised baseline algorithm as well as re-trained version of Pulsar Image Classification System (PICS) for all combinations of labelled datasets and based on all the metrics discussed in Section 2.7 including higher accuracy, precision, recall rates and a lower false positive rate
Summary
Some examples include the discovery of PSR B1257+12 (Wolszczan & Frail 1992), which led to the discovery of the first set of extrasolar planets. The discovery of the first pulsar triple system (a pulsar orbiting two white-dwarfs) led to one of the most stringent test of the Strong Equivalence Principle (SEP), a prediction of general relativity (Voisin et al 2020). PSR J1141-6545 was used to infer Lense–Thirring precession (relativistic frame-dragging), a prediction of General Relativity (Venkatraman Krishnan et al 2020). These examples are only some of the highlights that display the value of pulsar discoveries. In order to keep pushing the boundaries of fundamental physics, it is important that we continue the investigation of new techniques in order to enhance the discovery process
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
