Ensemble candidate classification for the LOTAAS pulsar survey

Abstract

One of the biggest challenges arising from modern large-scale pulsar surveys is the number of candidates generated. Here, we implemented several improvements to the machine learning classifier previously used by the LOFAR Tied-Array All-Sky Survey (LOTAAS) to look for new pulsars via filtering the candidates obtained during periodicity searches. To assist the machine learning algorithm, we have introduced new features which capture the frequency and time evolution of the signal and improved the signal-to-noise calculation accounting for broad profiles. We enhanced the machine learning classifier by including a third class characterising RFI instances, allowing candidates arising from RFI to be isolated, reducing the false positive return rate. We also introduced a new training dataset used by the machine learning algorithm that includes a large sample of pulsars misclassified by the previous classifier. Lastly, we developed an ensemble classifier comprised of five different Decision Trees. Taken together these updates improve the pulsar recall rate by 2.5 per cent, while also improving the ability to identify pulsars with wide pulse profiles, often misclassified by the previous classifier. The new ensemble classifier is also able to reduce the percentage of false positive candidates identified from each LOTAAS pointing from 2.5 per cent (~500 candidates) to 1.1 per cent (~220 candidates).

The LOFAR Tied-Array All-Sky Survey (LOTAAS): Survey overview and initial pulsar discoveries

“…Out of these 50M CPU-hours, 30M have been used to process all acquired data at the time of writing (January 2019), and reprocess the early LOTAAS data (acquired before May 2015), which had been processed only with the prototype LOTAAS pipeline. The LOTAAS processing pipeline is based on the presto 6 (Ransom 2001) pulsar search software suite, with additional code written by members of the LOTAAS group for single-pulse searches (Michilli et al 2018) and candidate classification (Lyon et al 2016;Tan et al 2018b). Currently it is a purely CPU-based pipeline; however, in the near future some parts will be replaced by GPU implementations.…”

Pulsars detection by machine learning with very few features

“…Further investigation in other features found that these mislabeled pulsars are almost on the boundary of pulsars and non-pulsars. Thus, to raise the recall rate of the ML, one can create new features just like what has been done in Tan et al (2018), or improve data quality from the survey.…”

Surveying the reach and maturity of machine learning and artificial intelligence in astronomy

“…For optical astronomers, a light curve is time‐based photometry, where the variation in intensity of a source over time helps with the identification and classification of a variety of variable star types (e.g., Cohen et al, ; Naul, Bloom, Pérez, & van der Walt, ; Papageorgiou, Catelan, Christopoulou, Drake, & Djorgovski, ) or indicates the presence of otherwise unseen objects, such as an extrasolar planet (e.g., Mislis, Pyrzas, & Alsubai, ; Pearson et al, ; Shallue & Vanderburg, ). In this review, other time‐based measurements in radio astronomy (e.g., pulsar and transient object searches—Connor & van Leeuwen, ; Farah et al, ; Michilli et al, ; Pang, Goseva‐Popstojanova, Devine, & McLaughlin, ; Tan et al, ), and the emerging field of observational gravitational wave astronomy (George & Huerta, , ; Powell et al, ; Zevin et al, ), are categorized as time series .…”