Artificial neural networks for selection of pulsar candidates from radio continuum surveys

It is very computationally expensive to search for pulsars using time-domain observations, and the volume of data will be enormous with next-generation telescopes such as the Square Kilometre Array. We use artificial neural networks (ANNs), a machine learning method, for the efficient selection of pulsar candidates from radio continuum surveys; this is much cheaper than using time-domain observations. With observed quantities such as radio fluxes, sky position and compactness as inputs, our ANNs output the 'score' that indicates the degree of likeliness that an object is a pulsar. We demonstrate ANNs based on existing survey data by the Tata Institute for Fundamental Research (TIFR) Giant Metrewave Radio Telescope (GMRT) Sky Survey (TGSS) and the National Radio Astronomy Observatory (NRAO) Very Large Array (VLA) Sky Survey (NVSS) and we test their performance. The precision, which is the ratio of the number of pulsars classified correctly as pulsars to the number of any objects classified as pulsars, is about 96 per cent. Finally, we apply the trained ANNs to unidentified radio sources and our fiducial ANN with five inputs (the galactic longitude and latitude, the TGSS and NVSS fluxes and compactness) generates 2436 pulsar candidates from 456 866 unidentified radio sources. We need to confirm whether these candidates are truly pulsars by using time-domain observations. More information, such as polarization, will narrow the number of candidates down further.