Artificial neural networks for selection of pulsar candidates from the radio continuum surveys. (arXiv:1811.11478v1 [astro-ph.IM])
<a href="http://arxiv.org/find/astro-ph/1/au:+Yonemaru_N/0/1/0/all/0/1">Naoyuki Yonemaru</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Takahashi_K/0/1/0/all/0/1">Keitaro Takahashi</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Kumamoto_H/0/1/0/all/0/1">Hiroki Kumamoto</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Dai_S/0/1/0/all/0/1">Shi Dai</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Yoshiura_S/0/1/0/all/0/1">Shintaro Yoshiura</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Ideguchi_S/0/1/0/all/0/1">Shinsuke Ideguchi</a>
Pulsar search with timing observation is very computationally expensive and
data volume will be enormous with the next generation telescopes such as SKA.
We develop artificial neural networks (ANNs), one of machine learning methods,
for efficient selection of pulsar candidates from radio continuum surveys,
which are much cheaper than timing observation. With observed quantities such
as radio fluxes, sky position and compactness as inputs, our ANNs output the
probability that the object is a pulsar. We demonstrate ANNs based on existing
survey data by TIFR GMRT Sky Survey and NRAO VLA Sky Survey and test their
performance. The ANNs work quite well and the false positive rate is $0.16%$
at best. Finally, we apply the ANN to unidentified radio sources and obtain
32,583 pulsar candidates. More information such as polarization will narrow the
candidates down further.
Pulsar search with timing observation is very computationally expensive and
data volume will be enormous with the next generation telescopes such as SKA.
We develop artificial neural networks (ANNs), one of machine learning methods,
for efficient selection of pulsar candidates from radio continuum surveys,
which are much cheaper than timing observation. With observed quantities such
as radio fluxes, sky position and compactness as inputs, our ANNs output the
probability that the object is a pulsar. We demonstrate ANNs based on existing
survey data by TIFR GMRT Sky Survey and NRAO VLA Sky Survey and test their
performance. The ANNs work quite well and the false positive rate is $0.16%$
at best. Finally, we apply the ANN to unidentified radio sources and obtain
32,583 pulsar candidates. More information such as polarization will narrow the
candidates down further.
http://arxiv.org/icons/sfx.gif
