A Machine Learning Based Morphological Classification of 14,251 Radio AGNs Selected From The Best-Heckman Sample. (arXiv:1812.07190v1 [astro-ph.GA])

A Machine Learning Based Morphological Classification of 14,251 Radio AGNs Selected From The Best-Heckman Sample. (arXiv:1812.07190v1 [astro-ph.GA])
<a href="http://arxiv.org/find/astro-ph/1/au:+Ma_Z/0/1/0/all/0/1">Zhixian Ma</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Xu_H/0/1/0/all/0/1">Haiguang Xu</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Zhu_J/0/1/0/all/0/1">Jie Zhu</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Hu_D/0/1/0/all/0/1">Dan Hu</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Li_W/0/1/0/all/0/1">Weitian Li</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Shan_C/0/1/0/all/0/1">Chenxi Shan</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Zhu_Z/0/1/0/all/0/1">Zhenghao Zhu</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Gu_L/0/1/0/all/0/1">Liyi Gu</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Li_J/0/1/0/all/0/1">Jinjin Li</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Liu_C/0/1/0/all/0/1">Chengze Liu</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Wu_X/0/1/0/all/0/1">Xiangping Wu</a>

We present a morphological classification of 14,251 radio active galactic
nuclei (AGNs) into six types, i.e., typical Fanaroff-Riley Class I / II
(FRI/II), FRI/II-like bent tailed (BT), X-shaped radio galaxy (XRG), and
ring-like radio galaxy (RRG), by designing a convolutional neural network (CNN)
based autoencoder (CAE), namely MCRGNet, and applying it to a labeled radio
galaxy (LRG) sample containing 1,442 AGNs and an unlabled radio galaxy (unLRG)
sample containing 14,251 unlabeled AGNs selected from the Best-Heckman sample.
We train the MCRGNet and implement the classification task by a three-step
strategy, i.e., pre-training, fine-tuning, and classification, which combines
both unsupervised and supervised learnings. A four-layer dichotomous tree is
designed to classify the radio AGNs, which leads to a significantly better
performance than the direct six-type classification. On the LRG sample, our
MCRGNet achieves a total precision of $sim 93%$ and an averaged sensitivity
of $sim 87%$, which are better than those obtained in previous works. On
the unLRG sample, whose labels have been human-inspected, the neural network
achieves a total precision of $sim 84%$. Also, by using the Sloan Digital
Sky Survey (SDSS) Data Release 7 (DR7) to calculate the $r$-band absolute
magnitude ($M_mathrm{opt}$), and using the flux densities to calculate the
radio luminosity ($L_mathrm{radio}$), we find that the distributions of the
unLRG sources on the $L_mathrm{radio}$-$M_mathrm{opt}$ plane do not show
an apparent redshift evolution, and could confirm with a sufficiently large
sample that there could not exist an abrupt separation between FRIs and FRIIs
as reported in some previous works.

We present a morphological classification of 14,251 radio active galactic
nuclei (AGNs) into six types, i.e., typical Fanaroff-Riley Class I / II
(FRI/II), FRI/II-like bent tailed (BT), X-shaped radio galaxy (XRG), and
ring-like radio galaxy (RRG), by designing a convolutional neural network (CNN)
based autoencoder (CAE), namely MCRGNet, and applying it to a labeled radio
galaxy (LRG) sample containing 1,442 AGNs and an unlabled radio galaxy (unLRG)
sample containing 14,251 unlabeled AGNs selected from the Best-Heckman sample.
We train the MCRGNet and implement the classification task by a three-step
strategy, i.e., pre-training, fine-tuning, and classification, which combines
both unsupervised and supervised learnings. A four-layer dichotomous tree is
designed to classify the radio AGNs, which leads to a significantly better
performance than the direct six-type classification. On the LRG sample, our
MCRGNet achieves a total precision of $sim 93%$ and an averaged sensitivity
of $sim 87%$, which are better than those obtained in previous works. On
the unLRG sample, whose labels have been human-inspected, the neural network
achieves a total precision of $sim 84%$. Also, by using the Sloan Digital
Sky Survey (SDSS) Data Release 7 (DR7) to calculate the $r$-band absolute
magnitude ($M_mathrm{opt}$), and using the flux densities to calculate the
radio luminosity ($L_mathrm{radio}$), we find that the distributions of the
unLRG sources on the $L_mathrm{radio}$-$M_mathrm{opt}$ plane do not show
an apparent redshift evolution, and could confirm with a sufficiently large
sample that there could not exist an abrupt separation between FRIs and FRIIs
as reported in some previous works.

http://arxiv.org/icons/sfx.gif