SuperRAENN: A Semisupervised Supernova Photometric Classification Pipeline Trained on Pan-STARRS1 Medium-Deep Survey Supernovae
SuperRAENN: A Semisupervised Supernova Photometric Classification Pipeline Trained on Pan-STARRS1 Medium-Deep Survey Supernovae
复制标题
DOI：
10.3847/1538-4357/abc6fd
复制
发表时间：
2020-08
期刊：
影响因子：
--
通讯作者：
V. Villar;G. Hosseinzadeh;E. Berger;M. Ntampaka;David O. Jones;P. Challis;R. Chornock;M. Drout
中科院分区：
文献类型：
--
作者：
V. Villar;G. Hosseinzadeh;E. Berger;M. Ntampaka;David O. Jones;P. Challis;R. Chornock;M. Drout
基于光学光度学光变曲线信息的超新星（SNe）自动分类在即将到来的宽场时域调查时代是必不可少的，例如鲁宾天文台进行的时空遗产调查（LSST）。光度分类可以实时识别感兴趣的事件，用于扩展的多波长随访以及存档人口研究。在这里，我们提出了5243“SN样”光变曲线（在gP 1 rP 1 iP 1 zP 1）从泛星1中深调查（PS1-MDS）的完整样本。PS1-MDS在节奏、过滤器和深度方面与计划中的LSST Wide-Fast-Deep调查相似，使其成为社区有用的培训集。使用这个数据集，我们训练了一种新的半监督机器学习算法，对2315个新的类SN光变曲线与宿主星系光谱红移进行光度分类。我们的算法包括一个RF监督分类步骤和一个新的无监督步骤，其中我们引入了一个经常性的自编码器神经网络（RAENN）。我们的最终管道被称为SuperRAENN，在五个SN类别（Ia，Ibc，II，IIn，SLSN-I）中的准确性为87%，宏观平均纯度和完整性分别为66%和69%。我们发现SNe Ia和SLSNe的准确率最高，SNe Ibc的准确率最低。我们完整的光谱和光度分类样本分为62.0% Ia型（1839个物体），19.8% II型（553个物体），4.8% IIn型（136个物体），11.7% Ibc型（291个物体）和1.6% I型SLSNe（54个物体）。
Automated classification of supernovae (SNe) based on optical photometric light-curve information is essential in the upcoming era of wide-field time domain surveys, such as the Legacy Survey of Space and Time (LSST) conducted by the Rubin Observatory. Photometric classification can enable real-time identification of interesting events for extended multiwavelength follow-up, as well as archival population studies. Here we present the complete sample of 5243 “SN-like” light curves (in gP1rP1iP1zP1) from the Pan-STARRS1 Medium-Deep Survey (PS1-MDS). The PS1-MDS is similar to the planned LSST Wide-Fast-Deep survey in terms of cadence, filters, and depth, making this a useful training set for the community. Using this data set, we train a novel semisupervised machine learning algorithm to photometrically classify 2315 new SN-like light curves with host galaxy spectroscopic redshifts. Our algorithm consists of an RF supervised classification step and a novel unsupervised step in which we introduce a recurrent autoencoder neural network (RAENN). Our final pipeline, dubbed SuperRAENN, has an accuracy of 87% across five SN classes (Type Ia, Ibc, II, IIn, SLSN-I) and macro-averaged purity and completeness of 66% and 69%, respectively. We find the highest accuracy rates for SNe Ia and SLSNe and the lowest for SNe Ibc. Our complete spectroscopically and photometrically classified samples break down into 62.0% Type Ia (1839 objects), 19.8% Type II (553 objects), 4.8% Type IIn (136 objects), 11.7% Type Ibc (291 objects), and 1.6% Type I SLSNe (54 objects).