Separating the EoR signal with a convolutional denoising autoencoder: a deep-learning-based method

doi:10.1093/mnras/stz582

CORC > 国家天文台 > 中国科学院国家天文台

	Separating the EoR signal with a convolutional denoising autoencoder: a deep-learning-based method
	Li, Weitian 2; Xu, Haiguang 2,3; Ma, Zhixian 1; Zhu, Ruimin 4; Hu, Dan 2; Zhu, Zhenghao 2; Gu, Junhua 5; Shan, Chenxi 2; Zhu, Jie 1; Wu, Xiang-Ping 5
刊名	MONTHLY NOTICES OF THE ROYAL ASTRONOMICAL SOCIETY
	2019-05-01
卷号	485 期号:2 页码:2628-2637
关键词	methods: data analysis techniques: interferometric dark ages, reionization, first stars radio continuum: general
ISSN号	0035-8711
DOI	10.1093/mnras/stz582
英文摘要	When applying the foreground removal methods to uncover the faint cosmological signal from the epoch of reionization (EoR), the foreground spectra are assumed to be smooth. However, this assumption can be seriously violated in practice since the unresolved or missubtracted foreground sources, which are further complicated by the frequency-dependent beam effects of interferometers, will generate significant fluctuations along the frequency dimension. To address this issue, we propose a novel deep-learning-based method that uses a nine-layer convolutional denoising autoencoder (CDAE) to separate the EoR signal. After being trained on the SKA images simulated with realistic beam effects, the CDAE achieves excellent performance as the mean correlation coefficient ((rho) over bar) between the reconstructed and input EoR signals reaches 0.929 +/- 0.045. In comparison, the two representative traditional methods, namely the polynomial fitting method and the continuous wavelet transform method, both have difficulties inmodelling and removing the foreground emission complicated with the beameffects, yielding only (rho) over bar poly = 0.296 +/- 0.121 and (rho) over bar cwt = 0.198 +/- 0.160, respectively. We conclude that, by hierarchically learning sophisticated features through multiple convolutional layers, the CDAE is a powerful tool that can be used to overcome the complicated beam effects and accurately separate the EoR signal. Our results also exhibit the great potential of deeplearning- based methods in future EoR experiments.
资助项目	Ministry of Science and Technology of China[2018YFA0404601] ; Ministry of Science and Technology of China[2017YFF0210903] ; National Natural Science Foundation of China[11433002] ; National Natural Science Foundation of China[11621303] ; National Natural Science Foundation of China[11835009] ; National Natural Science Foundation of China[61371147]
WOS关键词	FOREGROUND REMOVAL ; COSMIC DAWN ; REIONIZATION ; SKY ; EPOCH ; REPRESENTATIONS ; DIMENSIONALITY ; SIMULATIONS ; SUBTRACTION ; CALIBRATION
WOS研究方向	Astronomy & Astrophysics
语种	英语
出版者	OXFORD UNIV PRESS
WOS记录号	WOS:000474886200083
资助机构	Ministry of Science and Technology of China ; Ministry of Science and Technology of China ; National Natural Science Foundation of China ; National Natural Science Foundation of China ; Ministry of Science and Technology of China ; Ministry of Science and Technology of China ; National Natural Science Foundation of China ; National Natural Science Foundation of China ; Ministry of Science and Technology of China ; Ministry of Science and Technology of China ; National Natural Science Foundation of China ; National Natural Science Foundation of China ; Ministry of Science and Technology of China ; Ministry of Science and Technology of China ; National Natural Science Foundation of China ; National Natural Science Foundation of China
内容类型	期刊论文
源URL	[http://ir.bao.ac.cn/handle/114a11/26939]
专题	中国科学院国家天文台
通讯作者	Li, Weitian; Xu, Haiguang
作者单位	1.Shanghai Jiao Tong Univ, Dept Elect Engn, 800 Dongchuan Rd, Shanghai 200240, Peoples R China 2.Shanghai Jiao Tong Univ, Sch Phys & Astron, 800 Dongchuan Rd, Shanghai 200240, Peoples R China 3.Shanghai Jiao Tong Univ, IFSA Collaborat Innovat Ctr, Tsung Dao Lee Inst, 800 Dongchuan Rd, Shanghai 200240, Peoples R China 4.Northwestern Univ, Dept Stat, 2006 Sheridan Rd, Evanston, IL 60208 USA 5.Chinese Acad Sci, Natl Astron Observ, 20A Datun Rd, Beijing 100012, Peoples R China
推荐引用方式 GB/T 7714	Li, Weitian,Xu, Haiguang,Ma, Zhixian,et al. Separating the EoR signal with a convolutional denoising autoencoder: a deep-learning-based method[J]. MONTHLY NOTICES OF THE ROYAL ASTRONOMICAL SOCIETY,2019,485(2):2628-2637.
APA	Li, Weitian.,Xu, Haiguang.,Ma, Zhixian.,Zhu, Ruimin.,Hu, Dan.,...&Wu, Xiang-Ping.(2019).Separating the EoR signal with a convolutional denoising autoencoder: a deep-learning-based method.MONTHLY NOTICES OF THE ROYAL ASTRONOMICAL SOCIETY,485(2),2628-2637.
MLA	Li, Weitian,et al."Separating the EoR signal with a convolutional denoising autoencoder: a deep-learning-based method".MONTHLY NOTICES OF THE ROYAL ASTRONOMICAL SOCIETY 485.2(2019):2628-2637.