Cosmological parameter estimation from large-scale structure deep learning

作     者：ShuYang Pan MiaoXin Liu Jaime Forero-Romero Cristiano G.Sabiu ZhiGang Li HaiTao Miao Xiao-Dong Li ShuYang Pan;MiaoXin Liu;Jaime Forero-Romero;Cristiano G.Sabiu;ZhiGang Li;HaiTao Miao;Xiao-Dong Li

作者机构：School of Physics and AstronomySun Yat-Sen UniversityGuangzhou 510297China Departamento de FísicaUniversidad de los AndesBogotáCP 111711Colombia Department of AstronomyYonsei UniversitySeoul 03722Korea College of Physics and Electronic EngineeringNanyang Normal UniversityNanyang 473061China 

出 版 物：《Science China(Physics,Mechanics & Astronomy)》 (中国科学：物理学、力学、天文学（英文版）)

年 卷 期：2020年第63卷第11期

页      面：36-50页

核心收录：

学科分类：07[理学] 070401[理学-天体物理] 0805[工学-材料科学与工程（可授工学、理学学位）] 0704[理学-天文学] 

基　　金：support from the National Natural Science Foundation of China(Grant No.11803094) the Science and Technology Program of Guangzhou,China(Grant No.202002030360) support from COLCIENCIAS(Contract No.287-2016,Project 1204-712-50459) support from the National Research Foundation(Grant Nos.2017R1D1A1B03034900,2017R1A2B2004644,and 2017R1A4A1015178) support from the Project for New Faculty of Shanghai JiaoTong University(Grant No.AF0720053) the National Science Foundation of China(Grant Nos.11533006,and 11433001) the National Basic Research Program of China(Grant No.2015CB857000) 

主　　题：large-scale structure of Universe cosmological parameters dark energy machine learning 

摘      要：We propose a light-weight deep convolutional neural network(CNN)to estimate the cosmological parameters from simulated 3-dimensional dark matter distributions with high *** training set is based on 465 realizations of a cubic box with a side length of 256 h-1 Mpc,sampled with 1283 particles interpolated over a cubic grid of 1283 *** volumes have cosmological parameters varying within the flatΛCDM parameter space of 0.16≤?m≤0.46 and 2.0≤109 As≤*** neural network takes as an input cubes with 32^3 oxels and has three convolution layers,three dense layers,together with some batch normalization and pooling *** the final predictions from the network we find a 2.5%bias on the primordial amplitudeσ8 that cannot easily be resolved by continued *** correct this bias to obtain unprecedented accuracy in the cosmological parameter estimation with statistical uncertainties ofδ?m=0.0015 andδσ8=0.0029,which are several times better than the results of previous CNN *** with a 2-point analysis method using the clustering region of 0-130 and 10-130 h-1 Mpc,the CNN constraints are several times and an order of magnitude more precise,***,we conduct preliminary checks of the error-tolerance abilities of the neural network,and find that it exhibits robustness against smoothing,masking,random noise,global variation,rotation,reflection,and simulation *** effects are well understood in typical clustering analysis,but had not been tested before for the CNN *** work shows that CNN can be more promising than people expected in deriving tight cosmological constraints from the cosmic large scale structure.