|
|
|
|
|
|
Prediction of Stellar Spectrum Categories Based on Deep Residual Network |
WANG Tian-xiang1, 2, FAN Yu-feng1*, WANG Xiao-li1, LONG Qian1, WANG Chuan-jun1 |
1. Yunnan Observatories, Chinese Academy of Sciences, Kunming 650011, China
2. University of Chinese Academy of Sciences, Beijing 100049, China |
|
|
Abstract In recent years, the spectral data of celestial bodies observed have achieved a dramatic increase thanks to the successful implementation of various projects of spectral sky survey. Therefore, higher requirements for the automatic classification and analysis of spectrum are proposed for large-scale projects of spectral sky survey. The classification problem is transformed into a regression one in this paper, and a method of spectral category regression based on the residual depth network is put forward to conduct a prediction of MK spectral subtype on stellar spectrum. The network is mainly composed of 25 convolution layers, 1 maximum pooling layer, 1 average pooling layer, full connection layer and 12 residual structures. The maximum pooling layer is used to filter features, the convolution layer to extract features, and the average pooling layer to reduce parameters and improve efficiency. The residual structure can prevent the degradation of the network, extract high-dimensional abstract features by deepening the network and improve training speed. Considering the non-zero probability of data with false labels and corrupted data, Log-Cosh is adopted as a loss function in this paper to reduce the negative impact of bad samples. 80 000 spectra that are randomly selected from LAMOST DR5 are used as the experimental data. The spectra are divided into the training set, verification set and test set according to the proportion of 7∶1∶2 after eliminating the bad value and normalizing the flow. The experiment includes two parts. In the first part, the network is adopted to carry out a prediction on the spectral subtype, and the maximum absolute error, the average absolute error and the standard deviation are used to compare the performance of convolution kernels with different shapes. The predicted value is taken as the abscissa and the label as the ordinate, and the second-order nonlinear fitting is used for all sample points in the test set, a straight line that is coincident with y=x is obtained, proving that the model can predict the spectral subtype well. The second part is concerning the internal analysis of the model. The main characteristics of the model in predicting four types of spectra, A, F, G, K, are mainly explored with the method of category activation mapping, thus endowing the model with interpretability. In the text data set, 91.4% of the spectral prediction errors of this method are within 0.5 spectral subtypes, and the average absolute error of the prediction is 0.3 spectral subtypes. It is shown that the method proposed in this paper can better predict spectral subtypes with faster speed and higher accuracy according to the comparison of the same data set with nonparametric regression, Adaboost regression tree and K-means.
|
Received: 2020-03-15
Accepted: 2020-07-06
|
|
Corresponding Authors:
FAN Yu-feng
E-mail: fanyf@ynao.ac.cn
|
|
[1] Luo A-li,Zhao Yongheng,Zhao G,et al. RAA (Research in Astronomy and Astrophysics),2015,15:1095.
[2] Yi Zhenping, Pan Jingchang. Image and Signal Processing (CISP). 3rd International Congress,2010.
[3] Schierscher F, Paunzen E. Astronomische Nachrichten,2011, 332(6): 597.
[4] Liu Chao, Cui Wenyuan, Zhang Bo, et al. Research in Astronomy and Astrophysics, 2015,15:1137.
[5] Kaushal Sharma, Ajit Kembhavi, Aniruddha Kembhavi, et al. Monthly Notices of the Royal Astronomical Society, 2020, 491: 2280.
[6] Gray R O, Corbally C J. The Astronomical Journal,2014, 147(4): 80.
[7] LIU Rong, QIAO Xue-jun, ZHANG Jian-nan, et al(刘 蓉,乔学军,张健楠,等). Spectroscopy and Spectral Analysis(光谱学与光谱分析), 2017,37(5): 1553.
[8] Kheirdastan S, Bazarghan M. Astrophysics and Space Science, 2016, 361(9): 304.
[9] He K, Zhang X, Ren S, et al. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition,2016. 770.
[10] Zhou Bolei, Aditya Khosla, Agata Lapedriza, et al. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition,2016. 2921. |
[1] |
BAO Hao1, 2,ZHANG Yan1, 2*. Research on Spectral Feature Band Selection Model Based on Improved Harris Hawk Optimization Algorithm[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2024, 44(01): 148-157. |
[2] |
LI Hu1, ZHONG Yun1, 2, FENG Ya-ting1, LIN Zhen1, ZHU Shi-jiang1, 2*. Multi-Vegetation Index Soil Moisture Inversion Model Based on UAV
Remote Sensing[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2024, 44(01): 207-214. |
[3] |
CHENG Hui-zhu1, 2, YANG Wan-qi1, 2, LI Fu-sheng1, 2*, MA Qian1, 2, ZHAO Yan-chun1, 2. Genetic Algorithm Optimized BP Neural Network for Quantitative
Analysis of Soil Heavy Metals in XRF[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2023, 43(12): 3742-3746. |
[4] |
LI Qi-chen1, 2, LI Min-zan1, 2*, YANG Wei2, 3, SUN Hong2, 3, ZHANG Yao1, 3. Quantitative Analysis of Water-Soluble Phosphorous Based on Raman
Spectroscopy[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2023, 43(12): 3871-3876. |
[5] |
HUANG Hua1, LIU Ya2, KUERBANGULI·Dulikun1, ZENG Fan-lin1, MAYIRAN·Maimaiti1, AWAGULI·Maimaiti1, MAIDINUERHAN·Aizezi1, GUO Jun-xian3*. Ensemble Learning Model Incorporating Fractional Differential and
PIMP-RF Algorithm to Predict Soluble Solids Content of Apples
During Maturing Period[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2023, 43(10): 3059-3066. |
[6] |
LI Wen-wen1, 2, LONG Chang-jiang1, 2, 4*, LI Shan-jun1, 2, 3, 4, CHEN Hong1, 2, 4. Detection of Mixed Pesticide Residues of Prochloraz and Imazalil in
Citrus Epidermis by Surface Enhanced Raman Spectroscopy[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2023, 43(10): 3052-3058. |
[7] |
CHEN Jia-wei1, 2, ZHOU De-qiang1, 2*, CUI Chen-hao3, REN Zhi-jun1, ZUO Wen-juan1. Prediction Model of Farinograph Characteristics of Wheat Flour Based on Near Infrared Spectroscopy[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2023, 43(10): 3089-3097. |
[8] |
XUE Fang-jia, YU Jie*, YIN Hang, XIA Qi-yu, SHI Jie-gen, HOU Di-bo, HUANG Ping-jie, ZHANG Guang-xin. A Time Series Double Threshold Method for Pollution Events Detection in Drinking Water Using Three-Dimensional Fluorescence Spectroscopy[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2023, 43(10): 3081-3088. |
[9] |
WANG Yu-chen1, 2, KONG Ling-qin1, 2, 3*, ZHAO Yue-jin1, 2, 3, DONG Li-quan1, 2, 3*, LIU Ming1, 2, 3, HUI Mei1, 2. Hyperspectral Reconstruction From RGB Images for Tissue Oxygen
Saturation Assessment[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2023, 43(10): 3193-3201. |
[10] |
FENG Hai-kuan1, 2, FAN Yi-guang1, TAO Hui-lin1, YANG Fu-qin3, YANG Gui-jun1, ZHAO Chun-jiang1, 2*. Monitoring of Nitrogen Content in Winter Wheat Based on UAV
Hyperspectral Imagery[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2023, 43(10): 3239-3246. |
[11] |
GUO Zhou-qian1, 2, LÜ Shu-qiang1, 2, HOU Miao-le1, 2*, SUN Yu-tong1, 2, LI Shu-yang1, 2, CUI Wen-yi1. Inversion of Salt Content in Simulated Mural Based on Hyperspectral
Mural Salt Index[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2023, 43(10): 3272-3279. |
[12] |
AN Bai-song1, 2, WANG Xue-mei1, 2*, HUANG Xiao-yu1, 2, KAWUQIATI Bai-shan1, 2. Hyperspectral Estimation of Soil Lead Content Based on Random Frog Band Selection Algorithm[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2023, 43(10): 3302-3309. |
[13] |
LI Xin1, LIU Jiang-ping1, 2*, HUANG Qing1, HU Peng-wei1, 2. Optimization of Prediction Model for Milk Fat Content Based on Improved Whale Optimization Algorithm[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2023, 43(09): 2779-2784. |
[14] |
WU Yong-qing1, 2, TANG Na1, HUANG Lu-yao1, CUI Yu-tong1, ZHANG Bo1, GUO Bo-li1, ZHANG Ying-quan1*. Model Construction for Detecting Water Absorption in Wheat Flour Using Vis-NIR Spectroscopy and Combined With Multivariate Statistical #br#
Analyses[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2023, 43(09): 2825-2831. |
[15] |
KANG Ying1, ZHUO Kun1, LIAO Yu-kun1, MU Bing1, QIN Ping2, LI Qian1, LUAN Xiao-ning1*. Quantitative Determination of Alcohol Concentration in Liquor Based on Polarized Raman Spectroscopy[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2023, 43(09): 2768-2774. |
|
|
|
|