标题：Spectral Recognition of M Giant Based on Deep Convolutional Neural Network
作者：Qin, Zheyun; Yi, Zhenping; Lu, Yuxiang; Pan, Jingchang; Bu, Yude
作者机构：[Qin, Zheyun; Yi, Zhenping; Lu, Yuxiang; Pan, Jingchang] Shandong Univ Weihai, Sch Mech Elect & Informat Engn, Weihai 264209, Shandong, Peoples R Chin 更多
会议名称：2nd International Conference on Big Data Research (ICBDR)
会议日期：OCT 27-29, 2018
来源：PROCEEDINGS OF THE 2018 2ND INTERNATIONAL CONFERENCE ON BIG DATA RESEARCH (ICBDR 2018)
关键词：M giant; Depth feature extraction; Convolutional neural network;; Spectral classification
摘要：Because of its similarity with the M dwarf spectrum, the M giant spectrum is prone to be mixed in M dwarfs spectra, which leads to the difficulty to select pure M giant or M dwarf samples. Convolutional Neural Network(CNN) is an effective technique to solve this problem. In this study, two CNN models, CNN_1dim and CNN_2dim, were constructed to classify M giants and M dwarfs. The experimental results show that the accuracy of the two models can reach 98.910% and 98.956% respectively, both are better than the performance of Random Forest (RF) and Gradient Boosting Decision Tree (GBDT), while using the same spectral data. This indicates that the deep convolutional neural network has an obvious advantage over the problem of identifying M giants from the M Type spectra.