恒星大气物理参数估计中的稀疏特征提取

发布时间：2018-08-07 17:37

【摘要】：随着斯隆数字巡天项目(SDSS)、欧空局GAIA和我国大天区面积多目标光纤光谱天文望远镜(LAMOST)等项目的相继实施,拥有的恒星光谱数据量急速增加,由此导致基于光谱的恒星大气物理参数自动测量方法的研究成为天文光谱分析的重要课题之一[1]。探讨了恒星光谱特征提取的方法(Haar+lasso),其基本思想是首先使用Haar小波包对原始光谱进行多尺度分解,去除高频系数,选取低频系数作为光谱信息的描述;再采用lasso算法提取最优的特征;最后将最优特征输入非参数回归模型中对恒星大气参数进行自动测量。Haar小波可以较好地去除原始光谱信号中的高频噪声,对全频谱数据进行降维。lasso算法可以进一步剔除数据冗余,提取与物理参数相关度较强的特征。Haar+lasso方法提高了物理参数自动测量的准确性和运行效率。我们采用本文方案对SLOAN发布的40 000个恒星子样本的物理参数进行测量,三个物理参数的平均绝对误差为:log Teff:0.007 1dex,log g:0.225 2dex和[Fe/H]:0.199 6dex。同现有相关文献的实验结果相比,该方案可以获得更准确的物理参数。
[Abstract]:With the implementation of the Sloan Digital Sky Survey Project (SDSS),) ESA GAIA and the large Sky area Multitarget Optical Fiber Spectral Astronomical Telescope (LAMOST) in China, the amount of stellar spectral data has increased rapidly. As a result, the research on automatic measurement of atmospheric physical parameters of stars based on spectrum has become one of the most important topics in astronomical spectral analysis [1]. The basic idea of star spectral feature extraction method (Haar lasso),) is to use Haar wavelet packet to decompose the original spectrum, remove the high frequency coefficient, and select the low frequency coefficient as the description of spectral information. Lasso algorithm is used to extract the optimal features, and the optimal features are input into the non-parametric regression model to measure the atmospheric parameters of stars automatically. Haar wavelet can remove the high frequency noise from the original spectral signals. The dimensionality reduction. Lasso algorithm for full spectrum data can further eliminate data redundancy, and extract the feature with strong correlation with physical parameters. Haar lasso method improves the accuracy and efficiency of automatic measurement of physical parameters. In this paper, we measure the physical parameters of 40,000 star sub-samples published by SLOAN. The average absolute errors of the three physical parameters are: 1% log Teff:0.007 1 dextral log g: 0.225 2dex and [Fe/H]: 0.199 6 dexter. Compared with the experimental results of related literatures, the proposed scheme can obtain more accurate physical parameters.
【作者单位】：华南师范大学数学科学学院;
【基金】：国家自然科学基金项目(61273248,61075033) 广东省自然科学基金项目(S2011010003348)资助
【分类号】：P144.1

【参考文献】