The ability of spectroscopic sky survey to obtain spectra is increasingly growing with the development of telescope technology. The Sloan Digital Sky Survey (SDSS) and the Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST) are two of the most important multi-object fiber spectroscopic sky surveys. Number of fiber of SDSS has been increased up to 1000 from 640 and the SDSS has released more than 4,000,000 spectra. The LAMOST can observe 4000 objects one time and has released more than 2,000,000 spectra for only two years. There are many interesting and rare stellar objects, and it is quite significant for the followed research and analysis to search for these special objects from massive spectra database automatically and fast using machine learning algorithm. Our main contributions of this thesis are as follows. 1) We have applied the label propagation algorithm to search for carbon stars and DZ white dwarfs from Data Release Eight (DR8) of the Sloan Digital Sky Survey (SDSS), which is verified to be quite efficient. From nearly two million spectra including stars, galaxies, and QSOs, we have found 260 new carbon stars in which 96 stars have been identified as dwarfs and 7 identified as giants, and 11 composition spectrum systems (each of them consists of a white dwarf and a carbon star). Similarly, using the label propagation method, we have obtained 29 new DZ white dwarfs from SDSS DR8. Compared with spectra reconstructed by PCA, the 29 new findings are typical DZ white dwarfs. We have also investigated their proper motions by comparing them with proper motion distribution of 9,374 white dwarfs, and found that they have similar proper motions with the current observed white dwarfs by SDSS. In addition, we have estimated their effective temperatures by fitting the polynomial relationship between effective temperature and g-r color of known DZ white dwarfs, and found 12 of the 29 new DZ white dwarfs are cool, in which nine are between 6000K and 6600K, and three are below 6000K. 2）We have applied efficient manifold ranking learning algorithm to search for carbon stars in the LAMOST pilot survey. The performance of the efficient manifold ranking in searching for carbon stars from SDSS DR8 stellar spectra is verified comprehensively using variety of features and parameter values. It is shown that the algorithm is robust to parameters with various values, and the performance can be significantly improved if we use median filtered spectra f...
修改评论