Stellar spectral subclasses classification based on Isomap and SVM

Abstract

Isometric feature map (Isomap), a nonlinear dimension reduction technique, can preserve both the local and global structure of the data when embed the original data into much lower dimensional space. In this paper we will investigate the performance of Isomap+SVM in classifying the stellar spectral subclasses. We first reduce the dimension of spectra data by PCA and Isomap respectively. Then we apply support vector machine (SVM) to classify the 4 subclasses of K-type spectra from Sloan Digital Sky Survey (SDSS). The experiment result shows that Isomap-based SVM (IS) perform better than PCA-based SVM (PS) with the default γ in SVM, except on the spectra whose SNRs are between 5 and 10 in our experiment. The performance of PS and IS both change in a larger range with the increase of signal-to-noise ratio of the spectra.