Yi-Long Chiu, Shih-Ping Lai, Gi-Ting Ho, Daw-Wei Wang
Accurate measurements of statistical properties, such as the star formation rate and the lifetime of young stellar objects (YSOs) in different stages, is essential for constraining the star formation theories. However, it is a difficult task to separate galaxies and YSOs based on SEDs alone, since their SEDs both contain various amount of stellar and dust thermal emission. Here we attempt to solve this issue by identifying galaxies and YSOs with deep learning. Deep learning is widely used to figure out patterns and classify items into categories in many subjects. We use the established YSO lists from Evans et al. (2009) and SWIRE survey data to provide knowledge bases for deep learning. We will examine how much the deep learning program can self-confirme the classification of galaxies and YSOs, and we will further examine whether there are additional YSOs in the sources catalogs of Spitzer telescope. If the deep learning program proved to be efficient, we can classify the all sky data and obtain the most complete YSO catalog with the available data.