Searching for Young Stellar Objects with Convolutional Neural Network

Chi-Ting Ho^1*, Yi-Lung Chiu³, Daw-Wei Wang^1,2, Shih-Ping Lai^1,3

¹Physics Department, National Tsing Hua University, Hsinchu, Taiwan
²Physics Division, National center for Theoretical Science, Hsinchu, Taiwan
³Institute of Astronomy, National Tsing Hua University, Hsinchu, Taiwan

* Presenter:Chi-Ting Ho, email:aeio6646@yahoo.com.tw

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 spectral energy distributions (SEDs) alone, since their SEDs both contain various amount of stellar and dust thermal emission. Most of the time, one may distinguish YSOs from galaxies from the total flux of SEDs, reflecting the effects of distance.
Here we develop a machine learning tool using Convolutional Neural Network (CNN) to classify regular stars, galaxies, and YSOs, solely based on their normalized SEDs where the distance factor is completely removed. This tool is solely trained by labeled data without any priori theoretical knowledge. We show that the trained CNN can provide very good recall and precision rates ( > 92% for YSOs ), thus it is a very good robot to classify YSOs and to study the physical reasons behind it. For example, we find the most important contribution comes from the data in the low energy region, where the extinction effect is much less influential.

Keywords: Young Stellar Objects, Convolutional Neural Network