Identifying Intrinsic X-ray Source Population Groupings in External Galaxies: Astronomical Uses for Machine Learning
Abstract
This study examines the X-ray and optical properties of X-ray sources in order to perform classification and clustering with machine learning algorithms. We used an archival survey of X-ray emitting sources in galaxies within 15 Mpc, excluding the Milky Way and its satellites, as detected by the Chandra X-ray Observatory. The purpose of our research was to develop a grouping scheme and discover patterns related to X-ray color. We experimented with variables such as X-ray color, luminosity, and galaxy morphological type as dimensions for unsupervised machine learning algorithms and achieved the best results with just X-ray Hard and Soft color ratios. We also began to explore adding optical color as another variable. We applied supervised learning with a small training set in order to look at model performances with respect to known classification labels. While we were unable to identify a perfect algorithm, both K-Neighbors and Gaussian Mixture Modeling produced promising results.
- Publication:
-
American Astronomical Society Meeting Abstracts #233
- Pub Date:
- January 2019
- Bibcode:
- 2019AAS...23334813C