Skip to main content
SearchLoginLogin or Signup

Classifying Serendipitous X-ray Sources from Chandra Source Catalog using Machine Learning

Presentation #404.06 in the session The Chandra Source Catalog version 2.1: New Avenues for Discovery in X-ray Datasets.

Published onJul 01, 2023
Classifying Serendipitous X-ray Sources from Chandra Source Catalog using Machine Learning

Modern X-ray observatories (e.g., Chandra, eROSITA) have detected millions of X-ray sources most of which are discovered serendipitously. Machine Learning (ML) algorithms enable timely classifications to explore the astrophysical nature of large numbers of sources, allowing quick follow-up studies on interesting sources and population studies of various kinds. The multiwavelength ML classification pipeline (MUWCLASS) that our group developed, uses the supervised ML random forest algorithm. It takes into account the measurement uncertainties and absorption/extinction biases for active galactic nuclei in our training data set. We apply the pipeline to Chandra Source Catalog version 2.0 (CSCv2) as it provides a wealth of information including photometric, spectral, and variability properties with excellent subarcsec localizations, reducing greatly the level of confusion when crossmatching to multiwavelength counterparts. We have classified more than 66,000 CSCv2 sources, consisting ~21% of the CSCv2.0, based on a training data set of 2941 X-ray sources of confidently established classes. Several narrow-focused studies on high-mass X-ray binary candidates and TeV source identification will also be discussed, among a wide range of other potential applications that would benefit the broad astrophysics community.

Comments
0
comment
No comments here