Classifying Unidentified X-Ray Sources in the Chandra Source Catalog Using a Multiwavelength Machine-learning Approach

Abstract

The rapid increase in serendipitous X-ray source detections requires the development of novel approaches to efficiently explore the nature of X-ray sources. If even a fraction of these sources could be reliably classified, it would enable population studies for various astrophysical source types on a much larger scale than currently possible. Classification of large numbers of sources from multiple classes characterized by multiple properties (features) must be done automatically and supervised machine learning (ML) seems to provide the only feasible approach. We perform classification of Chandra Source Catalog version 2.0 (CSCv2) sources to explore the potential of the ML approach and identify various biases, limitations, and bottlenecks that present themselves in these kinds of studies. We establish the framework and present a flexible and expandable Python pipeline, which can be used and improved by others. We also release the training data set of 2941 X-ray sources with confidently established classes. In addition to providing probabilistic classifications of 66,369 CSCv2 sources (21% of the entire CSCv2 catalog), we perform several narrower-focused case studies (high-mass X-ray binary candidates and X-ray sources within the extent of the H.E.S.S. TeV sources) to demonstrate some possible applications of our ML approach. We also discuss future possible modifications of the presented pipeline, which are expected to lead to substantial improvements in classification confidences.

Population of X-Ray Sources in the Intermediate-age Cluster NGC 3532: a Test Bed for Machine-learning Classification

“…The same confidence threshold in Equation (2) was recalculated to evaluate the confident classifications in the 5-class scheme. The performance evaluation of the pipeline using the 5-class scheme is shown in the lower panel in Figure 22 in Appendix A. classification mostly consistent with the results of Yang et al (2022). 16 These include 19 LM-STARs, 6 AGNs, 4 YSOs, 1 HM-STAR, and 1 LMXB.…”

“…CSC2 provides the mode (F mode ), as well as the lower and upper limits at 1σ confidence (F lo and F hi ) to the mode to characterize the flux distribution for each source in the catalog. We calculate the mean and the variance, using the same equation from Yang et al (2022), i.e., assuming the flux distribution to be the Fechner distribution with the equations from Possolo et al (2019).…”

“…We run MUWCLASS 1000 times, each time sampling features from their probability distribution functions, and each time producing classification probabilities for each class, based on the percent of trees in the random forest that predict that class. 15 After 1000 samplings, the mean probability (P class ) of a source belonging to each class, and its standard deviation (ΔP class ; hereafter the classification probability uncertainty) that characterizes the width of the P class distribution, can thus be calculated by incorporating uncertainty information for each feature (see Yang et al 2022 for further details).…”

“…Among the confidently classified in this work sources, only 3 sources classified as LM-STARs were classified as HM-STARs inYang et al (2022) albeit at lower confidence.17 Values taken from http://svo2.cab.inta-csic.es/svo/theory/fps3/index. php?mode=browse&gname=GAIA&asttype=.…”

“…Note that the pipeline described inYang et al (2022) did not use distances and luminosities. We added these features in this work.…”