By mapping the colours of galactic sources onto the Hubble filter system, this pipeline classifies extragalactic X-ray sources into seven classes with 99.6 per cent accuracy.
Most of the X-ray sources spotted by the Chandra Observatory in nearby galaxies have never been classified. Researchers propose XClass, a machine learning chain that divides them into seven families: active galactic nuclei, low- and high-mass X-ray binaries, cataclysmic variables, low- and high-mass foreground stars, and supernova remnants.
Bypass the filter problem
The difficulty lies in the training data: the known sources are mostly galactic and measured by wide-field surveys such as Pan-STARRS and 2MASS, while extragalactic targets only have Hubble images, taken in other filters. The team therefore fits a spectral model to each training source to deduce synthetic magnitudes in the Hubble filters. A two-stage random forest first sorts the broad categories, then distinguishes between the two types of X-ray binaries.
On a reference sample of 11,374 sources, the system achieves 99.6% accuracy and a balanced accuracy of 0.90, with very well-calibrated probabilities. It will be applied to the Andromeda (M31) and Triangulum (M33) galaxies in a second paper.
Read the full article on the original source
Go to the source →