Classifying Compact X-ray Binaries in Globular Clusters with Machine Learning Techniques

Owing to the high stellar density in globular clusters, different classes of compact X-ray binaries (e.g. millisecond pulsars, cataclysmic variables ... etc) can be formed through dynamical interactions. Traditionally, they are classified by their distribution in a 2-D X-ray colour-luminosity diagram. However, with data obtained from the broadband X-ray observations with Chandra, there are many other features, such as spatial distribution and colours in different sub-bands, which have not been fully utilized in the traditional method. Here we report our attempt in classifying these X-ray binaries in a multi-dimensional feature space with machine learning techniques.