Applying machine-learning to understand the nature of gamma-ray sources

New generation gamma-ray observatories (such as Fermi-LAT, H.E.S.S., and VERITAS) have discovered hundreds of Galactic sources, many of which are extended and unidentified. Multiwavelength observations help to constrain the nature of these sources. We aim to classify the X-ray sources residing within the extent of the unidentified gamma-ray sources using a machine-learning approach. By compiling mutiwavelength information (X-ray, optical, infrared) from existing catalogs, a training data set has been created and used to classify the unidentified sources with several machine-learning algorithms. We will present the methods and the results of our classification. We will also discuss the application of these methods in the light of upcoming observatories (e.g., CTA, Astro-H, eROSITA).