A comprehensive search for X-ray detected intermediate-mass black holes

Intermediate-mass black holes (IMBH) are thought to be the seeds which give rise to the formation of supermassive black holes. The observational evidince for these seeds remains elusive, with the possible exception of the recently discovered gravitational wave event GW190521, massive black holes seen in dwarf galaxies and HLX-1. As a member of the sparse class of hyperluminous X-ray sources (HLX, brighter than 10⁴¹ erg/s in X-rays and located outside the nucleus of a massive galaxy), and showing a tremendous variability which suggests partial tidal disruption events, HLX-1 is the perfect illustration of two methods to probe for IMBH in large X-ray surveys. However, this data mining task first requires a reliable identification of large fractions of existing X-ray detections in the archives, still unexplored to a large extent. To this end, I will present a general-purpose, probabilistic approach to classify X-ray sources found in catalogues, in which sources are classified based on their spatial, spectral and variability properties on different timescales and their multi-wavelength counterparts. I will show how this optimized Naive Bayes classifier can be applied to clean up samples of ULXs and HLXs, and demonstrate its reliability and suitability to data mining purposes for the Swift, XMM and Chandra catalogues. An outlier measure is used to identify rare objects, including tidal disruption events. Last but not least, I will show recent results from applying this classification to the X-ray populations in dwarf galaxies.