|
The Large Area Telescope (LAT) on board the Fermi gamma-ray space telescope (launched in 2008) has revolutionized the field of gamma-ray astronomy, producing a plethora of scientific results over a broad range of areas, from Galactic (e.g. pulsars, SNRs) to extra-Galactic (e.g. GRBs, AGN) astronomy. The number of (> 100 MeV) gamma-ray sources has increased by over an order of magnitude, to over 3,000, in just the first four years of the LAT mission. The origin of >1000 of these sources, however, remains a mystery; these are the so-called unassociated gamma-ray sources, with no known astrophysical counterpart at any other wavelength. While many LAT unassociated sources likely belong to known classes of objects (e.g. pulsars), a more exotic origin for some of these sources is also possible (e.g. dark matter annihilation). Using the properties of known gamma-ray sources, one can train a computer, via machine learning algorithms, to automatically rank and classify unassociated sources, according to their most likely class. I will present our efforts to apply a number of popular algorithms (e.g. logistic regression, random forests) to the sources in the most recent LAT catalog (3FGL). These results will aid further studies and follow-up observations of LAT unassociated sources. By prioritizing the most promising candidates in certain categories (e.g. pulsars), we will optimize the search strategy and reduce the amount of time and effort required to identify the nature of these sources. |
|