The expectation-maximization algorithm applied to the search of point sources of astroparticles

The expectation-maximization algorithm, widely employed in cluster and pattern recognition analysis, is proposed in this article for the search of point sources of astroparticles. We show how to adapt the method for the particular case in which a faint source signal over a large background is expected. In particular, the method is applied to the point source search in neutrino telescopes. A generic neutrino telescope of an area of 1 km2 located in the Mediterranean Sea has been simulated. Results in terms of minimum detectable number of events are given and the method is compared advantageously with the results of a classical method with binning.