| Abstract |
Over the last decade, machine learning algorithms have become standard analysis tools in astroparticle physics, used by a variety of instruments and for an even larger variety of analyses. While a few characteristic patterns can be observed, the portability of established machine learning-based analysis chains from one experiment to another, remains challenging, as instrument-specific prerequisites and adjustments need to be addressed prior to the application. The use Boosted Decision Trees and other tree-based ensemble methods, has been established, but also recently been challenged by the overall success of Deep Neural Networks. Machine learning has been applied for particle selection and parameter reconstruction, as well as for the extraction of energy spectra. This paper aims at summarizing some of the most common approaches on the application of machine learning in astroparticle physics and at providing brief overview on how they have been applied in practice.
|
