Reconstruction of IACT events using deep learning techniques with CTLearn

Arrays of imaging atmospheric Cherenkov telescopes (IACT)
are superb instruments to probe the very-high-energy gamma-ray
sky. This type of telescope focuses the Cherenkov light emitted from
air showers, initiated by very-high-energy gamma rays and cosmic rays,
onto the camera plane. Then, a fast camera digitizes the longitudinal
development of the air shower, recording its spatial, temporal, and
calorimetric information. The properties of the primary very-high-energy
particle initiating the air shower can then be inferred from those
images: the primary particle can be classified as a gamma ray or a
cosmic ray and its energy and incoming direction can be estimated.
This so-called full-event reconstruction, crucial to the sensitivity
of the array to gamma rays, can be assisted by machine learning
techniques. We present a deep-learning driven, full-event
reconstruction applied to simulated, IACT events using
CTLearn. CTLearn is a Python package that includes modules for loading
and manipulating IACT data and for running deep learning models with
TensorFlow, using pixel-wise camera data as input.

Theme – Machine Learning, Statistics, and Algorithms