Keras and deep learning are helping to reconstruct cosmic signals

An article on Google Developers explains how neural networks use detector tracks to study particles from space.

In an article published on August 27, Google Developers presents the uses of deep learning in astroparticle physics. Observatories often detect the indirect effects of a cosmic particle, such as showers produced in the atmosphere, rather than the original particle.

The methods described exploit the spatial and temporal structure of the measurements to reconstruct the nature, energy and direction of events. The article mentions Pierre Auger, IceCube and the Cherenkov telescopes. It is a question of shedding light on the analysis tools and their possibilities, and not of announcing a new astronomical discovery.