Neural triggering system operating on high resolution calorimetry information

This paper presents an electron/jet discriminator system for operating at the Second Level Trigger of ATLAS. The system processes calorimetry data and organizes the regions of interest in the calorimeter in the form of concentric ring sums of energy deposition, so that both signal compaction and high performance can be achieved. The ring information is fed into a feed forward neural discriminator. This implementation resulted on a 97% electron detection efficiency for a false alarm of 3%. The full discrimination chain could still be executed in less than 500μs.