An online neural network triggering system for the Tile Calorimeter

For the hadronic calorimeter of ATLAS, TileCal, neural processing is used to establish an efficient methodology for the online particle identification in beam tests of calorimeter prototypes. Although beam purity is usually very good for a selected particle type, background from wrong-type particles cannot be avoided and is routinely identified in the offline analysis. The proposed neural system is trained online to identify electrons, pions, and muons at different energy levels and it achieves more than 90% efficiency in terms of particle identification. The neural system is being implemented by integrating it to the readout drive (ROD) of the TileCal.