Online Particle Detection by Neural Networks Based on Topologic Calorimetry Information

The neural ringer is an alternative algorithm (for both feature extraction and hypothesis testing) for electron identification at the ATLAS L2 calorimetry trigger. The feature extraction consists on calculating concentric energetic rings at each calorimeter layer. For each layer, the first ring is the energy from the hottest cell, and the energy of the outer cells are summed up forming the second ring (and sequentially for the other rings). A feedforward MLP neural network operates over the extracted rings performing particle identification. This study shows the later resuls considering improvements on the HLT implementation and performance evaluation over pileup from Monte Carlo proton-proton collisions simulations of 14 TeV at 2e34 luminosity.