Neural network based cluster reconstruction in the ATLAS pixel detector

Dense jet environments are frequent signatures in 8 TeV proton–proton collisions, currently occurring at the LHC. These are characterised by small spatial track separations in the innermost detector layers, which can lead to the creation of shared clusters in the pixel detector. To cope with this challenging environment we present a neural network based cluster reconstruction algorithm that can identify overlapping clusters and improves the overall particle position resolution.