A recurrent neural network for track reconstruction in the LHCb Muon System

In this work we describe an algorithm for muon track reconstruction in the LHCb Muon System. The algorithm is based on a recursive neural network known as Hopfield Network. The algorithm is particularly suitable in the harsh LHCb environment, where the expected number of hits per event in the muon detector is of the order of 2–3*10³. The algorithm has been tested successfully on cosmic ray tracks, collected during the commissioning phase of the LHCb Muon System, and on simulated proton-proton collision events, showing excellent performances. The algorithm is currently used for standalone muon reconstruction in the Muon System online monitoring.