Abstract
In this paper we propose the Nearest-Neighbor Technique (NNT), a classical nonparametric method of Pattern Recognition, for addressing Particle Identification (PID) problems. Given a set of correctly identified events and a distance measure between events, the NNT assigns a previously unseen event the same class of the nearest event in the set. The NNT can be applied to correlated data with virtually no preprocessing and does not require assumptions on the shape of the various probability distributions. We compare the performance of NNT on electron identification in experiment E760 at FNAL against a standard likelihood technique, and find that NNT gives better results, particularly in the rejection rates. We conclude that NNT, which is conceptually clear and easy to implement and use, can be very useful for PID.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call