Jet identification based on probability calculations using Bayes’ theorem

The problem of identifying jets at CERN LEP and DESY HERA is studied. Identification using jet energies and fragmentation properties is treated separately in order to investigate the degree of quark-gluon separation that can be achieved by either of these approaches. In the case of the fragmentation-based identification, a neural network is used, and a test of the dependence on the jet production process and the fragmentation model is done. Instead of working with the separation variables directly, these are used to calculate probabilities of having a specific type of jet, according to Bayes’ theorem. This offers a direct interpretation of the performance of the jet identification and provides a simple means of combining the results of the energy- and fragmentation-based identifications.