Identification of pp → K±π±K0(K0) events using artificial neural networks

An artificial neural network classifier has been developed in order to separate on-line events containing a decaying neutral kaon. The proposed system performs better than classical selection methods in real-time applications. The biases due to the on-line selection are smaller than few per cent and are known. The system is robust against calibration variations and missing information. A hardware implementation of the algorithm allows an event selection in less than 40 μs and provides a considerable on-line rate reduction in tagged neutral kaon experiments. This classifier has been developed within the framework of the CPLEAR experiment.