MINOS: Neural Network Analysis

Description of implemented Neural Nets

Neural Network for NC / CC Neutrino Event Classification

Inputs: Event length Maximum event width Number of hit strips Total pulse height ANN Architecture: 4 (inputs) - 6 - 1 (output) see: Selection of Optimum Neural Net Architecture Advantages: Event classification does not depend on reconstructed information. It is therefore in-sensitive to reconstruction deficiencies. There are only 4 inputs of rather simple and characteristic event features. It is therefore simpler to check/ensure that the MC matches real data in these 4 event features.

Neural Net Input Distributions

Correlations between Neural Net Inputs