Speaker
Description
Detection of rare neutrino interactions requires a large target mass. One of commonly used solutions is a detector composed of liquid scintillator surrounded by an array of photo-multiplier tubes. Its large dimensions and a variety of optical effects lead to non-uniformity of light collection, which significantly complicates deriving the neutrino energy from the raw photo-multiplier signals. Machine learning techniques may be a good alternative to the traditional methods usually used for this task. This talk discusses advantages and drawbacks of machine learning techniques for the energy and vertex reconstruction for two cases of near future detectors: a small scale detector prototype at Baksan Neutrino Observatory and a large 20-kiloton detector of the JUNO experiment.
