Convolutional Neural Networks for event classification

Abstract

Cutting-edge Artificial Intelligence is being implemented in a wide range of tasks in High Energy Physics (HEP) in order to facilitate the analysis of large datasets. However, visual recognition has not been explored as much in HEP for event classification. This study shows how Convolutional Neural Networks could be applied for such an important task, for which a novel method to represent the event information in images is explored. This technique is applied for a classification problem corresponding to a search for Dark Matter in proton-proton collisions. The results obtained with this technique are also compared with the performance of a Boosted Decision Tree.