Abstract

Experimental particle physics uses machine learning for many of tasks, whereone application is to classify signal and background events. The classificationcan be used to bin an analysis region to enhance the expected significance fora mass resonance search. In natural language processing, one of the leadingneural network architectures is the transformer. In this work, an eventclassifier transformer is proposed to bin an analysis region, in which thenetwork is trained with special techniques. The techniques developed here canenhance the significance and reduce the correlation between the network'soutput and the reconstructed mass. It is found that this trained network canperform better than boosted decision trees and feed-forward networks.

 

Quick Read (beta)