Abstract
Deep neural networks (DNNs) have been applied to the fields of computer vision and natural language processing with great success in recent years. The success of these applications has hinged on the development of specialized DNN architectures that take advantage of specific characteristics of the problem to be solved, namely convolutional neural networks for computer vision and recurrent neural networks for natural language processing. This research explores whether a neural network architecture specific to the task of identifying t → Wb decays in particle collision data yields better performance than a generic, fully-connected DNN. Although applied here to resolved top quark decays, this approach is inspired by an DNN technique for tagging boosted top quarks, which consists of defining custom neural network layers known as the combination and Lorentz layers. These layers encode knowledge of relativistic kinematics applied to combinations of particles, and the output of these specialized layers can then be fed into a fully connected neural network to learn tasks such as classification. This research compares the performance of these physics inspired networks to that of a generic, fully-connected DNN, to see if there is any advantage in terms of classification performance, size of the network, or ease of training.
Highlights
The process of reconstructing top quarks from collision data is an important first step in studying rare top quark processes such as the ttH process
The physics inspired Lorentz neural network was shown to provide a small performance boost on a top quark reconstruction task compared to a standard fully connected, feed-forward deep neural network
While this performance gain was modest, the differences in network output plotted against physics variables shows that there are important differences in the ways that networks learn when encoded with existing knowledge of the problem
Summary
The process of reconstructing top quarks from collision data is an important first step in studying rare top quark processes such as the ttH process. This task can grow difficult when looking to study a process that can produce as many as ten jets in a single event. Increasing the performance of top reconstruction methods would improve ability to measure rare processes at the LHC that produce many jets. This study aims to improve reconstruction techniques for resolved top quarks rather than focusing on the boosted regime, since much effort has already gone into reconstructing boosted top quarks.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have