Application of deep learning in top pair and single top quark production at the LHC

We demonstrate the performance of a very efficient top tagger applies on hadronically decaying boosted top quark pairs as signal based on deep neural network algorithms working with Lorentz Layer and the Minkowskimetric. Due to limited computing resources, we could show only the receiver ordering characteristic curve, accuracy and loss which illustrates the trade-off between signal acceptance against huge QCD multi-jet background acceptance. Alternatively, we also report the modern machine learning approaches and applymultivariate technique on single top quark production through weak interaction at v root s = 14 TeV proton-proton Collider to demonstrate its observability against the most relevant Standard Model backgrounds through the techniques of boosted decision tree (BDT), likelihood and multilayer perceptron (MLP). The analysis is trained to observe the performance of classifiers in comparison with the conventional cut based and counting approach.

Automated summary

This article demonstrates the performance of an efficient top tagger for hadronically decaying boosted top quark pairs using deep neural network algorithms and Lorentz Layer with the Minkowskimetric. Due to limited computing resources, the analysis shows the receiver operating characteristic curve, accuracy, and loss, highlighting the trade-off between signal acceptance and QCD multi-jet background acceptance. In addition, modern machine learning techniques such as boosted decision tree (BDT), likelihood, and multilayer perceptron (MLP) are used to observe the observability of single top quark production through weak interaction against relevant Standard Model backgrounds, comparing their performance with conventional cut based and counting approaches.