Machine learning-based identification for ttH→invisible (upgrade)

This project focuses in a search for Higgs boson invisible decay modes has been carried out in events where the Higgs boson is produced in association with a top quark-antiquark pair (ttH). The study will focus on production modes that include only hadronic objects (jets), that drive the sensitivity in the Higgs to invisible channel. Deep Machine Learning (ML) techniques is used to classify events between multiple categories: signal regions adapted to the four dominant Higgs production processes; and categories adapted to the estimation of background sources. Three models were used in this project, the first one is recurrent neural network, the second is multilayer perceptron and the final model is more complicated model mixing between the two previous ones. The results indicates that the final model construct in this project is capable of correctly classifying the events to a certain limit. There was a remarkable issue in distinguishing between signal (ttH) and the background (𝑡𝑡̅) arises from the similarity of their variable distribution. The code for this project is written in python language using Keras and scikit-learn models and TensorFlow as a backend. All the data used are collected from a Monte Carlo simulation. This project is based on previous summer student project (note: the link to the previous summer student report is added in the references section).