Machine learning-based identification for ttH→invisible

To measure the Higgs→invisible BR in the ttH channel, we wish to improve signal to background ratio as larger as possible. One method worth considering is using machine learning to build a multi-class classifier to identify different types of events(ttH, 𝑡𝑡̅ and QCD are considered in this project). A multi-layer perceptron(MLP) was first introduced for it is simplicity and reliability. And the results show that the MLP classifier performs well to identify the signal (ttH) from the background(𝑡𝑡̅ and QCD), but was less effective at distinguishing different backgrounds. After that, recurrent neural network(RNN) was operated to our problem. The results indicate that the RNN can reliably discriminate different backgrounds as well. Therefore, machine learning may be a promising method in the research of ttH→invisible.