Study of the Higgs Boson in the ZH → l⁺l⁻bb̄ Process Using Various Machine Learning Algorithms

This is the authors' abstract. We don't add key points for this paper.

A study of the Higgs boson in associated production with the Z boson and decay into b quarks (H→ bb̄ ) is performed using various machine learning algorithms. Only the decay of the Z boson into charged leptons, an electron-positron pair, or a muon-antimuon pair was selected. Signal-background classification was carried out using three machine learning algorithms: boosted decision tree, artificial neural network, and deep neural network. Signal and background events used for training were generated by the Monte Carlo event generators Powheg and Sherpa. About a million signal and background events were used to train machine learning algorithms. After optimizing all three algorithms, it was found that the best results in terms of performance or signal significance were obtained using artificial neural network. Switching from TMVA to TensorFlow, leveraging parallelism, has significantly enhanced neural network training performance. This advancement has effectively resolved the previously long training times in artificial neural networks, making them comparable to or better than those of boosted decision trees.