Deep Neural Network and Deep Learning Algorithms for the Classification of ggF and VBF Di-Higgs Production

Di-Higgs production, a rare process predicted by the Standard Model, allows for a deeper study of properties of the Higgs boson. Specifically, the classification of gluon-gluon fusion (ggF) and vector boson fusion (VBF) di-Higgs production modes enables precise measurements of the Higgs self-coupling as well as its coupling strength to the weak vector bosons. Successful separation ability of these processes is essential for further testing of Standard Model predictions, refining our understanding of electroweak symmetry break, and probing for potential physics beyond the Standard Model.

This talk will investigate the implementation and performance of Deep Neural Networks (DNN) and Deep Learning (DL) algorithms to classify these production modes in the $ HH \to b b \tau^+ \tau^- $ decay channel. A comparison will be made with the established Boosted Decision Tree (BDT) method and the more advanced Graph Neural Network (GNN) approach. Results indicate that DNN and DL methods offer improved separation of ggF and VBF events compared to BDT, though the GNN model shows the highest performance. Future work will focus on integrating GNN models into the current experimental framework to further increase measurement sensitivity.