Search for the standard model Higgs boson in the bbH production mode with Deep Neural Networks at CMS

Supervisors: Prof. Elisabetta Gallo (DESY, UHH), Dr. Rainer Mankel (DESY), Dr. Roberval Walsh (DESY), Prof. Peer Stelldinger (HAW)

The ATLAS and CMS experiments at Large Hadron Collider (LHC) at CERN are very successfully performing tests of the Standard Model (SM) of elementary particles and their interactions. This success is particularly highlighted by the discovery in 2012 of the SM Higgs boson, which has been extensively studied in the last years. The associated bbH production, due to the low rate and overwhelming background, has not been studied yet and could be enhanced by beyond the SM physics. This project aims at looking at associated bbH production with the CMS data taken in Run 2. Machine learning techniques will have to be applied in order to improve the sensitivity. The classification task is to be solved by a feedforward neural network. State of the art learning algorithms still tend to get stuck during learning, especially in case of a very low signal to background ratio like at the CMS data. While this is partly solved by modern approaches like batch normalization, these methods only take into account the variance of the propagated net activation and not the variance of the loss or the weight update itself. It is aimed to solve this by using new weight rebalancing methods.