A bb-Tagger for light resonances with Interaction Networks

This project is based on the search for Higgs bosons decaying in the bb¯-channel in pp-collisions with the CMS-detector at the Large Hadron Collider (short LHC) at CERN. Compared with other decay channels, this channel provides relatively high statistics, as the Higgs decay into bb¯ has, with about 6 × 10−1, the largest branching ratio. On the other hand, the QCD multijet events represent an important background to the H → bb¯ process. The signal-to-background ratio can be improved considering Higgs Bosons produced with large transverse momentum (pT). In this case, their decay products are highly boosted which leads to the overlapping of the b-jets. This fact allows a good discrimination of the background. A key ingredient to exploit this signal topology is a H → bb jet tagger. As discussed in [6] this could be built training an interaction network. The networks are trained with simulated data of the H → bb¯ process with a Higgs mass of 125 GeV. This could be generalised to a generic resonance decaying to two b jets. In this case, the mass of the decaying particle is not known. Therefore the network can not be trained on the actual mass. In this project the tagging performance of the interaction networks is to be tested for a training with several different resonance masses.