Search for Heavy Higgs Bosons and Axion-Like Particles with the CMS Experiment via Deep Neural Networks

Supervisors: Prof. Christian Schwanenberger (DESY, UHH), Dr. Alexander Grohsjean (DESY): Prof. Peer Stelldinger (HAW)

The big unknown in particle physics is Dark Matter. To detect the presence of Dark Matter, we use data coming from the CMS experiment at the Large Hadron Collider at CERN. In this project we search both for Dark Matter particles such as axions, and for the particles that mediate their interaction with known particles, such as exotic heavy Higgs bosons. To classify the data into potential Dark Matter candidates and known elementary particles, a deep neural network is implemented, trained, and fine-tuned. There is a special challenge here, since the new physics interferes with the known processes in theory leading to a complicated expected signal+background topology that has to be separated from a backgroundonly hypothesis.
To improve the learning properties of deep neural networks in particular in analyses with such large interferences, new paths are explored to create weight rebalancing methods taking into account the variance of the loss of the weight update itself rather than just the variance of the propagated net activation. Even more important, new combinatorial approaches are investigated to handle negative event weights due to interferences consistently which will provide an answer to a problem that is currently unsolved in high energy physics.