Research
Data Analysis at the CMS Experiment
The CMS experiment is a large particle detector at the LHC accelerator of the European research center CERN in Geneva, Switzerland. Together with researchers from more than 50 countries, we analyze the enormous datasets that are taken with this detector. When the particles collide in the center of the detector, particles with larger mass are created. An important particle is the Higgs boson, which was discovered at the Large Hadron Collider in 2012. We aim measure the properties of the Higgs boson with ever higher precision, to better determine the fundamental properties of the already discovered elementary particles and their interactions. An overarching question is whether the properties of elementary particles may differ from the predictions in our current best model and whether this can lead us to the discovery of new particles and interactions.
Ongoing projects (selection):
- Differential measurements in the decay channel H → γγ
- Search for rare Higgs boson processes (tH, HH) in the decay channel H → γγ
Deep Learning for Particle Physics
Analyzing large datasets requires optimized algorithms. Deep learning allows to train artificial neural networks on simulated or real data in order to represent complex relations in the experimental data. We use the neural network for a variety of applications, for example in the analyses of Higgs boson events in the CMS detector, to improve the precision of the data analysis. Examples are the classification of particle signatures in the detector, the simulation of these signatures, the determination of particle properties and the estimate of parameters of the theory of particle physics from the experimental data.
Ongoing projects (selection):
- Graph networks for photon identification
- Super resolution for calorimeter images
Data Analysis for the Einstein Telescope
The first experimental evidence of gravitational waves in 2015 has opened the field of gravitational-wave astronomy. With the Einstein Telescope, an experiment is planned that is much more sensitive than current detectors. The larger number of expected gravitational-wave signals however results in new challenges for the analysis of the data. We address these challenges by developing more efficient algorithms for the measurement of gravitational-wave signals using deep learning.
Ongoing projects:
- Determining gravitational-wave parameters with likelihood-free inference