Project Description
Triggers in high energy physics first reconstruct objects (e.g. jets or tracks), and then perform a selection on these objects. Event triggers select based on overall (global) event properties, avoid time-intensive reconstruction, and permit searching for new physics in areas where traditional object-based selections are too slow. In order to implement event triggers, the PhD student in this project will receive training in deep learning of complex systems, as well as other modern ML and AI methods, and apply these methods to global analyses of HEP collisions in RTA. The PhD student will focus on deep learning techniques such as recurrent neural networks, which have shown great success in speech recognition or translation, and use them to improve global analysis tools.
The first objective will be to analyse the overall pattern of detector hits to identify events enhanced in interesting physical processes, building on promising initial studies by the Dortmund group. The PhD student will design a global trigger selection and benchmark its performance against a traditional object-based approach. This will be done in collaboration with the PhD student within the network at Nikhef. The experience in global event trigger selections will be used to collaborate with IBM France’s PhD student, comparing a symbolic approach with a purely stochastic ML one. With the internship at Point 8, the PhD student will work on a real-time data-analysis project in German industry that requires a global analysis of their manufacturing processes, for example by analysing client-provided data on real-time manufacturing prediction for industrial companies, applying similar techniques as in HEP. The physics analysis focus will be Lepton Flavour Universality violation in semileptonically decaying beauty mesons, which are abundant enough to allow the event trigger to be tested against more traditional approaches.
Host country: Germany
Host beneficiary: TU Dortmund
PhD-awarding institution: TU Dortmund
Experiment affiliation: LHCb
Planned collaborations: Point 8, Nikhef