Deep Learning on the LHCb detector at CERN

Student: Phillip Marshall
Supervisor: Themis Bowcock

The application of Machine Learning techniques to analyse data from Particle Physics experiments has already had a large impact across a wide range of physics topics. Deep Learning on the LHCb Detector at CERN will apply Deep Learning to flavour physics at the LHC. LHCb raw data rates can be as high as many 10s of TBytes/s, and both real time data reduction and analysis represent major technical undertakings. Serious challenges will be presented as the data volume increases over the next decade. Deep learning techniques  will be used to investigate the long term aim of performing standard calibration style tasks more efficiently, improving the resources required for data reduction, and with the aim of removing the quasi-arbitrariness introduced by the human analyst in searching for new physics.  Working with LBDN and Fondazione Bruno Kessler (FBK), who are experts in Deep Learning in large scale engineering systems, existing DL techniques will be applied to the alignment of the tracker system. Novel and dynamic application of DL to dynamic event triggering for Lepton Flavour Violation in B decays will also be studied. This promises breakthrough in performance and methodology of the experiment. The information to be used to drive the analyses will be both real and simulated (MC) data from LHCb.