LAMBDA Seminar on 'Pattern Recognition Techniques for Finding Very Rare Events in COMET Experiment'

Ewen Gillies, PhD student at the Imperial College of London, COMET experiment member, LAMBDA Research Assistant presented the report 'Pattern Recognition Techniques for Finding Very Rare Events in COMET Experiment' at LAMBDA seminar.

Abstract:

The Coherent Muon to Electron Transition (COMET) experiment is an upcoming experiment probing for as of yet unseen physics.  It aim is to test one of the most fundamental conservation laws in particle physics, charged lepton flavour conservation, at unprecedented sensitivities.  Due to the accuracy of measurement required and volume of data analysed, machine learning and computer vision techniques have been applied to extract key measurements. Specifically, a series of gradient boosted decision trees have been used in conjunction with Hough transforms to separate background from signal events by using both local and shape based features. This method outperforms typical Hough transform and machine learning techniques used thus far on COMET.

Working materials: 

 COMET Experiment (PDF, 3.52 Мб)