The Tracking Machine Learning Challenge: Accuracy Phase

This paper reports the results of an experiment in high energy physics: usingthe power of the "crowd" to solve difficult experimental problems linked totracking accurately the trajectory of particles in the Large Hadron Collider(LHC). This experiment took the form of a machine learning challenge organizedin 2018: the Tracking Machine Learning Challenge (TrackML). Its results werediscussed at the competition session at the Neural Information ProcessingSystems conference (NeurIPS 2018). Given 100.000 points, the participants hadto connect them into about 10.000 arcs of circles, following the trajectory ofparticles issued from very high energy proton collisions. The competition wasdifficult with a dozen front-runners well ahead of a pack. The singlecompetition score is shown to be accurate and effective in selecting the bestalgorithms from the domain point of view. The competition has exposed adiversity of approaches, with various roles for Machine Learning, a number ofwhich are discussed in the document