This paper reports the results of an experiment in high energy physics

Abstract. Charged particle track reconstruction is a major component of data-processing in high-energy physics experiments such as those at the Large Hadron Collider (LHC), and is foreseen to become more and more challenging with higher collision rates. A simplified two-dimensional version of the track reconstruction problem is set up on a collaborative platform, RAMP, in order for the developers to prototype and test new ideas. A small-scale competition was held during the Connecting The Dots / Intelligent Trackers 2017 (CTDWIT 2017) workshop. Despite the short time scale, a number of different approaches have been developed and compared along a single score metric, which was kept generic enough to accommodate a summarized performance in terms of both efficiency and fake rates. Data challenges and RAMPAdvances in machine-learning technology is moving towards a new paradigm in data processing design, in which the role of the individual will have to be redefined. In order to integrate the human creativity with the computational resources in an efficient way, various data-challenge platforms have been emerging, which provide a competitive and/or collaborative environment to crowd-source the analysis expertise.

We propose a novel approach to charged particle tracking at high intensity particle colliders based on Approximate Nearest Neighbors search. With hundreds of thousands of measurements per collision to be reconstructed e.g. at the High Luminosity Large Hadron Collider, the currently employed combinatorial track finding approaches become inadequate. Here, we use hashing techniques to separate measurements into buckets of 20-50 hits and increase their purity using metric learning. Two different approaches are studied to further resolve tracks inside buckets: Local Fisher Discriminant Analysis and Neural Networks for triplet similarity learning. We demonstrate the proposed approach on simulated collisions and show significant speed improvement with bucket tracking efficiency of 96 % and a fake rate of 8 % on unseen particle events. * Corresponding author 2 In the contrast to cars, however, we have quasi-deterministic equations of motions that govern the particle trajectory.

The High-Luminosity LHC (HL-LHC) is expected to reach unprecedented collision intensities, which in turn will greatly increase the complexity of tracking within the event reconstruction. To reach out to computer science specialists, a tracking machine learning challenge (TrackML) was set up on Kaggle by a team of ATLAS, CMS, and LHCb physicists tracking experts and computer scientists building on the experience of the successful Higgs Machine Learning challenge in 2014. A training dataset based on a simulation of a generic HL-LHC experiment tracker has been created, listing for each event the measured 3D points, and the list of 3D points associated to a true track.The participants to the challenge should find the tracks in the test dataset, which means building the list of 3D points belonging to each track.The emphasis is to expose innovative approaches, rather than hyper-optimising known approaches. A metric reflecting the accuracy of a model at finding the proper associations that matter most to physics analysis will allow to select good candidates to augment or replace existing algorithms.

This paper reports on the second “Throughput” phase of the Tracking Machine Learning (TrackML) challenge on the Codalab platform. As in the first “Accuracy” phase, the participants had to solve a difficult experimental problem linked to tracking accurately the trajectory of particles as e.g. created at the Large Hadron Collider (LHC): given $$O(10^5)$$ O ( 10 5 ) points, the participants had to connect them into $$O(10^4)$$ O ( 10 4 ) individual groups that represent the particle trajectories which are approximated helical. While in the first phase only the accuracy mattered, the goal of this second phase was a compromise between the accuracy and the speed of inference. Both were measured on the Codalab platform where the participants had to upload their software. The best three participants had solutions with good accuracy and speed an order of magnitude faster than the state of the art when the challenge was designed. Although the core algorithms were less diverse than in the first phase, a diversity of techniques have been used and are described in this paper. The performance of the algorithms is analysed in depth and lessons derived.

No abstract

No abstract

No abstract