• A
  • A
  • A
  • ABC
  • ABC
  • ABC
  • А
  • А
  • А
  • А
  • А
Regular version of the site

News

TrackML challenge

Tracking Machine Learning challenge (TrackML) has been organized throughout 2018 and 2019 by a team of particle physicists and tracking experts to reach out to Computer Science and Machine Learning expertise of the wider community.

TrackML challenge

Tracking Machine Learning challenge (TrackML) has been organized throughout 2018 and 2019 by a team of particle physicists and tracking experts to reach out to Computer Science and Machine Learning expertise of the wider community.

 

DATA: an accurate simulation of an LHC-like detector that includes the measured 3D hits for each collision event, and the ground-truth label of those hits belonging to particle tracks.

The challenge was split into 2 phases:

1) ”Accuracy” phase: focus on the accuracy of the model at finding the proper point to track association. Such efficiency is essential to most of the physics analyses.

2) ”Throughput” phase: optimization of the algorithm inference speed for two CPU core configuration. This phase adds more realism to the challenge since all tracking algorithms have to process millions of events per second at LHC.

The top “Throughput” phase participants’ submissions are faster than state of the art by order of magnitude.