Speaker
Description
The CMS tracking follows an iterative approach. Tracks are reconstructed in multiple passes starting from the ones that are easiest to find and moving to the ones with more complex topologies (lower transverse momentum, high displacement). A track classification is applied after each iteration using multivariate analysis and several selection requirements are defined: loose, tight, and high purity. If a track meets the high purity requirement, its hits are removed from the hit collection, thus simplifying the later iterations. The track classification and selection is therefore an integral part of the reconstruction process. Tracks passing loose selections are saved for physics analysis. The high purity selection criterion is also commonly used in the particle flow and b-jet reconstruction. The CMS experiment improved the track classification starting from a parametric selection used in Run 1, moving to a BDT in Run 2, and finally to a DNN in Run3. An overview of the DNN training and current performance is presented in this poster. The DNN improves the tracking performance both for the legacy Combinatorial Kalman Filter tracking algorithm and for the newly introduced parallelized/vectorized Kalman Filter mkFit algorithm, after being trained specifically on each algorithm.
| Consider for long presentation | No |
|---|
