Potentiality of automatic parameter tuning suite available in ACTS track reconstruction software framework

• DOI: 10.48550/arxiv.2309.12422
• Publication Date: 2023-01-01
• Authors: R. B. Garg; C. Allaire; A. Salzburger; Hadrien Grasland; L. Tompkins; E. Hofgard
• Keywords: Maintainability; Tracking (education); Bayesian Optimization; Software suite

Particle tracking is among the most sophisticated and complex part of full event reconstruction chain. A number algorithms work in a sequence to build these trajectories from detector hits. These use many configuration parameters that need be fine-tuned properly account for detector/experimental setup, available CPU budget desired physics performance. The popular method tune hand-tuning using brute-force techniques. techniques can inefficient raise issues long-term maintainability such algorithms. open-source track software framework known as "A Common Tracking Framework (ACTS)" offers an alternative solution parameter tuning through automatic optimization ACTS comes equipped with auto-tuning suite provides necessary setup performing input belonging user choose tunable flexible way define cost/benefit function optimizing fast execution speed allows run several iterations within reasonable time bracket. performance optimizers has been demonstrated on different trajectory seed selection, particle vertex generation simplified material map, geometries Generic Detector Open Data (ODD). We aim bring this approach all aspects by having more integration ACTS.