Tracking in nuclear physics experiments is the most computationally extensive
procedure. In CLAS12 detector data reconstruction tracking takes 94% of
computational time. Due to high particle multiplicity a lot of time is spend
on combinatorial search through track candidates.
In this work we present our use of Neural Networks to identify best track
candidates based on segments information from 6 super-layers of drift chambers.
The developed Neural Network achieves 96% accuracy of track identification
and improves the reconstruction speed by factor of 3-6.
|Second most appropriate track (if necessary)||Architectures and techniques for real-time tracking and fast track reconstruction|
|Consider for young scientist forum (Student or postdoc speaker)||No|