Rapid Data Processing for Track Reconstruction : Filtration of simulated particle collision using Spiking Neural Networks with cone segmentation

Sammanfattning

The increasing data rates produced by modern particle detectors require innovative strategies for efficient data processing and reduction. The Large Hadron Collider (LHC) at CERN represents a prime example of this ongoing development. This study investigates the combination of charged particle dynamics and movement inside a detector with a magnetic field with neuromorphic network tools for reducing data volume in particle collision. A local data-handling method in the form of cone-regions pointing out into space, enclosing a subset of data for smaller handling. The local data is then used together with a spiking neural network leaky integrate-and-fire neuron to detect if particles of interest lie inside the region. In this study, particles of interest are the ones with a high transverse momentum over 10 GeV and a high hit count, Nhit > 10. The results demonstrate that the usage of a single spiking neuron, when applied to simulated particle collision data, can reduce the volume significantly while retaining high-momentum particles. Since the method uses information about the particles, it cannot be used on raw data from a detector and should therefore be considered a proof-of-concept study.