- Andrea De Vita
- Enrico Lupi
Final project for Neural Networks and Deep Learning course, Physics of Data degree at University of Padova.
Triggering is a crucial aspect in high energy physics collider experiments, as it would be simply impossible to store all the raw amount of data produced. In this paper, we investigate the prospects of applying neuromorphic computing spiking neural network models to filter data coming from the Muon Chamber detectors in the CMS experiment at LHC. We first develop a model that discriminates between signal and noise based only on the position and timing of the hits. Later, we study the feasibility of a completely on-line filtering system that lets the signal go thorough to the downstream electronics while completely blocking the noise. We present our findings on the different system design choices, from net architecture and neuron type to optimal loss function to adopt.