Skip to content

Applications of Convolutional Neural Networks in Experimental Particle Physics

Notifications You must be signed in to change notification settings

IaBrilliant/Particle-Identification

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

11 Commits
 
 
 
 
 
 

Repository files navigation

Particle-Identification

The project was completed in 2020 jointly with Benjamin Forster, supervised by Prof. Morgan Wascko. Personal responsibilities included generation of samples and CNN development.

Convolutional Neural Networks (CNNs) have been widely applied as a high-accuracy visual recognition tool to solve various classification problems. We studied the performance of a multi-layer CNN employed as a classifier for a commonly used particle identification technique - Cherenkov Ring-imaging.

Cherenkov

Cherenkov Rings (Adopted from Belle II Experiment)

The operating principle relies on measuring the Cherenkov angle when a charged particle traverses a dispersive medium. In order to generate patterns that closely resemble raw experimental data, dimensional parameters of a real proximity focusing detector (ARICH) were integrated into simulation.

Pion:Kaon

Simulated Pion and Kaon rings with the color scale representing the number of photon hits

The network was trained to classify three categories of patterns produced by pion, kaon, and proton beams with 5 GeV/c momenta. The CNN performed exceptionally well when differentiating between proton and kaon patterns, achieving 100% accuracy after a single epoch, however, it did not reach its full potential when classifying pion and kaon patterns - particles with small angular difference. To improve the efficiency we introduced changes to the network’s architecture, adding another convolutional layer with greater number of filters.

Architecture

Architecture of the Convolutional Neural Network in use

Results

Improved training and testing accuracy upon introduction of an additional layer

We found that this tuning produced a significant effect on the network’s accuracy and time of execution. We discuss the results of this investigation as well as give recommendations for improvement of simulation model.

About

Applications of Convolutional Neural Networks in Experimental Particle Physics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages