🎉 New Accepted Solution: Top Quark Tagging using Deep CNN

Submitted by Adit Smoak from [School/University Unknown].

Domain: Particle Physics Method: Deep CNN Result: 90.87% Accuracy Project Description

Adit Smoak tackles this challenge by applying a custom convolutional neural network built in MATLAB’s Deep Network Designer to the problem of identifying top‐quark jets in high-energy physics data. Using a multichannel "jet image" representation derived from the CERN Zenodo dataset (including energy, momentum X/Y/Z for 200 jet constituents) and embedding residual grouped-convolution + SE (Squeeze-and-Excite) blocks, Adit’s model achieved ≈ 90.87 % testing accuracy (and ≈ 90.33 % validation accuracy) on the 90k-sample training set. (github.com) Through this work, Adit demonstrates how modern deep-learning techniques can be adapted to the "jet-tagging" domain — offering a promising pathway toward real-time event filtering or trigger-level deployment in detector systems. Congratulations to Adit for delivering a creative, technically well-documented, and high‐impact solution!

Open in MATLAB Online View Repository

At a glance

• Challenge: Discriminate top-quark versus background jets using deep learning and large-scale physics data. (github.com)
• Dataset/Models: CERN Zenodo dataset (Energy, pX, pY, pZ for up to 200 jet constituents; processed to images) (github.com)
• Toolboxes/Methods: MATLAB Deep Learning Toolbox, Deep Network Designer, grouped convolution, SE-blocks, residual connections
• Key results: 90.87 % test accuracy (with ~90k training samples) and 90.33 % validation accuracy (github.com)
• Deployment: HDL code generation for FPGA, deployment scripts included in ‘deploy’ folder (github.com)

Accepted to the MATLAB & Simulink Challenge Project Hub. Congratulations again to Adit Smoak for their remarkable contribution!