Particle Physics and Machine Learning

Welcome to the ParticlePhysicsAndMachineLearning repository! 🚀 This project explores the intersection of Particle Physics and Machine Learning (ML) by leveraging deep learning models for physics event classification.

🔬 Project Overview

This repository aims to apply modern deep learning techniques to analyze and classify particle physics events using publicly available datasets. The primary focus is on electron-photon using models like ResNet.

⚡ Getting Started

1️⃣ Clone the Repository

The actual data is extremely large and requires to be downloaded through curl.

git clone https://github.com/MandaKausthubh/ParticlePhysicsAndMachineLearning.git
cd ParticlePhysicsAndMachineLearning
mkdir data
cd data
curl -o Photon.hdf5 https://cernbox.cern.ch/remote.php/dav/public-files/AtBT8y4MiQYFcgc/SinglePhotonPt50_IMGCROPS_n249k_RHv1.hdf5
curl -o Electron.hdf5 https://cernbox.cern.ch/remote.php/dav/public-files/FbXw3V4XNyYB3oA/SingleElectronPt50_IMGCROPS_n249k_RHv1.hdf5
cd ..

⚙️ Creating an Environment

Please note that this uses conda to create and manage the environment.

conda env create -f environment.yml
conda activate ML4Sci

🏋️ Downloading weights directly:

Please note the direct weights are available in the file : ModelWeights/BestModelElectronPhoton.pth
These can be directly loaded into a pytorch model without the prior architecture and be used.

Achievemets:

This successfully replicates the predictions of the paper: E2E CMS paper. Achieved an AOC Score of 79.02% (the paper predicts 79%). This is one of the best results in this problem setup it the world!!

Acknowledgement: CMS Data CERN

📩 Contact

For questions or collaborations, reach out to Manda Kausthubh.

---

Let's push the boundaries of Particle Physics & Machine Learning together! ⚛️🤖