Deep Learning Tutorials for Experimental Nuclear Physics

Welcome to the DNP 2025 Deep Learning Tutorials! This GitBook provides comprehensive tutorials on applying deep learning techniques to experimental nuclear physics, specifically for the Forward Calorimeter (FCAL) at GlueX.

Overview

These tutorials are designed for physicists and researchers working on the GlueX experiment at Jefferson Lab. We focus on two main applications of deep learning:

1. CNN-based Classification: Using Convolutional Neural Networks to classify FCAL showers and distinguish between photons and splitoffs
2. Generative AI: Building generative models to simulate FCAL showers for different particle types (π⁺, π⁻, and photons) based on their kinematics and particle identification (PID)

Tutorial Goals

By the end of these tutorials, you will be able to:

• Understand the physics behind FCAL showers and the importance of accurate classification
• Prepare and preprocess FCAL data for deep learning applications
• Build and train CNN models for binary classification (photons vs splitoffs)
• Develop generative models (GANs, VAEs, or diffusion models) for FCAL shower simulation
• Evaluate model performance using physics-informed metrics
• Apply these techniques to your own nuclear physics research

Target Audience

These tutorials are intended for:

• Graduate students and postdocs in nuclear physics
• Researchers working on calorimeter systems
• Scientists interested in applying ML to experimental physics
• Anyone wanting to learn about deep learning in the context of particle physics

Prerequisites

• Basic understanding of nuclear physics and particle detectors
• Python programming experience
• Familiarity with NumPy and basic data analysis
• (Optional) Prior exposure to machine learning concepts

Tutorial Structure

The tutorials are organized into several chapters, progressing from fundamental concepts to advanced applications. Each chapter includes:

• Theoretical background
• Hands-on code examples
• Exercises and challenges
• References for further reading

Event Information

DNP 2025 Tutorial Session
Date: October 17, 2025
Location: Chicago, IL

For questions or feedback, please open an issue on our GitHub repository.

---

This tutorial is part of the AI4EIC collaboration's effort to bring modern machine learning techniques to experimental nuclear physics.