Neural Networks for Scientific Data Analysis in Particle Physics

by Dr Artemiy Belousov

Wednesday 10 Mar 2027, 10:00 → 17:00 Europe/Zurich

N260/3.13

Description

This one-day hands-on workshop introduces the fundamental concepts of neural networks, with a focus on convolutional neural networks (CNNs), and their application to scientific data. Examples come from modern analysis workflows of the CBM experiment, where CNNs are used to classify heavy-ion collision events and identify signatures of quark–gluon plasma formation.

Participants will learn how to prepare scientific datasets for machine learning and how to train and validate models on simulated data generated with transport models such as PHSD. They will then apply trained networks to realistic detector data and learn how to evaluate and interpret model performance.

The workshop builds on "Monte Carlo Simulation and Data Analysis with ROOT in Particle Physics" (March 9). Participation in that workshop is strongly recommended, as its datasets and analysis concepts serve as the starting point here.

Learning Goals

Participants will learn how to:

• Understand neural network architectures, including CNNs
• Prepare scientific datasets for machine learning
• Train and validate CNN models
• Apply machine learning methods to detector and simulation data

Program

Morning

• Introduction to machine learning in particle physics
• Fundamentals of neural networks
• Convolutional neural networks (CNNs)
• Dataset preparation and feature engineering

Afternoon

• CNN training and validation
• Event classification examples from CBM
• Performance evaluation and interpretation
• Hands-on ML exercises

Target audience: Students and researchers interested in applying machine learning to scientific data. Basic Python experience is required. Prior attendance of Workshop 1 or equivalent familiarity with simulation and detector data is recommended.