Generative AI for Correcting Detector Effects in Particle and Nuclear Physics: A Comprehensive Survey
College
College of Sciences
Department
Computer Science
Graduate Level
Doctoral
Graduate Program/Concentration
AI/ML
Presentation Type
Poster Presentation
Abstract
In particle and nuclear physics, experimental data are often distorted by detector effects such as smearing, acceptance, and inefficiencies, necessitating correction through unfolding techniques. Recent advancements in generative AI have introduced powerful approaches for addressing these challenges, offering scalability to high-dimensional data, the ability to model complex detector responses, and support for event-level analysis. This survey explores the application of generative AI models—such as Generative Adversarial Networks, Variational Autoencoders, Normalizing Flows, and Diffusion Models—for unfolding detector effects. We provide a systematic review of existing frameworks, methodologies, and implementations, highlighting key developments, practical challenges, and areas for further improvement. Additionally, we discuss critical aspects such as latent space representations, uncertainty quantification, physics-informed constraints, and background subtraction. By synthesizing recent progress in the field, we illustrate how generative AI can significantly enhance data reconstruction and analysis in particle and nuclear physics experiments.
Keywords
Generative AI, Detector Effects, High Energy Physics (HEP), Unfolding
Generative AI for Correcting Detector Effects in Particle and Nuclear Physics: A Comprehensive Survey
In particle and nuclear physics, experimental data are often distorted by detector effects such as smearing, acceptance, and inefficiencies, necessitating correction through unfolding techniques. Recent advancements in generative AI have introduced powerful approaches for addressing these challenges, offering scalability to high-dimensional data, the ability to model complex detector responses, and support for event-level analysis. This survey explores the application of generative AI models—such as Generative Adversarial Networks, Variational Autoencoders, Normalizing Flows, and Diffusion Models—for unfolding detector effects. We provide a systematic review of existing frameworks, methodologies, and implementations, highlighting key developments, practical challenges, and areas for further improvement. Additionally, we discuss critical aspects such as latent space representations, uncertainty quantification, physics-informed constraints, and background subtraction. By synthesizing recent progress in the field, we illustrate how generative AI can significantly enhance data reconstruction and analysis in particle and nuclear physics experiments.