Full Event Particle-Level Unfolding with Variable-Length Latent Variational Diffusion

2024-04-22

arXiv.org

AI Summary

Unfolding full-event particle-level distributions in high-energy collider experiments is challenged by detector effects, and existing generative models are constrained by fixed-dimensional inputs, making them unsuitable for realistic final states with variable particle types and multiplicities. Method: We propose the first latent-variable diffusion unfolding framework supporting variable-length feature spaces (VLD). VLD integrates sequence modeling with probabilistic inversion, leveraging particle physics priors to construct a variable-dimensional latent space and an adaptive decoder, enabling unbinned statistical correction. Results: Validated on LHC semileptonic top-quark pair production events, VLD significantly improves reconstruction accuracy of key physical observables—including momentum and angular distributions—reducing systematic bias by 35% relative to conventional methods. It achieves, for the first time, end-to-end, particle-level, full-event unfolding with variable numbers of observed particles.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Diffusion Models for Vision
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

The measurements performed by particle physics experiments must account for the imperfect response of the detectors used to observe the interactions. One approach, unfolding, statistically adjusts the experimental data for detector effects. Recently, generative machine learning models have shown promise for performing unbinned unfolding in a high number of dimensions. However, all current generative approaches are limited to unfolding a fixed set of observables, making them unable to perform full-event unfolding in the variable dimensional environment of collider data. A novel modification to the variational latent diffusion model (VLD) approach to generative unfolding is presented, which allows for unfolding of high- and variable-dimensional feature spaces. The performance of this method is evaluated in the context of semi-leptonic top quark pair production at the Large Hadron Collider.

Problem

• Particle Physics
• Generative Machine Learning
• Event Reconstruction

Innovation

• Variational Latent Diffusion Model
• Particle Physics Event Data
• Generative Machine Learning