Pile-Up Mitigation using Attention

Abstract

Particle production from secondary proton-proton collisions, commonly referred to as pile-up, impair the sensitivity of both new physics searches and precision measurements at LHC experiments. We propose a novel algorithm, PUMA, for identifying pile-up objects with the help of deep neural networks based on sparse transformers. These attention mechanisms were developed for natural language processing but have become popular in other applications. In a realistic detector simulation, our method outperforms classical benchmark algorithms for pile-up mitigation in key observables. It provides a perspective for mitigating the effects of pile-up in the high luminosity era of the LHC, where up to 200 proton-proton collisions are expected to occur simultaneously.

17 pages, 6 figures, final published version

References in corpus (7)

• PyTorch: An Imperative Style, High-Performance Deep Learning Library
• An Introduction to PYTHIA 8.2
• Particle-flow reconstruction and global event description with the CMS detector
• Generating Long Sequences with Sparse Transformers
• Pileup Per Particle Identification
• Interleaved Parton Showers and Tuning Prospects
• Pileup Mitigation with Machine Learning (PUMML)

Cited by in corpus (10)

• Re-Simulation-based Self-Supervised Learning for Pre-Training Foundation Models
• Optimizing Observables with Machine Learning for Better Unfolding
• Learning Uncertainties the Frequentist Way: Calibration and Correlation in High Energy Physics
• Semi-supervised Graph Neural Networks for Pileup Noise Removal
• Bias and Priors in Machine Learning Calibrations for High Energy Physics
• Optimal transport for a novel event description at hadron colliders
• MACK: Mismodeling Addressed with Contrastive Knowledge
• Learnable cut flow for high energy physics
• Distilling particle knowledge for fast reconstruction at high-energy physics experiments
• Variational inference for pile-up removal at hadron colliders with diffusion models