Pile-Up Mitigation using Attention

TL;DR

This paper introduces PUMA, a deep learning algorithm using sparse transformers to effectively identify and mitigate pile-up effects in LHC proton-proton collision data, enhancing analysis sensitivity in high-luminosity conditions.

Contribution

The paper presents a novel pile-up mitigation algorithm based on attention mechanisms, outperforming classical methods in realistic simulations for LHC experiments.

Findings

Outperforms classical pile-up mitigation algorithms

Effective in high pile-up scenarios with up to 200 simultaneous collisions

Utilizes sparse transformer-based neural networks for particle identification

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.