Particle identification with machine learning from incomplete data in the ALICE experiment

Aug 4, 2026

Upload by Ezra

The ALICE experiment at the LHC studies strongly interacting matter produced in ultrarelativistic heavy-ion collisions, where precise particle identification (PID) is essential. ALICE derives PID from multiple detectors covering momenta from about 100 MeV/c to 20 GeV/c, while traditional rectangular cuts are limited. The work develops an ML-based PID strategy using several neural networks as binary classifiers. It further adds Feature Set Embedding and an attention mechanism to train on incomplete samples, and discusses integration into the ALICE analysis software and domain adaptation between simulation and real data.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages10

Reading Time25 min

AccessFree to download

Introduction

ALICE and the need for PID

Detector-based PID observables

PID with machine learning

Neural network approach

Integration with the O2 framework

Feature Set Embedding and the attention mechanism

Domain Adversarial Neural Networks

Conclusions and outlook

arXiv :2403 . 17436v3 [hep-ex] 25 Jul 2024 Prepared for submission to JINST 3rd Artificial Intelligence for the Electron Ion Collider workshop–AI4EIC2023 November 28-December 1, 2023 Catholic University of America, Washington (D.C.), USA Particle identification with machine learning from incomplete data in the ALICE experiment Maja Karwowska, 􀀰,􀀱,1 Łukasz Graczykowski 􀀰 Kamil Deja􀀲,􀀳 Miłosz Kasak􀀲 and Małgorzata Janik􀀰 on behalf of the ALICE collaboration 􀀰 Faculty of Physics, Warsaw University of Technology Koszykowa 75, 00-662 Warsaw, Poland 􀀱 CERN– European Organization for Nuclear Research Espl. des Particules 1, 1211 Geneva, Switzerland 􀀲 Faculty of Electronics and Information Technology, Warsaw University of Technology Nowowiejska 15/19, 00-665 Warsaw, Poland 􀀳 IDEAS NCBR Chmielna 69, 00-801 Warsaw, Poland E-mail: [maja.karwowska@cern.ch](maja.karwowska@cern.ch) Abstract: The ALICE experiment atthe LHC measures properties of the strongly interacting matter formed in ultrarelativistic heavy-ion collisions. Such studies require accurate particle identification (PID) . ALICE provides PID information via several detectors for particles with momentum from about 100 MeV/c up to 20 GeV/c. Traditionally, particles are selected with rectangular cuts. A much better performance can be achieved with machine learning (ML) methods. Our solution uses multiple neural networks (NN) serving as binary classifiers. Moreover, we extended our particle classifier with Feature Set Embedding and attention in order to train on data withincomplete samples. We also present the integration of the ML project with the ALICE analysis software, and we discuss domain adaptation, the ML technique needed to transfer the knowledge between simulated and real experimental data. Keywords: Particle identification methods, Analysis and statistical methods, Data processing methods 1Corresponding author. Contents 1 Introduction 1 2 PID with machine learning 2 2.1 Neural network approach 2 2.2 Integration of PID ML with the O2 framework 3 3 Feature Set Embedding and the attention mechanism 4 4 Domain Adversarial Neural Networks 6 5 Conclusions and outlook 7 1 Introduction ALICE (A Large Ion Collider Experiment) [1] is one of the four major detectors located at the Large Hadron Collider (LHC) at CERN [2] . The main goal of ALICE is to measure the properties of the quark–gluon plasma (QGP), a deconfined state of quarks and gluons, theorized to exist in the early Universe [3] . Detailed studies of QGP require very precise particle identification (PID), i.e., the ability to discriminate between different particle species produced during the collision. High PID accuracy distinguishes ALICE from other LHC experiments. It also allows for selecting a subset of particles required for specific analysis. Thanks to several detectors operating concurrently, various types of particles can be separated over a wide range of momentum from just around 100 MeV/c up to around 10 GeV/c. Figure 1 presents a scheme of the ALICE detectors as used during the Run 1 and Run 2 LHC data-taking periods. In particular, particle identification over the full azimuthal angle uses information from three detectors: Time Projection Chamber [5](TPC), Time-of-Flight [6](TOF), Transition Radiation Detector [7](TRD) . The TPC is one of the most important ALICE detectors as it records the 3D trajectory of charged particles. It also measures particle-specific ionization energy loss, which is essential for PID. The TOF takes measurements of particle time offlight from the collision vertex to the detector. Particle velocity and mass can be further computed from the time offlight and the track information. The TRD records transition radiation, the emission of photons by electrons traversing the boundaries of a radiator. It enables to distinguish electrons from other charged particles. Since all the aforementioned detectors detect particles carrying a non-zero ele

FAQs

What particle identification challenge motivates using machine learning in ALICE?

Different particle species have overlapping detector responses and combining multiple detector observables becomes complex, limiting traditional selection based on fixed cuts.

How does the proposed method structure the machine learning model for PID?

It uses multiple neural networks acting as binary classifiers to improve separation performance compared with rectangular-cut approaches.

How is learning performed when the available data are incomplete?

The work extends the particle classifier with Feature Set Embedding and an attention mechanism to train effectively using incomplete samples.

Aug 4, 2026

Upload by Ezra