Insights into Dark Matter Direct Detection Experiments: Decision Trees versus Deep Learning

Abstract

The detection of Dark Matter (DM) remains a significant challenge in particle physics. This study exploits advanced machine learning models to improve detection capabilities of liquid xenon time projection chamber experiments, utilizing state-of-the-art transformers alongside traditional methods like Multilayer Perceptrons and Convolutional Neural Networks. We evaluate various data representations and find that simplified feature representations, particularly corrected S1 and S2 signals as well as a few shape-related features including the time difference between signals, retain critical information for classification. Our results show that while transformers offer promising performance, simpler models like XGBoost can achieve comparable results with optimal data representations. We also derive exclusion limits in the cross-section versus DM mass parameter space, showing minimal differences between XGBoost and the best performing deep learning models. The comparative analysis of different machine learning approaches provides a valuable reference for future experiments by guiding the choice of models and data representations to maximize detection capabilities.

Version published in JCAP. Discussions expanded, two appendices and references added. 31 pages, 10 figures, 2 tables

References in corpus (55)

• Scikit-learn: Machine Learning in Python
• Asymptotic formulae for likelihood-based tests of new physics
• Machine learning and the physical sciences
• Dark Matter Search Results from a One Tonne$\times$Year Exposure of XENON1T
• First Dark Matter Search Results from the XENON1T Experiment
• First Dark Matter Search Results from the LUX-ZEPLIN (LZ) Experiment
• WIMP dark matter candidates and searches - current status and future prospects
• Dark Matter Search Results from the PandaX-4T Commissioning Run
• Direct Detection of WIMP Dark Matter: Concepts and Status
• First Dark Matter Search with Nuclear Recoils from the XENONnT Experiment
• Constraining the Milky Way potential with a 6-D phase-space map of the GD-1 stellar stream
• Deep Learning and its Application to LHC Physics
• DARWIN: towards the ultimate dark matter detector
• The RAVE survey: the Galactic escape speed and the mass of the Milky Way
• Search for New Physics in Electronic Recoil Data from XENONnT
• Projected WIMP sensitivity of the LUX-ZEPLIN (LZ) dark matter experiment
• Projected WIMP Sensitivity of the XENONnT Dark Matter Experiment
• TabTransformer: Tabular Data Modeling Using Contextual Embeddings
• Search for dark matter with a 231-day exposure of liquid argon using DEAP-3600 at SNOLAB
• Dark matter direct search sensitivity of the PandaX-4T experiment
• A Next-Generation Liquid Xenon Observatory for Dark Matter and Neutrino Physics
• Machine and Deep Learning Applications in Particle Physics
• The SABRE project and the SABRE PoP
• The COSINUS project - perspectives of a NaI scintillating calorimeter for dark matter search
• Search for light dark matter with ionization signals in the PandaX-4T Experiment
• A Living Review of Machine Learning for Particle Physics
• XENON1T Dark Matter Data Analysis: Signal & Background Models, and Statistical Inference
• Modern Machine Learning and Particle Physics
• Three-year annual modulation search with COSINE-100
• PandaX-xT: a Multi-ten-tonne Liquid Xenon Observatory at the China Jinping Underground Laboratory
• A search for new physics in low-energy electron recoils from the first LZ exposure
• Better plain ViT baselines for ImageNet-1k
• Convolutional Neural Networks for Direct Detection of Dark Matter
• Sensitivity of future liquid argon dark matter search experiments to core-collapse supernova neutrinos
• An induced annual modulation signature in COSINE-100 data by DAMA/LIBRA's analysis method
• Dark matter search project PICO-LON
• Review on dark matter searches
• A semi-supervised approach to dark matter searches in direct detection data with machine learning
• Deep-underground dark matter search with a COSINUS detector prototype
• Improving ANAIS-112 sensitivity to DAMA/LIBRA signal with machine learning techniques
• Simulation and background characterisation of the SABRE South experiment
• Towards an automated data cleaning with deep learning in CRESST
• A method for approximating optimal statistical significances with machine-learned likelihoods
• The DarkSide-20k experiment
• Towards a method to anticipate dark matter signals with deep learning at the LHC
• Fast and Flexible Analysis of Direct Dark Matter Search Data with Machine Learning
• Improved Sensitivity of the DRIFT-IId Directional Dark Matter Experiment using Machine Learning
• A machine learning-based methodology for pulse classification in dual-phase xenon time projection chambers
• DarkSide-20k: Next generation Direct Dark Matter searches with liquid Argon
• Deep learning for Directional Dark Matter search
• Machine-learning techniques applied to three-year exposure of ANAIS-112
• Insights into Dark Matter Direct Detection Experiments: Decision Trees versus Deep Learning
• Bayesian technique to combine independently-trained Machine-Learning models applied to direct dark matter detection
• FormerTime: Hierarchical Multi-Scale Representations for Multivariate Time Series Classification
• Present status of PICOLON project

Cited by in corpus (3)

• Insights into Dark Matter Direct Detection Experiments: Decision Trees versus Deep Learning
• Shedding Light on Dark Matter at the LHC with Machine Learning
• State of the Art in Direct Dark Matter Detectors: Technologies, Performance, and Future Directions