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

TL;DR

This paper compares traditional machine learning models and deep learning techniques, including transformers, for improving dark matter detection in liquid xenon experiments, highlighting the importance of data representation and model choice.

Contribution

It provides a comprehensive evaluation of various ML models and data representations, demonstrating that simpler models with optimized features can perform comparably to complex deep learning approaches.

Findings

Transformers show promising performance but are comparable to XGBoost with optimal features.

Simplified feature representations retain critical information for classification.

Minimal differences in exclusion limits between XGBoost and deep learning models.

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.