Xenon Signal Denoising via Supervised, Semi-Supervised, and Unsupervised Models

2026/03/31 by Grant Kendrick Parker, Jason Brodsky, Indra Chakraborty
Physics and Astronomy · #physics.ins-det #hep-ex #nucl-ex

paper · pdf · doi:10.1140/epjc/s10052-026-16114-z

published as Eur. Phys. J. C, vol. 86, no. 7, July 2026, p. 875 · 10 pages, 5 figures, updated to correct typos and formatting

arxiv created 2026/08/04 · arxiv updated 2026/08/06

Abstract

This study presents a denoising algorithm trained using machine learning to improve the energy resolution of a single-phase liquid xenon time projection chamber for neutrinoless double beta decay detection. Supervised, unsupervised, and semi-supervised models are demonstrated to significantly remove noise from simulated measurements while preserving signal information. The supervised model achieves an energy resolution of <1%, while the semi-supervised models achieve energy resolutions of ∼ 1%, and the unsupervised model performance is ∼ 1.5%. This work is evidence that machine learning denoising can improve energy resolution compared to traditional algorithms, even when experimentalists lack perfect a priori knowledge of the signals. Such models provide a realistic path toward next-generation sensitivity in 0νββ searches.

Citations

• On the discriminability of self-supervised representation learning
• Attention U-Net for all-sky continuous gravitational wave searches
• Neural Network Reconstruction of Non-Gaussian Initial Conditions from Dark Matter Halos
• A new method for structural diagnostics with muon tomography and deep learning
• Generalized Recorrupted-to-Recorrupted: Self-Supervised Learning Beyond Gaussian Noise
• Robust multi-coil MRI reconstruction via self-supervised denoising
• UNSURE: self-supervised learning with Unknown Noise level and Stein's Unbiased Risk Estimate
• On the Discriminability of Self-Supervised Representation Learning
• Investigating Self-Supervised Image Denoising with Denaturation
• Analyzing the Sample Complexity of Self-Supervised Image Reconstruction Methods
• nEXO: neutrinoless double beta decay search beyond 10 28 year half-life sensitivity
• Noisier2Noise: Learning to Denoise from Unpaired Noisy Data
• Simulation of charge readout with segmented tiles in nEXO
• Unsupervised Learning with Stein's Unbiased Risk Estimator
• Noise2Noise: Learning Image Restoration without Clean Data
• Neutrinoless Double Beta Decay and the Baryon Asymmetry of the Universe
• Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising
• U-Net: Convolutional Networks for Biomedical Image Segmentation
• Ideal spatial adaptation by wavelet shrinkage
• Estimation of the Mean of a Multivariate Normal Distribution
• Smoothing and differentiation of data by simplified least square procedure