Understanding PDF Uncertainties with Neural Networks
Published:Dec 30, 2025 09:53
•1 min read
•ArXiv
Analysis
This paper addresses the crucial need for robust Parton Distribution Function (PDF) determinations with reliable uncertainty quantification in high-precision collider experiments. It leverages Machine Learning (ML) techniques, specifically Neural Networks (NNs), to analyze the training dynamics and uncertainty propagation in PDF fitting. The development of a theoretical framework based on the Neural Tangent Kernel (NTK) provides an analytical understanding of the training process, offering insights into the role of NN architecture and experimental data. This work is significant because it provides a diagnostic tool to assess the robustness of current PDF fitting methodologies and bridges the gap between particle physics and ML research.
Key Takeaways
- •Applies Machine Learning (ML) and Neural Networks (NNs) to improve PDF determination.
- •Develops a theoretical framework based on the Neural Tangent Kernel (NTK) for analyzing training dynamics.
- •Provides a quantitative understanding of uncertainty propagation in PDF fitting.
- •Offers a diagnostic tool to assess the robustness of PDF fitting methodologies.
Reference
“The paper develops a theoretical framework based on the Neural Tangent Kernel (NTK) to analyse the training dynamics of neural networks, providing a quantitative description of how uncertainties are propagated from the data to the fitted function.”