Physics-driven learning for inverse problems in quantum chromodynamics

2025-01-06

Nature Reviews Physics

AI Summary

This work addresses key inverse problems in quantum chromodynamics (QCD), such as hadron spectrum inversion. Methodologically, it introduces a physics-informed deep learning framework: for the first time, QCD effective theories—specifically, the heavy-quarkonium potential model—are embedded into neural networks, yielding a physics-informed neural network (PINN) that jointly incorporates differential equation constraints, variational Bayesian inference, and lattice-QCD-derived priors; an attention mechanism is further integrated to enhance feature representation. The principal contribution lies in establishing an interpretable and verifiable paradigm for physics-constrained learning, rigorously enforcing fundamental QCD properties—including asymptotic freedom and confinement—within the model architecture. Evaluated on bottomonium energy spectrum inversion, the proposed method reduces relative error by 62% compared to purely data-driven approaches, markedly improving both physical consistency and predictive accuracy.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Relational Probabilistic Models
• Search and Optimization: Learning to Search

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Large pretrained models with web data

Problem

• Quantum Chromodynamics
• Inverse Problem
• Physical Property Extraction

Innovation

• Physics-driven Learning
• Deep Learning Integration
• Inverse Problems in QCD