Using AI for Efficient Statistical Inference of Lattice Correlators Across Mass Parameters

TL;DR

This paper introduces a machine learning approach to efficiently infer lattice QCD correlation functions across different mass parameters, reducing computational costs and improving uncertainty estimation.

Contribution

It presents a novel supervised learning method for lattice correlator inference and a data separation technique for bias correction and uncertainty estimation.

Findings

ML models outperform simple ratio methods in accuracy

The approach reduces computational expense in lattice QCD analysis

Effective uncertainty estimation through data separation

Abstract

Lattice QCD is notorious for its computational expense. Modern lattice simulations require large-scale computational resources to handle the large number of Dirac operator inversions used to construct correlation functions. Machine learning (ML) techniques that can increase, at the analysis level, the information inferred from the correlation functions would therefore be beneficial. We apply supervised learning to infer two-point lattice correlation functions at different target masses. Our work proposes a new method for separating data into training and bias correction subsets for efficient uncertainty estimation. We also benchmark our ML models against a simple ratio method.