Generalized Parton Distributions from Symbolic Regression
Authors:
Anusha Reddy Singireddy,
Andrew Dotson,
Zaki Panjsheeri,
Douglas Q. Adams,
Emmanuel Ortiz-Pacheco,
Marija Cuic,
Yaohang Li,
Huey-Wen Lin,
Simonetta Liuti,
Matthew D. Sievert,
Marie Boer,
Gia-Wei Chern,
Michael Engelhardt,
Gary R. Goldstein
Abstract:
AI/ML informed Symbolic Regression is the next stage of scientific modeling. We utilize a highly customizable symbolic regression package "PySR" to model the x and t dependence of the flavor isovector combination $H_{u-d}(x,t,ζ,Q^2)$ at $ζ=0$ and $Q^2$= 4 GeV$^2$. These PySR models were trained on GPD pseudodata provided by both Lattice QCD and phenomenological sources GGL, GK, and VGG. We demonst…
▽ More
AI/ML informed Symbolic Regression is the next stage of scientific modeling. We utilize a highly customizable symbolic regression package "PySR" to model the x and t dependence of the flavor isovector combination $H_{u-d}(x,t,ζ,Q^2)$ at $ζ=0$ and $Q^2$= 4 GeV$^2$. These PySR models were trained on GPD pseudodata provided by both Lattice QCD and phenomenological sources GGL, GK, and VGG. We demonstrate, for the first time, the consistency and systematic convergence of Symbolic Regression by quantifying the disparate models through their Taylor expansion coefficients. In addition to PySR penalizing models with higher complexity and mean-squared error, we implement schemes that test specific physics hypotheses, including FF$x$ and $t$ dependence and Regge behavior in PySR GPDs. We show that PySR can identify factorizing GPD sources based on their response to the Force-Factorized model. Knowing the precise behavior of the GPDs, and their uncertainties in in a wide range in $x$ and $t$, crucially impacts our ability to concretely and quantitatively predict hadronic spatial distributions and their derived quantities.
△ Less
Submitted 17 April, 2025;
originally announced April 2025.
