Dimensionally Constrained Symbolic Regression

TL;DR

This paper introduces dimensionally constrained symbolic regression (DCSR) for high-energy physics, improving convergence and solution quality in mass measurement problems by restricting term combinations based on physical dimensions.

Contribution

The paper presents a novel DCSR method that enhances symbolic regression performance and discovers new solutions in high-energy physics mass measurement tasks.

Findings

DCSR improves convergence behavior in symbolic regression.

DCSR yields solutions with better fitness than unconstrained methods.

DCSR can find novel solutions in complex variable spaces.

Abstract

We describe dimensionally constrained symbolic regression which has been developed for mass measurement in certain classes of events in high-energy physics (HEP). With symbolic regression, we can derive equations that are well known in HEP. However, in problems with large number of variables, we find that by constraining the terms allowed in the symbolic regression, convergence behavior is improved. Dimensionally constrained symbolic regression (DCSR) finds solutions with much better fitness than is normally possible with symbolic regression. In some cases, novel solutions are found.