Work

Genetic Programming-Based Modeling of Systems of Differential Equations

Year: 2026

Type: conference-paper

Abstract: In this work, we show how recent advancements make GP a competitive method compared to the current dominant SINDy approach for modeling dynamical systems governed by differential equations. Specifically, we leverage Pareto tournament selection, model curvature optimization, time-series modeling, and GP linear model optimization to find solutions to a diverse set of differential equations in both clean and noisy environments. We discuss key benefits of using genetic programming-based symbolic regression over traditional linear modeling approaches such as SINDy.

Source: Genetic and Evolutionary Computation Conference Companion (GECCO Companion)

DOI: 10.1145/3795101.3805380

Authors: Nathan Haut, Stu Card, Mark Kotanchek

Institutions: Michigan State University, Critical Technologies (United States), Evolved Analytics (United States)

Language: English

Topic: Evolutionary Algorithms and Applications

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: hybrid

OpenAlex ID: W7202348416

FWCI:

Cites: 5

Cited by:

Related to: 10