Automatically discovering ordinary differential equations from data with sparse regression
Published in Communications Physics, 2024
This paper introduces ARGOS, a method for recovering the governing equations of a dynamical system from limited and noisy observations. ARGOS combines denoising, sparse regression, and bootstrap confidence intervals to identify which terms in a candidate equation are supported by the data. Across benchmark systems, it recovered the correct terms more consistently than the widely used SINDy framework and identified how much data, and how little noise, reliable recovery requires.
ARGOS is available as an open-source R package on CRAN.
Recommended citation: Egan, K., Li, W. & Carvalho, R. (2024). "Automatically discovering ordinary differential equations from data with sparse regression." Communications Physics. 7, 20.
Download Paper