Abstract:This work proposes a generalized methodology for sparse identification of dynamical systems (SID), utilizing reweighted l1 -regularized least absolute deviation (LAD) regression to recover the governing equation of dynamical systems. To programmatically implement the proposed method, a stabilized and efficient resolution scheme is developed by combining the solution method of the classical LAD-based least absolute shrinkage and selection operator (LAD-lasso) and the threshold iterative idea of the sequentially thresholded least-squares (STLS) method. The new method relaxes the assumption of normality for the error terms in the linear regression model employed by traditional SID techniques. Additionally, it overcomes the low identification accuracy deficiency of the recently proposed regularized least absolute deviation-based sparse identification of dynamics method (RLAD-SID) under some poor data conditions. The performance of the proposed approach is investigated by simulations of two well-known dynamical systems, that is, the Lorenz system and the Duffing system. The reliability of this approach in recovering the governing equation of dynamical systems from noise-contaminated datasets is firstly demonstrated. Furthermore, the advantage of the proposed method is illustrated by comparing it with four traditional techniques in relatively poor data environments. Finally, we display the robustness of the new method when the time derivative dataset is affected by different types of noise, and reveal its superiority compared to the traditional RLAD-SID.

Discovering New Interpretable Conservation Laws as Sparse Invariants

Interpretable conservation laws as sparse invariants

Extracting Conservative Equations from Nonconservative State Data

Data-Driven Automated Discovery of Variational Laws Hidden in Physical Systems

Discovering governing equations from data: Sparse identification of nonlinear dynamical systems

Discovering governing equations from data by sparse identification of nonlinear dynamical systems

Data-Driven Discovery of Conservation Laws from Trajectories via Neural Deflation

Discovering Conservation Laws using Optimal Transport and Manifold Learning

Discovering Sparse Interpretable Dynamics from Partial Observations

Learning conservation laws in unknown quantum dynamics

Deep Conservation: A latent-dynamics model for exact satisfaction of physical conservation laws

A Computational Approach to Polynomial Conservation Laws

Towards Robust Data-Driven Automated Recovery of Symbolic Conservation Laws from Limited Data

Sparse Identification of Dynamical Systems by Reweighted l1 -regularized Least Absolute Deviation Regression

Machine learning independent conservation laws through neural deflation

Automatically discovering ordinary differential equations from data with sparse regression

Sparse Identification of Nonlinear Dynamics with Side Information (SINDy-SI)

Invariant Reduction for Partial Differential Equations. I: Conservation Laws and Systems with Two Independent Variables

Governing equation discovery of a complex system from snapshots

Numerical preservation of multiple local conservation laws

Abide by the Law and Follow the Flow: Conservation Laws for Gradient Flows