The Multiverse of Dynamic Mode Decomposition Algorithms

Matthew J. Colbrook

2023-12-21

Abstract:Dynamic Mode Decomposition (DMD) is a popular data-driven analysis technique used to decompose complex, nonlinear systems into a set of modes, revealing underlying patterns and dynamics through spectral analysis. This review presents a comprehensive and pedagogical examination of DMD, emphasizing the role of Koopman operators in transforming complex nonlinear dynamics into a linear framework. A distinctive feature of this review is its focus on the relationship between DMD and the spectral properties of Koopman operators, with particular emphasis on the theory and practice of DMD algorithms for spectral computations. We explore the diverse "multiverse" of DMD methods, categorized into three main areas: linear regression-based methods, Galerkin approximations, and structure-preserving techniques. Each category is studied for its unique contributions and challenges, providing a detailed overview of significant algorithms and their applications as outlined in Table 1. We include a MATLAB package with examples and applications to enhance the practical understanding of these methods. This review serves as both a practical guide and a theoretical reference for various DMD methods, accessible to both experts and newcomers, and enabling readers to delve into their areas of interest in the expansive field of DMD.

Dynamical Systems,Machine Learning,Numerical Analysis,Spectral Theory

What problem does this paper attempt to address?

The problems that this paper attempts to solve mainly focus on the application and theoretical development of Dynamic Mode Decomposition (DMD) technology. Specifically, the paper aims to:

1. **Comprehensively review the DMD algorithm**: The paper provides a comprehensive and in - depth review, covering the basic principles of DMD, its different variants and their practical applications. Special emphasis is placed on the relationship between DMD and the Koopman operator, and how to transform complex nonlinear dynamic systems into a linear framework for analysis through the Koopman operator.

2. **Classify DMD methods**: The paper classifies DMD methods into three categories:
- **Methods based on linear regression**: These include methods for improving robustness to noise (such as forward - backward DMD, total least squares DMD, optimized DMD, etc.), compression and random linear algebra methods (such as compressed sensing DMD, random DMD, etc.), multi - resolution DMD, and control - related DMD methods.
- **Methods based on Galerkin approximation**: These include extended DMD (for handling nonlinear observable variables), time - delay embedding methods (such as Hankel DMD, HAVOK, etc.), and methods for controlling projection errors (such as residual DMD).
- **Methods for maintaining structure or symmetry**: These include physically - informed DMD, measure - preserving extended DMD, compactification methods for continuous - time systems, etc.

3. **Explore the theoretical basis of DMD**: The paper details the definition, spectral properties of the Koopman operator and its application in DMD. Through these theoretical bases, it explains how DMD can extract the dynamic modes and characteristics of a dynamic system from data.

4. **Solve challenges in practical applications**: The paper discusses the main challenges faced by DMD in practical applications, such as the influence of noise, computational efficiency, multi - scale dynamic analysis, separation of control systems, etc., and proposes corresponding solutions.

5. **Provide practical tools**: The paper is accompanied by a MATLAB package that contains implementations and examples of various DMD methods, helping readers better understand and apply these methods.

In summary, this paper aims to provide a comprehensive guide for DMD researchers and practitioners, covering not only the basic theory and algorithms of DMD, but also deeply exploring its application prospects in various complex systems.