Abstract:The framework of Koopman operator theory is discussed along with its connections to Dynamic Mode Decomposition (DMD) and (Kernel) Extended Dynamic Mode Decomposition (EDMD). This paper provides a succinct overview with consistent notation. The authors hope to provide an exposition that more naturally emphasizes the connections between theory and algorithms which may result in a sense of clarity on the subject.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to provide a concise and consistent framework for Koopman operator theory and its connection with dynamic mode decomposition (DMD) and extended dynamic mode decomposition (EDMD). Specifically, the author hopes to solve related problems through the following points:
1. **Unifying symbols and expressions**: By using a consistent symbol system, the relationship between Koopman operator theory and related algorithms becomes clearer, thus helping readers better understand these complex concepts.
2. **Combination of theory and algorithm**: Emphasize the connection between theory and practical algorithms, enabling readers to more naturally understand how to apply theory to practical calculations. This includes explaining the application of Koopman operators in different situations, such as linear autonomous systems, non - autonomous systems, and partially observable systems.
3. **Introduction of approximation methods**: Detail how to approximate Koopman operators in practice, especially through algorithms such as DMD and EDMD. These methods can help deal with the analysis problems of high - dimensional and complex systems.
### Overview of main content
- **Koopman operator theory**: The Koopman operator can transform a nonlinear dynamical system into a globally linear infinite - dimensional representation, thus allowing the use of linear system tools for analysis.
- **Koopman decomposition**: Describes the eigenvalues, eigenfunctions, and modes of the Koopman operator, and shows how to represent the observation function with these components.
- **Approximating Koopman operator**: Discusses how to approximate Koopman operators in practice, including the application of DMD, EDMD, and the kernel trick. These methods aim to reduce computational complexity and improve the applicability of the model.
- **Non - autonomous systems and partially observable systems**: Explores how to apply the Koopman operator to non - autonomous systems with control inputs and disturbances, as well as the case where the state is partially observable.
Through these contents, the author hopes to provide readers with a clear and easy - to - understand framework of Koopman operator theory and show its potential in practical applications.
### Formula summary
- **Koopman operator acting on the observation function**:
\[
K g_j(x_t)=g_j(F_{\Delta t}(x_t))
\]
- **Koopman eigenvalues and eigenfunctions**:
\[
K \phi_i(x_t)=\lambda_i \phi_i(x_t)
\]
- **Expansion of the observation function**:
\[
g_j(x_t)=\sum_{i = 1}^{\infty}\phi_i(x_t)c_i
\]
- **Expansion of the vector - valued observation function**:
\[
g(x_t)=\sum_{i = 1}^{\infty}\phi_i(x_t)v_i
\]
- **Koopman decomposition**:
\[
K g(x_t)=\sum_{i = 1}^{\infty}\lambda_i \phi_i(x_t)v_i
\]
- **Core equation of DMD**:
\[
X' = X C
\]
- **Dictionary function matrix of EDMD**:
\[
X = [\theta(x_t, u_t, d_t),\theta(x_{t + 1}, u_{t + 1}, d_{t + 1}),\dots,\theta(x_{t + n}, u_{t + n}, d_{t + n})]
\]
- **Inner product in the kernel trick**:
\[
A_{ij}=\langle\theta(z_{i - h:i}, u_{i - h:i}, d_{i - h:i}),\theta(z_{j - h:j}, u_{j - h:j}, d_{j - h:j})\rangle
\]
Through these formulas and theoretical introductions, the author hopes to provide readers with a comprehensive and in - depth guide to understanding Koopman operator theory.