Abstract:In this work, we develop a method to identify continuous-time nonlinear networked dynamics via the Koopman operator framework. The proposed technique consists of two steps: the first step identifies the neighbors of each node, and the second step identifies the local dynamics at each node from a predefined set of dictionary functions. The technique can be used to either identify the Boolean network of interactions (first step) or to solve the complete network identification problem that amounts to estimating the local node dynamics and the nature of the node interactions (first and second steps). Under a sparsity assumption, the data required to identify the complete network dynamics is significantly less than the total number of dictionary functions describing the dynamics. This makes the proposed approach attractive for identifying large dimensional networks with sparse interconnections. The accuracy and performance of the proposed identification technique are demonstrated with several examples.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to identify the dynamic characteristics of nonlinear networks. Specifically, the author has developed a method based on the Koopman operator framework for identifying continuous - time nonlinear network dynamics. This method is divided into two steps:
1. **Identify the neighbors of each node**: By estimating the interactions between nodes, determine which nodes are connected.
2. **Identify the local dynamics of each node**: Identify the local dynamics of each node from a predefined set of dictionary functions.
Under the sparse assumption (that is, most nodes in the network are not directly connected), the amount of data required to identify the complete network dynamics is significantly less than the total number of dictionary functions describing the dynamics. This makes the proposed method very attractive for the identification of large - scale networks with sparse connections.
### Mathematical formula representation
To understand this method more clearly, here are some key mathematical formulas:
- Network dynamics model:
\[
\dot{x}_i = F_i(x_i) + \sum_{k \in M_i} G_{ik}(u_k) + \sum_{k \in N_i} H_{ik}(x_k),
\]
where \( x_i \in \mathbb{R}^{n_i} \) is the state of node \( i \), \( u_k \in \mathbb{R}^{m_k} \) is the input, \( M_i \subseteq \{1, \dots, M\} \) is the input set of node \( i \), and \( N_i \subseteq \{1, \dots, N\} \) is the neighbor set of node \( i \).
- Koopman operator definition:
\[
K_t h(x) = h(\Phi_t(x)),
\]
where \( \Phi_t(x) \) is the flow mapping of system (1), and \( h \in \mathcal{F} \) is a set of functions defined on the state space.
- For non - autonomous systems, the Koopman operator is extended as:
\[
K_t h(x, \bar{u}) = h(\Phi_t(x, \bar{u}), T_t(\bar{u})),
\]
where \( T_t \) is the left - shift operator, defined as \( T_t(\bar{u}(\tau)) := \bar{u}(t + \tau) \).
### Method advantages
The main advantages of this method are:
- **Less data required**: Under the sparse assumption, the amount of training data required is much less than the total number of network parameters.
- **Applicable to large - scale networks**: It can effectively handle large - scale networks with sparse connections.
- **High accuracy**: Verified by experiments, this method shows high accuracy in Boolean network reconstruction and dynamics identification.
In conclusion, this paper proposes a Koopman - operator - based nonlinear network identification method, aiming to overcome the limitations of existing methods in large - scale and complex networks and provide an efficient and accurate solution.