New approach to synchronization analysis of linearly coupled ordinary differential systems (Q2492256): Difference between revisions

The authors consider the following linearly coupled system of ordinary differential equations \[ \frac{dx_i(t)}{dt} = f(x_i(t),t)+\sum_{j=1}^{n}a_{ij}\Gamma x_j(t), \quad i=1,\dots,n, \] where \(x_i(t)\in \mathbb{R}^n\), \(A=(a_{ij})\in \mathbb{R}^{n\times n}\) is the coupling matrix with \(a_{ij}\geq 0\) for \(i\neq j\), and \(\Gamma = \mathrm{diag}\{ \gamma_1,\gamma_2,\dots,\gamma_n\}\) with \(\gamma_i>0\). The main result of the paper is a number of criteria for synchronization of the above system, i.e., for the stability of the invariant subspace \(S=\{x: x_i=x_j, i\neq j\}\). Using Lyapunov function approach, conditions for global synchronization are given. Presented criteria for local synchronization are mathematically rigorous reformulation of the so-called master stability function approach [\textit{L. M. Pecora, T. L. Carroll, G. A. Johnson, D. J. Mar} and \textit{J. F. Heagy}, Chaos 7, 520--543 (1997; Zbl 0933.37030)]. The paper presents two examples of linearly coupled neural networks.