Structure-based node selection for capturing complex system states

Accurately characterizing the state of complex networked dynamical systems is crucial for understanding, prediction, and control of the systems, yet often hindered by the inaccessibility of the states of all nodes in the system. In this work, we introduce centrality-guided node selection strategies, enabling precise approximation of the system-wide average state based on the states of the selected nodes. Remarkably, our approach harnesses solely the structural information of the underlying network and does not require knowledge of the detailed node dynamics. Comprehensive simulations on various empirical networks and dynamic models demonstrate that eigenvector centrality, in particular, offers a robust choice. It yields node collections that reliably capture the global system state with exceptionally low errors, outperforming known strategies based on other measures across interaction topologies, network dynamics, sizes of the observed node set, and different regimes of systems’ dynamics in the parameter space. This work therefore provides a structure-based node selection strategy for efficient and accurate estimations of the global state of complex networked dynamical systems.