Optimal Control of Information Diffusion in Temporal Networks

Abstract

Effective control of information diffusion is quite a significant and universal task in information science and control engineering. The diffusion process and network evolution often share a common time scale. However, previous studies mainly focus on fixed topological structures, ignoring the temporal property of real networks. To fill this gap, we comprehensively investigate the optimal control of information diffusion in temporal networks. First, activity-driven networks (ADNs) are introduced to represent realistic temporal networks. Meanwhile, the node-based susceptible-infected-recovered-susceptible (SIRS) model is employed to describe the diffusion dynamics from a microcosmic perspective. Second, we propose two synergistic optimal control approaches to suppress negative diffusion and enhance positive diffusion, respectively. Then, the optimal control problems of minimizing cumulative system cost and maximizing total system revenue are formulated. Subsequently, the optimal control signals are obtained using the optimal control theory and Pontryagin’s minimum principle. Third, we exploit an implementation framework with a multi-layer hybrid feedback control scheme for large-scale temporal networks. This framework utilizes centralized and distributed techniques to implement different control strategies. Finally, experimental results on various real-world datasets validate the efficiency of the proposed synergistic optimal control approaches and implementation framework.