Big-Data-Driven Control Strategies for Complex Networks

Abstract

Spurred by the growth of the World Wide Web and the Internet, and their similarity to numerous other large, dynamic, real-life networks, a truly cross-disciplinary science of complex networks has emerged in the past 15 years [1-12]. Hundreds of studies have been conducted and papers published exploring properties of complex networks—such as, their size, diameter, degree distribution, pairwisedistance distribution, cliques, communities, clustering coefficient, and the like [5-11]. Several growth models of these evolving networks have been proposed and studied [5,6]. However, an area of active interest, which has not been studied adequately, is that of designing control policies to steer the evolution of such a network towards a desired goal [8]. In practical situations, such as controlling the spread of diseases or the formation of opinions, topological properties, of the network, may need to be controlled. It would, therefore, be valuable to have automated synthesis [7] of strategies for controlling relevant topological properties of complex networks described by, for example, a preferentialattachment and preferential-deletion model [6]. We propose that an automated optimal control of evolving complex networks be achieved by leveraging recent results in developing stochastic models for the evolution of complex networks in combination with the classical results on dynamic-programming algorithms for optimal control [2-4].