Constructing Robust and Influential Networks Against Cascading Failures via a Multi-objective Evolutionary Algorithm

The influence maximization problem and robustness optimization of networks are hot-spots in the current research. Existing studies have rather investigated the problems of network robustness optimization or influence maximization separately, without a comprehensive consideration of the possible co-optimization. Meanwhile, few studies have touched upon the problem of constructing networks’ typologies to improve the robust influential ability given a specific seed set. Therefore, in this paper, the correlation between the network robustness and the robust influential ability has been verified to be contradictory through empirical analyses. A multi-objective evolutionary algorithm is employed to construct comprehensively robust networks through topological rewiring, termed the MOEAMIM. Experiments on several synthetic networks demonstrate the effectiveness of the proposed algorithm, and networks with diverse preferences towards the structural or the influential robustness can be obtained.