Research On Interdependent Network Robustness Optimization Based On Memetic Algorithm

With the development of human civilization and modern technology,an increasing number of complex systems have emerged in contemporary society.They are interdependent and interactive,which provide convenience for human life together.However,when a complex system is attacked,some nodes in the system will fail,which may cause a large area of collapse and lead to incalculable economic losses.Therefore,researchers abstract this kind of complex system as an interdependent networks model and study the relationship between structural topology and network robustness.Network robustness refers to the ability of a network to maintain its functional integrity in the event of an attack or failure.Although previous scientists have proposed many effective robust optimization methods for isolated networks,due to the interdependence of modern network systems,the previous optimization strategies cannot be directly applied to interdependent networks.Thus,it is necessary to develop new structure adjustment methods according to the structural characteristics of interdependent networks.Aiming at the problem of inefficient and incomplete optimization in the existing robust optimization methods of interdependent networks based on malicious attack,this paper analyzes the network characteristics from the perspectives of interlayer structure,intra-layer structure,and the whole structure,and proposes corresponding optimization methods according to its characteristics.The specific work is divided into the following three parts:1.A memetic algorithm for optimizing inter-links to enhance the robustness of interdependent networks against malicious attacksIn order to solve the problem that there are few kinds of research based on malicious attack and the intra-layer structure are not fully used as the prior information to assist the search,based on memetic algorithm framework,we design a partial mapping crossover operator based on the preservation of local structure and a local search operator based on the recombination of neighboring nodes to improve the quality of candidate solutions and accelerate the convergence of the algorithm.In addition,the proposed algorithm is applied to synthetic networks and real coupled networks and compared with the existing methods,the experimental results show the performance advantages in enhancing network robustness for the proposed method.2.A multiagent genetic algorithm for designing robust interdependent networks against malicious attacksIn order to solve the problem that the existing methods ignore the inter-layer structure of the interdependent network and do not consider the characteristics of different layer networks under the attack model,based on the framework of multiagent genetic optimization,we design a neighborhood crossover operator based on network structure migration and a selflearning operator based on layered local search and adopt an onion structure search strategy and structural similarity learning strategy for the attacked layer network and the affected layer network respectively.In subsequent experiments,we test the proposed algorithm with a variety of synthetic networks and real coupled networks,and the results obtained by the proposed algorithm are better than that of the two classical heuristic optimization algorithms.Moreover,we analyze the optimized network structure and further discuss the relationship between structural characteristics and network robustness.3.A cooperative coevolutionary algorithm for robust optimization of interdependent networks against malicious attacksIn order to solve the problem that the existing methods only adjust the inter-layer structure or intra-layer structure,which leads to the incomplete optimization of network structure,we decomposed the structural optimization problem of interdependent networks into two separate sub-problems,namely inter-layer structural optimization and intra-layer structural optimization.The cooperative coevolution algorithm is used to realize the independent search of the sub-problems and the information exchange among different sub-problems,so as to promote the fast convergence of the algorithm.Then we perform simulation experiments on synthetic networks and real coupled networks and compare the proposed algorithm with the serial optimization method and the heuristic method.The results demonstrate that the proposed method can optimize the robustness of interdependent networks more effectively and comprehensively.