Complex Network-Based Genetic Algorithm for Optimization

Genetic algorithm (GA) is a classical swarm intelligence algorithm, where selection, crossover and mutation are operated on a population to simulate the evolution of genes. However, it faces the limitations of premature convergence and falling into local optimal solutions. To overcome this problem, a complex network-based genetic algorithm (CNGA) is proposed by thoroughly analyzing the characteristics of complex network and exploiting the similarity between complex network and swarm intelligence. Firstly, to reduce the possibility of falling into local optimal solutions, the complex network structure is designed for decentralized local search. Secondly, to improve the accuracy and convergence speed, the evolutionary game strategies are proposed for real-time update. To test the feasibility and superiority of the proposed CNGA, six benchmark functions and five other swarm intelligence algorithms have been utilized. The results show that the CNGA has superior performance in terms of accuracy, stability and convergence speed. This provides a new idea for the development of swarm intelligence algorithms.