Complex network analysis based adaptive differential evolution

Differential Evolution is a powerful stochastic population-based evolutionary algorithm for continuous functions optimisation. Unfortunately, it is not free of problems of possible premature convergence and stagnation. Many attempts have been made to remedy these issues and improve the performance and reliability through either self-adaptive parameters and strategies, or by controlling the population topology. In this paper, the adaptive approach based on analysis of complex network modelling the exchange of information in the population is presented. Two variants of Adaptive DE algorithm based on this mechanism are introduced and their performance compared against original DE, showing that Adaptive DE outperforms DE in many of the benchmark problems.