Characterizing Evolutionary Algorithm Using Complex Networks Theory: A Case Study

Abstract

Evolutionary algorithms (EAs) are a type of complex systems which mimic biological evolution in nature to solve real world problems. In this paper, we propose to use complex networks theory to characterize the topological properties of evolutionary algorithms (EAs). A case study on Guo's algorithm is given as an example to show how to use our method. In our method, we represent the evolutionary process of Guo's algorithm as a directed network, directed evolutionary algorithm network (DEAN). Many aspects of DEAN are analyzed, such as degree distribution, average path length, assortativity coefficient, and clustering coefficient. Our results imply that DEAN is a small-world and scare-free type network. Our results give great insight into the underlining regularities in EAs.