On the application of complex network analysis for metaheuristics

Abstract

This contribution deals with the hybridisation of complex network frameworks and metaheuristic algorithms. The population is visualised as an evolving complex network that exhibits non-trivial features. It briefly investigates the time and structure development of a complex network within a run of selected metaheuristic algorithms – i.e., PSO and Differential Evolution (DE). Two different approaches for the construction of complex networks are presented herein. It also briefly discusses the possible utilisation of complex network attributes. These attributes include an adjacency graph that depicts interconnectivity, while centralities provide an overview of convergence and stagnation, and clustering encapsulates the diversity of the population, whereas other attributes show the efficiency of the network. The experiments were performed for one selected DE/PSO strategy and one simple test function. © Proceedings of the 7th International Conference on Bioinspired Optimization Methods and their Applications, BIOMA 2016. All rights reserved.