Complex Network Analysis of Evolutionary Algorithms Applied to Combinatorial Optimization Problem.

Abstract

This thesis explores the connection between Evolutionary Algorithms (EA's) and Complex Networks (CN's). EA's are bio-inspired algorithms which mimic naturally occurring phenomena in order to model and solve complex engineering tasks. One of its features is its population based paradigm. The behaviour of the population over the iterations is analysed in this thesis using CN analysis tools. Four distinct broad attributes are analysed; adjacency matrix, centralities, cliques and communities. Using these attributes, a number of experimentations and analysis were conducted, from which interesting information regarding population development, stagnation, network interconnection and hierarchical development was obtained. These data supported the concept of population dynamics and furthermore could be used for population and evolution control.