To address the topic of varied rates of code growth, Chapter 6 investigated the change of population dynamics occurring under increasingly difficult problem instances. Using previously researched problem instances and several randomly generated instances, Chapter 6 showed how problem difficulty induces different kinds of population dynamics. These dynamics affect both selection pressure and diversity, and subsequently the rate of code growth. A causal model was also supported by a constructed problem and related literature. Results suggested that controlling the rate of code growth can be achieved by considering the variability of selection pressure and the presence of diversity. Whereas previous methods are often heavy-handed in penalising the large individuals, this chapter outlined alternative ways in which code growth might be controlled without destructively effecting fitness improvements. Also, this chapter further highlighted the different types of behaviours of increasingly difficult problem instances.
Given the diversity results of Chapter 4, the results of controlling diversity in Chapter 5 and the effects of diversity on code growth and selection in Chapter 6, the next line of research carried out was to analyse the contribution of dissimilar individuals during search. Distributed models are frequently suggested as the remedy for convergence. As these models typically move migrants between subpopulations, this study allowed the role of these migrants to be examined.