next up previous contents
Next: 2 Remarks and Problem Up: 1 Contributions Previous: 5 A model was   Contents

6 Summary

The research presented in this thesis provides an analysis of several key issues in genetic programming. First, selection pressure and subtree crossover success depend on the type of fitness-based and genetic diversity in the population. Secondly, the rates of code growth could be controlled by adaptively controlling both fitness-based and genetic diversity. Thirdly, the type of search genetic programming carries out can be described using a metaphor of hill-climbing and some knowledge of the problem domain. Finally, the inability of dissimilar and fit individuals, and similar and equally well fit individuals, to produce good offspring was shown.

S Gustafson 2004-05-20