A survey and analysis of diversity measures in genetic programming
Edmund Burke, Steven Gustafson, Graham Kendall · 2002
This paper presents a survey and comparison of the significant diversity measures in the genetic programming literature. The overall aim and motivation behind this study is to attempt to gain a deeper understanding of genetic programming dynamics and the conditions under which genetic programming works well. Three benchmark problems (Artificial Ant, Symbolic Regression and Even5-parity) are used to illustrate different diversity measures and to analyse their correlation with performance. The results show that diversity is not an absolute indicator of performance and that phenotypic measures appear superior to genotypic ones. Finally we conclude that interesting potential exists with tracking ancestral lineages.