7. Neutrality, Robustness, and Evolvability in Genetic Programming

Redundant mapping from genotype to phenotype is common in evolutionary algorithms, especially in genetic programming (GP). Such a redundancy leads to neutrality, a situation where mutations to a genotype may not alter its phenotypic outcome. The effects of neutrality can be better understood by quantitatively analyzing its two observed properties, robustness and evolvability. In this chapter, we summarize our previous work on this topic in examining a compact Linear GP algorithm. Due to the choice of this particular system we can characterize its entire genotype, phenotype, and fitness networks, and quantitatively measure robustness and evolvability at the genotypic, phenotypic, and fitness levels. We then investigate the relationship between robustness and evolvability at those different levels. Technically, we use an ensemble of random walkers and hill climbers to study how robustness and evolvability are related to the structure of genotypic, phenotypic, and fitness networks and influence the evolutionary search process.