An investigation of redundant genotype-phenotype mappings and their role in evolutionary search

Normalization in Genetic Algorithms

“…It is related to the previous studies in the aspect of using the neutral paths created by redundancy [12] [13]. However, normalization attempts to alleviate the negative effects of neutral paths.…”

“…In particular, Shipman, Shackleton, and Ebner [12] [13] showed that some redundant encodings are useful to help mutation-based hill-climbing search by constructing many different neutral paths in the genotype space that enable moves at the same fitness level. However, the tackled problems were of small sizes and only mutation operators were considered.…”

The Effects of Constant and Bit-Wise Neutrality on Problem Hardness, Fitness Distance Correlation and Phenotypic Mutation Rates

“…In [22], [23], Shackleton et al artificially added neutrality to evolutionary search with the use of five different genotype-phenotype mappings. The static random mapping consisted in defining a genotype of length 30 which was mapped to a phenotype of 16 bits.…”

“…In particular, they focused their attention on the RBN mapping proposed and studied in [22]- [24] and measured its influence on evolution using the rate of fitness increase. EAs and Hill-Climbers were used on three different problems to compare the performance obtained with and without the RBN mapping.…”

“…They identified some properties that are useful to characterise redundant representations: (a) a redundant representation is uniform if all phenotypes can be obtained by the same number of genotypes, (b) a representation is synonymously redundant if the genotypes that map to the same phenotype are part of a neutral network (i.e., they are close to each other), 3 Fonseca and Correia [28] developed two redundant representations using approaches based on mathematical 3 An example of this type of redundancy is the trivial voting mapping proposed in [22], [23].…”