Choice and development

The process of development creates a phenotype from one or more genotypes of an individual through interaction with an environment. The opportunity for development to choose a phenotype from a set of alternatives made possible by the individual’s genotype(s) has not been widely considered in evolutionary computation. We briefly review recent research on developmental learning, dominance, and hybrid genetic algorithms that has investigated the role of choice in development. A new model of probabilistic development is presented based upon genotypes that encode the probabilities that the various alleles are expressed in the phenotype. The model outperforms a standard, binary haploid model on two families of single-peaked fitness functions in terms of average fitness. The standard model performed better on multi-peaked MAXSAT environments. More research is needed to fully evaluate the new model. Categories and Subject Descriptors I.2 [Artificial Intelligence]: Problem Solving, Searc...

• Type: Conference
• Year: 2007
• Where: GECCO
• Authors: Arthur M. Farley
• Keywords: Binary Haploid Model, GECCO 2007, Hybrid Genetic Algorithms, Phenotype