Probabilistic Grammatical Evolution

AbstractGrammatical Evolution (GE) is one of the most popular Genetic Programming (GP) variants, and it has been used with success in several problem domains. Since original proposal, many enhancements have proposed to GE order address some its main issues improve performance.In this paper we propose Probabilistic Grammatical (PGE), which introduces a new genotypic representation mapping mechanism for GE. Specifically, resort Context-Free Grammar (PCFG) where probabilities are adapted during evolutionary process, taking into account productions chosen construct fittest individual. The genotype list real values, each value represents likelihood selecting derivation rule. We evaluate performance PGE two regression problems compare Structured (SGE).The results show that better than GE, statistically significant differences, achieved similar when comparing SGE.KeywordsGenetic ProgrammingGrammatical EvolutionProbabilistic GrammarProbabilistic EvolutionGenotype-to-Phenotype Mapping