Publicação

Non-geometric pulse: An adaptive geometricity approach for Genetic Algorithms

Ver documento

Detalhes bibliográficos
Resumo:Evolutionary algorithms (EAs) are a family of algorithms inspired by the Darwinian theory of evolution. Mathematics, particularly geometry and topology, allows for the possibility of developing a general geometrical framework and consequently generating deeper insights that may be shared among the different EAs. Genetic Algorithm (GA), inspired by Darwin’s theory of natural selection, is a popular algorithm among EAs. This is a population-based, fitness-oriented algorithm that performs a convex heuristic search to optimize a plethora of problems. Common limitations of GA as well as other EAs have geometrical origins like premature convergence, where the final population’s convex-hull might not include the best solution, called Global Optima. Population diversity maintenance is a key idea that tries to tackle this problem but is often performed through geometrical methods that constantly diminish the search space’s area. In this work, a self-adaptive geometricity approach will be presented. In particular, the non-geometric crossover is strategically employed in a symbiotic relation with geometric crossover, maintaining diversity in a logical way from a geometric/topological grammar standpoint. A comparison with well-known diversity maintenance methods is provided, using common benchmarks that serve as general testing ground for the considered techniques.
Autores principais:Ferreira, José Pedro Mendes Ribeiro do Vale
Assunto:Convex Search Evolutionary Algorithms Genetic Algorithms Geometric semantic operators Diversity maintenance
Ano:2022
País:Portugal
Tipo de documento:dissertação de mestrado
Tipo de acesso:acesso aberto
Instituição associada:Universidade Nova de Lisboa
Idioma:inglês
Origem:Repositório Institucional da UNL
Descrição
Resumo:Evolutionary algorithms (EAs) are a family of algorithms inspired by the Darwinian theory of evolution. Mathematics, particularly geometry and topology, allows for the possibility of developing a general geometrical framework and consequently generating deeper insights that may be shared among the different EAs. Genetic Algorithm (GA), inspired by Darwin’s theory of natural selection, is a popular algorithm among EAs. This is a population-based, fitness-oriented algorithm that performs a convex heuristic search to optimize a plethora of problems. Common limitations of GA as well as other EAs have geometrical origins like premature convergence, where the final population’s convex-hull might not include the best solution, called Global Optima. Population diversity maintenance is a key idea that tries to tackle this problem but is often performed through geometrical methods that constantly diminish the search space’s area. In this work, a self-adaptive geometricity approach will be presented. In particular, the non-geometric crossover is strategically employed in a symbiotic relation with geometric crossover, maintaining diversity in a logical way from a geometric/topological grammar standpoint. A comparison with well-known diversity maintenance methods is provided, using common benchmarks that serve as general testing ground for the considered techniques.