Skip to content

Application of Biodiversity Statistical Indexes to Control the Diversity of Solutions in Multiobjective Genetic Algorithms

Notifications You must be signed in to change notification settings

JeanCarloEM/nsgan

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Scientific Initiation Data Set (Academic Work)

This is the source repository (dataset) of academic paper "Application Of Biodiversity Statistical Indexes To Control The Diversity Of Solutions In Multiobjective Genetic Algorithms (Aplicação De Índices Estatísticos De Biodiversidade No Controle De Diversidade Das Soluções Em Algoritmos Genéticos Multiobjetivo)". The full paper can be accessed here.

A work of

Publishers

Abstract

In the last decades, different optimization methodologies through evolutionary algorithms have been proposed, explored and successfully applied to a wide range of problems. Each of these methodologies has distinct characteristics, typical advantages and characteristic disadvantages. However, there is a problem that is shared by almost all of them: controlling the diversity of solutions. While natural selection favors variations toward greater divergence, in artificial evolution candidate solutions are homogenized. The diversity of solutions directly affects algorithmic performance, especially in multiobjective problems. Many authors have argued that promoting diversity would be beneficial in evolutionary optimization processes and that this could help prevent premature convergence into suboptimal solutions. In this paper we analyze some statistical indexes of biodiversity widely used in ecology and their impact when inserted in evolutionary algorithms. We also suggest practical ways to measure and promote diversity in multiobjective genetic algorithms.

  • Key-words: Multiobjective Genetic Algorithm. Diversity. Pareto Frontier.

Resumo

Nas últimas décadas, diferentes metodologias de otimização por meio de algoritmos evolutivos foram propostas, exploradas e aplicadas com sucesso a uma ampla gama de problemas. Cada uma destas metodologias possui características distintas, vantagens típicas e desvantagens características. No entanto, existe um problema que é compartilhado por quase todas elas: o controle da diversidade das soluções. Enquanto a seleção natural favorece variações em direção a uma maior divergência, na evolução artificial as soluções candidatas se homogeneízam. A diversidade das soluções afeta diretamente o desempenho algorítmico, principalmente em problemas multiobjetivos. Muitos autores argumentaram que a promoção da diversidade seria benéfica nos processos de otimização evolutiva e que isso poderia ajudar a evitar a convergência prematura em soluções abaixo do ideal. Neste trabalho analisamos alguns índices estatísticos de biodiversidade amplamente utilizados na ecologia e qual o seu impacto quando inseridos os algoritmos evolutivos. Sugerimos, também, formas práticas de medir e promover a diversidade em algoritmos genéticos multiobjetivo.

  • Palavras-chaves: Algoritmo Genético Multiobjetivo. Diversidade. Fronteira de Pareto.

About

Application of Biodiversity Statistical Indexes to Control the Diversity of Solutions in Multiobjective Genetic Algorithms

Topics

Resources

Stars

Watchers

Forks