@InProceedings{OliveiraLorePretStep:2004:AdHiFa,
author = "Oliveira, Alexandre Cesar Muniz and Lorena, Luiz Antonio Nogueira
and Preto, Airam Jonatas and Stephani, Stephan",
affiliation = "Federal University of Maranh{\~a}o, Department of Informatic and
Universidade Federal do Maranh{\~a}o, Departamento de
Inform{\'a}tica (UFMA) and National Institute for Space Research,
Computing and Applied Mathematics Laboratory and Instituto
Nacional de Pesquisas Espaciais, Laborat{\'o}rio de Associado de
Computa{\c{c}}{\~a}o e Matem{\'a}tica Aplicada (INPE. LAC)",
title = "An adaptive hierarchical fair competition genetic algorithm for
large-scale numerical optimization",
booktitle = "Proceedings...",
year = "2004",
pages = "6",
organization = "Brazilian Symposium in Neural Networks, 4. (SBRN).",
publisher = "INPE",
keywords = "COMPUTER SCIENCE, Genetic algorithms, Hierarchical Fair
Competition (HFC), Computer systems performance, Message
processing, Message Passing Interface (MPI), Optimization,
COMPUTA{\C{C}}{\~A}O APLICADA, Algoritmos gen{\'e}ticos,
Desempenho computacional, Processamento de mensagem, Interace de
transmiss{\~a}o de mensagem, Competi{\c{c}}{\~a}o
hierarquica.",
abstract = "Genetic algorithms, inspired by the theory of evolution of
species, are intended to be unfair. Individuals compete against
each other and the best-adapted ones prevail. Unfairness is due to
big dierences of skills, generally evaluated by a fitness measure,
in a population of individuals competing for survival. However,
population diversity is important to preserve some features that
are not always associated to high ranked skills. Such diversity
can be achieved by imposing fairness rules to the competition. The
adaptive hierarchical fair competition genetic algorithm has been
proposed to comply with this feature by segregating individuals in
casts or demes, according to their fitness. This work proposes a
parallel implementation that enhances the capabilities and
computational performance of an adaptive hierarchical fair
competition genetic algorithm. The code was parallelized using the
MPI (Message Passing Interface) communication library and executed
in a distributed memory parallel machine, a PC cluster. Test
results are shown for standard numerical optimization problems
presenting hundreds of variables.",
conference-location = "S{\~a}o Luiz",
conference-year = "29 Sept. - 01 Oct.",
copyholder = "SID/SCD",
language = "en",
ibi = "6qtX3pFwXQZsFDuKxG/EC3F7",
url = "http://urlib.net/ibi/6qtX3pFwXQZsFDuKxG/EC3F7",
targetfile = "fair.pdf",
urlaccessdate = "02 maio 2024"
}