Estudo de Parâmetros Ótimos em Algoritmos Genéticos Elitistas
Cadeias de Markov. Simulação. Otimização. Algoritmos Evolutivos. Seleção de Parâmetros.
The genetic algorithm is a search process used to find the global maximum of
functions. This algorithm rests on naturalistic firmaments that evaluate a sample of
global maximum candidates in each iteration. This evolution is consequence of three
operators (Selection, Mutation and Crossover) exploiting the domain of the function
at the same time that select the best candidates found. In this study we will show
a Markov chain to fit this algorithm’s evolution. We will simulate a model to fit
the effect of this algorithm’s parametrization in its hate of convergence, estimated by
the number of iterations until the global maximum is reached. In the simulations we
observed this effect at functions such as: unidimensional, bidimensional, with only one
local maximum (the global one) and with many local maxima. Finally, this work shows
results that put in question the crossover operator’s relevance in the studied functions
and arguments to believe that the mutation and crossover operators, that makes the
convergence’s hate optimum, depend on the function.