From Recombination of Genes to the Estimation of Distributions II. Continuous Parameters

摘要

The Breeder Genetic Algorithm (BGA) is based on the equation for the response to selection. In order to use this equation for prediction, the variance of the fitness of the population has to be estimated. For the usual sexual recombination this can be difficult. In this paper the new points (offspring) are generated from distributions, a uniform distribution and a distribution generated by univariate marginal distributions. For a class of unimodal fitness functions the performance of the BGA is analytically computed. The results are compared to gene recombination methods. The uniform distribution is approximately generated by line recombination; recombination methods acting independently on each gene approximate the second distribution.