Gene pool recombination and utilization of covariances for the Breeder Genetic Algorithm
H. Voigt, Heinz Mühlenbein · 2002
New recombination operators based on fuzzy and gene pool recombination are introduced for continuous genes. The performance of the operators is analyzed by means of the equation describing the response to selection. The operators are evaluated according to a new design criterion: maximizing the product of heritability and standard deviation. By means of a transformation to Eigenspace unimodal functions having arbitrary shaped level sets can be optimized. A prediction model based on the response to selection equation is derived which shows good agreement with the experimental results. 1. Introduction Let an optimization problem be given on a domain G ae R n f = f(x ) = min x2G f(x); G ae R n : (1) We make no assumptions concerning the convexity and differentiability of the function f(x). For the minimization a number of algorithms have been proposed. In this paper we apply the Breeder Genetic Algorithm BGA [2] to obtain approximations. The BGA uses a continuous representati...