Global optimization by evolution strategies. A technique using the genetic of diploid individuals. (Optimisation globale par stratégie d’évolution. Technique utilisant la génétique des individus diploïdes.) (French) Zbl 0876.90079

Summary: A new algorithm for global optimization, based on genetic algorithms und evolution strategies, is presented. This class of algorithms is characterized by a stochastic search on sets of points and uses natural adaptive population ability. The proposed algorithm follows one which was first developed in the laboratory und used a genetic model for the optimization problem. The main difference lies in the modelling of individuals which first used the haploid model. In the present work, a more evolved model is used, consisting of a diploid one. Following the description of the algorithm, a demonstration of asymptotic convergence is provided. Then, the influence of the parameters is evaluated showing the great importance of homozygocity rate and the nature of the dominance. A maximisation problem is finally carried out, and the performances with the two types of dominance are compared with these obtained through the intermediary of a classical und a hybrid genetic algorithm. In conclusion, this study shows the efficiency und potentialities of such an algorithm.


90C27 Combinatorial optimization
68T05 Learning and adaptive systems in artificial intelligence
Full Text: DOI EuDML