The number of offspring as strategy parameter in hierarchically organized evolution strategies

The evolution strategy, an algorithm imitating the effect of the biological evolution, has already been presented in this journal [Bäck92]. The basic algorithm described there uses a single population with a fixed number of parents and offspring. The present paper shows an extended algorithm with a higher imitative level of biological evolution. In this hierarchically organized evolution strategy there exist several populations being in competition with each other. These populations differ from each other by varied strategy parameters. The number of offspring is such a strategy parameter which is different in each population and it is shown that it can be adjusted adaptively with the proposed algorithm.After derivating the theoretical values for the optimal number of offspring for the two exemplary testfunctions hyperplane and hypersphere the simulation results for these testfunctions are shown.