EVOLUTION STRATEGIES

This is developed through natural selection as inspired from the theory of adaptation and evolution. It deals with micro-​ or genomic-​level data. It is a global optimization algorithm for regulating the distribution of mutations, it utilizes self-​adaptive mechanisms. These self-​adaptive mechanisms involve search progress optimization by developing solutions for the problem and also a few parameters for transforming the solutions. Evolution strategies use different selection and sampling schemes such as (1 + 1), (μ + λ) and (μ, λ) evaluation strategies. In (1 + 1) strategy, one of its parent’s real-​evaluated vectors of object variables is formed by applying mutation to each object variable with an equal standard deviation. The resulting individual is then compared and contrasted with its parent and the better one succeeds to the next generation as a parent and the remaining solutions are left. In (μ + λ) strategy, by selecting μ parents from the current generation, λ children are generated. These λ children are generated through some mutation operators on selected parents. From (μ + λ) population, that is, μ parents and λ children, only the best μ succeeds to the next generation. In (μ, λ) evaluation strategy, λ children (with λ ≥ μ) are generated by selecting current generation of μ parents and only the best μ children are selected to the further generation, and parents are discarded completely.