A polymorphical mutation operator for evolution strategies

Since the advent of evolution strategies many attempts have been done to improve the mutation operator [7]. The first enhancement could be seen in the introduction of a flexible step size. As a logical consequence later n step sizes, one per variable, were used. Highest flexibility is achieved with the correlated mutation but also quadratic growth of the number of strategy parameters [8]. A newer approach introduces asymmetry to the mutation operator getting by with linear growth of the strategy parameters [3]. The Polymorphical Mutation Operator, a new asymmetrical mutation operator, will be presented in the following. It overcomes some of the disadvantages of the one mentioned before.