Evolution strategy (ES)

variant of EA, which generally operates on the ``natural'' problem representation (no genotype-phenotype mapping for object parameters). The basic types of ES are classified using the $(\mu/\rho \stackrel{+}{,} \lambda)$-ES (pronounced: ``mu-slash-rho-plus-(or)-comma-lambda-ES'') notation. Here $\mu$ is the number of parents, $\lambda$ is the number of offspring and $\rho$ is the mixing number (the number of parents which take part in the (multi-) recombination process to produce one offspring); ``$+$'' or ``,'' determines the type of truncation selection. An individual consists of a set of object parameters, the corresponding value of the objective function and a set of (endogenous) strategy parameters. The ES employs mutation and recombination (if $\rho > 1$) as variation operators.