Mutation/ mutation operator

variation operator, which changes the information contained in the genome of a parent according to a given probability distribution. In the case of bit strings, this is realized by random negation (bit mutation, mutation rate) of single bits. For real-valued object parameters a random number from a suitable distribution is added (mutation strength) and for mutative self-adaptation strategies (ES/EP) the strategy parameter is multiplied with a random number (learning parameter). In real-valued $n$-dimensional search spaces, isotropic vs. non-isotropic and correlated vs. uncorrelated mutations can be distinguished, which are often generated from a normal distribution. In GP, mutation is usually achieved by the deletion of a sub-tree, the addition of a randomly generated sub-tree or by the exchange of a random block (node). If necessary, the individual has to be repaired after mutation in order to obtain a consistent program.