Representation

also coding or problem representation, the mapping of the optimization problem to the genome level. Problem-dependent (also called ``natural'') representations and problem-independent representations have to be distinguished. For parameter optimization problems genotype and phenotype are identical when using the ``natural'' representation (see ES or real-coded GA). For combinatorial or structure optimization problems the ``natural'' representation cannot be defined in general. The canonical GA typically operates on a problem-independent representation, i.e., all object parameter spaces are mapped onto the space of bit-strings (e.g., binary or Gray code for the representation of real-valued numbers). If every property of the phenotype has a unique state in the genotype, the representation is called direct, otherwise indirect. If parts of the genotype do not have any influence on the phenotype, the coding is called redundant.