An Analysis of Recombination in Some Simple Landscapes
- Cite this paper as:
- Rosenblueth D.A., Stephens C.R. (2009) An Analysis of Recombination in Some Simple Landscapes. In: Aguirre A.H., Borja R.M., Garciá C.A.R. (eds) MICAI 2009: Advances in Artificial Intelligence. MICAI 2009. Lecture Notes in Computer Science, vol 5845. Springer, Berlin, Heidelberg
Recombination is an important operator in the evolution of biological organisms and has also played an important role in Evolutionary Computation. In neither field however, is there a clear understanding of why recombination exists and under what circumstances it is useful. In this paper we consider the utility of recombination in the context of a simple Genetic Algorithm (GA). We show how its utility depends on the particular landscape considered. We also show how the facility with which this question may be addressed depends intimately on the particular representation used for the population in the GA, i.e., a representation in terms of genotypes, Building Blocks or Walsh modes. We show how, for non-epistatic landscapes, a description in terms of Building Blocks manifestly shows that recombination is always beneficial, leading to a “royal road” towards the optimum, while the contrary is true for highly epistatic landscapes such as “needle-in-a-haystack”.
Unable to display preview. Download preview PDF.