On the use of stochastic ranking for parent selection in differential evolution for constrained optimization

Year
2017
Type(s)
Author(s)
Gregorio Toscano and Giomara Larraga, Ricardo Landa and Guillermo Leguizamon
Source
Soft Computing, 21(16): 4617—4633, 2017
Url
https://doi.org/10.1007/s00500-016-2073-6

The role of parent selection is to distinguish between individuals based on their quality. Parent selection has been a key component in the design of evolutionary algorithms, since it is partially responsible for improving the quality of the population. Stochastic ranking is quite a successful approach for evolutionary constrained optimization, and has usually been employed during the survival selection process. This paper provides the first insight into the use of the stochastic ranking procedure during the parent selection mechanism in the design of new evolutionary algorithms for constraint handling. We adopted differential evolution as the base algorithm, mainly because of its outstanding performance in continuous optimization found in literature. We undertake seven experiments in order to validate our proposal. The results indicate that our proposed approach is able to find solutions that are competitive with respect to other recently proposed approaches, but uses a fraction of the required computational effort. Furthermore, our proposal can easily be incorporated into any probabilistic evolutionary algorithm that is based on parent selection.