Package org.opt4j.optimizers.ea.aeseh

Package for the classes of the Adaptive ε-Sampling ε-Hood MOEA. This evolutionary algorithm performs both the survivor and the parent selection based on ε-sampling.

The EpsilonSamplingSelector uses the ESamplingSurvivorGeneration to pick the individuals that form the pool of possible parents. Hereby, ε-dominant Individuals are preferred. The check for ε-dominance is implemented by enhancing the Objectives of the considered individual using the EpsilonMapping. For the parent selection, the EpsilonNeighborhoodCoupler divides the survivor pool based on ε-dominance among the survivors. A pair of parents is then always picked from the same neighborhood, while the arbitration of the neighborhoods to pick from is handled by the NeighborhoodScheduler.

Throughout the exploration, the ε-values used for the survivor generation and the parent selection are adjusted in order to pick a number of ε-dominant survivors that equals the number of individuals that are to be generated in each generation and to create a number of neighborhoods close to a number provided by the user. The adaptation of the ε-values is hereby managed by the EpsilonAdaptation.

In the Opt4J GUI, the AeSeH is configured using the AeSeHModule and the EvolutionaryAlgorithmModule.