Algorithm

Contents

Algorithm#

Evolutionary Algorithm defined in the experiment. An algorithm can operate on a single Subpopulation, or multiple Subpopulations (coevolution, island model).

The following psuedocode demonstrates the main flow of the algorithm:

1. create initial population
2. while termination condition not met do:
    a. evaluate the fitness functions of the current generation
    b. select parents from the current generation
    c. perform crossover and mutation operators

Each step is performed by an object that is invoked by the algorithm:

  • The creation of the initial population is done by the Creator.

  • The termination condition check is done by the TerminationChecker.

  • The fitness evaluation of the population is done by the Evaluator.

  • The selection, crossover and mutation is done by the Breeder.

Algorithm types#

Algorithm is an abstract class. The most straightforward concrete Algorithm class is SimpleEvolution. SimpleEvolution assumes there is only one population in the evolutionary experiment. Another Algorithm subclass is NSGA2Evolution, used for multi-objective evolution (MOE).