16.1 Overview

The genetics package provides a framework and implementations for genetic algorithms.

16.2 GA Framework

GeneticAlgorithm provides an execution framework for Genetic Algorithms (GA). Populations, consisting of Chromosomes are evolved by the GeneticAlgorithm until a StoppingCondition is reached. Evolution is determined by SelectionPolicy, MutationPolicy and Fitness.

The GA itself is implemented by the evolve method of the GeneticAlgorithm class, which looks like this:

public Population evolve(Population initial, StoppingCondition condition) {
    Population current = initial;
    while (!condition.isSatisfied(current)) {
        current = nextGeneration(current);
    }
    return current;
}
          

1. Get nextGeneration population to fill from current generation, using its nextGeneration method
2. Loop until new generation is filled:
• Apply configured SelectionPolicy to select a pair of parents from current
• With probability = getCrossoverRate(), apply configured CrossoverPolicy to parents
• With probability = getMutationRate(), apply configured MutationPolicy to each of the offspring
• Add offspring individually to nextGeneration, space permitting
3. Return nextGeneration

16.3 Implementation

Here is an example GA execution:

// initialize a new genetic algorithm
GeneticAlgorithm ga = new GeneticAlgorithm(
    new OnePointCrossover<Integer>(),
    1,
    new RandomKeyMutation(),
    0.10,
    new TournamentSelection(TOURNAMENT_ARITY)
);

// initial population
Population initial = getInitialPopulation();

// stopping condition
StoppingCondition stopCond = new FixedGenerationCount(NUM_GENERATIONS);

// run the algorithm
Population finalPopulation = ga.evolve(initial, stopCond);

// best chromosome from the final population
Chromosome bestFinal = finalPopulation.getFittestChromosome();