genetic_algorithm.py

import random
import numpy

LOW_BOUND = -10.0
HIGH_BOUND = 10.0

POPULATION_SIZE = 100
MAX_GENERATIONS = 1000


def calculate_fitness(gene):
    variables = [6, -1, 4, 3, 19, -3]
    return numpy.sum(numpy.multiply(gene, variables))


def crossover(gene1, gene2):
    position = numpy.random.randint(len(gene1))
    new_gene1 = gene1[:position] + gene2[position:]
    new_gene2 = gene2[:position] + gene1[position:]
    return new_gene1, new_gene2


def mutate(gene):
    position = numpy.random.randint(len(gene))
    new_gene = gene
    new_gene[position] = numpy.random.uniform(low=LOW_BOUND, high=HIGH_BOUND)
    return new_gene


def run_evolution():
    fittest_gene = None
    genes = numpy.random.uniform(
            low=LOW_BOUND, high=HIGH_BOUND, size=(POPULATION_SIZE, 6),
    ).tolist()
    for generation in xrange(MAX_GENERATIONS):
        # Evaluate fitness of genes
        fitnesses = map(calculate_fitness, genes)
        ranked_genes = sorted(zip(fitnesses, genes))

        generation_fittest = ranked_genes[-1]
        if not fittest_gene or generation_fittest[0] > fittest_gene[0]:
            fittest_gene = generation_fittest
            print "Generation {0}: {1}".format(generation, fittest_gene)

        # Survival of the fittest, get the top half of the population
        survivors = ranked_genes[len(ranked_genes) // 2:]
        genes = [survivor[1] for survivor in survivors]

        survivor_count = len(survivors)

        for _ in xrange(survivor_count // 2):
            # Crossover
            gene1 = random.choice(survivors)[1]
            gene2 = random.choice(survivors)[1]
            new_genes = crossover(gene1, gene2)
            for new_gene in new_genes:
                roll = random.random()
                # Mutation
                if roll > 0.1:
                    new_gene = mutate(new_gene)
                genes.append(new_gene)

    return fittest_gene


if __name__ == "__main__":
    print run_evolution()