Genetic Algorithm

GA.jl

import Base: <,>,==

const MR = 0.3
const MA = 0.2
const XR = 0.5
const T_SIZE = 20
const L = 30
const S = 200
const G = 100
const best = minimum
const absMax = 10

const SILENT = true

function sprintln(args...)
    if !SILENT
        println(args...)
    end
end

mutable struct Individual
    chromosome
    fitness::Float32
    function Individual(size::Int, fitness)
        this = new()
        this.chromosome = rand(size) * absMax*2 - absMax
        this.fitness = fitness(this)
        return this
    end
    function Individual(data)
        this = new()
        this.chromosome = data
        this.fitness = 0.0
        return this
    end
end

<(x::Individual, y::Individual) = x.fitness < y.fitness
>(x::Individual, y::Individual) = x.fitness > y.fitness
==(x::Individual, y::Individual) = x.fitness == y.fitness

function mutate!(input::Individual)
    mut_amp = MA
    mut_rate = MR
    # mutator = () -> rand(Float32) < mut_rate ? rand(Float32)*mut_amp*2 - mut_amp : 0
    # input.chromosome .+= mutator.()
    for i = 1:length(input.chromosome)
        input.chromosome[i] += rand(Float32) < mut_rate ? rand(Float32)*mut_amp*2 - mut_amp : 0
    end
end

function crossover(parent1::Individual, parent2::Individual)
    xover_chance = XR
    if rand() < xover_chance
        newChromosome = zeros(Float32, length(parent1.chromosome))
        for i = 1:length(newChromosome)
            @inbounds newChromosome[i] = (rand() < 0.5)? parent1.chromosome[i]:parent2.chromosome[i]
        end
        return Individual(newChromosome)
    else
        return parent1
    end
end

function select(pop)
    tournament = pop[rand(1:length(pop),T_SIZE)]
    return best(tournament)
end

function genIndividual(pop)
    newInd = crossover(select(pop), select(pop))
    mutate!(newInd)
    newInd.fitness = sphereFitness(newInd)
    return newInd
end

# sphereFitness(input::Individual) = sum(input.chromosome .* input.chromosome)

function sphereFitness(input::Individual)
    total = 0.0f0
    for gene in input.chromosome
        total += gene*gene
    end
    return total
end

function main()
    length = L
    size = S
    gen = G
    pop = [Individual(length,sphereFitness) for _ in 1:size]
    for i = 1:gen
        newpop = Array{Individual}(size)
        newpop[1] = minimum(pop)
        for j = 2:size
           newpop[j] = genIndividual(pop)
        end
        pop = newpop
        sprintln(minimum(pop).fitness)
    end
    println(minimum(pop).fitness)
    return minimum(pop)
end

main()
@time main()