Genetic Programming Implementation

Question

I am familiar with the generalities of genetic programming but am wondering where i might find something that shows me details of implementing genetic programming. I use C# and .NET 3.5, and I would like to put to use genetic programming for things like pathfinding, and generally just want to see what it can do. EDIT: I should probably clarify what I'm looking for: I'm interested in what sort of data structures would be used to store the syntax trees, how a breeding operation might be performed, that sort of thing.

Answer 1

Here is a quick rewrite of one of C++ HelloWorld examples that helped me learn genetic programming:

using ga_vector = List<ga_struct>;

class ga_struct
{
    public ga_struct(string str, uint fitness)
    {
        Str = str;
        Fitness = fitness;
    }

    public string Str { get; set; }
    public uint Fitness { get; set; }
}

class Program
{

    private const int GA_POPSIZE = 2048;
    private const int GA_MAXITER = 16384;
    private const float GA_ELITRATE = 0.10f;
    private const float GA_MUTATIONRATE = 0.25f;
    private const float GA_MUTATION = 32767 * GA_MUTATIONRATE;
    private const string GA_TARGET = "Hello world!";

    private static readonly Random random = new Random((int)DateTime.Now.Ticks);

    static void Main(string[] args)
    {
        ga_vector popAlpha = new ga_vector();
        ga_vector popBeta = new ga_vector();

        InitPopulation(ref popAlpha, ref popBeta);
        ga_vector population = popAlpha;
        ga_vector buffer = popBeta;

        for (int i = 0; i < GA_MAXITER; i++)
        {
            CalcFitness(ref population);
            SortByFitness(ref population);
            PrintBest(ref population);

            if (population[0].Fitness == 0) break;

            Mate(ref population, ref buffer);
            Swap(ref population, ref buffer);
        }

        Console.ReadKey();
    }

    static void Swap(ref ga_vector population, ref ga_vector buffer)
    {
        var temp = population;
        population = buffer;
        buffer = temp;
    }

    static void InitPopulation(ref ga_vector population, ref ga_vector buffer)
    {
        int tsize = GA_TARGET.Length;
        for (int i = 0; i < GA_POPSIZE; i++)
        {
            var citizen = new ga_struct(string.Empty, 0);

            for (int j = 0; j < tsize; j++)
            {
                citizen.Str += Convert.ToChar(random.Next(90) + 32);
            }

            population.Add(citizen);
            buffer.Add(new ga_struct(string.Empty, 0));
        }
    }

    static void CalcFitness(ref ga_vector population)
    {
        const string target = GA_TARGET;
        int tsize = target.Length;

        for (int i = 0; i < GA_POPSIZE; i++)
        {
            uint fitness = 0;
            for (int j = 0; j < tsize; j++)
            {
                fitness += (uint) Math.Abs(population[i].Str[j] - target[j]);
            }

            population[i].Fitness = fitness;
        }
    }

    static int FitnessSort(ga_struct x, ga_struct y)
    {
        return x.Fitness.CompareTo(y.Fitness);
    }

    static void SortByFitness(ref ga_vector population)
    {
        population.Sort((x, y) => FitnessSort(x, y));
    }

    static void Elitism(ref ga_vector population, ref ga_vector buffer, int esize)
    {
        for (int i = 0; i < esize; i++)
        {
            buffer[i].Str = population[i].Str;
            buffer[i].Fitness = population[i].Fitness;
        }
    }

    static void Mutate(ref ga_struct member)
    {
        int tsize = GA_TARGET.Length;
        int ipos = random.Next(tsize);
        int delta = random.Next(90) + 32;

        var mutated = member.Str.ToCharArray();
        Convert.ToChar((member.Str[ipos] + delta)%123).ToString().CopyTo(0, mutated, ipos, 1);
        member.Str = mutated.ToString();
    }

    static void Mate(ref ga_vector population, ref ga_vector buffer)
    {
        const int esize = (int) (GA_POPSIZE*GA_ELITRATE);
        int tsize = GA_TARGET.Length, spos, i1, i2;

        Elitism(ref population, ref buffer, esize);

        for (int i = esize; i < GA_POPSIZE; i++)
        {
            i1 = random.Next(GA_POPSIZE/2);
            i2 = random.Next(GA_POPSIZE/2);
            spos = random.Next(tsize);

            buffer[i].Str = population[i1].Str.Substring(0, spos) + population[i2].Str.Substring(spos, tsize - spos);

            if (random.Next() < GA_MUTATION)
            {
                var mutated = buffer[i];
                Mutate(ref mutated);
                buffer[i] = mutated;
            }
        }
    }

    static void PrintBest(ref ga_vector gav)
    {
        Console.WriteLine("Best: " + gav[0].Str + " (" + gav[0].Fitness + ")");
    }

There might be some minor errors but otherwise it looks it's working ok. Also it could be written better in spirit of C# but those are just details. :)

Answer 2

Roger Alsing's Mona Lisa project is a quite good example. http://rogeralsing.com/2008/12/07/genetic-programming-evolution-of-mona-lisa/

EDIT: The reason I like the example is because it rather small and easy to understand. Its a quick and easy way to grasp the concept of genetic programming.

Answer 3

You could look at Survival of the Fittest: Natural Selection with Windows Forms.

EDIT: See this previous SO question, which I just found. It's pretty much a duplicate. Sorry you don't understand the link (it's good to mention such things in the question). Also, the other question is still open for more answers/edits, even though an answer has been accepted.

Answer 4

You can try this C# .NET 4.0 port of Sean Luke's ECJ (Evolutionary Computation in Java):

http://branecloud.codeplex.com

It is very flexible and powerful software! But it is also relatively easy to get started because it includes many working console samples out-of-the-box (and many helpful unit tests that were developed during the conversion).

Ben