public T GetChildren <T>(double mutationRate, int amount) where T : GeneralizedCreature, new()
{
if (CreatureBrain == null)
{
return(default(T));
}
if (!CreatureBrain.IsMeshed)
{
return(default(T));
}
DNA _DNAChild = DNA.CrossOver(CreatureBrain.DNA, CreatureBrain.DNA);
_DNAChild.Mutate(mutationRate, amount);
T _NewChild = new T()
{
CreatureBrain =
new Brain(
CreatureBrain.Layers[0].NeuroneCount,
CreatureBrain.Layers[1].NeuroneCount,
CreatureBrain.Layers[2].NeuroneCount,
CreatureBrain.Layers[3].NeuroneCount,
_DNAChild)
};
_NewChild.CreatureBrain.Mesh();
return((T)_NewChild);
}
private void CreatenextGenerationCars(int[] indexProbability, List <GameObject> nextGenerationCars)
{
int totalScore = ActualCars.Sum(x => x.GetComponent <Car>().Score);
//PrintScores();
for (int i = 0; i < this.Population; i++)
{
int firstCarIndex = indexProbability[Random.Range(0, totalScore - 1)];
int secondCarIndex = indexProbability[Random.Range(0, totalScore - 1)];
DNA first = ActualCars[firstCarIndex].GetComponent <Car>().DNA;
DNA second = ActualCars[secondCarIndex].GetComponent <Car>().DNA;
DNA crossOver = first.CrossOver(first, second);
crossOver.Mutate();
float xRandom = Random.Range(InitPosition.x - 1f, InitPosition.x + 1f);;
float yRandom = Random.Range(InitPosition.y - 0.5f, InitPosition.y + 0.5f);
GameObject newCar = Instantiate(Prefab, new Vector2(xRandom, yRandom), Rotation);
Car car = newCar.GetComponent <Car>();
car.Initialize(crossOver);
car.CarEvent += ReducePopulation;
nextGenerationCars.Add(newCar);
}
}
public void Should_CrossOver(string str1, string str2, string expected)
{
DNA parent1 = new DNA(str1.Length);
parent1.Word = new StringBuilder(str1);
DNA parent2 = new DNA(str2.Length);
parent2.Word = new StringBuilder(str2);
DNA child = parent1.CrossOver(parent2);
Assert.Equal(child.Word.ToString(), expected);
}
// Creating a new Generation
public void Generate()
{
for (int i = 0; i < popullations.Length; i++)
{
DNA randomChild_1 = mattingPool [Random.Range(0, mattingPool.Count)];
DNA randomChild_2 = mattingPool [Random.Range(0, mattingPool.Count)];
DNA newChild = randomChild_1.CrossOver(randomChild_2);
newChild.Mutate(mutationRate);
this.popullations [i] = newChild;
}
CalculateFitness();
generation++;
}
public void NewGeneration()
{
if (population.Count <= 0)
{
return;
}
CalculateFitness();
List <Challenger> newPopulation = new List <Challenger>();
// for (int i = 0; i < amountOfChallengers; i++)
// {
// Challenger c = Instantiate(challengerPrefab, challengerPrefab.transform.position, Quaternion.identity);
// c.dna = new DNA<Direction>(genesLength, random, GetRandomDirection, FitnessFunction, true);
// c.transform.name = "Challenger " + i;
// newPopulation.Add(c);
// }
//crossovers
for (int i = 0; i < population.Count; i++)
{
DNA <Direction> parent1 = ChooseParent();
DNA <Direction> parent2 = ChooseParent();
DNA <Direction> child = null;
Challenger c = Instantiate(challengerPrefab, challengerPrefab.transform.position, Quaternion.identity);
if (parent1 != null && parent2 != null)
{
child = parent1.CrossOver(parent2);
child.Mutate(MutationRate);
c.dna = child;
}
else
{
c.dna = new DNA <Direction>(genesLength, random, GetRandomDirection, FitnessFunction, true);
}
newPopulation.Add(c);
}
foreach (var item in population)
{
Destroy(item.gameObject);
}
population.Clear();
population = newPopulation;
Generation++;
}
public void NewGeneration(int numNewDNA = 0, bool crossoverNewDna = false)
{
int finalCount = Population.Count + numNewDNA;
if (finalCount <= 0)
{
return;
}
if (Population.Count > 0)
{
CalculateFitness();
Population.Sort(CompareDNA);
}
newPopulation.Clear();
for (int i = 0; i < finalCount; i++)
{
if (i < Elitism && i < Population.Count)
{
newPopulation.Add(Population[i]);
}
else if (i < Population.Count || crossoverNewDna)
{
DNA <T> parent1 = ChooseParent();
DNA <T> parent2 = ChooseParent();
DNA <T> child = parent1.CrossOver(parent2);
child.Mutate(MutationRate);
newPopulation.Add(child);
}
else
{
newPopulation.Add(new DNA <T>(dnaSize, random, getRandomGene, fitnessFunction, shouldInitGenes: true));
}
}
List <DNA <T> > tmpList = Population;
Population = newPopulation;
newPopulation = tmpList;
Generation++;
}
static void Main(string[] args)
{
var neural1 = new NeuralNetwork(new int[] { 4, 4, 2 });
var neural2 = new NeuralNetwork(new int[] { 4, 4, 2 });
Console.WriteLine("Valeurs d'entrées : 0 1 0 1");
Console.WriteLine("-- --- --- --- --- --- --- --");
/* Test de la propagation des données, on doit obtenir les mêmes résultats si on mets les mêmes entrées! */
Console.WriteLine("Reseau neuronal 1");
for (int i = 0; i < 2; i++)
{
var results = neural1.Propagate(new float[] { 0, 1, 0, 1 });
Console.Write(string.Format("n{0}\n", i));
foreach (var result in results)
{
Console.WriteLine(result.ToString());
}
Console.Write("\n");
}
Console.WriteLine("-- --- --- --- --- --- --- --");
Console.WriteLine("Reseau neuronal 2");
for (int i = 0; i < 2; i++)
{
var results = neural2.Propagate(new float[] { 0, 1, 0, 1 });
Console.Write(string.Format("n{0} : \n", i));
foreach (var result in results)
{
Console.WriteLine(result.ToString());
}
Console.Write("\n");
}
Console.WriteLine("-- --- --- --- --- --- --- --");
Console.WriteLine("CrossOver");
/* Test du cross-over */
DNA parent1 = new DNA(neural1);
DNA parent2 = new DNA(neural2);
DNA child1 = parent1.CrossOver(parent2);
DNA child2 = parent2.CrossOver(parent1);
Console.WriteLine("\nSans mutation\n");
for (int i = 0; i < parent1.genes.Length; i++)
{
Console.Write(string.Format("P1 : {0} | P2 : {1} | C1 : {2}{4} | C2 : {3}{5}\n",
Math.Round(parent1.genes[i], 5),
Math.Round(parent2.genes[i], 5),
Math.Round(child1.genes[i], 5),
Math.Round(child2.genes[i], 5),
child1.genes[i] == parent1.genes[i] ? " (P1)" : child1.genes[i] == parent2.genes[i] ? " (P2)" : " (MU)",
child2.genes[i] == parent1.genes[i] ? " (P1)" : child2.genes[i] == parent2.genes[i] ? " (P2)" : " (MU)"
)
);
}
int mutationRate = 5;
child1.Mutate(mutationRate);
child2.Mutate(mutationRate);
Console.WriteLine(string.Format("\nAvec un taux de mutation a {0}/100 par gènes, soit une déterioration de l'ADN de {1}% max, ce qui fait que l'ADN contient au min {2}% des gènes des parents.\n", mutationRate, (mutationRate / 100.0) * 100, 100.0 - (mutationRate)));
for (int i = 0; i < parent1.genes.Length; i++)
{
Console.Write(string.Format("P1 : {0} | P2 : {1} | C1 : {2}{4} | C2 : {3}{5}\n",
Math.Round(parent1.genes[i], 5),
Math.Round(parent2.genes[i], 5),
Math.Round(child1.genes[i], 5),
Math.Round(child2.genes[i], 5),
child1.genes[i] == parent1.genes[i] ? " (P1)" : child1.genes[i] == parent2.genes[i] ? " (P2)" : " (MU)",
child2.genes[i] == parent1.genes[i] ? " (P1)" : child2.genes[i] == parent2.genes[i] ? " (P2)" : " (MU)"
)
);
}
var childNeural1 = new NeuralNetwork(child1);
var childNeural2 = new NeuralNetwork(child2);
var child1Result = childNeural1.Propagate(new float[] { 0, 1, 0, 1 });
var child2Result = childNeural2.Propagate(new float[] { 0, 1, 0, 1 });
Console.WriteLine("Resultat du child 1 : ");
foreach (var result in child1Result)
{
Console.Write(result.ToString() + "\n");
}
Console.WriteLine("\nResultat du child 2 : ");
foreach (var result in child2Result)
{
Console.Write(result.ToString() + "\n");
}
}
