protected override IList <IChromosome> PerformCross(IList <IChromosome> parents)
{
ExpressionChromosome parent1 = parents[0] as ExpressionChromosome;
ExpressionChromosome parent2 = parents[1] as ExpressionChromosome;
int index1 = (int)Math.Floor(RandomizationProvider.Current.GetDouble() * parent1.Length);
int index2 = (int)Math.Floor(RandomizationProvider.Current.GetDouble() * parent2.Length);
Gene[] subtree1 = GetSubtree(parent1.GetGenes(), index1);
Gene[] subtree2 = GetSubtree(parent2.GetGenes(), index2);
// Copy subtree2 to parent1 at index1.
ExpressionChromosome child1 = parent1.CreateNew() as ExpressionChromosome;
Gene[] result = GetNewGenes(parent1.GetGenes(), index1, subtree1, subtree2);
child1.SetGenes(result);
// Copy subtree1 to parent2 at index2.
ExpressionChromosome child2 = parent2.CreateNew() as ExpressionChromosome;
result = GetNewGenes(parent2.GetGenes(), index2, subtree2, subtree1);
child2.SetGenes(result);
List <IChromosome> childs = new List <IChromosome>();
if (child1.Length < maxLength)
{
childs.Add(child1);
}
if (child2.Length < maxLength)
{
childs.Add(child2);
}
return(childs);
}
public double Evaluate(IChromosome chromosome)
{
ExpressionChromosome ch = chromosome as ExpressionChromosome;
if (ch == null)
{
throw new ArgumentException("Chromosome type is not ExpressionChromosome");
}
double fitness = 0;
foreach (InputFunction func in data)
{
double result = ch.Evaluate(func.Parameters);
fitness += Math.Abs(result - func.Result);
}
if (double.IsNaN(fitness))
{
return(-1000);
}
return(fitness * -1);
}
static void Main(string[] args)
{
string s1 = "c-5*(a*(b+5)-1)/c+(a*b/c)";
string s2 = "a*b*c+(a/2)*b-(c/(b+9))";
int variableNumber = 3;
string prefix1 = PrefixHelper.InfixToPrefix(s1);
string prefix2 = PrefixHelper.InfixToPrefix(s2);
List <InputFunction> data = new List <InputFunction>();
int sampleDataSize = 20;
using (StreamWriter sw = new StreamWriter("InputData.txt"))
{
for (int i = 0; i < sampleDataSize; i++)
{
sw.Write("(");
double[] values = new double[variableNumber];
for (int j = 0; j < variableNumber; j++)
{
values[j] = RandomizationProvider.Current.GetDouble(-10, 10);
sw.Write(values[j]);
if (j < variableNumber - 1)
{
sw.Write(',');
}
}
sw.Write(")(");
double result1 = PrefixHelper.EvaluatePrefix(prefix1, values);
double result2 = PrefixHelper.EvaluatePrefix(prefix2, values);
sw.WriteLine($"{result1},{result2})");
InputFunction input = new InputFunction(result1, values);
data.Add(input);
}
}
#region old sample data
// Sample data for 5A-7B
//List<InputFunction> data = new List<InputFunction>()
//{
// new InputFunction(-15, 4,5),
// new InputFunction(-2,1,1),
// new InputFunction(8,-4,-4),
// new InputFunction(-31,-2,3)
//};
//int funcLength = 2;
// Sample data for 5A-7B-4C+2D
//List<InputFunction> data = new List<InputFunction>()
//{
// new InputFunction(11, 4,5,-5,3),
// new InputFunction(12,1,1,0,7),
// new InputFunction(0,-4,-4,3,2),
// new InputFunction(-9,-2,3,-7,-3)
//};
//int funcLength = 4;
// Sample data for 5A-7B-4C+2D+0E+1F+9G
//List<InputFunction> data = new List<InputFunction>()
//{
// new InputFunction(152,4,5,-5,3,8,6,15),
// new InputFunction(221,1,1,0,7,-25,-7,24),
// new InputFunction(-269,-4,-4,3,2,-1,55,-36),
// new InputFunction(384,-2,3,-7,-3,0,-12,45)
//};
//int funcLength = 7;
// Custom data input
//Console.WriteLine("Enter function length:");
//int funcLength = int.Parse(Console.ReadLine());
//int[] parameters = new int[funcLength];
//while (true)
//{
// for (int i = 0; i < funcLength; i++)
// {
// Console.WriteLine("Enter #{0} param:", i);
// parameters[i] = int.Parse(Console.ReadLine());
// }
// Console.WriteLine("Enter result:");
// int result = int.Parse(Console.ReadLine());
// data.Add(new InputFunction(result, parameters));
// Console.WriteLine("press enter to continue, or write 'done' to start the ga");
// string input = Console.ReadLine();
// if (input == "done")
// break;
//}
#endregion
int maxLength = 6 + 5 * variableNumber;
IChromosome chromosome = new ExpressionChromosome(variableNumber, maxLength);
IPopulation population = new Population(8500, 10000, chromosome);
population.GenerationStrategy = new PerformanceGenerationStrategy();
IFitness fitness = new ExpressionFitness(data.ToArray());
ISelection selection = new EliteSelection();
ICrossover crossover = new ExpressionCrossover(maxLength);
IMutation mutation = new ExpressionMutation(variableNumber);
GeneticAlgorithm ga = new GeneticAlgorithm(population, fitness, selection, crossover, mutation);
ga.Termination = new ExpressionTermination(200);
ga.MutationProbability = .5f;
double latestFitness = double.MinValue;
ga.GenerationRan += (sender, e) =>
{
Console.Title = ga.TimeEvolving.ToString();
//Console.Title = ga.Population.CurrentGeneration.Chromosomes.Count.ToString();
ExpressionChromosome bestChromosome = ga.BestChromosome as ExpressionChromosome;
double bestFitness = bestChromosome.Fitness.Value;
if (bestFitness != latestFitness)
{
latestFitness = bestFitness;
Console.WriteLine("\n--------\n");
Console.WriteLine("Generation: {0}", ga.Population.GenerationsNumber);
//Console.WriteLine("Time: {0}", ga.TimeEvolving);
Console.WriteLine("Fitness: {0}", bestFitness.ToString("0.00"));
Console.WriteLine("Best: {0}", bestChromosome.ToString());
}
};
ga.Start();
Console.WriteLine("\nTime: {0}", ga.TimeEvolving);
double sum = 0;
double[] diffs = new double[sampleDataSize];
for (int i = 0; i < sampleDataSize; i++)
{
double resultActual = PrefixHelper.EvaluatePrefix(prefix1, data[i].Parameters);
double resultOur = PrefixHelper.EvaluatePrefix(ga.BestChromosome.GetGenes(), data[i].Parameters);
double diff = Math.Abs(resultActual - resultOur);
Console.WriteLine($"Actual: {resultActual.ToString("0.00")} | Our: {resultOur.ToString("0.00")} | Diff: {diff.ToString("0.00")}");
sum += diff;
diffs[i] = diff;
}
double avgDiff = sum / sampleDataSize;
Console.WriteLine($"Average diff: {avgDiff.ToString("0.00")} | Median: {GetMedian(diffs).ToString("0.00")}");
Console.WriteLine("\n ---- New Samples ---- \n");
for (int i = 0; i < sampleDataSize; i++)
{
double[] values = new double[variableNumber];
for (int j = 0; j < variableNumber; j++)
{
values[j] = RandomizationProvider.Current.GetDouble(-10, 10);
}
double result = PrefixHelper.EvaluatePrefix(prefix1, values);
double our = PrefixHelper.EvaluatePrefix(ga.BestChromosome.GetGenes(), values);
double diff = Math.Abs(result - our);
Console.WriteLine($"Actual: {result.ToString("0.00")} | Our: {our.ToString("0.00")} | Diff: {diff.ToString("0.00")}");
sum += diff;
diffs[i] = diff;
}
avgDiff = sum / sampleDataSize;
Console.WriteLine($"Average diff: {avgDiff.ToString("0.00")} | Median: {GetMedian(diffs).ToString("0.00")}");
Console.ReadKey();
}