public static void RouletteWheelSelection <T>(Chromosome <T>[] Chs, double[] Fit)
{
Chromosome <T>[] selection = new Chromosome <T> [Chs.Length];
double total = 0;
for (int i = 0; i < Chs.Length; i++)
{
total += 1 / Fit[i];
}
double[] sectors = new double[Chs.Length];
for (int i = 0; i < Chs.Length; i++)
{
sectors[i] = (1 / Fit[i]) / total * 100;
}
//randomly selecting crossing chromosomes according to their weights
for (int i = 0; i < Chs.Length; i++)
{
double num = Rng.NextDouble() * 100, seek = 0;
for (int j = 0; j < Chs.Length; j++)
{
seek += sectors[j];
if (num < seek)
{
selection[i] = Chs[j];
break;
}
}
}
selection.CopyTo(Chs, 0);
}
public static void Crossover <T>(Chromosome <T>[] Chs)
{
Chromosome <T>[] crossed = new Chromosome <T> [Chs.Length];
//forming pairs
int[] pairs = new int[Chs.Length];
for (int i = 0; i < Chs.Length; i++)
{
pairs[i] = i;
}
//shuffling array of pairs to make them random
int n = pairs.Length;
while (n > 1)
{
int k = Rng.Next(n--);
int t = pairs[n];
pairs[n] = pairs[k];
pairs[k] = t;
}
//crossing
for (int i = 0; i < pairs.Length / 2; i++)
{
int point = Rng.Next(0, Chs[0].ParamNum - 1);
T[] Ch1 = new T[Chs[0].ParamNum], Ch2 = new T[Chs[0].ParamNum];
for (int j = 0; j < Chs[0].ParamNum; j++)
{
if (j <= point)
{
Ch1[j] = Chs[pairs[2 * i]].Genes[j];
Ch2[j] = Chs[pairs[2 * i + 1]].Genes[j];
}
else
{
Ch1[j] = Chs[pairs[2 * i + 1]].Genes[j];
Ch2[j] = Chs[pairs[2 * i]].Genes[j];
}
}
crossed[2 * i] = new Chromosome <T>(Ch1);
crossed[2 * i + 1] = new Chromosome <T>(Ch2);
}
crossed.CopyTo(Chs, 0);
}
private void GenerateNextPopulation()
{
Random rnd = new Random();
Chromosome oldBestSolution = new Chromosome(ProblemParameters.Dimension);
Chromosome newBestSolution, newWrostSolution;
// Store best found solution
_bestOfCurrentPopulation.CopyTo(oldBestSolution);
//
#region Selection (Algorithm: Tournament)
int tourSize; // Tournament size
int bestChromosomeIndex = 0;
double bestFitness;
if (PopulationSize < 32)
{
tourSize = 3;
}
else
{
tourSize = (int)Math.Log(PopulationSize, 2) - 1;
}
for (int i = 0; i < PopulationSize; i++)
{
bestFitness = Double.NegativeInfinity;
for (int j = 0; j < tourSize; j++)
{
int rndIndex;
rndIndex = rnd.Next(PopulationSize);
if (_currentPopulation[rndIndex].FitnessValue > bestFitness)
{
bestFitness = _currentPopulation[rndIndex].FitnessValue;
bestChromosomeIndex = rndIndex;
}
}
// Copy the "winner of tournament" to Intermediate Population
_currentPopulation[bestChromosomeIndex].CopyTo(_intermediatePopulation[i]);
}
#endregion
#region Crossover (Algorithm: Arithmetic Crossover)
for (int i = 0; i < PopulationSize / 2; i++)
{
if (rnd.NextDouble() < ProbabilityOfCrossover)
{
// Do Crossover
Chromosome.DoCrossover(_intermediatePopulation[i], _intermediatePopulation[i + 1],
_currentPopulation[i], _currentPopulation[i + 1]);
}
else
{
// Do NOT Crossover (Copy parents to the new population without any change)
_intermediatePopulation[i].CopyTo(_currentPopulation[i]);
_intermediatePopulation[i + 1].CopyTo(_currentPopulation[i + 1]);
}
}
#endregion
#region Mutation
newBestSolution = newWrostSolution = _currentPopulation[0];
for (int i = 0; i < PopulationSize; i++)
{
if (rnd.NextDouble() < ProbabilityOfMutation)
{
_currentPopulation[i].DoMutation();
}
// Find wrost of this (new) population
if (_currentPopulation[i].FitnessValue < newWrostSolution.FitnessValue)
{
newWrostSolution = _currentPopulation[i];
}
// Find best of this (new) population
if (_currentPopulation[i].FitnessValue > newBestSolution.FitnessValue)
{
newBestSolution = _currentPopulation[i];
}
}
#region Elitism
oldBestSolution.CopyTo(newWrostSolution); // After this statement, "newWrostSolution" will point to "Old Best Solution"!
// Now we must find best of current and old populations
_bestOfCurrentPopulation = (oldBestSolution.FitnessValue < newBestSolution.FitnessValue)
? newBestSolution : newWrostSolution;
#endregion
#endregion
}
