public PopulationNovelty(int size,
NoveltyChromosome ancestor,
IFitnessFunction fitnessFunction,
IDistanceFunction distanceFunction,
ISelectionMethod selectionMethod,
int kNearestNeighbours) : base(size, ancestor, fitnessFunction, selectionMethod)
{
if (kNearestNeighbours < 2)
{
throw new ArgumentException("Too small nearest neighbours was specified.");
}
_distanceFunction = distanceFunction;
_kNearestNeighbours = kNearestNeighbours;
}
private void CalculateSparseness()
{
population = population.OrderByDescending(x => x.Fitness).ToList();
// CALCULATE ALL SPARSENESS HERE
Parallel.For(0, population.Count, i =>
{
double[] distances = new double[population.Count + _archive.Count];
NoveltyChromosome auxChromosome = (NoveltyChromosome)this[i];
for (int j = 0; j < population.Count; j++)
{
distances[j] = findDistance(auxChromosome, this[j]);
}
for (int j = 0; j < _archive.Count; j++)
{
distances[population.Count + j] = findDistance(auxChromosome, _archive[j]);
}
double distSum = 0;
double[] closestDistances = new double[_kNearestNeighbours];
double[] closestFitness = new double[_kNearestNeighbours];
for (int k = 0; k < _kNearestNeighbours; k++)
{
closestDistances[k] = int.MaxValue;
}
for (int j = 0; j < population.Count + _archive.Count; j++)
{
double currentDistance = distances[j];
//
// If it's better than the worst saved distances
//
if (currentDistance < closestDistances[_kNearestNeighbours - 1] && (i != j))
{
bool shift = false;
double auxDistance1 = 0;
double auxDistance2 = 0;
double auxFitness1 = 0;
double auxFitness2 = 0;
for (int k = 0; k < _kNearestNeighbours; k++)
{
//
// Save the new distance at its proper place in the closestDistances array
//
if (shift == false)
{
if (currentDistance < closestDistances[k])
{
shift = true;
auxDistance1 = closestDistances[k];
auxFitness1 = closestFitness[k];
closestDistances[k] = currentDistance;
if (j < population.Count)
{
closestFitness[k] = this[j].Fitness;
}
else if (auxChromosome != _archive[j - population.Count])
{
closestFitness[k] = _archive[j - population.Count].Fitness;
}
}
}
//
// And shift the rest keeping the array closestDistances always ordered
//
else
{
auxDistance2 = closestDistances[k];
closestDistances[k] = auxDistance1;
auxDistance1 = auxDistance2;
if (j < population.Count || auxChromosome != _archive[j - population.Count])
{
auxFitness2 = closestFitness[k];
closestFitness[k] = auxFitness1;
auxFitness1 = auxFitness2;
}
}
}
}
}
auxChromosome.LocalCompetition = 0;
//
// Make a sum of all those _nearestNeighbours distances.
//
for (int k = 0; k < _kNearestNeighbours; k++)
{
distSum += closestDistances[k];
if (auxChromosome.Fitness > closestFitness[k])
{
auxChromosome.LocalCompetition++;
}
}
//
// Claculate the average
//
auxChromosome.Sparseness = distSum / KNearestNeighbours;
if (auxChromosome.Fitness < _minimalCriteria)
{
auxChromosome.LocalCompetition = 0;
}
});
for (int i = 0; i < population.Count; i++)
{
NoveltyChromosome currentChromosome = (NoveltyChromosome)this[i];
if (currentChromosome.Sparseness > _sparsenessThreshold)
{
_archive.Add(currentChromosome);
//
// Count as added to archive.
//
_addedToArchiveCount++;
if (_archive.Count > _archiveSizeLimit)
{
_archive.RemoveAt(0);
}
}
}
}