/// <summary>
/// Selects the specified number of <see cref="GeneticEntity"/> objects from <paramref name="population"/>.
/// </summary>
/// <param name="entityCount">Number of <see cref="GeneticEntity"/> objects to select from the population.</param>
/// <param name="population"><see cref="Population"/> containing the <see cref="GeneticEntity"/> objects from which to select.
/// objects from which to select.</param>
/// <returns>The <see cref="GeneticEntity"/> object that was selected.</returns>
protected override IEnumerable <GeneticEntity> SelectEntitiesFromPopulation(int entityCount, Population population)
{
if (population == null)
{
throw new ArgumentNullException(nameof(population));
}
this.AssertIsInitialized();
GeneticEntity[] sortedEntities = population.Entities.GetEntitiesSortedByFitness(
this.SelectionBasedOnFitnessType,
this.Algorithm !.FitnessEvaluator !.EvaluationMode).ToArray();
List <WheelSlice> wheelSlices = new List <WheelSlice>(sortedEntities.Length);
for (int i = 0; i < sortedEntities.Length; i++)
{
wheelSlices.Add(new WheelSlice(sortedEntities[i], i + 1));
}
List <GeneticEntity> result = new List <GeneticEntity>();
for (int i = 0; i < entityCount; i++)
{
result.Add(RouletteWheelSampler.GetEntity(wheelSlices));
}
return(result);
}
/// <summary>
/// Selects the specified number of <see cref="GeneticEntity"/> objects from <paramref name="population"/>.
/// </summary>
/// <param name="entityCount">Number of <see cref="GeneticEntity"/> objects to select from the population.</param>
/// <param name="population"><see cref="Population"/> containing the <see cref="GeneticEntity"/> objects from which to select.
/// objects from which to select.</param>
/// <returns>The <see cref="GeneticEntity"/> object that was selected.</returns>
protected override IEnumerable <GeneticEntity> SelectEntitiesFromPopulation(int entityCount, Population population)
{
if (population == null)
{
throw new ArgumentNullException(nameof(population));
}
double totalSubVals = 0;
foreach (GeneticEntity entity in population.Entities)
{
totalSubVals += Math.Pow(Math.E, entity.GetFitnessValue(this.SelectionBasedOnFitnessType) / this.CurrentTemperature);
if (Double.IsInfinity(totalSubVals))
{
throw new OverflowException(StringUtil.GetFormattedString(Resources.ErrorMsg_BoltzmannTotalOverflow, this.GetType().Name));
}
}
double meanSubVals = totalSubVals / population.Entities.Count;
List <WheelSlice> wheelSlices = new List <WheelSlice>();
foreach (GeneticEntity entity in population.Entities)
{
double expectedValue = Math.Pow(Math.E, entity.GetFitnessValue(this.SelectionBasedOnFitnessType) / this.CurrentTemperature) / meanSubVals;
wheelSlices.Add(new WheelSlice(entity, expectedValue));
}
List <GeneticEntity> result = new List <GeneticEntity>();
for (int i = 0; i < entityCount; i++)
{
result.Add(RouletteWheelSampler.GetEntity(wheelSlices));
}
return(result);
}
/// <summary>
/// Selects the specified number of <see cref="GeneticEntity"/> objects from <paramref name="population"/>.
/// </summary>
/// <param name="entityCount">Number of <see cref="GeneticEntity"/> objects to select from the population.</param>
/// <param name="population"><see cref="Population"/> containing the <see cref="GeneticEntity"/> objects from which to select.
/// objects from which to select.</param>
/// <returns>The <see cref="GeneticEntity"/> object that was selected.</returns>
protected override IEnumerable <GeneticEntity> SelectEntitiesFromPopulation(int entityCount, Population population)
{
if (population == null)
{
throw new ArgumentNullException(nameof(population));
}
this.AssertIsInitialized();
FitnessEvaluationMode evaluationMode = this.Algorithm !.FitnessEvaluator !.EvaluationMode;
List <TemporaryWheelSlice> tempSlices = new List <TemporaryWheelSlice>();
// If smaller fitness values are better, we need to inverse the wheel slice size distribution so that
// the entity with the smallest fitness value gets the largest wheel slice size. We do this by
// using the largest fitness value as the wheel slice size for the entity with the smallest fitness
// value.
if (evaluationMode == FitnessEvaluationMode.Minimize)
{
GeneticEntity[] entities = population.Entities.GetEntitiesSortedByFitness(this.SelectionBasedOnFitnessType, evaluationMode).ToArray();
int descendingIndex = entities.Length - 1;
for (int i = 0; i < entities.Length; i++)
{
tempSlices.Add(
new TemporaryWheelSlice(
entities[i],
entities[descendingIndex].GetFitnessValue(this.SelectionBasedOnFitnessType)
));
descendingIndex--;
}
}
else
{
// calculate percentage ranges
foreach (GeneticEntity entity in population.Entities)
{
tempSlices.Add(
new TemporaryWheelSlice(
entity,
entity.GetFitnessValue(this.SelectionBasedOnFitnessType)
));
}
}
double minSize = tempSlices.Min(slice => slice.Size);
// Ensure that all wheel slice sizes are above 0 by shifting all values to be above 0.
if (minSize <= 0)
{
double offset = Math.Abs(minSize);
foreach (TemporaryWheelSlice slice in tempSlices)
{
slice.Size += offset + 1;
}
}
List <WheelSlice> wheelSlices = new List <WheelSlice>(tempSlices.Select(slice => new WheelSlice(slice.Entity, slice.Size)));
List <GeneticEntity> result = new List <GeneticEntity>();
for (int i = 0; i < entityCount; i++)
{
result.Add(RouletteWheelSampler.GetEntity(wheelSlices));
}
return(result);
}
