public BinaryPopulation(int populationSize, int genotypeSize, string selectionProtocol,
AbstractFitnessEvaluator evaluator, AbstractTranslator translator, int intervalStart,
int intervalEnd)
{
_random = new Random();
Evaluator = evaluator;
Translator = translator;
var individuals = new List<AbstractPhenotype>();
for (int i = 0; i < populationSize; i++)
{
var genotype = new BitVector();
for (int j = 0; j < genotypeSize; j++)
{
int number = _random.Next(intervalStart, intervalEnd);
string binary = Convert.ToString(number, 2);
string binaryDigits = Convert.ToString(intervalEnd - 1, 2);
while (binary.Length < binaryDigits.Length)
{
binary = "0" + binary;
}
for (int k = 0; k < binary.Length; k++)
{
genotype.Vector.Add(Convert.ToInt32(binary[k].ToString(), 2));
}
}
individuals.Add(Translator.Translate(genotype));
}
CurrentPopulation = individuals;
Offsprings = new List<BitVector>();
SelectionProtocol = selectionProtocol;
}
public override AbstractPhenotype Translate(BitVector genom)
{
//NB: Requires explicit size genom size of 33
double a = binaryToDoubleRange(genom.GetRange(0, 7), 0.001, 0.2); //200 values : 8 bit
double b = binaryToDoubleRange(genom.GetRange(7, 7), 0.01, 0.3); // 30 values : 5 bit
double c = binaryToDoubleRange(genom.GetRange(14, 7), -80.0, -30.0); // 51 values : 6 bit
double d = binaryToDoubleRange(genom.GetRange(21, 7), 0.1, 10.0); // 100 values : 7 bit
double k = binaryToDoubleRange(genom.GetRange(28, 7), 0.01, 1.0); // 100 values : 7 bit
return new IzhikevichPhenotype(a, b, c, d, k) {Genotype = genom, Fitness = 0.0, RouletteProportion = 0.0};
}
public override List<BitVector> Crossover(AbstractPhenotype parent1, AbstractPhenotype parent2,
double crossoverRate)
{
var par1 = (BitVector) parent1.Genotype;
var par2 = (BitVector) parent2.Genotype;
var offspring = new List<BitVector>();
int randomIndex = _random.Next(1, par1.Vector.Count);
double randomDouble = _random.NextDouble();
var child1 = new BitVector();
var child2 = new BitVector();
if (randomDouble <= crossoverRate)
{
for (int i = 0; i < par1.Vector.Count; i++)
{
//for (int j = i; j < i + 8; j++)
//{
// if (_random.NextDouble() > 0.5)
// {
// child1.Vector.Add(par1.Vector[i]);
// child2.Vector.Add(par2.Vector[i]);
// }
// else
// {
// child1.Vector.Add(par2.Vector[i]);
// child2.Vector.Add(par1.Vector[i]);
// }
//}
if (i <= randomIndex)
{
child1.Vector.Add(par1.Vector[i]);
child2.Vector.Add(par2.Vector[i]);
}
else
{
child1.Vector.Add(par2.Vector[i]);
child2.Vector.Add(par1.Vector[i]);
}
}
}
else
{
child1.Vector = par1.Vector;
child2.Vector = par2.Vector;
}
offspring.Add(child1);
offspring.Add(child2);
return offspring;
}
public override AbstractPhenotype Translate(BitVector genom)
{
var battles = new List<double>();
for (int i = 0; i < genom.Vector.Count; i = i + 4)
{
string binaryStrength = genom.Vector[i].ToString() + genom.Vector[i + 1].ToString() +
genom.Vector[i + 2].ToString() + genom.Vector[i + 3].ToString();
var armyStrength = (double) Convert.ToInt32(binaryStrength, 2);
battles.Add(armyStrength);
}
double sum = battles.Sum();
List<double> army = battles.Select(battleStrength => (battleStrength/sum)).ToList();
var cbp = new ColonelBlottoPhenotype {Genotype = genom, Army = army, Strength = 1.0, Wins = 0, Ties = 0};
cbp.CalculateEntropy();
return cbp;
}
public override void Mutate(double mutationRate, BitVector genotype)
{
double randomDouble = _random.NextDouble();
if (randomDouble <= mutationRate)
{
int randomIndex = _random.Next(0, genotype.Vector.Count);
genotype.Vector[randomIndex] = (genotype.Vector[randomIndex] == 0 ? 1 : 0);
}
//for (int i = 0; i + 8 < genotype.Vector.Count; i += 8)
//{
// if (_random.NextDouble() <= mutationRate)
// {
// int randomIndex = _random.Next(0, 8);
// genotype.Vector[randomIndex] = (genotype.Vector[randomIndex + i] == 0 ? 1 : 0);
// }
//}
}
public override AbstractPhenotype Translate(BitVector genom)
{
Network network = new Network();
network.createGraph();
int index = 0;
foreach (Node hiddenNode in network.HiddenNodes)
{
foreach (var neighbour in hiddenNode.UpstreamConnections)
{
if (neighbour.left == network.BiasNode)
neighbour.right = binaryToDoubleRange(genom.GetRange(index++*8, 8), -10.0, 0.0);
else
neighbour.right = binaryToDoubleRange(genom.GetRange(index++*8, 8), -5.0, 5.0);
}
hiddenNode.Gain = binaryToDoubleRange(genom.GetRange(index++*8, 8), 1.0, 5.0);
hiddenNode.TimeConstant = binaryToDoubleRange(genom.GetRange(index++*8, 8), 1.0, 2.0);
}
foreach (Node outputNode in network.OutputNodes)
{
foreach (var neighbour in outputNode.UpstreamConnections)
{
if (neighbour.left == network.BiasNode)
neighbour.right = binaryToDoubleRange(genom.GetRange(index++*8, 8), -10.0, 0.0);
else
neighbour.right = binaryToDoubleRange(genom.GetRange(index++*8, 8), -5.0, 5.0);
}
outputNode.Gain = binaryToDoubleRange(genom.GetRange(index++*8, 8), 1.0, 5.0);
outputNode.TimeConstant = binaryToDoubleRange(genom.GetRange(index++*8, 8), 1.0, 2.0);
}
MinCogPhenotype pheno = new MinCogPhenotype(network.InputNodes, network.HiddenNodes, network.OutputNodes, network.BiasNode)
{
Genotype = genom,
Fitness = 0.0,
RouletteProportion = 0.0
};
return pheno;
}
public abstract void Mutate(double rate, BitVector genotype);
public abstract AbstractPhenotype Translate(BitVector genom);
public OneMaxPhenotype(BitVector genom)
{
Genotype = genom;
}
public override AbstractPhenotype Translate(BitVector genom)
{
return new OneMaxPhenotype(genom);
}