public static IEnumerable<Z2<float>> TestData()
{
var g = new MathNet.Numerics.Distributions.Normal(0.0, 1.0);
var randy = new MathNet.Numerics.Random.MersenneTwister();
g.RandomSource = randy;
var dblsX = new double[100000];
var dblsY = new double[100000];
g.Samples(dblsX);
g.Samples(dblsY);
return dblsX.Select((d,i) =>
new Z2<float>((float)d, (float)dblsY[i]));
}
public double TakeSamples()
{
var dateTimeElapsed = 0.0;
var dateTime = DateTime.Now;
if (this.DistributionName == "Binomial")
{
fullName = $"{DistributionName}-Samples:{SamplesNumber}-Trials:{TrialsNumber}";
var binomaial = new MathNet.Numerics.Distributions.Binomial(0.5, this.TrialsNumber);
var generatedsamples = binomaial.Samples().Take(SamplesNumber).ToArray();
}
else if (this.DistributionName == "Geometric")
{
fullName = $"{DistributionName}-Samples:{SamplesNumber}";
var geometric = new MathNet.Numerics.Distributions.Geometric(0.5);
var generatedsamples = geometric.Samples().Take(SamplesNumber).ToArray();
}
else if (this.DistributionName == "Poisson")
{
fullName = $"{DistributionName}-Samples:{SamplesNumber}";
var poisson = new MathNet.Numerics.Distributions.Poisson(0.5);
var generatedsamples = poisson.Samples().Take(SamplesNumber).ToArray();
}
else if (this.DistributionName == "Normal")
{
fullName = $"{DistributionName}-Samples:{SamplesNumber}";
var normal = new MathNet.Numerics.Distributions.Normal(0.5, 2);
var generatedsamples = normal.Samples().Take(SamplesNumber).ToArray();
}
dateTimeElapsed = (DateTime.Now - dateTime).TotalMilliseconds;
return(dateTimeElapsed);
}
/// <summary>
/// Gets a new Decision Vector, based on the PCX logic.
/// </summary>
/// <param name="parents">A list of parent <see cref="DecisionVector"/>s.</param>
/// <returns>A new <see cref="DecisionVector"/>.</returns>
/// <exception cref="ArgumentOutOfRangeException">
/// Thrown if:
/// - there are less than two parents; or
/// - the parents have different length or zero length decision vectors; or
/// - any of the parents have non-continuous Decision Vector elements.
/// </exception>
public DecisionVector Operate(params DecisionVector[] parents)
{
if (parents.Length < 2)
{
throw new ArgumentOutOfRangeException(nameof(parents),
"There must be at least two parents.");
}
// TODO: These calls to .Any() are slow - can we remove the error checking?
if (parents.Any(p => p.GetContinuousElements().Count == 0))
{
throw new ArgumentOutOfRangeException(nameof(parents),
"Parents must have non-zero length decision vectors.");
}
if (parents.Any(p => p.GetContinuousElements().Count != parents.First().Count))
{
throw new ArgumentOutOfRangeException(nameof(parents),
"Parents must have the same length and fully continuous decision vectors.");
}
// 1: Pre-process
var parentDVs = Matrix <double> .Build.DenseOfColumns(parents.Select(dv => dv.Select(d => (double)d)));
var motherDV = Vector <double> .Build.DenseOfArray(parents.ElementAt(0).Select(d => (double)d).ToArray());
// 1a: centroid of all parents
var centroid = parentDVs.RowSums().Divide(parents.Count());
// 1b: vector distance from centroid to mother (following Deb's C code, not paper)
var motherCentroidVectorDistance = centroid - motherDV;
var motherCentroidAbsoluteDistance = motherCentroidVectorDistance.L2Norm();
if (motherCentroidAbsoluteDistance < 1e-20)
{
return(DecisionVector.CreateForEmpty());
}
// 1c: vector distance from other parents to mother
var otherParentDVs = parentDVs.RemoveColumn(0);
var parentMotherVectorDistances = otherParentDVs.EnumerateColumns()
.Select(v => v - motherDV).ToArray();
var parentMotherAbsoluteDistances = parentMotherVectorDistances.Select(v => v.L2Norm()).ToArray();
if (parentMotherAbsoluteDistances.Any(d => d < 1e-20))
{
return(DecisionVector.CreateForEmpty());
}
// 1d: perpendicular distances from other parents to centroid-mother vector
var orthogonalDistances = parentMotherVectorDistances
.Select((v, i) => parentMotherAbsoluteDistances.ElementAt(i) *
Math.Sqrt(1.0 - Math.Pow(
v.DotProduct(motherCentroidVectorDistance) /
(parentMotherAbsoluteDistances.ElementAt(i) * motherCentroidAbsoluteDistance),
2.0)));
var meanOrthogonalDistance = orthogonalDistances.Mean();
// 2: Now create a new individual
var normRnd = new MathNet.Numerics.Distributions.Normal(rngManager.Rng);
var samplesEta = new double[motherDV.Count];
normRnd.Samples(samplesEta);
var newRandomDv = Vector <double> .Build.DenseOfArray(samplesEta)
.Multiply(sigmaEta * meanOrthogonalDistance);
//Remove component of randomness in direction of ?
var offset1 = motherCentroidVectorDistance
.Multiply(newRandomDv.DotProduct(motherCentroidVectorDistance))
.Divide(Math.Pow(motherCentroidAbsoluteDistance, 2.0));
newRandomDv -= offset1;
var offset2 = motherCentroidVectorDistance
.Multiply(sigmaZeta * normRnd.Sample());
newRandomDv += offset2;
// Modification of Deb2002 which should maintain stability.
var finalDv = motherDV +
newRandomDv.Divide(Math.Sqrt(motherDV.Count));
return(DecisionVector.CreateFromArray(parents.First().GetDecisionSpace(), finalDv.ToArray()));
}
public double TakeSamples()
{
var dateTimeElapsed = 0.0;
var dateTime = DateTime.Now;
//IEnumerable<int> generatedSamplesEnumerable = Enumerable.Empty<int>();
//IEnumerable<double> generatedSamplesDoubleEnumerable = Enumerable.Empty<double>();
if (this.DistributionName == "Binomial")
{
fullName = $"{DistributionName}-Samples:{SamplesNumber}-Trials:{TrialsNumber}";
var binomaial = new MathNet.Numerics.Distributions.Binomial(0.5, this.TrialsNumber);
var generatedsamples = binomaial.Samples().Take(SamplesNumber).ToArray();
//generatedSamplesEnumerable = binomaial.Samples().Take(SamplesNumber);
//foreach (var item in generatedSamplesEnumerable)
//{
// var test = item;
//}
}
else if (this.DistributionName == "Geometric")
{
fullName = $"{DistributionName}-Samples:{SamplesNumber}";
var geometric = new MathNet.Numerics.Distributions.Geometric(0.5);
var generatedsamples = geometric.Samples().Take(SamplesNumber).ToArray();
//generatedSamplesEnumerable = geometric.Samples().Take(SamplesNumber);
//foreach (var item in generatedSamplesEnumerable)
//{
// var test = item;
//}
}
else if (this.DistributionName == "Poisson")
{
fullName = $"{DistributionName}-Samples:{SamplesNumber}";
var poisson = new MathNet.Numerics.Distributions.Poisson(0.5);
var generatedsamples = poisson.Samples().Take(SamplesNumber).ToArray();
//generatedSamplesEnumerable = poisson.Samples().Take(SamplesNumber);
//foreach (var item in generatedSamplesEnumerable)
//{
// var test = item;
//}
}
else if (this.DistributionName == "Normal")
{
fullName = $"{DistributionName}-Samples:{SamplesNumber}";
var normal = new MathNet.Numerics.Distributions.Normal(0.5, 2);
var generatedsamples = normal.Samples().Take(SamplesNumber).ToArray();
//generatedSamplesDoubleEnumerable = normal.Samples().Take(SamplesNumber);
//foreach(var item in generatedSamplesDoubleEnumerable)
//{
// var test = item;
//}
}
dateTimeElapsed = (DateTime.Now - dateTime).TotalMilliseconds;
return(dateTimeElapsed);
}
/// <summary>
/// Se estiman las propiedades estáticas en función de la profundidad del core
/// </summary>
public void Estimar()
{
// se toma la informacion de segmentacion vs areas de interes y se estiman las propiedades petrofisicas estaticas
// se preparan los generadores de CT aleatorios para cada phantom
var phantom1High = new MathNet.Numerics.Distributions.Normal(padre.actual.phantom1.mediaHigh, padre.actual.phantom1.desvHigh);
var phantom1Low = new MathNet.Numerics.Distributions.Normal(padre.actual.phantom1.mediaLow, padre.actual.phantom1.desvLow);
var phantom2High = new MathNet.Numerics.Distributions.Normal(padre.actual.phantom2.mediaHigh, padre.actual.phantom2.desvHigh);
var phantom2Low = new MathNet.Numerics.Distributions.Normal(padre.actual.phantom2.mediaLow, padre.actual.phantom2.desvLow);
var phantom3High = new MathNet.Numerics.Distributions.Normal(padre.actual.phantom3.mediaHigh, padre.actual.phantom3.desvHigh);
var phantom3Low = new MathNet.Numerics.Distributions.Normal(padre.actual.phantom3.mediaLow, padre.actual.phantom3.desvLow);
// se prepara un vector de valores CT high y low para cada phantom
// este vector representa un slide y solo se guarda un valor promedio del slide
double[] temp = new double[padre.actual.datacuboHigh.dataCube[0].segCore.Count];
MathNet.Numerics.Statistics.DescriptiveStatistics stats;
// se preparan los vectores para densidad y zeff
this.Dfm = new double[padre.actual.datacuboHigh.dataCube.Count];
this.Zfme = new double[padre.actual.datacuboHigh.dataCube.Count];
this.Pefm = new double[padre.actual.datacuboHigh.dataCube.Count];
double ctP1High, ctP2High, ctP3High, ctP1Low, ctP2Low, ctP3Low;
double A, B, C, D, E, F;
List<double> Df, Zf, Zeff, Pef;
int iarea;
// se empieza a recorrer cada slide que se encuentre dentro de las areas de interes
// para cada slide se genera una colección de datos aleatorios simulando cada phantom, y se estima la media para cada phantom en cada slide
bool slide = false;
for (int i = 0; i < padre.actual.datacuboHigh.dataCube.Count; i++)
{
slide = false;
iarea = -1;
for (int j = 0; j < padre.actual.areasCore.Count; j++)
{
// se busca si este slide esta dentro de al menos un area de interes
if ((i >= padre.actual.areasCore[j].ini) & (i <= padre.actual.areasCore[j].fin))
{
slide = true;
iarea = j;
}
}
if (slide)
{
// el slide pertenece al menos a un area de interes, por tanto se procede a calcular la densidad y zeff para este slide
// se generan los valores CT para cada phantom y se toma su media
phantom1High.Samples(temp);
stats = new MathNet.Numerics.Statistics.DescriptiveStatistics(temp);
ctP1High = stats.Mean;
phantom2High.Samples(temp);
stats = new MathNet.Numerics.Statistics.DescriptiveStatistics(temp);
ctP2High = stats.Mean;
phantom3High.Samples(temp);
stats = new MathNet.Numerics.Statistics.DescriptiveStatistics(temp);
ctP3High = stats.Mean;
phantom1Low.Samples(temp);
stats = new MathNet.Numerics.Statistics.DescriptiveStatistics(temp);
ctP1Low = stats.Mean;
phantom2Low.Samples(temp);
stats = new MathNet.Numerics.Statistics.DescriptiveStatistics(temp);
ctP2Low = stats.Mean;
phantom3Low.Samples(temp);
stats = new MathNet.Numerics.Statistics.DescriptiveStatistics(temp);
ctP3Low = stats.Mean;
// se resuelve el sistema lineal para obtener las constantes A,B,C,D,E,F
var matriz = MathNet.Numerics.LinearAlgebra.Matrix<double>.Build.DenseOfArray(new double[,] { { ctP1Low, ctP1High, 1 }, { ctP2Low, ctP2High, 1 }, { ctP3Low, ctP3High, 1 } });
var sol = MathNet.Numerics.LinearAlgebra.Vector<double>.Build.Dense(new double[] { padre.actual.phantom1.densidad, padre.actual.phantom2.densidad, padre.actual.phantom3.densidad });
var x = matriz.Solve(sol);
A = x[0];
B = x[1];
C = x[2];
sol = MathNet.Numerics.LinearAlgebra.Vector<double>.Build.Dense(new double[] { padre.actual.phantom1.zeff, padre.actual.phantom2.zeff, padre.actual.phantom3.zeff });
x = matriz.Solve(sol);
D = x[0];
E = x[1];
F = x[2];
// se empieza a recorrer cada voxel, en la segmentacion del actual i-slide, se revisa que este dentro del area de interes
Df = new List<double>();
Zf = new List<double>();
Zeff = new List<double>();
Pef = new List<double>();
int jkindex = 0;
double dx;
double dy;
double tDf, tZf, tZeff, tPef;
// dado que se recorre fila a fila, entonces el indice j corresponde al eje Y y el indice k al eje X
for (int j = 0; j < padre.actual.datacuboHigh.widthSeg; j++)
{
for (int k = 0; k < padre.actual.datacuboHigh.widthSeg; k++)
{
// se calcula la distancia de la posicion (j,k) al centro del area de interes
// si la distancia es menor que el radio entonces se agrega al calculo, sino no
dx = k - padre.actual.areasCore[iarea].x;
dx = dx * dx;
dy = j - padre.actual.areasCore[iarea].y;
dy = dy * dy;
if (Math.Sqrt(dx + dy) <= padre.actual.datacuboHigh.widthSeg)
{
// la coordenada (j,k) esta dentro del area de interes
// se calculan las propiedades estaticas
tDf = A * padre.actual.datacuboLow.dataCube[i].segCore[jkindex] + B * padre.actual.datacuboHigh.dataCube[i].segCore[jkindex] + C;
Df.Add(tDf);
tZf = D * padre.actual.datacuboLow.dataCube[i].segCore[jkindex] + E * padre.actual.datacuboHigh.dataCube[i].segCore[jkindex] + F;
Zf.Add(tZf);
//tZeff = Math.Pow(Math.Pow((tZf / (0.9342 * tDf + 0.1759)), 10), 1 / 36);
tZeff = Math.Pow((tZf / (0.9342 * tDf + 0.1759)), 1/3.6);
Zeff.Add(tZeff);
tPef = Math.Pow(Math.Pow((tZeff / 10), 36), 0.1);
Pef.Add(tPef);
}
jkindex++;
}
}
stats = new MathNet.Numerics.Statistics.DescriptiveStatistics(Df);
Dfm[i] = stats.Mean;
stats = new MathNet.Numerics.Statistics.DescriptiveStatistics(Zeff);
Zfme[i] = stats.Mean;
stats = new MathNet.Numerics.Statistics.DescriptiveStatistics(Pef);
Pefm[i] = stats.Mean;
}
else
{
// se llenan los vectores de densidad y zeff con valores -1... si el valor es -1 entonces no se grafica
this.Dfm[i] = -1;
this.Zfme[i] = -1;
this.Pefm[i] = -1;
}
}
DateTime fin = DateTime.Now;
}
