public void TestLogLogLogisticFeaturedLL()
{
DataGen dg = new DataGen();
dg.GenTrivFeaturesData();
double shapeMax = 5.0;
double scaleMax = 900.0;
LogLogistic modelLogLogisticFeatured = new LogLogistic(dg.organicRecoveryDurations,
dg.inorganicRecoverydurations,
dg.fSamples, dg.fCensored);
modelLogLogisticFeatured.ShapeUpperBound = shapeMax;
modelLogLogisticFeatured.ScaleUpperBound = scaleMax;
double[,] warr = new double[2, 2] {
{ 1, 1 },
{ 1, 1 }
};
Matrix <double> w = Matrix <double> .Build.DenseOfArray(warr);
var loglik = modelLogLogisticFeatured.LogLikelihood(w, dg.fSamples, dg.fCensored);
System.Console.WriteLine("LogLikelihood is: " + loglik.ToString());
Assert.IsTrue(Math.Abs(loglik + 55.83229) < 1e-3);
}
public void TestLogLogisticFeaturedGrad()
{
DataGen dg = new DataGen();
dg.GenTrivFeaturesData();
double shapeMax = 5.0;
double scaleMax = 900.0;
LogLogistic modelLogLogisticFeatured = new LogLogistic(dg.organicRecoveryDurations,
dg.inorganicRecoverydurations,
dg.fSamples, dg.fCensored);
modelLogLogisticFeatured.ShapeUpperBound = shapeMax;
modelLogLogisticFeatured.ScaleUpperBound = scaleMax;
double[,] warr = new double[2, 2] {
{ 1, 1 },
{ 1, 1 }
};
Matrix <double> w = Matrix <double> .Build.DenseOfArray(warr);
var grd = modelLogLogisticFeatured.GradLL2(w, dg.fSamples, dg.fCensored);
var grd_numr = modelLogLogisticFeatured.NumericalGradLL(w);
System.Console.WriteLine("Gradient first component:" + grd[0, 0].ToString());
Assert.IsTrue(Math.Abs(grd[0, 0] - grd_numr[0, 0]) < 1e-3);
}
public void GenLogLogisticWFeatures()
{
LogLogistic ll = new LogLogistic(1.2, 300.0);
List <double> llSamples = ll.GenerateSample(2000);
var censorLvl = llSamples.Sum() / llSamples.Count();
var t = CensorList(llSamples, censorLvl);
int n_x = llSamples.Count(ti => ti > censorLvl);
var x = NPOnes(n_x, censorLvl);
List <double[]> fSamplesData = new List <double[]>();
List <double[]> fCensoredData = new List <double[]>();
for (int i = 0; i < t.Count(); i++)
{
fSamplesData.Add(new double[] { 1.0, 2.0, 3.0 });
}
for (int i = 0; i < x.Count(); i++)
{
fCensoredData.Add(new double[] { 1.0, 2.0, 3.0 });
}
ll = new LogLogistic(0.7, 80.0);
llSamples = ll.GenerateSample(2000);
censorLvl = llSamples.Sum() / llSamples.Count();
var t1 = CensorList(llSamples, censorLvl);
AppendToLst(t, t1);
n_x = llSamples.Count(ti => ti > censorLvl);
var x1 = NPOnes(n_x, censorLvl);
AppendToLst(x, x1);
for (int i = 0; i < t1.Count(); i++)
{
fSamplesData.Add(new double[] { 1.0, 4.0, 2.0 });
}
for (int i = 0; i < x1.Count(); i++)
{
fCensoredData.Add(new double[] { 1.0, 4.0, 2.0 });
}
this.organicRecoveryDurations = t;
this.inorganicRecoverydurations = x;
Matrix <double> fSamples = Matrix <double> .Build.DenseOfArray(CreateRectangularArray(fSamplesData));
Matrix <double> fCensored = Matrix <double> .Build.DenseOfArray(CreateRectangularArray(fCensoredData));
this.fSamples = fSamples;
this.fCensored = fCensored;
}
public void TestLogLogisticWFeatures()
{
DataGen dg = new DataGen();
dg.GenLogLogisticWFeatures();
double shapeMax = 5.0;
double scaleMax = 500.0;
double[] arr = new double[] { 1.0, 150.0 };
Vector <double> init = Vector <double> .Build.DenseOfArray(arr);
LogLogistic modelLogLogistic = new LogLogistic(dg.organicRecoveryDurations,
dg.inorganicRecoverydurations);
modelLogLogistic.GradientDescent(init);
Console.WriteLine("LL without features is " +
modelLogLogistic.LogLikelihood(modelLogLogistic.Kappa, modelLogLogistic.Lambda) +
" with Kappa " + modelLogLogistic.Kappa + " and Lambda " + modelLogLogistic.Lambda);
double[,] warr = new double[2, dg.fCensored.ColumnCount];
warr[0, 0] = Sigmoid.InverseSigmoid(modelLogLogistic.Kappa, shapeMax);
warr[1, 0] = Sigmoid.InverseSigmoid(modelLogLogistic.Lambda, scaleMax);
Matrix <double> w = Matrix <double> .Build.DenseOfArray(warr);
LogLogistic modelLogLogisticFeatured = new LogLogistic(dg.organicRecoveryDurations,
dg.inorganicRecoverydurations,
dg.fSamples, dg.fCensored);
modelLogLogisticFeatured.ShapeUpperBound = shapeMax;
modelLogLogisticFeatured.ScaleUpperBound = scaleMax;
Matrix <double> logLogisticParameters = modelLogLogisticFeatured.GradientDescent(w, 2001);
Vector <double> frstSample = Vector <double> .Build.DenseOfArray(
new double[] { 1.0, 2.0, 3.0 });
Vector <double> scndSample = Vector <double> .Build.DenseOfArray(
new double[] { 1.0, 4.0, 2.0 });
Vector <double> res = logLogisticParameters.Multiply(frstSample);
var alpha_shape = res[0];
var shape = Sigmoid.Transform(alpha_shape, shapeMax);
var alpha_scale = res[1];
var scale = Sigmoid.Transform(alpha_scale, scaleMax);
res = logLogisticParameters.Multiply(scndSample);
alpha_shape = res[0];
shape = Sigmoid.Transform(alpha_shape, shapeMax);
alpha_scale = res[1];
scale = Sigmoid.Transform(alpha_scale, scaleMax);
Assert.IsTrue(Math.Abs(scale - 80.0) < 2.0);
}
public void TestMixedLogLogisticFitting()
{
LogLogistic lg = new LogLogistic(1.2, 300.0);
List <double> llSamples = lg.GenerateSample(5000);
LogLogistic lg1 = new LogLogistic(0.7, 80.0);
List <double> llSamples1 = lg1.GenerateSample(5000);
DataGen.AppendToLst(llSamples, llSamples1);
var ti = llSamples;
List <double> xi = new List <double> {
.1
};
LogLogistic llModel = new LogLogistic(ti, xi);
double[] arr = new double[] { 0.8, 150.0 };
Vector <double> init = Vector <double> .Build.DenseOfArray(arr);
llModel.GradientDescent(init);
Debug.WriteLine("Model completed!");
Assert.IsTrue(Math.Abs(llModel.Kappa - 0.83) < 1e-1);
}
public void TestLogLogisticFitting()
{
LogLogistic lg = new LogLogistic(1.2, 300);
var lgSamples = lg.GenerateSample(3000);
var lgCensored = new List <double> {
0.1, 0.1
};
double[] arr = new double[] { 0.5, 12.0 };
Vector <double> init = Vector <double> .Build.DenseOfArray(arr);
LogLogistic lg2 = new LogLogistic(lgSamples, lgCensored);
lg2.GradientDescent(init);
Console.WriteLine(lg2.Kappa);
//The un-biased estimator of the LogLogistic distribution
//has a high variance. So, we have to set the bar low.
Assert.IsTrue(Math.Abs(lg2.Kappa - 1.2) < 1e-1);
}
