public void CombineTest2()
{
Accord.Math.Random.Generator.Seed = 0;
double[,] S = Matrix.Random(5000, 2);
double[,] A =
{
{ 0.25, 0.25 },
{ -0.25, 0.75 },
};
double[,] X = Matrix.Multiply(S, A);
var ica = new IndependentComponentAnalysis(X);
ica.Compute(2);
double[,] result = ica.Result;
float[][] expected = ica.Combine(result).ToSingle().ToJagged(true);
float[][] actual = ica.Combine(result.ToSingle().ToJagged(true));
Assert.IsTrue(expected.IsEqual(actual, atol: 1e-4f));
}
public void CombineTest()
{
Accord.Math.Tools.SetupGenerator(0);
double[,] S = Matrix.Random(5000, 2);
double[,] A =
{
{ 0.25, 0.25 },
{ -0.25, 0.75 },
};
double[,] X = S.Multiply(A);
IndependentComponentAnalysis ica = new IndependentComponentAnalysis(X);
ica.Compute(2);
var result = ica.Result;
var expected = Accord.Statistics.Tools.ZScores(X);
var actual = Accord.Statistics.Tools.ZScores(ica.Combine(result));
Assert.IsTrue(expected.IsEqual(actual, 1e-4));
}
public void CombineTest2()
{
Accord.Math.Tools.SetupGenerator(0);
double[,] S = Matrix.Random(5000, 2);
double[,] A =
{
{ 0.25, 0.25 },
{ -0.25, 0.75 },
};
double[,] X = S.Multiply(A);
IndependentComponentAnalysis ica = new IndependentComponentAnalysis(X);
ica.Compute(2);
double[,] result = ica.Result;
float[][] expected = ica.Combine(result).ToSingle().ToArray(true);
float[][] actual = ica.Combine(result.ToSingle().ToArray(true));
Assert.IsTrue(expected.IsEqual(actual, 1e-4f));
}
public void ComputeTest()
{
Accord.Math.Tools.SetupGenerator(0);
// Let's create a random dataset containing
// 5000 samples of two dimensional samples.
//
double[,] source = Matrix.Random(5000, 2);
// Now, we will mix the samples the dimensions of the samples.
// A small amount of the second column will be applied to the
// first, and vice-versa.
//
double[,] mix =
{
{ 0.25, 0.25 },
{ -0.25, 0.75 },
};
// mix the source data
double[,] input = source.Multiply(mix);
// Now, we can use ICA to identify any linear mixing between the variables, such
// as the matrix multiplication we did above. After it has identified it, we will
// be able to revert the process, retrieving our original samples again
// Create a new Independent Component Analysis
var ica = new IndependentComponentAnalysis(input);
Assert.AreEqual(IndependentComponentAlgorithm.Parallel, ica.Algorithm);
// Compute it
ica.Compute();
// Now, we can retrieve the mixing and demixing matrices that were
// used to alter the data. Note that the analysis was able to detect
// this information automatically:
double[,] mixingMatrix = ica.MixingMatrix; // same as the 'mix' matrix
double[,] revertMatrix = ica.DemixingMatrix; // inverse of the 'mix' matrix
double[,] result = ica.Result;
// Verify mixing matrix
mixingMatrix = mixingMatrix.Divide(mixingMatrix.Sum().Sum());
Assert.IsTrue(mix.IsEqual(mixingMatrix, 0.008));
// Verify demixing matrix
double[,] expected =
{
{ 0.75, -0.25 },
{ 0.25, 0.25 },
};
revertMatrix = revertMatrix.Divide(revertMatrix.Sum().Sum());
Assert.IsTrue(expected.IsEqual(revertMatrix, 0.008));
}
public void ComputeTest()
{
Accord.Math.Tools.SetupGenerator(0);
// Let's create a random dataset containing
// 5000 samples of two dimensional samples.
//
double[,] source = Matrix.Random(5000, 2);
// Now, we will mix the samples the dimensions of the samples.
// A small amount of the second column will be applied to the
// first, and vice-versa.
//
double[,] mix =
{
{ 0.25, 0.25 },
{ -0.25, 0.75 },
};
// mix the source data
double[,] input = source.Multiply(mix);
// Now, we can use ICA to identify any linear mixing between the variables, such
// as the matrix multiplication we did above. After it has identified it, we will
// be able to revert the process, retrieving our original samples again
// Create a new Independent Component Analysis
var ica = new IndependentComponentAnalysis(input);
Assert.AreEqual(IndependentComponentAlgorithm.Parallel, ica.Algorithm);
// Compute it
ica.Compute();
// Now, we can retrieve the mixing and demixing matrices that were
// used to alter the data. Note that the analysis was able to detect
// this information automatically:
double[,] mixingMatrix = ica.MixingMatrix; // same as the 'mix' matrix
double[,] revertMatrix = ica.DemixingMatrix; // inverse of the 'mix' matrix
double[,] result = ica.Result;
// Verify mixing matrix
mixingMatrix = mixingMatrix.Divide(mixingMatrix.Sum().Sum());
Assert.IsTrue(mix.IsEqual(mixingMatrix, 0.008));
// Verify demixing matrix
double[,] expected =
{
{ 0.75, -0.25 },
{ 0.25, 0.25 },
};
revertMatrix = revertMatrix.Divide(revertMatrix.Sum().Sum());
Assert.IsTrue(expected.IsEqual(revertMatrix, 0.008));
}
public void SerializeTest()
{
Accord.Math.Tools.SetupGenerator(0);
double[,] source = Matrix.Random(5000, 2);
double[,] mix =
{
{ 0.25, 0.25 },
{ -0.25, 0.75 },
};
double[,] input = source.Dot(mix);
var ica = new IndependentComponentAnalysis(input);
ica.Compute();
MemoryStream stream = new MemoryStream();
{
BinaryFormatter b = new BinaryFormatter();
b.Serialize(stream, ica);
}
stream.Seek(0, SeekOrigin.Begin);
{
BinaryFormatter b = new BinaryFormatter();
ica = (IndependentComponentAnalysis)b.Deserialize(stream);
}
Assert.AreEqual(IndependentComponentAlgorithm.Parallel, ica.Algorithm);
double[][] mixingMatrix = ica.MixingMatrix; // same as the 'mix' matrix
double[][] revertMatrix = ica.DemixingMatrix; // inverse of the 'mix' matrix
double[,] result = ica.Result;
mixingMatrix = mixingMatrix.Divide(mixingMatrix.Sum());
Assert.IsTrue(mix.IsEqual(mixingMatrix, atol: 0.008));
double[,] expected =
{
{ 0.75, -0.25 },
{ 0.25, 0.25 },
};
revertMatrix = revertMatrix.Divide(revertMatrix.Sum());
Assert.IsTrue(expected.IsEqual(revertMatrix, atol: 0.008));
}
public void ComputeTest2()
{
Accord.Math.Tools.SetupGenerator(0);
double[,] S = Matrix.Random(5000, 2);
double[,] A =
{
{ 1, 1 },
{ -1, 3 },
};
A = A.Divide(Norm.Norm1(A));
double[,] X = S.Multiply(A);
IndependentComponentAnalysis ica = new IndependentComponentAnalysis(X, IndependentComponentAlgorithm.Deflation);
Assert.AreEqual(IndependentComponentAlgorithm.Deflation, ica.Algorithm);
ica.Compute(2);
var result = ica.Result;
var mixingMatrix = ica.MixingMatrix;
var revertMatrix = ica.DemixingMatrix;
// Verify mixing matrix
mixingMatrix = mixingMatrix.Divide(Norm.Norm1(mixingMatrix));
Assert.IsTrue(A.IsEqual(mixingMatrix, 0.05));
// Verify demixing matrix
double[,] expected =
{
{ 3, -1 },
{ 1, 1 },
};
expected = expected.Divide(Norm.Norm1(expected));
revertMatrix = revertMatrix.Divide(Norm.Norm1(revertMatrix));
Assert.IsTrue(expected.IsEqual(revertMatrix, 0.05));
var reverted = Accord.Statistics.Tools.ZScores(result).Abs();
var original = Accord.Statistics.Tools.ZScores(S).Abs();
Assert.IsTrue(reverted.IsEqual(original, 0.1));
}
public void ConvergenceTest()
{
IndependentComponentAnalysis ica;
// https://github.com/accord-net/framework/issues/225
var mixedData = LoadData();
ica = new IndependentComponentAnalysis(mixedData, AnalysisMethod.Standardize);
ica.Overwrite = false;
ica.Iterations = 1000;
ica.Algorithm = IndependentComponentAlgorithm.Parallel;
ica.Contrast = new Kurtosis();
ica.Compute();
Assert.AreEqual(3.2178976535060348, ica.WhiteningMatrix.Sum());
Assert.AreEqual(1, ica.MixingMatrix.Sum(), 1e-7);
}
public void ComputeTest()
{
Accord.Math.Tools.SetupGenerator(0);
double[,] S = Matrix.Random(5000, 2);
double[,] A =
{
{ 0.25, 0.25 },
{ -0.25, 0.75 },
};
double[,] X = S.Multiply(A);
IndependentComponentAnalysis ica = new IndependentComponentAnalysis(X, IndependentComponentAlgorithm.Parallel);
Assert.AreEqual(IndependentComponentAlgorithm.Parallel, ica.Algorithm);
ica.Compute(2);
var result = ica.Result;
var mixingMatrix = ica.MixingMatrix;
var revertMatrix = ica.DemixingMatrix;
// Verify mixing matrix
mixingMatrix = mixingMatrix.Divide(mixingMatrix.Sum().Sum());
Assert.IsTrue(A.IsEqual(mixingMatrix, 0.008));
// Verify demixing matrix
double[,] expected =
{
{ 0.75, -0.25 },
{ 0.25, 0.25 },
};
revertMatrix = revertMatrix.Divide(revertMatrix.Sum().Sum());
Assert.IsTrue(expected.IsEqual(revertMatrix, 0.008));
}
public void ComputeTest()
{
Accord.Math.Tools.SetupGenerator(0);
double[,] S = Matrix.Random(5000, 2);
double[,] A =
{
{ 0.25, 0.25 },
{ -0.25, 0.75 },
};
double[,] X = S.Multiply(A);
IndependentComponentAnalysis ica = new IndependentComponentAnalysis(X, IndependentComponentAlgorithm.Parallel);
Assert.AreEqual(IndependentComponentAlgorithm.Parallel, ica.Algorithm);
ica.Compute(2);
var result = ica.Result;
var mixingMatrix = ica.MixingMatrix;
var revertMatrix = ica.DemixingMatrix;
// Verify mixing matrix
mixingMatrix = mixingMatrix.Divide(mixingMatrix.Sum().Sum());
Assert.IsTrue(A.IsEqual(mixingMatrix, 0.008));
// Verify demixing matrix
double[,] expected =
{
{ 0.75, -0.25 },
{ 0.25, 0.25 },
};
revertMatrix = revertMatrix.Divide(revertMatrix.Sum().Sum());
Assert.IsTrue(expected.IsEqual(revertMatrix, 0.008));
}
private void btnRunAnalysis_Click(object sender, EventArgs e)
{
// Retrieve the input data as a double[,] matrix
double[][] data = input.Transpose().ToDouble();
// Create a new Independent Component Analysis
ica = new IndependentComponentAnalysis(AnalysisMethod.Standardize)
{
Overwrite = true
};
// Compute the analysis
var demixer = ica.Learn(data);
// Separate the input signals
demix = demixer.Transform(data).Transpose().ToSingle();
btnSource1.Enabled = true;
btnSource2.Enabled = true;
}
private void btnRunAnalysis_Click(object sender, EventArgs e)
{
// Retrieve the input data as a double[,] matrix
double[,] data = input.ToMatrix(true).ToDouble();
// Create a new Independent Component Analysis
ica = new IndependentComponentAnalysis(data,
AnalysisMethod.Standardize)
{
Overwrite = true
};
// Compute the analysis
ica.Compute();
// Separate the input signals
demix = ica.Separate(input);
btnSource1.Enabled = true;
btnSource2.Enabled = true;
}
public void SeparateTest()
{
Accord.Math.Tools.SetupGenerator(0);
double[,] S = Matrix.Random(5000, 2);
double[,] A =
{
{ 0.25, 0.25 },
{ -0.25, 0.75 },
};
double[,] X = Matrix.Multiply(S, A);
var ica = new IndependentComponentAnalysis(X);
ica.Compute(2);
var expected = ica.Result;
var actual = ica.Separate(X);
Assert.IsTrue(expected.IsEqual(actual, 1e-4));
}
public void SeparateTest2()
{
Accord.Math.Tools.SetupGenerator(0);
double[,] S = Matrix.Random(5000, 2);
double[,] A =
{
{ 0.25, 0.25 },
{ -0.25, 0.75 },
};
double[,] X = S.Dot(A);
var ica = new IndependentComponentAnalysis(X);
ica.Compute(2);
var expected = ica.Result.ToSingle().ToJagged(true);
var actual = ica.Separate(X.ToSingle().ToJagged(true));
Assert.IsTrue(expected.IsEqual(actual, 1e-4f));
}
public void learn_test()
{
Accord.Math.Random.Generator.Seed = 0;
#region doc_learn
// Let's create a random dataset containing
// 5000 samples of two dimensional samples.
//
double[][] source = Jagged.Random(5000, 2);
// Now, we will mix the samples the dimensions of the samples.
// A small amount of the second column will be applied to the
// first, and vice-versa.
//
double[][] mix =
{
new double[] { 0.25, 0.25 },
new double[] { -0.25, 0.75 },
};
// mix the source data
double[][] input = source.Multiply(mix);
// Now, we can use ICA to identify any linear mixing between the variables, such
// as the matrix multiplication we did above. After it has identified it, we will
// be able to revert the process, retrieving our original samples again
// Create a new Independent Component Analysis
var ica = new IndependentComponentAnalysis()
{
Algorithm = IndependentComponentAlgorithm.Parallel,
Contrast = new Logcosh()
};
// Learn the demixing transformation from the data
MultivariateLinearRegression demix = ica.Learn(input);
// Now, we can retrieve the mixing and demixing matrices that were
// used to alter the data. Note that the analysis was able to detect
// this information automatically:
double[][] mixingMatrix = ica.MixingMatrix; // same as the 'mix' matrix
double[][] revertMatrix = ica.DemixingMatrix; // inverse of the 'mix' matrix
// We can use the regression to recover the separate sources
double[][] result = demix.Transform(input);
#endregion
// Verify mixing matrix
mixingMatrix = mixingMatrix.Divide(mixingMatrix.Sum());
Assert.IsTrue(mix.IsEqual(mixingMatrix, atol: 0.008));
Assert.IsTrue(revertMatrix.IsEqual(demix.Weights, atol: 0.008));
var dm = demix.Inverse().Weights;
dm = dm.Divide(dm.Sum());
Assert.IsTrue(mixingMatrix.IsEqual(dm, atol: 0.008));
// Verify demixing matrix
double[,] expected =
{
{ 0.75, -0.25 },
{ 0.25, 0.25 },
};
Assert.AreEqual(IndependentComponentAlgorithm.Parallel, ica.Algorithm);
Assert.AreEqual(ica.Contrast.GetType(), typeof(Logcosh));
revertMatrix = revertMatrix.Divide(revertMatrix.Sum());
Assert.IsTrue(expected.IsEqual(revertMatrix, atol: 0.008));
}
public void ConvergenceTest()
{
IndependentComponentAnalysis ica;
// https://github.com/accord-net/framework/issues/225
var mixedData = LoadData();
ica = new IndependentComponentAnalysis(mixedData, AnalysisMethod.Standardize);
ica.Overwrite = false;
ica.Iterations = 1000;
ica.Algorithm = IndependentComponentAlgorithm.Parallel;
ica.Contrast = new Kurtosis();
ica.Compute();
Assert.AreEqual(3.2178976535060348, ica.WhiteningMatrix.Sum());
Assert.AreEqual(1, ica.MixingMatrix.Sum(), 1e-7);
}
// THE MAIN CARDIO COMPUTATION:
private double HeartComputation(Bitmap image)
{
// ComputeAndStorAverageRGB(image);
ComputeAndStorAverageRGB(image);
if (isDataReady)
{
////////////////////////////////
// NORMALIZE THE RGB SIGNALS BY SUBTRACTING THE MEAN:
//build color matrix
double[][] colors =
{
red,
green,
blue
};
///////////////////////////////////
// ADD A HIGH PASS FILTERING STEP:
var testFilter = new Accord.Audio.Filters.HighPassFilter(0.1f);
float RC = 1 / (2 * (float)(3.142) * (float)(0.1));
float dt = 1 / (float)(30);
float ALPHA = dt / (dt + RC);
testFilter.Alpha = ALPHA;
var floatMtx = new float[colors.Length][];
for (int i = 0; i < colors.Length; i++)
{
floatMtx[i] = new float[colors[i].Length];
for (int j = 0; j < colors[i].Length; j++)
{
floatMtx[i][j] = (float)colors[i][j];
}
}
Accord.Audio.Signal target = Accord.Audio.Signal.FromArray(floatMtx[0], 1, Accord.Audio.SampleFormat.Format32BitIeeeFloat);
Accord.Audio.Signal target_out = testFilter.Apply(target);
target_out.CopyTo(floatMtx[0]);
target = Accord.Audio.Signal.FromArray(floatMtx[1], 1, Accord.Audio.SampleFormat.Format32BitIeeeFloat);
target_out = testFilter.Apply(target);
target_out.CopyTo(floatMtx[1]);
target = Accord.Audio.Signal.FromArray(floatMtx[2], 1, Accord.Audio.SampleFormat.Format32BitIeeeFloat);
target_out = testFilter.Apply(target);
target_out.CopyTo(floatMtx[2]);
for (int i = 0; i < floatMtx.Length; i++)
{
for (int j = 0; j < floatMtx[i].Length; j++)
{
colors[i][j] = (double)floatMtx[i][j];
}
}
///////////////////////////////////
///////////////////////////////////
// Z-SCORE:
double[][] norm_colorsT;
double[][] norm_colors;
double[][] colorsT = colors.Transpose();
norm_colorsT = Accord.Statistics.Tools.ZScores(colorsT);
norm_colors = norm_colorsT.Transpose();
///////////////////////////////////
///////////////////////////////////
// PERFORM ICA:
var ica = new IndependentComponentAnalysis()
{
Algorithm = IndependentComponentAlgorithm.Parallel,
Contrast = new Logcosh(),
Iterations = 1000,
NumberOfInputs = 3,
NumberOfOutputs = 3
};
ica.Iterations = 1000000;
MultivariateLinearRegression demix = ica.Learn(norm_colorsT);
double[][] resultT = demix.Transform(norm_colorsT);
double[][] result = resultT.Transpose();
///////////////////////////////////
///////////////////////////////////
// ADD A Low PASS FILTERING STEP:
var LPFilter = new Accord.Audio.Filters.LowPassFilter(0.1f);
RC = 1 / (2 * (float)(3.142) * (float)(0.1));
dt = 1 / (float)(30);
ALPHA = dt / (dt + RC);
testFilter.Alpha = ALPHA;
var LPMtx = new float[result.Length][];
for (int i = 0; i < result.Length; i++)
{
LPMtx[i] = new float[result[i].Length];
for (int j = 0; j < result[i].Length; j++)
{
LPMtx[i][j] = (float)result[i][j];
}
}
target = Accord.Audio.Signal.FromArray(LPMtx[0], 30, Accord.Audio.SampleFormat.Format32BitIeeeFloat);
target_out = testFilter.Apply(target);
target_out.CopyTo(LPMtx[0]);
target = Accord.Audio.Signal.FromArray(LPMtx[1], 30, Accord.Audio.SampleFormat.Format32BitIeeeFloat);
target_out = testFilter.Apply(target);
target_out.CopyTo(LPMtx[1]);
target = Accord.Audio.Signal.FromArray(LPMtx[2], 30, Accord.Audio.SampleFormat.Format32BitIeeeFloat);
target_out = testFilter.Apply(target);
target_out.CopyTo(LPMtx[2]);
/*
* for (int i = 0; i < LPMtx.Length; i++)
* {
* for (int j = 0; j < LPMtx[i].Length; j++)
* result[i][j] = (double)LPMtx[i][j];
* }
*/
///////////////////////////////////
///////////////////////////////////
// CALCULATE THE FFT OF ECH CHANNEL:
int length = red.Length;
double[] imagR = new double[length];
double[] imagG = new double[length];
double[] imagB = new double[length];
red_norm = norm_colors[0];
green_norm = norm_colors[1];
blue_norm = norm_colors[2];
ica_source_1 = result[0];
ica_source_2 = result[1];
ica_source_3 = result[2];
Accord.Math.Transforms.FourierTransform2.FFT(result[0], imagR, FourierTransform.Direction.Forward);
Accord.Math.Transforms.FourierTransform2.FFT(result[1], imagG, FourierTransform.Direction.Forward);
Accord.Math.Transforms.FourierTransform2.FFT(result[2], imagB, FourierTransform.Direction.Forward);
double[] magR = GetMag(result[0], imagR);
double[] magG = GetMag(result[1], imagG);
double[] magB = GetMag(result[2], imagB);
for (int i = 0; i < timeWindows / 2; i++)
{
red_norm_FFT[i] = magR[i];
green_norm_FFT[i] = magG[i];
blue_norm_FFT[i] = magB[i];
}
///////////////////////////////////
///////////////////////////////////
// FIND THE MAX FREQUENCY AND POWER FROM THE FFTs:
double[] freq = Vector.Interval((double)0, (double)(timeWindows - 1));
for (int i = 0; i < freq.Length; i++)
{
freq[i] = freq[i] / (double)timeWindows * FrameRate;
}
int MaxRedIndex;
double MaxRedValue;
GetMax(out MaxRedIndex, out MaxRedValue, freq, magR, 0.45, 3);
int MaxGreenIndex;
double MaxGreenValue;
GetMax(out MaxGreenIndex, out MaxGreenValue, freq, magG, 0.45, 3);
int MaxBlueIndex;
double MaxBlueValue;
GetMax(out MaxBlueIndex, out MaxBlueValue, freq, magB, 0.45, 3);
//Get the max power
if ((MaxRedValue > MaxGreenValue) && (MaxRedValue > MaxBlueValue))
{
//Red is the Max
HeartBeatFrequency = freq[MaxRedIndex];
}
if ((MaxGreenValue > MaxRedValue) && (MaxGreenValue > MaxBlueValue))
{
//Green is the Max
HeartBeatFrequency = freq[MaxGreenIndex];
}
if ((MaxBlueValue > MaxRedValue) && (MaxBlueValue > MaxGreenValue))
{
//Blue is the Max
HeartBeatFrequency = freq[MaxBlueIndex];
}
///////////////////////////////////
HeartBeatFrequency = HeartBeatFrequency * 60;
isDataReady = false;
return(HeartBeatFrequency);
}
return(HeartBeatFrequency);
}
private void btnRunAnalysis_Click(object sender, EventArgs e)
{
// Retrieve the input data as a double[,] matrix
double[][] data = input.Transpose().ToDouble();
// Create a new Independent Component Analysis
ica = new IndependentComponentAnalysis(AnalysisMethod.Standardize)
{
Overwrite = true
};
// Compute the analysis
var demixer = ica.Learn(data);
// Separate the input signals
demix = demixer.Transform(data).Transpose().ToSingle();
btnSource1.Enabled = true;
btnSource2.Enabled = true;
}
public void SerializeTest()
{
Accord.Math.Tools.SetupGenerator(0);
double[,] source = Matrix.Random(5000, 2);
double[,] mix =
{
{ 0.25, 0.25 },
{ -0.25, 0.75 },
};
double[,] input = source.Multiply(mix);
var ica = new IndependentComponentAnalysis(input);
ica.Compute();
MemoryStream stream = new MemoryStream();
{
BinaryFormatter b = new BinaryFormatter();
b.Serialize(stream, ica);
}
stream.Seek(0, SeekOrigin.Begin);
{
BinaryFormatter b = new BinaryFormatter();
ica = (IndependentComponentAnalysis)b.Deserialize(stream);
}
Assert.AreEqual(IndependentComponentAlgorithm.Parallel, ica.Algorithm);
double[,] mixingMatrix = ica.MixingMatrix; // same as the 'mix' matrix
double[,] revertMatrix = ica.DemixingMatrix; // inverse of the 'mix' matrix
double[,] result = ica.Result;
mixingMatrix = mixingMatrix.Divide(mixingMatrix.Sum().Sum());
Assert.IsTrue(mix.IsEqual(mixingMatrix, 0.008));
double[,] expected =
{
{ 0.75, -0.25 },
{ 0.25, 0.25 },
};
revertMatrix = revertMatrix.Divide(revertMatrix.Sum().Sum());
Assert.IsTrue(expected.IsEqual(revertMatrix, 0.008));
}
public void CombineTest()
{
Accord.Math.Tools.SetupGenerator(0);
double[,] S = Matrix.Random(5000, 2);
double[,] A =
{
{ 0.25, 0.25 },
{ -0.25, 0.75 },
};
double[,] X = S.Multiply(A);
IndependentComponentAnalysis ica = new IndependentComponentAnalysis(X);
ica.Compute(2);
var result = ica.Result;
var expected = Accord.Statistics.Tools.ZScores(X);
var actual = Accord.Statistics.Tools.ZScores(ica.Combine(result));
Assert.IsTrue(expected.IsEqual(actual, 1e-4));
}
public void SeparateTest2()
{
Accord.Math.Tools.SetupGenerator(0);
double[,] S = Matrix.Random(5000, 2);
double[,] A =
{
{ 0.25, 0.25 },
{ -0.25, 0.75 },
};
double[,] X = S.Multiply(A);
IndependentComponentAnalysis ica = new IndependentComponentAnalysis(X);
ica.Compute(2);
var expected = ica.Result.ToSingle().ToArray(true);
var actual = ica.Separate(X.ToSingle().ToArray(true));
Assert.IsTrue(expected.IsEqual(actual, 1e-4f));
}
public void learn_test()
{
Accord.Math.Random.Generator.Seed = 0;
#region doc_learn
// Let's create a random dataset containing
// 5000 samples of two dimensional samples.
//
double[][] source = Jagged.Random(5000, 2);
// Now, we will mix the samples the dimensions of the samples.
// A small amount of the second column will be applied to the
// first, and vice-versa.
//
double[][] mix =
{
new double[] { 0.25, 0.25 },
new double[] { -0.25, 0.75 },
};
// mix the source data
double[][] input = source.Dot(mix);
// Now, we can use ICA to identify any linear mixing between the variables, such
// as the matrix multiplication we did above. After it has identified it, we will
// be able to revert the process, retrieving our original samples again
// Create a new Independent Component Analysis
var ica = new IndependentComponentAnalysis()
{
Algorithm = IndependentComponentAlgorithm.Parallel,
Contrast = new Logcosh()
};
// Learn the demixing transformation from the data
MultivariateLinearRegression demix = ica.Learn(input);
// Now, we can retrieve the mixing and demixing matrices that were
// used to alter the data. Note that the analysis was able to detect
// this information automatically:
double[][] mixingMatrix = ica.MixingMatrix; // same as the 'mix' matrix
double[][] revertMatrix = ica.DemixingMatrix; // inverse of the 'mix' matrix
// We can use the regression to recover the separate sources
double[][] result = demix.Transform(input);
#endregion
// Verify mixing matrix
mixingMatrix = mixingMatrix.Divide(mixingMatrix.Sum());
Assert.IsTrue(mix.IsEqual(mixingMatrix, atol: 0.008));
Assert.IsTrue(revertMatrix.IsEqual(demix.Weights, atol: 0.008));
var dm = demix.Inverse().Weights;
dm = dm.Divide(dm.Sum());
Assert.IsTrue(mixingMatrix.IsEqual(dm, atol: 0.008));
// Verify demixing matrix
double[,] expected =
{
{ 0.75, -0.25 },
{ 0.25, 0.25 },
};
Assert.AreEqual(IndependentComponentAlgorithm.Parallel, ica.Algorithm);
Assert.AreEqual(ica.Contrast.GetType(), typeof(Logcosh));
revertMatrix = revertMatrix.Divide(revertMatrix.Sum());
Assert.IsTrue(expected.IsEqual(revertMatrix, atol: 0.008));
}
private void btnRunAnalysis_Click(object sender, EventArgs e)
{
// Retrieve the input data as a double[,] matrix
double[,] data = input.ToMatrix(true).ToDouble();
// Create a new Independent Component Analysis
ica = new IndependentComponentAnalysis(data,
AnalysisMethod.Standardize) { Overwrite = true };
// Compute the analysis
ica.Compute();
// Separate the input signals
demix = ica.Separate(input);
btnSource1.Enabled = true;
btnSource2.Enabled = true;
}
public void CombineTest2()
{
Accord.Math.Random.Generator.Seed = 0;
double[,] S = Matrix.Random(5000, 2);
double[,] A =
{
{ 0.25, 0.25 },
{ -0.25, 0.75 },
};
double[,] X = Matrix.Multiply(S, A);
var ica = new IndependentComponentAnalysis(X);
ica.Compute(2);
double[,] result = ica.Result;
float[][] expected = ica.Combine(result).ToSingle().ToJagged(true);
float[][] actual = ica.Combine(result.ToSingle().ToJagged(true));
Assert.IsTrue(expected.IsEqual(actual, atol: 1e-4f));
}
public void ComputeTest2()
{
Accord.Math.Tools.SetupGenerator(0);
double[,] S = Matrix.Random(5000, 2);
double[,] A =
{
{ 1, 1 },
{ -1, 3 },
};
A = A.Divide(Norm.Norm1(A));
double[,] X = S.Multiply(A);
IndependentComponentAnalysis ica = new IndependentComponentAnalysis(X, IndependentComponentAlgorithm.Deflation);
Assert.AreEqual(IndependentComponentAlgorithm.Deflation, ica.Algorithm);
ica.Compute(2);
var result = ica.Result;
var mixingMatrix = ica.MixingMatrix;
var revertMatrix = ica.DemixingMatrix;
// Verify mixing matrix
mixingMatrix = mixingMatrix.Divide(Norm.Norm1(mixingMatrix));
Assert.IsTrue(A.IsEqual(mixingMatrix, 0.05));
// Verify demixing matrix
double[,] expected =
{
{ 3, -1 },
{ 1, 1 },
};
expected = expected.Divide(Norm.Norm1(expected));
revertMatrix = revertMatrix.Divide(Norm.Norm1(revertMatrix));
Assert.IsTrue(expected.IsEqual(revertMatrix, 0.05));
var reverted = Accord.Statistics.Tools.ZScores(result).Abs();
var original = Accord.Statistics.Tools.ZScores(S).Abs();
Assert.IsTrue(reverted.IsEqual(original, 0.1));
}
