/// <summary>
/// Performs training iteration.
/// </summary>
public void Iteration()
{
//Next epoch
++_epoch;
//Noise intensity
double intensity = _alphas[Math.Min(_maxEpoch, _epoch) - 1];
//Adjusted predictors
Matrix predictors = PreparePredictors((double)intensity);
//Decomposition
QRD decomposition = new QRD(predictors);
//New waights
double[] newWaights = new double[_net.NumOfWeights];
//Regression for each output neuron
for (int outputIdx = 0; outputIdx < _net.NumOfOutputValues; outputIdx++)
{
//Regression
Matrix solution = decomposition.Solve(_outputSingleColMatrixCollection[outputIdx]);
//Store weights
//Input weights
for (int i = 0; i < solution.NumOfRows - 1; i++)
{
newWaights[outputIdx * _net.NumOfInputValues + i] = solution.Data[i][0];
}
//Bias weight
newWaights[_net.NumOfOutputValues * _net.NumOfInputValues + outputIdx] = solution.Data[solution.NumOfRows - 1][0];
}
//Set new weights and compute error
_net.SetWeights(newWaights);
_mse = _net.ComputeBatchErrorStat(_inputVectorCollection, _outputVectorCollection).MeanSquare;
return;
}
/// <summary>
/// Performs training iteration.
/// </summary>
public void Iteration()
{
//Next epoch
++_epoch;
//Noise intensity
double intensity = _alphas[Math.Min(_maxEpoch, _epoch) - 1];
//Adjusted predictors
Matrix predictors = PreparePredictors((double)intensity);
//Decomposition
QRD decomposition = null;
bool useableQRD = true;
try
{
//Try to create QRD. Any exception signals numerical unstability
decomposition = new QRD(predictors);
}
catch
{
//Creation of QRD object throws exception. QRD object is not ready for use.
useableQRD = false;
if (_epoch == 1)
{
//No previous successful epoch so stop training
throw;
}
}
if (useableQRD)
{
//QRD is ready for use (low probability of numerical unstability)
//New weights
double[] newWeights = new double[_net.NumOfWeights];
//Regression for each output neuron
for (int outputIdx = 0; outputIdx < _net.NumOfOutputValues; outputIdx++)
{
//Regression
Matrix solution = decomposition.Solve(_outputSingleColMatrixCollection[outputIdx]);
//Store weights
//Input weights
for (int i = 0; i < solution.NumOfRows - 1; i++)
{
newWeights[outputIdx * _net.NumOfInputValues + i] = solution.Data[i][0];
}
//Bias weight
newWeights[_net.NumOfOutputValues * _net.NumOfInputValues + outputIdx] = solution.Data[solution.NumOfRows - 1][0];
}
//Set new weights and compute error
_net.SetWeights(newWeights);
_mse = _net.ComputeBatchErrorStat(_inputVectorCollection, _outputVectorCollection).MeanSquare;
}
return;
}
/// <summary>
/// Performs training iteration.
/// </summary>
public bool Iteration()
{
//Fetch next noise intensity
double newMaxNoise = _maxNoiseSeeker.Next;
//Check continue conditions
if (AttemptEpoch == MaxAttemptEpoch || Math.Abs(_currMaxNoise - newMaxNoise) < StopNoiseDifference)
{
//Try new attempt
if (!NextAttempt())
{
//Next attempt is not available
return(false);
}
else
{
newMaxNoise = _maxNoiseSeeker.Next;
}
}
//Next epoch
++AttemptEpoch;
_currMaxNoise = newMaxNoise;
InfoMessage = $"maxNoise={_currMaxNoise.ToString(CultureInfo.InvariantCulture)}";
//Adjusted predictors
Matrix predictors = PreparePredictors(_currMaxNoise);
//Decomposition
QRD decomposition = null;
bool useableQRD = true;
try
{
//Try to create QRD. Any exception signals numerical unstability
decomposition = new QRD(predictors);
}
catch
{
//Creation of QRD object throws exception. QRD object is not ready for use.
useableQRD = false;
if (AttemptEpoch == 1)
{
//No previous successful epoch so stop training
throw;
}
}
if (useableQRD)
{
//QRD is ready for use (low probability of numerical unstability)
//New weights
double[] newWeights = new double[_net.NumOfWeights];
//Regression for each output neuron
for (int outputIdx = 0; outputIdx < _net.NumOfOutputValues; outputIdx++)
{
//Regression
Matrix solution = decomposition.Solve(_outputSingleColMatrixCollection[outputIdx]);
//Store weights
//Input weights
for (int i = 0; i < solution.NumOfRows - 1; i++)
{
newWeights[outputIdx * _net.NumOfInputValues + i] = solution.Data[i][0];
}
//Bias weight
newWeights[_net.NumOfOutputValues * _net.NumOfInputValues + outputIdx] = solution.Data[solution.NumOfRows - 1][0];
}
//Set new weights and compute error
_net.SetWeights(newWeights);
MSE = _net.ComputeBatchErrorStat(_inputVectorCollection, _outputVectorCollection).MeanSquare;
}
_maxNoiseSeeker.ProcessError(MSE);
return(true);
}
//Methods
public void Run()
{
//Linear algebra test
double[] flatData =
{
0.2, 5, 17.3, 1.01, 54, 7,
2.2, 5.5, 12.13, 11.57, 5.71, -85,
-70.1, 15, -18.3, 0.3, 42, -6.25,
0.042, 1, 7.75, -81.01, -21.29, 5.44,
0.1, 4, -4.3, 18.01, 7.12, -3.14,
-80.1, 24.4, 4.3, 12.03, 2.789, -13
};
Matrix testM = new Matrix(6, 6, flatData);
/*
* //Inversion test
* Matrix resultM = new Matrix(testM);
* resultM.SingleThreadInverse();
*/
/*
* //Transpose test
* Matrix resultM = testM.Transpose();
*/
/*
* //Multiply test
* Matrix resultM = Matrix.Multiply(testM, testM);
* for (int i = 0; i < resultM.NumOfRows; i++)
* {
* Console.WriteLine($"{resultM.Data[i][0]}; {resultM.Data[i][1]}; {resultM.Data[i][2]}; {resultM.Data[i][3]}; {resultM.Data[i][4]}; {resultM.Data[i][5]}");
* }
*/
;
int numOfweights = 4;
int xIdx, dIdx = 0;
double[][] data = new double[5][];
data[dIdx] = new double[numOfweights];
xIdx = -1;
data[dIdx][++xIdx] = 1; //Bias
data[dIdx][++xIdx] = 2;
data[dIdx][++xIdx] = 1;
data[dIdx][++xIdx] = 3;
++dIdx;
data[dIdx] = new double[numOfweights];
xIdx = -1;
data[dIdx][++xIdx] = 1; //Bias
data[dIdx][++xIdx] = 1;
data[dIdx][++xIdx] = 3;
data[dIdx][++xIdx] = -3;
++dIdx;
data[dIdx] = new double[numOfweights];
xIdx = -1;
data[dIdx][++xIdx] = 1; //Bias
data[dIdx][++xIdx] = -2;
data[dIdx][++xIdx] = 4;
data[dIdx][++xIdx] = 4;
++dIdx;
data[dIdx] = new double[numOfweights];
xIdx = -1;
data[dIdx][++xIdx] = 1; //Bias
data[dIdx][++xIdx] = -5;
data[dIdx][++xIdx] = 7;
data[dIdx][++xIdx] = 6;
++dIdx;
data[dIdx] = new double[numOfweights];
xIdx = -1;
data[dIdx][++xIdx] = 1; //Bias
data[dIdx][++xIdx] = 1;
data[dIdx][++xIdx] = 12;
data[dIdx][++xIdx] = 5;
Matrix M = new Matrix(data, true);
Vector desired = new Vector(5);
dIdx = -1;
desired.Data[++dIdx] = 8;
desired.Data[++dIdx] = 13;
desired.Data[++dIdx] = 5;
desired.Data[++dIdx] = 7;
desired.Data[++dIdx] = 10;
Vector weights = Matrix.RidgeRegression(M, desired, 0);
//Display results
for (int i = 0; i < data.Length; i++)
{
double result = 0;
for (int j = 0; j < weights.Length; j++)
{
result += data[i][j] * weights[j];
}
Console.WriteLine($"Computed {result}, Desired {desired.Data[i]}");
}
for (int i = 0; i < weights.Length; i++)
{
Console.WriteLine($"Weight[{i}] = {weights[i]}");
}
Console.WriteLine();
//QRD
QRD decomposition = new QRD(M);
Matrix B = new Matrix(desired.Length, 1, desired.Data);
Matrix W = decomposition.Solve(B);
//Display results
for (int i = 0; i < data.Length; i++)
{
double result = 0;
for (int j = 0; j < W.Data.Length; j++)
{
result += data[i][j] * W.Data[j][0];
}
Console.WriteLine($"Computed {result}, Desired {desired.Data[i]}");
}
for (int i = 0; i < W.Data.Length; i++)
{
Console.WriteLine($"Weight[{i}] = {W.Data[i][0]}");
}
;
//TimeSeriesGenerator.SaveTimeSeriesToCsvFile("MackeyGlass_big.csv", "Value", TimeSeriesGenerator.GenMackeyGlassTimeSeries(16000), CultureInfo.InvariantCulture);
MackeyGlassGeneratorSettings modSettings = new MackeyGlassGeneratorSettings(18, 0.1, 0.2);
IGenerator generator = new MackeyGlassGenerator(modSettings);
int steps = 100;
for (int i = 0; i < steps; i++)
{
Console.WriteLine(generator.Next());
}
Console.ReadLine();
generator.Reset();
for (int i = 0; i < steps; i++)
{
Console.WriteLine(generator.Next());
}
Console.ReadLine();
///*
SimpleIFSettings settings = new SimpleIFSettings(1,
new RandomValueSettings(15, 15),
new RandomValueSettings(0.05, 0.05),
new RandomValueSettings(5, 5),
new RandomValueSettings(20, 20),
0
);
IActivationFunction af = ActivationFactory.Create(settings, new Random(0));
//*/
TestActivation(af, 800, 0.15, 10, 600);
return;
}
