/// <summary>
/// Perform the split.
/// </summary>
/// <param name="dataDivisionList">The list of data divisions.</param>
/// <param name="dataset">The dataset to split.</param>
/// <param name="inputCount">The input count.</param>
/// <param name="idealCount">The ideal count.</param>
public void Perform(IList <DataDivision> dataDivisionList, MatrixMLDataSet dataset,
int inputCount, int idealCount)
{
GenerateCounts(dataDivisionList, dataset.Data.Length);
GenerateMasks(dataDivisionList);
if (_shuffle)
{
PerformShuffle(dataDivisionList, dataset.Data.Length);
}
CreateDividedDatasets(dataDivisionList, dataset, inputCount, idealCount);
}
/// <summary>
/// Perform the split.
/// </summary>
/// <param name="dataDivisionList">The list of data divisions.</param>
/// <param name="dataset">The dataset to split.</param>
/// <param name="inputCount">The input count.</param>
/// <param name="idealCount">The ideal count.</param>
public void Perform(IList<DataDivision> dataDivisionList, MatrixMLDataSet dataset,
int inputCount, int idealCount)
{
GenerateCounts(dataDivisionList, dataset.Data.Length);
GenerateMasks(dataDivisionList);
if (_shuffle)
{
PerformShuffle(dataDivisionList, dataset.Data.Length);
}
CreateDividedDatasets(dataDivisionList, dataset, inputCount, idealCount);
}
/// <summary>
/// Create the datasets that we will divide into.
/// </summary>
/// <param name="dataDivisionList">The list of divisions.</param>
/// <param name="parentDataset">The data set to divide.</param>
/// <param name="inputCount">The input count.</param>
/// <param name="idealCount">The ideal count.</param>
private void CreateDividedDatasets(IEnumerable <DataDivision> dataDivisionList,
MatrixMLDataSet parentDataset, int inputCount, int idealCount)
{
foreach (DataDivision division in dataDivisionList)
{
var dataset = new MatrixMLDataSet(parentDataset.Data, inputCount,
idealCount, division.Mask)
{
LagWindowSize = parentDataset.LagWindowSize,
LeadWindowSize = parentDataset.LeadWindowSize
};
division.Dataset = dataset;
}
}
private void BuildSets(IList <int[]> foldContents)
{
_folds.Clear();
for (int i = 0; i < K; i++)
{
// first calculate the size
int trainingSize = 0;
int validationSize = 0;
for (int j = 0; j < foldContents.Count; j++)
{
int foldSize = foldContents[j].Length;
if (j == i)
{
validationSize += foldSize;
}
else
{
trainingSize += foldSize;
}
}
// create the masks
var trainingMask = new int[trainingSize];
var validationMask = new int[validationSize];
int trainingIndex = 0;
for (int j = 0; j < foldContents.Count; j++)
{
int[] source = foldContents[j];
if (j == i)
{
EngineArray.ArrayCopy(source, 0, validationMask, 0, source.Length);
}
else
{
EngineArray.ArrayCopy(source, 0, trainingMask, trainingIndex, source.Length);
trainingIndex += source.Length;
}
}
// Build the set
var training = new MatrixMLDataSet(_baseDataset, trainingMask);
var validation = new MatrixMLDataSet(_baseDataset, validationMask);
_folds.Add(new DataFold(training, validation));
}
}
public KFoldCrossvalidation(MatrixMLDataSet theBaseDataset, int theK) : this()
{
_baseDataset = theBaseDataset;
_k = theK;
}
public KFoldCrossvalidation(MatrixMLDataSet theBaseDataset, int theK) : this()
{
_baseDataset = theBaseDataset;
_k = theK;
}
private void BuildSets(IList<int[]> foldContents)
{
_folds.Clear();
for (int i = 0; i < K; i++)
{
// first calculate the size
int trainingSize = 0;
int validationSize = 0;
for (int j = 0; j < foldContents.Count; j++)
{
int foldSize = foldContents[j].Length;
if (j == i)
{
validationSize += foldSize;
}
else
{
trainingSize += foldSize;
}
}
// create the masks
var trainingMask = new int[trainingSize];
var validationMask = new int[validationSize];
int trainingIndex = 0;
for (int j = 0; j < foldContents.Count; j++)
{
int[] source = foldContents[j];
if (j == i)
{
EngineArray.ArrayCopy(source, 0, validationMask, 0, source.Length);
}
else
{
EngineArray.ArrayCopy(source, 0, trainingMask, trainingIndex, source.Length);
trainingIndex += source.Length;
}
}
// Build the set
var training = new MatrixMLDataSet(_baseDataset, trainingMask);
var validation = new MatrixMLDataSet(_baseDataset, validationMask);
_folds.Add(new DataFold(training, validation));
}
}
public DataFold(MatrixMLDataSet theTraining, MatrixMLDataSet theValidation)
{
_training = theTraining;
_validation = theValidation;
}
/// <summary>
/// Create the datasets that we will divide into.
/// </summary>
/// <param name="dataDivisionList">The list of divisions.</param>
/// <param name="parentDataset">The data set to divide.</param>
/// <param name="inputCount">The input count.</param>
/// <param name="idealCount">The ideal count.</param>
private void CreateDividedDatasets(IEnumerable<DataDivision> dataDivisionList,
MatrixMLDataSet parentDataset, int inputCount, int idealCount)
{
foreach (DataDivision division in dataDivisionList)
{
var dataset = new MatrixMLDataSet(parentDataset.Data, inputCount,
idealCount, division.Mask)
{
LagWindowSize = parentDataset.LagWindowSize,
LeadWindowSize = parentDataset.LeadWindowSize
};
division.Dataset = dataset;
}
}
public DataFold(MatrixMLDataSet theTraining, MatrixMLDataSet theValidation)
{
_training = theTraining;
_validation = theValidation;
}
