/// <summary>
/// Learns and extracts the meta features learned by the ensemble models on the provided indices
/// </summary>
/// <param name="observations"></param>
/// <param name="targets"></param>
/// <param name="indices"></param>
/// <returns></returns>
public F64Matrix LearnMetaFeatures(F64Matrix observations, double[] targets,
int[] indices)
{
var numberOfClasses = targets.Distinct().Count();
var cvRows = indices.Length;
var ensembleFeatures = m_learners.Length * numberOfClasses;
var cvCols = ensembleFeatures;
if (m_includeOriginalFeaturesForMetaLearner)
{
cvCols = cvCols + observations.ColumnCount;
}
var cvPredictions = new F64Matrix(cvRows, cvCols);
var modelPredictions = new ProbabilityPrediction[cvRows];
for (int i = 0; i < m_learners.Length; i++)
{
Trace.WriteLine("Training model: " + (i + 1));
var learner = m_learners[i];
m_crossValidation.CrossValidate(learner, observations, targets,
indices, modelPredictions);
for (int j = 0; j < modelPredictions.Length; j++)
{
var probabilities = modelPredictions[j].Probabilities.Values.ToArray();
for (int k = 0; k < probabilities.Length; k++)
{
cvPredictions[j, i *numberOfClasses + k] = probabilities[k];
}
}
}
if (m_includeOriginalFeaturesForMetaLearner)
{
for (int i = 0; i < cvRows; i++)
{
for (int j = 0; j < observations.ColumnCount; j++)
{
var value = cvPredictions[i, j + ensembleFeatures];
cvPredictions[i, j + ensembleFeatures] = observations[indices[i], j];
}
}
}
return(cvPredictions);
}
/// <summary>
/// Learns and extracts the meta features learned by the ensemble models on the provided indices
/// </summary>
/// <param name="observations"></param>
/// <param name="targets"></param>
/// <param name="indices"></param>
/// <returns></returns>
public ProbabilityPrediction[][] LearnMetaFeatures(F64Matrix observations, double[] targets, int[] indices)
{
var cvRows = indices.Length;
var cvCols = m_learners.Length;
var ensemblePredictions = new ProbabilityPrediction[cvCols][];
for (int i = 0; i < m_learners.Length; i++)
{
Trace.WriteLine("Training model: " + (i + 1));
var learner = m_learners[i];
var learnerPredictions = new ProbabilityPrediction[cvRows];
m_crossValidation.CrossValidate(learner, observations, targets, indices, learnerPredictions);
ensemblePredictions[i] = learnerPredictions;
}
return(ensemblePredictions);
}
/// <summary>
/// Learns and extracts the meta features learned by the ensemble models on the provided indices
/// </summary>
/// <param name="observations"></param>
/// <param name="targets"></param>
/// <param name="indices"></param>
/// <returns></returns>
public F64Matrix LearnMetaFeatures(F64Matrix observations, double[] targets, int[] indices)
{
var cvRows = indices.Length;
var cvCols = m_learners.Length;
var cvPredictions = new F64Matrix(cvRows, cvCols);
var modelPredictions = new double[cvRows];
for (int i = 0; i < m_learners.Length; i++)
{
Trace.WriteLine("Training model: " + (i + 1));
var learner = m_learners[i];
m_crossValidation.CrossValidate(learner, observations, targets, indices, modelPredictions);
for (int j = 0; j < modelPredictions.Length; j++)
{
cvPredictions[j, i] = modelPredictions[j];
}
}
return(cvPredictions);
}
/// <summary>
/// Learns and extracts the meta features learned by the ensemble models on the provided indices
/// </summary>
/// <param name="observations"></param>
/// <param name="targets"></param>
/// <param name="indices"></param>
/// <returns></returns>
public F64Matrix LearnMetaFeatures(F64Matrix observations, double[] targets,
int[] indices)
{
var cvRows = indices.Length;
var cvCols = m_learners.Length;
if (m_includeOriginalFeaturesForMetaLearner)
{
cvCols = cvCols + observations.ColumnCount;
}
var cvPredictions = new F64Matrix(cvRows, cvCols);
var modelPredictions = new double[cvRows];
for (int i = 0; i < m_learners.Length; i++)
{
Trace.WriteLine("Training model: " + (i + 1));
var learner = m_learners[i];
m_crossValidation.CrossValidate(learner, observations, targets,
indices, modelPredictions);
for (int j = 0; j < modelPredictions.Length; j++)
{
cvPredictions[j, i] = modelPredictions[j];
}
}
if (m_includeOriginalFeaturesForMetaLearner)
{
for (int i = 0; i < cvRows; i++)
{
for (int j = 0; j < observations.ColumnCount; j++)
{
cvPredictions[i, j + m_learners.Length] = observations[indices[i], j];
}
}
}
return(cvPredictions);
}