public Datasets(CommonParameters cp, DatasetParameters dp, DataSparse trainData, DataSparse validData)
{
Common = cp;
Dataset = dp;
Training = LoadTrainingData(trainData);
if (validData != null)
{
Validation = LoadValidationData(Training, validData);
}
}
private Dataset LoadTrainingData(DataSparse trainData)
{
if (trainData == null)
{
throw new ArgumentNullException(nameof(trainData));
}
trainData.Validate();
// TODO: not parallelised, better off to concat data and pass in as a single matrix?
Dataset dtrain = CreateDatasetFromSamplingData(trainData, Common, Dataset);
return(dtrain);
}
/// <summary>
/// Create a dataset from the sampling data.
/// </summary>
private Dataset CreateDatasetFromSamplingData(DataSparse data,
CommonParameters cp,
DatasetParameters dp)
{
var dataset = new Dataset(data.Features
, data.NumColumns
, cp
, dp
, data.Labels
, data.Weights
, data.Groups
);
return(dataset);
}
private Dataset LoadValidationData(Dataset dtrain, DataSparse validData)
{
if (validData == null)
{
throw new ArgumentNullException(nameof(validData));
}
validData.Validate();
var dvalid = new Dataset(validData.Features
, validData.NumColumns
, Common
, Dataset
, validData.Labels
, validData.Weights
, validData.Groups
, dtrain
);
return(dvalid);
}
