Configuration structure used for all kinds of SampleProviders including default values. More...

#include <DataSourceConfig.hpp>

Public Attributes
size_t	batchSize = 0

datadriven::DataTransformationConfig	dataTransformationConfig

size_t	epochs = 1
	The number of epochs to train on. More...

std::string	filePath = ""
	Valid path to a file on disk. More...

DataSourceFileType	fileType = DataSourceFileType::NONE
	Which type of input file are we dealing with? NONE for auto detection or generated artificial datasets. More...

bool	hasTargets = true
	whether the file has targets (i.e. More...

bool	isCompressed = false
	The dataset is gzip compressed. More...

size_t	numBatches = 1
	How many batches should the dataset be split into for batch learning - if 1, take the entire dataset. More...

int64_t	randomSeed = -1
	Seed for the shuffling prng. More...

std::vector< double >	readinClasses = std::vector<double>()
	Specifies the set of classes (targets) to be read-in from the data file Any line with a class not contained in this vector is skipped If hasTargets=false this is ignored If empty then all classes/targets are considered (default) More...

std::vector< size_t >	readinColumns = std::vector<size_t>()
	Specifies the set of columns (dimensions) to be read-in from the data file Starts at 0, order matters; Any column not contained in this vector is ignored as a dimension If empty, then all columns are read in (default) More...

size_t	readinCutoff = -1
	After how many (valid) lines of the sourcefile to stop reading. More...

DataSourceShufflingType	shuffling = DataSourceShufflingType::sequential
	The type of shuffling to be applied to the data. More...

double	validationPortion = 0.3

Detailed Description

Configuration structure used for all kinds of SampleProviders including default values.

Member Data Documentation

◆ batchSize

size_t sgpp::datadriven::DataSourceConfig::batchSize = 0

Referenced by sgpp::datadriven::DataMiningConfigParser::getDataSourceConfig(), sgpp::datadriven::DataSource::getNextSamples(), and sgpp::datadriven::DataSourceBuilder::withBatchSize().

◆ dataTransformationConfig

datadriven::DataTransformationConfig sgpp::datadriven::DataSourceConfig::dataTransformationConfig

Referenced by sgpp::datadriven::DataSource::DataSource(), sgpp::datadriven::DataMiningConfigParser::getDataSourceConfig(), and sgpp::datadriven::DataSource::getNextSamples().

◆ epochs

size_t sgpp::datadriven::DataSourceConfig::epochs = 1

The number of epochs to train on.

Referenced by sgpp::datadriven::DataMiningConfigParser::getDataSourceConfig().

◆ filePath

std::string sgpp::datadriven::DataSourceConfig::filePath = ""

Valid path to a file on disk.

Empty for generated artificial datasets

Referenced by sgpp::datadriven::DataSourceBuilder::crossValidationFromConfig(), sgpp::datadriven::DataSource::DataSource(), sgpp::datadriven::DataMiningConfigParser::getDataSourceConfig(), and sgpp::datadriven::DataSourceBuilder::withPath().

◆ fileType

DataSourceFileType sgpp::datadriven::DataSourceConfig::fileType = DataSourceFileType::NONE

Which type of input file are we dealing with? NONE for auto detection or generated artificial datasets.

Referenced by sgpp::datadriven::DataSourceBuilder::crossValidationAssemble(), sgpp::datadriven::DataSourceBuilder::crossValidationFromConfig(), sgpp::datadriven::DataMiningConfigParser::getDataSourceConfig(), sgpp::datadriven::DataSourceBuilder::splittingAssemble(), sgpp::datadriven::DataSourceBuilder::splittingFromConfig(), sgpp::datadriven::DataSourceBuilder::withFileType(), and sgpp::datadriven::DataSourceBuilder::withPath().

◆ hasTargets

bool sgpp::datadriven::DataSourceConfig::hasTargets = true

whether the file has targets (i.e.

supervised learning)

Referenced by sgpp::datadriven::DataMiningConfigParser::getDataSourceConfig().

◆ isCompressed

bool sgpp::datadriven::DataSourceConfig::isCompressed = false

The dataset is gzip compressed.

Referenced by sgpp::datadriven::DataSourceBuilder::crossValidationAssemble(), sgpp::datadriven::DataMiningConfigParser::getDataSourceConfig(), sgpp::datadriven::DataSourceBuilder::splittingAssemble(), and sgpp::datadriven::DataSourceBuilder::withCompression().

◆ numBatches

size_t sgpp::datadriven::DataSourceConfig::numBatches = 1

How many batches should the dataset be split into for batch learning - if 1, take the entire dataset.

Referenced by sgpp::datadriven::DataSource::end(), sgpp::datadriven::DataMiningConfigParser::getDataSourceConfig(), sgpp::datadriven::DataSource::getNextSamples(), and sgpp::datadriven::DataSourceBuilder::inBatches().

◆ randomSeed

int64_t sgpp::datadriven::DataSourceConfig::randomSeed = -1

Seed for the shuffling prng.

Referenced by sgpp::datadriven::DataShufflingFunctorFactory::buildDataShufflingFunctor(), and sgpp::datadriven::DataMiningConfigParser::getDataSourceConfig().

◆ readinClasses

std::vector<double> sgpp::datadriven::DataSourceConfig::readinClasses = std::vector<double>()

Specifies the set of classes (targets) to be read-in from the data file Any line with a class not contained in this vector is skipped If hasTargets=false this is ignored If empty then all classes/targets are considered (default)

Referenced by sgpp::datadriven::DataMiningConfigParser::getDataSourceConfig().

◆ readinColumns

std::vector<size_t> sgpp::datadriven::DataSourceConfig::readinColumns = std::vector<size_t>()

Specifies the set of columns (dimensions) to be read-in from the data file Starts at 0, order matters; Any column not contained in this vector is ignored as a dimension If empty, then all columns are read in (default)

Referenced by sgpp::datadriven::DataMiningConfigParser::getDataSourceConfig().