DataSourceSlitting is a high level interface to provide functionality for processing data epoch-wise with a validation set that is retrieved at initialization time using the first samples the sample provider provides. More...

#include <DataSourceSplitting.hpp>

Inheritance diagram for sgpp::datadriven::DataSourceSplitting:

Public Member Functions
	DataSourceSplitting (const DataSourceConfig &config, SampleProvider *sampleProvider)
	Constructor.

Dataset *	getValidationData () override
	Returns the data that is used for validation.

void	reset ()
	Resets the state of the the sample provider to begin a new training epoch.

	~DataSourceSplitting () override
	Clean up memory.

Public Member Functions inherited from sgpp::datadriven::DataSource
DataSourceIterator	begin ()
	Return an iterator object pointing to the first batch of this DataSource.

	DataSource (DataSourceConfig config, SampleProvider *sampleProvider)
	Constructor.

DataSourceIterator	end ()
	Return an iterator object pointing to the last possible batch of this DataSource.

virtual Dataset *	getAllSamples ()
	Request all data from the underlying SampleProvider object upon construction.

const DataSourceConfig &	getConfig () const
	Read only access to the configuration used by DataSource and underlying SampleProvider.

size_t	getCurrentIteration () const
	Return how many batches have already been requested from this DataSource.

virtual Dataset *	getNextSamples ()
	Request data from the underlying SampleProvider as specified in the provided configuration object upon construction.

virtual	~DataSource ()=default

Additional Inherited Members
Protected Attributes inherited from sgpp::datadriven::DataSource
DataSourceConfig	config
	Configuration file that determines all relevant properties of the object.

size_t	currentIteration
	counter variable if data is requested in batches.

DataTransformation *	dataTransformation
	pointer to DataTransformation to perform transformations on init.

std::unique_ptr< SampleProvider >	sampleProvider
	pointer to sample provider that actually handles data aquisition.

Detailed Description

DataSourceSlitting is a high level interface to provide functionality for processing data epoch-wise with a validation set that is retrieved at initialization time using the first samples the sample provider provides.

Constructor & Destructor Documentation

◆ DataSourceSplitting()

sgpp::datadriven::DataSourceSplitting::DataSourceSplitting	(	const DataSourceConfig &	config,
		SampleProvider *	sampleProvider
	)

Constructor.

Parameters

config	configuration object used for the data source
sampleProvider	the sample provider to operate on.

◆ ~DataSourceSplitting()

sgpp::datadriven::DataSourceSplitting::~DataSourceSplitting ( )

inlineoverride

Clean up memory.

Member Function Documentation

◆ getValidationData()

Dataset * sgpp::datadriven::DataSourceSplitting::getValidationData ( )

overridevirtual

Returns the data that is used for validation.

Returns: pointer to the validation dataset

Implements sgpp::datadriven::DataSource.

◆ reset()

void sgpp::datadriven::DataSourceSplitting::reset ( )

Resets the state of the the sample provider to begin a new training epoch.

References sgpp::datadriven::DataSource::config, sgpp::datadriven::DataSource::sampleProvider, and sgpp::datadriven::DataSourceConfig::validationPortion_.

The documentation for this class was generated from the following files:

datadriven/src/sgpp/datadriven/datamining/modules/dataSource/DataSourceSplitting.hpp
datadriven/src/sgpp/datadriven/datamining/modules/dataSource/DataSourceSplitting.cpp

Public Member Functions