TrainingDataManager Class Reference

TrainingDataManager class encloses GRT::TimeSeriesClassificationData and improves upon by adding utility functions that relabel, delete or trim some training samples. More...

#include <training-data-manager.h>

Public Member Functions
	TrainingDataManager (uint32_t num_classes)
bool	setNumDimensions (uint32_t dim)
bool	setDatasetName (const std::string name)
bool	setDatasetName (const char *const name)
GRT::TimeSeriesClassificationData	getAllData ()
uint32_t	getNumLabels ()
uint32_t	getTotalNumSamples ()
bool	setNameForLabel (const std::string name, uint32_t label)
	This will modify the default name for this label, changing it from "Label X" to name. More...
std::string	getLabelName (uint32_t label)
std::string	getSampleName (uint32_t label, uint32_t index)
	Format the sample name. Default label name is "Label X", and the sample name is "Label X [Y]" Default name can be changed by setNameForLabel. More...
bool	setSampleName (uint32_t, uint32_t, const std::string)
bool	addSample (uint32_t label, const GRT::MatrixDouble &sample)
	Add new sample. Returns false if the label is larger than configured number of classes. More...
uint32_t	getNumSampleForLabel (uint32_t label)
GRT::MatrixDouble	getSample (uint32_t label, uint32_t index)
	Get the sample by label and index. More...
bool	deleteSample (uint32_t label, uint32_t index)
	Remove sample by label and the index. More...
bool	deleteAllSamples ()
	Remove all samples. More...
bool	deleteAllSamplesWithLabel (uint32_t label)
	Remove all samples. More...
bool	relabelSample (uint32_t label, uint32_t index, uint32_t new_label)
	Relabel a sample from label to new_label. More...
bool	trimSample (uint32_t label, uint32_t index, uint32_t start, uint32_t end)
	Trim sample. What's left will be [start, end], closed interval. More...
bool	hasSampleScore (uint32_t label, uint32_t index)
double	getSampleScore (uint32_t label, uint32_t index)
bool	setSampleScore (uint32_t label, uint32_t index, double score)
bool	hasSampleClassLikelihoods (uint32_t label, uint32_t index)
vector< double >	getSampleClassLikelihoods (uint32_t label, uint32_t index)
bool	setSampleClassLikelihoods (uint32_t label, uint32_t index, vector< double > likelihoods)
bool	save (const std::string &filename)
bool	load (const std::string &filename)

Detailed Description

TrainingDataManager class encloses GRT::TimeSeriesClassificationData and improves upon by adding utility functions that relabel, delete or trim some training samples.

This class will take the sole mutable ownership of the enclosed data. All operations over the training data have to gone through this class.

Training data can be viewed as a collection of training samples, where each sample consists of a label (which class this sample belongs to) and the data (a time-series data).

A few key augmentation to the underlying TimeSeriesClassificationData:

1. Edit (relabel, delete or trim individual samples).
2. Name individual sample.

Each individual sample is addressable by (label, index) tuple. Label starts from 1 and index starts from 0.

Constructor & Destructor Documentation

TrainingDataManager::TrainingDataManager

(

uint32_t

num_classes

)

Member Function Documentation

bool TrainingDataManager::addSample	(	uint32_t	label,
		const GRT::MatrixDouble &	sample
	)

Add new sample. Returns false if the label is larger than configured number of classes.

bool TrainingDataManager::deleteAllSamples

(

)

Remove all samples.

bool TrainingDataManager::deleteAllSamplesWithLabel

(

uint32_t

label

)

Remove all samples.

bool TrainingDataManager::deleteSample	(	uint32_t	label,
		uint32_t	index
	)

Remove sample by label and the index.

GRT::TimeSeriesClassificationData TrainingDataManager::getAllData ( )

inline

std::string TrainingDataManager::getLabelName

(

uint32_t

label

)

uint32_t TrainingDataManager::getNumLabels ( )

inline

uint32_t TrainingDataManager::getNumSampleForLabel

(

uint32_t

label

)

GRT::MatrixDouble TrainingDataManager::getSample	(	uint32_t	label,
		uint32_t	index
	)

Get the sample by label and index.

std::vector< double > TrainingDataManager::getSampleClassLikelihoods	(	uint32_t	label,
		uint32_t	index
	)

std::string TrainingDataManager::getSampleName	(	uint32_t	label,
		uint32_t	index
	)

Format the sample name. Default label name is "Label X", and the sample name is "Label X [Y]" Default name can be changed by setNameForLabel.

double TrainingDataManager::getSampleScore	(	uint32_t	label,
		uint32_t	index
	)

uint32_t TrainingDataManager::getTotalNumSamples ( )

inline

bool TrainingDataManager::hasSampleClassLikelihoods	(	uint32_t	label,
		uint32_t	index
	)

bool TrainingDataManager::hasSampleScore	(	uint32_t	label,
		uint32_t	index
	)

bool TrainingDataManager::load

(

const std::string &

filename

)

bool TrainingDataManager::relabelSample	(	uint32_t	label,
		uint32_t	index,
		uint32_t	new_label
	)

Relabel a sample from label to new_label.

bool TrainingDataManager::save ( const std::string &  filename )

inline

bool TrainingDataManager::setDatasetName

(

const std::string

name

)

bool TrainingDataManager::setDatasetName

(

const char *const

name

)

bool TrainingDataManager::setNameForLabel	(	const std::string	name,
		uint32_t	label
	)

This will modify the default name for this label, changing it from "Label X" to name.

bool TrainingDataManager::setNumDimensions

(

uint32_t

dim

)

bool TrainingDataManager::setSampleClassLikelihoods	(	uint32_t	label,
		uint32_t	index,
		vector< double >	likelihoods
	)

bool TrainingDataManager::setSampleName	(	uint32_t	label,
		uint32_t	index,
		const std::string	new_name
	)

bool TrainingDataManager::setSampleScore	(	uint32_t	label,
		uint32_t	index,
		double	score
	)

bool TrainingDataManager::trimSample	(	uint32_t	label,
		uint32_t	index,
		uint32_t	start,
		uint32_t	end
	)

Trim sample. What's left will be [start, end], closed interval.

---

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

• src/training-data-manager.h
• src/training-data-manager.cpp