Settings for the training of the neural net.

Definition at line 736 of file NeuralNet.h.

Public Member Functions
	Settings (TString name, size_t _convergenceSteps=15, size_t _batchSize=10, size_t _testRepetitions=7, double _factorWeightDecay=1e-5, TMVA::DNN::EnumRegularization _regularization=TMVA::DNN::EnumRegularization::NONE, MinimizerType _eMinimizerType=MinimizerType::fSteepest, double _learningRate=1e-5, double _momentum=0.3, int _repetitions=3, bool _multithreading=true)
	c'tor More...

virtual	~Settings ()
	d'tor More...

void	addPoint (std::string histoName, double x)
	for monitoring More...

void	addPoint (std::string histoName, double x, double y)
	for monitoring More...

size_t	batchSize () const
	mini-batch size More...

void	clear (std::string histoName)
	for monitoring More...

virtual void	computeResult (const Net &, std::vector< double > &)
	callback for monitoring and loggging More...

size_t	convergenceCount () const
	returns the current convergence count More...

size_t	convergenceSteps () const
	how many steps until training is deemed to have converged More...

void	create (std::string histoName, int bins, double min, double max)
	for monitoring More...

void	create (std::string histoName, int bins, double min, double max, int bins2, double min2, double max2)
	for monitoring More...

virtual void	cycle (double progress, TString text)

virtual void	drawSample (const std::vector< double > &, const std::vector< double > &, const std::vector< double > &, double)
	callback for monitoring and loggging More...

const std::vector< double > &	dropFractions () const

size_t	dropRepetitions () const

virtual void	endTestCycle ()
	callback for monitoring and loggging More...

virtual void	endTrainCycle (double)
	callback for monitoring and logging More...

bool	exists (std::string histoName)
	for monitoring More...

double	factorWeightDecay () const
	get the weight-decay factor More...

virtual bool	hasConverged (double testError)
	has this training converged already? More...

double	learningRate () const
	get the learning rate More...

size_t	maxConvergenceCount () const
	returns the max convergence count so far More...

size_t	minError () const
	returns the smallest error so far More...

MinimizerType	minimizerType () const
	which minimizer shall be used (e.g. SGD) More...

double	momentum () const
	get the momentum (e.g. for SGD) More...

void	pads (int numPads)
	preparation for monitoring More...

void	plot (std::string histoName, std::string options, int pad, EColor color)
	for monitoring More...

EnumRegularization	regularization () const
	some regularization of the DNN is turned on? More...

int	repetitions () const
	how many steps have to be gone until the batch is changed More...

template<typename Iterator >
void	setDropOut (Iterator begin, Iterator end, size_t _dropRepetitions)
	set the drop-out configuration (layer-wise) More...

void	setMonitoring (std::shared_ptr< Monitoring > ptrMonitoring)
	prepared for monitoring More...

virtual void	setProgressLimits (double minProgress=0, double maxProgress=100)

virtual void	startTestCycle ()
	callback for monitoring and loggging More...

virtual void	startTrainCycle ()

virtual void	startTraining ()

virtual void	testIteration ()
	callback for monitoring and loggging More...

size_t	testRepetitions () const
	how often is the test data tested More...

virtual void	testSample (double, double, double, double)
	virtual function to be used for monitoring (callback) More...

bool	useMultithreading () const
	is multithreading turned on? More...

Public Attributes
size_t	count_dE

size_t	count_E

size_t	count_mb_dE

size_t	count_mb_E

double	fLearningRate

MinimizerType	fMinimizerType

double	fMomentum

int	fRepetitions

size_t	m_batchSize
	mini-batch size More...

size_t	m_convergenceCount

size_t	m_convergenceSteps
	number of steps without improvement to consider the DNN to have converged More...

std::vector< double >	m_dropOut

double	m_dropRepetitions

double	m_factorWeightDecay

size_t	m_maxConvergenceCount

double	m_maxProgress
	current limits for the progress bar More...

double	m_minError

double	m_minProgress
	current limits for the progress bar More...

EnumRegularization	m_regularization

size_t	m_testRepetitions

Timer	m_timer
	timer for monitoring More...

Protected Attributes
std::shared_ptr< Monitoring >	fMonitoring

bool	m_useMultithreading

#include <TMVA/NeuralNet.h>

Inheritance diagram for TMVA::DNN::Settings:

[legend]

Constructor & Destructor Documentation

◆ Settings()

TMVA::DNN::Settings::Settings	(	TString	name,
		size_t	_convergenceSteps = `15`,
		size_t	_batchSize = `10`,
		size_t	_testRepetitions = `7`,
		double	_factorWeightDecay = `1e-5`,
		TMVA::DNN::EnumRegularization	_regularization = `TMVA::DNN::EnumRegularization::NONE`,
		MinimizerType	_eMinimizerType = `MinimizerType::fSteepest`,
		double	_learningRate = `1e-5`,
		double	_momentum = `0.3`,
		int	_repetitions = `3`,
		bool	_multithreading = `true`
	)

c'tor

Definition at line 232 of file NeuralNet.cxx.

◆ ~Settings()

TMVA::DNN::Settings::~Settings ( )

virtual

d'tor

Definition at line 261 of file NeuralNet.cxx.

Member Function Documentation

◆ addPoint() [1/2]

void TMVA::DNN::Settings::addPoint	(	std::string	histoName,
		double	x
	)

inline

for monitoring

Definition at line 828 of file NeuralNet.h.

◆ addPoint() [2/2]

void TMVA::DNN::Settings::addPoint	(	std::string	histoName,
		double	x,
		double	y
	)

inline

for monitoring

Definition at line 829 of file NeuralNet.h.

◆ batchSize()

size_t TMVA::DNN::Settings::batchSize ( ) const

inline

mini-batch size

Definition at line 774 of file NeuralNet.h.

◆ clear()

void TMVA::DNN::Settings::clear ( std::string histoName )

inline

for monitoring

Definition at line 831 of file NeuralNet.h.

◆ computeResult()

virtual void TMVA::DNN::Settings::computeResult	(	const Net &	,
		std::vector< double > &
	)

inlinevirtual

callback for monitoring and loggging

Definition at line 816 of file NeuralNet.h.

◆ convergenceCount()

size_t TMVA::DNN::Settings::convergenceCount ( ) const

inline

returns the current convergence count

Definition at line 834 of file NeuralNet.h.

◆ convergenceSteps()

size_t TMVA::DNN::Settings::convergenceSteps ( ) const

inline

how many steps until training is deemed to have converged

Definition at line 773 of file NeuralNet.h.

◆ create() [1/2]

void TMVA::DNN::Settings::create	(	std::string	histoName,
		int	bins,
		double	min,
		double	max
	)

inline

for monitoring

Definition at line 826 of file NeuralNet.h.

◆ create() [2/2]

void TMVA::DNN::Settings::create	(	std::string	histoName,
		int	bins,
		double	min,
		double	max,
		int	bins2,
		double	min2,
		double	max2
	)

inline

for monitoring

Definition at line 827 of file NeuralNet.h.

◆ cycle()

virtual void TMVA::DNN::Settings::cycle	(	double	progress,
		TString	text
	)

inlinevirtual

Parameters

text	advance on the progress bar

Definition at line 806 of file NeuralNet.h.

◆ drawSample()

virtual void TMVA::DNN::Settings::drawSample	(	const std::vector< double > &	,
		const std::vector< double > &	,
		const std::vector< double > &	,
		double
	)

inlinevirtual

callback for monitoring and loggging

Definition at line 814 of file NeuralNet.h.

◆ dropFractions()

const std::vector< double > & TMVA::DNN::Settings::dropFractions ( ) const

inline

Definition at line 769 of file NeuralNet.h.

◆ dropRepetitions()

size_t TMVA::DNN::Settings::dropRepetitions ( ) const

inline

Definition at line 768 of file NeuralNet.h.

◆ endTestCycle()

virtual void TMVA::DNN::Settings::endTestCycle ( )

inlinevirtual

callback for monitoring and loggging

Reimplemented in TMVA::DNN::ClassificationSettings.

Definition at line 812 of file NeuralNet.h.

◆ endTrainCycle()

virtual void TMVA::DNN::Settings::endTrainCycle ( double )

inlinevirtual

callback for monitoring and logging

Reimplemented in TMVA::DNN::ClassificationSettings.

Definition at line 795 of file NeuralNet.h.

◆ exists()

bool TMVA::DNN::Settings::exists ( std::string histoName )

inline

for monitoring

Definition at line 832 of file NeuralNet.h.

◆ factorWeightDecay()

double TMVA::DNN::Settings::factorWeightDecay ( ) const

inline

get the weight-decay factor

Definition at line 776 of file NeuralNet.h.

◆ hasConverged()

bool TMVA::DNN::Settings::hasConverged ( double testError )

virtual

has this training converged already?

check for convergence

Definition at line 488 of file NeuralNet.cxx.

◆ learningRate()

double TMVA::DNN::Settings::learningRate ( ) const

inline

get the learning rate

Definition at line 778 of file NeuralNet.h.

◆ maxConvergenceCount()

size_t TMVA::DNN::Settings::maxConvergenceCount ( ) const

inline

returns the max convergence count so far

Definition at line 835 of file NeuralNet.h.

◆ minError()

size_t TMVA::DNN::Settings::minError ( ) const

inline

returns the smallest error so far

Definition at line 836 of file NeuralNet.h.

◆ minimizerType()

MinimizerType TMVA::DNN::Settings::minimizerType ( ) const

inline

which minimizer shall be used (e.g. SGD)

Definition at line 781 of file NeuralNet.h.

◆ momentum()

double TMVA::DNN::Settings::momentum ( ) const

inline

get the momentum (e.g. for SGD)

Definition at line 779 of file NeuralNet.h.

◆ pads()

void TMVA::DNN::Settings::pads ( int numPads )

inline

preparation for monitoring

Definition at line 825 of file NeuralNet.h.

◆ plot()

void TMVA::DNN::Settings::plot	(	std::string	histoName,
		std::string	options,
		int	pad,
		EColor	color
	)

inline

for monitoring

Definition at line 830 of file NeuralNet.h.

◆ regularization()

EnumRegularization TMVA::DNN::Settings::regularization ( ) const

inline

some regularization of the DNN is turned on?

Definition at line 820 of file NeuralNet.h.

◆ repetitions()

int TMVA::DNN::Settings::repetitions ( ) const

inline

how many steps have to be gone until the batch is changed

Definition at line 780 of file NeuralNet.h.

◆ setDropOut()

template<typename Iterator >

void TMVA::DNN::Settings::setDropOut	(	Iterator	begin,
		Iterator	end,
		size_t	_dropRepetitions
	)

inline

set the drop-out configuration (layer-wise)

Parameters

begin	begin of an array or vector denoting the drop-out probabilities for each layer
end	end of an array or vector denoting the drop-out probabilities for each layer
_dropRepetitions	denotes after how many repetitions the drop-out setting (which nodes are dropped out exactly) is changed

Definition at line 766 of file NeuralNet.h.

◆ setMonitoring()

void TMVA::DNN::Settings::setMonitoring ( std::shared_ptr< Monitoring > ptrMonitoring )

inline

prepared for monitoring

Definition at line 771 of file NeuralNet.h.

◆ setProgressLimits()

virtual void TMVA::DNN::Settings::setProgressLimits	(	double	minProgress = `0`,
		double	maxProgress = `100`
	)

inlinevirtual

Parameters

maxProgress for monitoring and logging (set the current "progress" limits for the display of the progress)

Definition at line 797 of file NeuralNet.h.

◆ startTestCycle()

virtual void TMVA::DNN::Settings::startTestCycle ( )

inlinevirtual

callback for monitoring and loggging

Reimplemented in TMVA::DNN::ClassificationSettings.

Definition at line 811 of file NeuralNet.h.

◆ startTrainCycle()

virtual void TMVA::DNN::Settings::startTrainCycle ( )

inlinevirtual

Reimplemented in TMVA::DNN::ClassificationSettings.

Definition at line 789 of file NeuralNet.h.

◆ startTraining()

virtual void TMVA::DNN::Settings::startTraining ( )

inlinevirtual

Definition at line 802 of file NeuralNet.h.

◆ testIteration()

virtual void TMVA::DNN::Settings::testIteration ( )

inlinevirtual

callback for monitoring and loggging

Reimplemented in TMVA::DNN::ClassificationSettings.

Definition at line 813 of file NeuralNet.h.

◆ testRepetitions()

size_t TMVA::DNN::Settings::testRepetitions ( ) const

inline

how often is the test data tested

Definition at line 775 of file NeuralNet.h.

◆ testSample()

virtual void TMVA::DNN::Settings::testSample	(	double	,
		double	,
		double	,
		double
	)

inlinevirtual

virtual function to be used for monitoring (callback)

Reimplemented in TMVA::DNN::ClassificationSettings.

Definition at line 788 of file NeuralNet.h.

◆ useMultithreading()

bool TMVA::DNN::Settings::useMultithreading ( ) const

inline

is multithreading turned on?

Definition at line 822 of file NeuralNet.h.

Member Data Documentation

◆ count_dE

size_t TMVA::DNN::Settings::count_dE

Definition at line 850 of file NeuralNet.h.

◆ count_E

size_t TMVA::DNN::Settings::count_E

Definition at line 849 of file NeuralNet.h.

◆ count_mb_dE

size_t TMVA::DNN::Settings::count_mb_dE

Definition at line 852 of file NeuralNet.h.

◆ count_mb_E

size_t TMVA::DNN::Settings::count_mb_E

Definition at line 851 of file NeuralNet.h.

◆ fLearningRate

double TMVA::DNN::Settings::fLearningRate

Definition at line 859 of file NeuralNet.h.

◆ fMinimizerType

MinimizerType TMVA::DNN::Settings::fMinimizerType

Definition at line 862 of file NeuralNet.h.

◆ fMomentum

double TMVA::DNN::Settings::fMomentum

Definition at line 860 of file NeuralNet.h.

◆ fMonitoring

std::shared_ptr<Monitoring> TMVA::DNN::Settings::fMonitoring

protected

Definition at line 872 of file NeuralNet.h.

◆ fRepetitions

int TMVA::DNN::Settings::fRepetitions

Definition at line 861 of file NeuralNet.h.

◆ m_batchSize

size_t TMVA::DNN::Settings::m_batchSize

mini-batch size

Definition at line 845 of file NeuralNet.h.

◆ m_convergenceCount

size_t TMVA::DNN::Settings::m_convergenceCount

Definition at line 864 of file NeuralNet.h.

◆ m_convergenceSteps

size_t TMVA::DNN::Settings::m_convergenceSteps

number of steps without improvement to consider the DNN to have converged

Definition at line 844 of file NeuralNet.h.

◆ m_dropOut

std::vector<double> TMVA::DNN::Settings::m_dropOut

Definition at line 857 of file NeuralNet.h.

◆ m_dropRepetitions

double TMVA::DNN::Settings::m_dropRepetitions

Definition at line 856 of file NeuralNet.h.

◆ m_factorWeightDecay

double TMVA::DNN::Settings::m_factorWeightDecay

Definition at line 847 of file NeuralNet.h.

◆ m_maxConvergenceCount

size_t TMVA::DNN::Settings::m_maxConvergenceCount

Definition at line 865 of file NeuralNet.h.

◆ m_maxProgress

double TMVA::DNN::Settings::m_maxProgress

current limits for the progress bar

Definition at line 841 of file NeuralNet.h.

◆ m_minError

double TMVA::DNN::Settings::m_minError

Definition at line 866 of file NeuralNet.h.

◆ m_minProgress

double TMVA::DNN::Settings::m_minProgress

current limits for the progress bar

Definition at line 840 of file NeuralNet.h.

◆ m_regularization

EnumRegularization TMVA::DNN::Settings::m_regularization

Definition at line 854 of file NeuralNet.h.

◆ m_testRepetitions

size_t TMVA::DNN::Settings::m_testRepetitions

Definition at line 846 of file NeuralNet.h.

◆ m_timer

Timer TMVA::DNN::Settings::m_timer

timer for monitoring

Definition at line 839 of file NeuralNet.h.

◆ m_useMultithreading

bool TMVA::DNN::Settings::m_useMultithreading

protected

Definition at line 870 of file NeuralNet.h.

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

tmva/tmva/inc/TMVA/NeuralNet.h
tmva/tmva/src/NeuralNet.cxx

Public Member Functions

Public Attributes

Protected Attributes

Constructor & Destructor Documentation

◆ Settings()

◆ ~Settings()

Member Function Documentation

◆ addPoint() [1/2]

◆ addPoint() [2/2]

◆ batchSize()

◆ clear()

◆ computeResult()

◆ convergenceCount()

◆ convergenceSteps()

◆ create() [1/2]

◆ create() [2/2]

◆ cycle()

◆ drawSample()

◆ dropFractions()

◆ dropRepetitions()

◆ endTestCycle()

◆ endTrainCycle()

◆ exists()

◆ factorWeightDecay()

◆ hasConverged()

◆ learningRate()

◆ maxConvergenceCount()

◆ minError()

◆ minimizerType()

◆ momentum()

◆ pads()

◆ plot()

◆ regularization()

◆ repetitions()

◆ setDropOut()

◆ setMonitoring()

◆ setProgressLimits()

◆ startTestCycle()

◆ startTrainCycle()

◆ startTraining()

◆ testIteration()

◆ testRepetitions()

◆ testSample()

◆ useMultithreading()

Member Data Documentation

◆ count_dE

◆ count_E

◆ count_mb_dE

◆ count_mb_E

◆ fLearningRate

◆ fMinimizerType

◆ fMomentum

◆ fMonitoring

◆ fRepetitions

◆ m_batchSize

◆ m_convergenceCount

◆ m_convergenceSteps

◆ m_dropOut

◆ m_dropRepetitions

◆ m_factorWeightDecay

◆ m_maxConvergenceCount

◆ m_maxProgress

◆ m_minError

◆ m_minProgress

◆ m_regularization

◆ m_testRepetitions

◆ m_timer

◆ m_useMultithreading