neural net

holds the structure of all layers and some data for the whole net does not know the layer data though (i.e. values of the nodes and weights)

Definition at line 1068 of file NeuralNet.h.

Public Types
typedef std::pair< iterator_type, iterator_type >	begin_end_type

typedef std::vector< double >	container_type

typedef container_type::iterator	iterator_type

Public Member Functions
	Net ()
	c'tor More...

	Net (const Net &other)
	d'tor More...

void	addLayer (Layer &layer)
	add a layer (layout) More...

void	addLayer (Layer &&layer)

template<typename Settings >
void	backPropagate (std::vector< std::vector< LayerData >> &layerPatternData, const Settings &settings, size_t trainFromLayer, size_t totalNumWeights) const

void	clear ()

template<typename Weights >
std::vector< double >	compute (const std::vector< double > &input, const Weights &weights) const
	compute the net with the given input and the given weights More...

template<typename ItWeight >
std::tuple< double, double >	computeError (const Settings &settings, std::vector< LayerData > &lastLayerData, Batch &batch, ItWeight itWeightBegin, ItWeight itWeightEnd) const

void	dE ()

template<typename WeightsType , typename DropProbabilities >
void	dropOutWeightFactor (WeightsType &weights, const DropProbabilities &drops, bool inverse=false)
	set the drop out configuration More...

double	E ()

template<typename Container , typename ItWeight >
double	errorFunction (LayerData &layerData, Container truth, ItWeight itWeight, ItWeight itWeightEnd, double patternWeight, double factorWeightDecay, EnumRegularization eRegularization) const
	computes the error of the DNN More...

template<typename OutputContainer >
void	fetchOutput (const LayerData &lastLayerData, OutputContainer &outputContainer) const

template<typename OutputContainer >
void	fetchOutput (const std::vector< LayerData > &layerPatternData, OutputContainer &outputContainer) const

template<typename LayerContainer , typename PassThrough , typename ItWeight , typename ItGradient , typename OutContainer >
double	forward_backward (LayerContainer &layers, PassThrough &settingsAndBatch, ItWeight itWeightBegin, ItWeight itWeightEnd, ItGradient itGradientBegin, ItGradient itGradientEnd, size_t trainFromLayer, OutContainer &outputContainer, bool fetchOutput) const
	main NN computation function More...

template<typename LayerContainer , typename LayerPatternContainer >
void	forwardBatch (const LayerContainer &_layers, LayerPatternContainer &layerPatternData, std::vector< double > &valuesMean, std::vector< double > &valuesStdDev, size_t trainFromLayer) const

template<typename LayerContainer >
void	forwardPattern (const LayerContainer &_layers, std::vector< LayerData > &layerData) const

template<typename OutIterator >
void	initializeWeights (WeightInitializationStrategy eInitStrategy, OutIterator itWeight)
	initialize the weights with the given strategy More...

size_t	inputSize () const
	input size of the DNN More...

const std::vector< Layer > &	layers () const
	returns the layers (structure) More...

std::vector< Layer > &	layers ()
	returns the layers (structure) More...

size_t	numNodes (size_t trainingStartLayer=0) const
	returns the number of nodes in this net More...

size_t	numWeights (size_t trainingStartLayer=0) const
	returns the number of weights in this net More...

template<typename Weights , typename PassThrough >
double	operator() (PassThrough &settingsAndBatch, const Weights &weights) const
	execute computation of the DNN for one mini-batch (used by the minimizer); no computation of gradients More...

template<typename Weights , typename PassThrough , typename OutContainer >
double	operator() (PassThrough &settingsAndBatch, const Weights &weights, ModeOutput eFetch, OutContainer &outputContainer) const
	execute computation of the DNN for one mini-batch; helper function More...

template<typename Weights , typename Gradients , typename PassThrough >
double	operator() (PassThrough &settingsAndBatch, Weights &weights, Gradients &gradients) const
	execute computation of the DNN for one mini-batch (used by the minimizer); returns gradients as well More...

template<typename Weights , typename Gradients , typename PassThrough , typename OutContainer >
double	operator() (PassThrough &settingsAndBatch, Weights &weights, Gradients &gradients, ModeOutput eFetch, OutContainer &outputContainer) const

size_t	outputSize () const
	output size of the DNN More...

template<typename LayerContainer , typename DropContainer , typename ItWeight , typename ItGradient >
std::vector< std::vector< LayerData > >	prepareLayerData (LayerContainer &layers, Batch &batch, const DropContainer &dropContainer, ItWeight itWeightBegin, ItWeight itWeightEnd, ItGradient itGradientBegin, ItGradient itGradientEnd, size_t &totalNumWeights) const

template<typename Minimizer >
void	preTrain (std::vector< double > &weights, std::vector< Pattern > &trainPattern, const std::vector< Pattern > &testPattern, Minimizer &minimizer, Settings &settings)
	pre-training for future use More...

void	removeLayer ()
	remove one layer More...

void	setErrorFunction (ModeErrorFunction eErrorFunction)
	which error function is to be used More...

void	setInputSize (size_t sizeInput)
	set the input size of the DNN More...

void	SetIpythonInteractive (IPythonInteractive fI, bool fE, UInt_t M, UInt_t C)

void	setOutputSize (size_t sizeOutput)
	set the output size of the DNN More...

template<typename Minimizer >
double	train (std::vector< double > &weights, std::vector< Pattern > &trainPattern, const std::vector< Pattern > &testPattern, Minimizer &minimizer, Settings &settings)
	start the training More...

template<typename Iterator , typename Minimizer >
double	trainCycle (Minimizer &minimizer, std::vector< double > &weights, Iterator itPatternBegin, Iterator itPatternEnd, Settings &settings, DropContainer &dropContainer)
	executes one training cycle More...

Protected Member Functions
void	fillDropContainer (DropContainer &dropContainer, double dropFraction, size_t numNodes) const
	prepare the drop-out-container (select the nodes which are to be dropped out) More...

Protected Attributes
bool *	fExitFromTraining = nullptr

IPythonInteractive *	fInteractive = nullptr

UInt_t *	fIPyCurrentIter = nullptr

UInt_t *	fIPyMaxIter = nullptr

Private Attributes
ModeErrorFunction	m_eErrorFunction
	denotes the error function More...

std::vector< Layer >	m_layers
	layer-structure-data More...

size_t	m_sizeInput
	input size of this DNN More...

size_t	m_sizeOutput
	outut size of this DNN More...

#include <TMVA/NeuralNet.h>

Member Typedef Documentation

◆ begin_end_type

typedef std::pair<iterator_type,iterator_type> TMVA::DNN::Net::begin_end_type

Definition at line 1074 of file NeuralNet.h.

◆ container_type

typedef std::vector<double> TMVA::DNN::Net::container_type

Definition at line 1072 of file NeuralNet.h.

◆ iterator_type

typedef container_type::iterator TMVA::DNN::Net::iterator_type

Definition at line 1073 of file NeuralNet.h.

Constructor & Destructor Documentation

◆ Net() [1/2]

TMVA::DNN::Net::Net ( )

inline

c'tor

Definition at line 1081 of file NeuralNet.h.

◆ Net() [2/2]

TMVA::DNN::Net::Net ( const Net & other )

inline

d'tor

Definition at line 1092 of file NeuralNet.h.

Member Function Documentation

◆ addLayer() [1/2]

void TMVA::DNN::Net::addLayer ( Layer & layer )

inline

add a layer (layout)

Definition at line 1101 of file NeuralNet.h.

◆ addLayer() [2/2]

void TMVA::DNN::Net::addLayer ( Layer && layer )

inline

Definition at line 1102 of file NeuralNet.h.

◆ backPropagate()

template<typename Settings >

void TMVA::DNN::Net::backPropagate	(	std::vector< std::vector< LayerData >> &	layerPatternData,
		const Settings &	settings,
		size_t	trainFromLayer,
		size_t	totalNumWeights
	)		const

Definition at line 1353 of file NeuralNet.icc.

◆ clear()

void TMVA::DNN::Net::clear ( )

inline

Definition at line 1258 of file NeuralNet.h.

◆ compute()

template<typename Weights >

std::vector< double > TMVA::DNN::Net::compute	(	const std::vector< double > &	input,
		const Weights &	weights
	)		const

compute the net with the given input and the given weights

compute the neural net

Parameters

input	the input data
weights	the weight data

Definition at line 1026 of file NeuralNet.icc.

◆ computeError()

template<typename ItWeight >

std::tuple< double, double > TMVA::DNN::Net::computeError	(	const Settings &	settings,
		std::vector< LayerData > &	lastLayerData,
		Batch &	batch,
		ItWeight	itWeightBegin,
		ItWeight	itWeightEnd
	)		const

Definition at line 1317 of file NeuralNet.icc.

◆ dE()

void TMVA::DNN::Net::dE ( )

◆ dropOutWeightFactor()

template<typename WeightsType , typename DropProbabilities >

void TMVA::DNN::Net::dropOutWeightFactor	(	WeightsType &	weights,
		const DropProbabilities &	drops,
		bool	inverse = `false`
	)

set the drop out configuration

compute the drop-out-weight factor

when using drop-out a fraction of the nodes is turned off at each cycle of the computation once all nodes are turned on again (for instances when the test samples are evaluated), the weights have to be adjusted to account for the different number of active nodes this function computes the factor and applies it to the weights

Definition at line 645 of file NeuralNet.icc.

◆ E()

double TMVA::DNN::Net::E ( )

◆ errorFunction()

template<typename Container , typename ItWeight >

double TMVA::DNN::Net::errorFunction	(	LayerData &	layerData,
		Container	truth,
		ItWeight	itWeight,
		ItWeight	itWeightEnd,
		double	patternWeight,
		double	factorWeightDecay,
		EnumRegularization	eRegularization
	)		const

computes the error of the DNN

compute the error function

Definition at line 1577 of file NeuralNet.icc.

◆ fetchOutput() [1/2]

template<typename OutputContainer >

void TMVA::DNN::Net::fetchOutput	(	const LayerData &	lastLayerData,
		OutputContainer &	outputContainer
	)		const

Definition at line 1287 of file NeuralNet.icc.

◆ fetchOutput() [2/2]

template<typename OutputContainer >

void TMVA::DNN::Net::fetchOutput	(	const std::vector< LayerData > &	layerPatternData,
		OutputContainer &	outputContainer
	)		const

Definition at line 1308 of file NeuralNet.icc.

◆ fillDropContainer()

void TMVA::DNN::Net::fillDropContainer	(	DropContainer &	dropContainer,
		double	dropFraction,
		size_t	_numNodes
	)		const

protected

prepare the drop-out-container (select the nodes which are to be dropped out)

prepare the drop-out container given the provided drop-fractions

Definition at line 553 of file NeuralNet.cxx.

◆ forward_backward()

template<typename LayerContainer , typename PassThrough , typename ItWeight , typename ItGradient , typename OutContainer >

double TMVA::DNN::Net::forward_backward	(	LayerContainer &	_layers,
		PassThrough &	settingsAndBatch,
		ItWeight	itWeightBegin,
		ItWeight	itWeightEnd,
		ItGradient	itGradientBegin,
		ItGradient	itGradientEnd,
		size_t	trainFromLayer,
		OutContainer &	outputContainer,
		bool	doFetchOutput
	)		const

main NN computation function

forward propagation and backward propagation

Definition at line 1403 of file NeuralNet.icc.

◆ forwardBatch()

template<typename LayerContainer , typename LayerPatternContainer >

void TMVA::DNN::Net::forwardBatch	(	const LayerContainer &	_layers,
		LayerPatternContainer &	layerPatternData,
		std::vector< double > &	valuesMean,
		std::vector< double > &	valuesStdDev,
		size_t	trainFromLayer
	)		const

Definition at line 1232 of file NeuralNet.icc.

◆ forwardPattern()

template<typename LayerContainer >

void TMVA::DNN::Net::forwardPattern	(	const LayerContainer &	_layers,
		std::vector< LayerData > &	layerData
	)		const

Definition at line 1212 of file NeuralNet.icc.

◆ initializeWeights()

template<typename OutIterator >

void TMVA::DNN::Net::initializeWeights	(	WeightInitializationStrategy	eInitStrategy,
		OutIterator	itWeight
	)

initialize the weights with the given strategy

initialization of the weights

Definition at line 1468 of file NeuralNet.icc.

◆ inputSize()

size_t TMVA::DNN::Net::inputSize ( ) const

inline

input size of the DNN

Definition at line 1105 of file NeuralNet.h.

◆ layers() [1/2]

const std::vector<Layer>& TMVA::DNN::Net::layers ( ) const

inline

returns the layers (structure)

Definition at line 1252 of file NeuralNet.h.

◆ layers() [2/2]

std::vector<Layer>& TMVA::DNN::Net::layers ( )

inline

returns the layers (structure)

Definition at line 1253 of file NeuralNet.h.

◆ numNodes()

size_t TMVA::DNN::Net::numNodes ( size_t trainingStartLayer = 0 ) const

returns the number of nodes in this net

Definition at line 537 of file NeuralNet.cxx.

◆ numWeights()

size_t TMVA::DNN::Net::numWeights ( size_t trainingStartLayer = 0 ) const

returns the number of weights in this net

compute the number of weights given the size of the input layer

Definition at line 521 of file NeuralNet.cxx.

◆ operator()() [1/4]

template<typename Weights , typename PassThrough >

double TMVA::DNN::Net::operator()	(	PassThrough &	settingsAndBatch,
		const Weights &	weights
	)		const

execute computation of the DNN for one mini-batch (used by the minimizer); no computation of gradients

Definition at line 1059 of file NeuralNet.icc.

◆ operator()() [2/4]

template<typename Weights , typename PassThrough , typename OutContainer >

double TMVA::DNN::Net::operator()	(	PassThrough &	settingsAndBatch,
		const Weights &	weights,
		ModeOutput	eFetch,
		OutContainer &	outputContainer
	)		const

execute computation of the DNN for one mini-batch; helper function

Definition at line 1068 of file NeuralNet.icc.

◆ operator()() [3/4]

template<typename Weights , typename Gradients , typename PassThrough >

double TMVA::DNN::Net::operator()	(	PassThrough &	settingsAndBatch,
		Weights &	weights,
		Gradients &	gradients
	)		const

execute computation of the DNN for one mini-batch (used by the minimizer); returns gradients as well

Definition at line 1078 of file NeuralNet.icc.

◆ operator()() [4/4]

template<typename Weights , typename Gradients , typename PassThrough , typename OutContainer >

double TMVA::DNN::Net::operator()	(	PassThrough &	settingsAndBatch,
		Weights &	weights,
		Gradients &	gradients,
		ModeOutput	eFetch,
		OutContainer &	outputContainer
	)		const

Definition at line 1088 of file NeuralNet.icc.

◆ outputSize()

size_t TMVA::DNN::Net::outputSize ( ) const

inline

output size of the DNN

Definition at line 1106 of file NeuralNet.h.

◆ prepareLayerData()

template<typename LayerContainer , typename DropContainer , typename ItWeight , typename ItGradient >

std::vector< std::vector< LayerData > > TMVA::DNN::Net::prepareLayerData	(	LayerContainer &	layers,
		Batch &	batch,
		const DropContainer &	dropContainer,
		ItWeight	itWeightBegin,
		ItWeight	itWeightEnd,
		ItGradient	itGradientBegin,
		ItGradient	itGradientEnd,
		size_t &	totalNumWeights
	)		const

Definition at line 1100 of file NeuralNet.icc.

◆ preTrain()

template<typename Minimizer >

void TMVA::DNN::Net::preTrain	(	std::vector< double > &	weights,
		std::vector< Pattern > &	trainPattern,
		const std::vector< Pattern > &	testPattern,
		Minimizer &	minimizer,
		Settings &	settings
	)

pre-training for future use

◆ removeLayer()

void TMVA::DNN::Net::removeLayer ( )

inline

remove one layer

Definition at line 1255 of file NeuralNet.h.

◆ setErrorFunction()

void TMVA::DNN::Net::setErrorFunction ( ModeErrorFunction eErrorFunction )

inline

which error function is to be used

Definition at line 1103 of file NeuralNet.h.

◆ setInputSize()

void TMVA::DNN::Net::setInputSize ( size_t sizeInput )

inline

set the input size of the DNN

Definition at line 1099 of file NeuralNet.h.

◆ SetIpythonInteractive()

void TMVA::DNN::Net::SetIpythonInteractive	(	IPythonInteractive *	fI,
		bool *	fE,
		UInt_t *	M,
		UInt_t *	C
	)

inline

Definition at line 1290 of file NeuralNet.h.

◆ setOutputSize()

void TMVA::DNN::Net::setOutputSize ( size_t sizeOutput )

inline

set the output size of the DNN

Definition at line 1100 of file NeuralNet.h.

◆ train()

template<typename Minimizer >

double TMVA::DNN::Net::train	(	std::vector< double > &	weights,
		std::vector< Pattern > &	trainPattern,
		const std::vector< Pattern > &	testPattern,
		Minimizer &	minimizer,
		Settings &	settings
	)

start the training

execute the training until convergence emerges

Parameters

weights	weight vector
trainPattern	training pattern
testPattern	test pattern
minimizer	use this minimizer for training (e.g. SGD)
settings	settings used for this training run
weights	the container with the weights (synapses)
trainPattern	the pattern for the training
testPattern	the pattern for the testing
minimizer	the minimizer (e.g. steepest gradient descent) to be used
settings	the settings for the training (e.g. multithreading or not, regularization etc.)

Definition at line 705 of file NeuralNet.icc.

◆ trainCycle()

template<typename Iterator , typename Minimizer >

double TMVA::DNN::Net::trainCycle	(	Minimizer &	minimizer,
		std::vector< double > &	weights,
		Iterator	itPatternBegin,
		Iterator	itPatternEnd,
		Settings &	settings,
		DropContainer &	dropContainer
	)

inline

executes one training cycle

execute a single training cycle

Parameters

minimizier	the minimizer to be used
weights	the weight vector to be used
itPatternBegin	the pattern to be trained with
itPatternEnd	the pattern to be trainied with
settings	the settings for the training
dropContainer	the configuration for DNN drop-out

uses multithreading if turned on

Parameters

minimizer	the minimizer to be used (e.g. SGD)
weights	the weight container with all the synapse weights
itPatternBegin	begin of the pattern container itPatternEnd the end of the pattern container
settings	the settings for this training (e.g. multithreading or not, regularization, etc.)
dropContainer	the data for dropping-out nodes (regularization technique)

Definition at line 928 of file NeuralNet.icc.

Member Data Documentation

◆ fExitFromTraining

bool* TMVA::DNN::Net::fExitFromTraining = nullptr

protected

Definition at line 1284 of file NeuralNet.h.

◆ fInteractive

IPythonInteractive* TMVA::DNN::Net::fInteractive = nullptr

protected

Definition at line 1283 of file NeuralNet.h.

◆ fIPyCurrentIter

UInt_t * TMVA::DNN::Net::fIPyCurrentIter = nullptr

protected

Definition at line 1285 of file NeuralNet.h.

◆ fIPyMaxIter

UInt_t* TMVA::DNN::Net::fIPyMaxIter = nullptr

protected

Definition at line 1285 of file NeuralNet.h.

◆ m_eErrorFunction

ModeErrorFunction TMVA::DNN::Net::m_eErrorFunction

private

denotes the error function

Definition at line 1276 of file NeuralNet.h.

◆ m_layers

std::vector<Layer> TMVA::DNN::Net::m_layers

private

layer-structure-data

Definition at line 1279 of file NeuralNet.h.

◆ m_sizeInput

size_t TMVA::DNN::Net::m_sizeInput

private

input size of this DNN

Definition at line 1277 of file NeuralNet.h.

◆ m_sizeOutput

size_t TMVA::DNN::Net::m_sizeOutput

private

outut size of this DNN

Definition at line 1278 of file NeuralNet.h.

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

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

Public Types

Public Member Functions

Protected Member Functions

Protected Attributes

Private Attributes

Member Typedef Documentation

◆ begin_end_type

◆ container_type

◆ iterator_type

Constructor & Destructor Documentation

◆ Net() [1/2]

◆ Net() [2/2]

Member Function Documentation

◆ addLayer() [1/2]

◆ addLayer() [2/2]

◆ backPropagate()

◆ clear()

◆ compute()

◆ computeError()

◆ dE()

◆ dropOutWeightFactor()

◆ E()

◆ errorFunction()

◆ fetchOutput() [1/2]

◆ fetchOutput() [2/2]

◆ fillDropContainer()

◆ forward_backward()

◆ forwardBatch()

◆ forwardPattern()

◆ initializeWeights()

◆ inputSize()

◆ layers() [1/2]

◆ layers() [2/2]

◆ numNodes()

◆ numWeights()

◆ operator()() [1/4]

◆ operator()() [2/4]

◆ operator()() [3/4]

◆ operator()() [4/4]

◆ outputSize()

◆ prepareLayerData()

◆ preTrain()

◆ removeLayer()

◆ setErrorFunction()

◆ setInputSize()

◆ SetIpythonInteractive()

◆ setOutputSize()

◆ train()

◆ trainCycle()

Member Data Documentation

◆ fExitFromTraining

◆ fInteractive

◆ fIPyCurrentIter

◆ fIPyMaxIter

◆ m_eErrorFunction

◆ m_layers

◆ m_sizeInput

◆ m_sizeOutput