NeuralNet.h File Reference

#include <vector>
#include <iostream>
#include <algorithm>
#include <iterator>
#include <functional>
#include <tuple>
#include <cmath>
#include <cassert>
#include <random>
#include <thread>
#include <future>
#include <type_traits>
#include <string>
#include <utility>
#include "Pattern.h"
#include "Monitoring.h"
#include "TApplication.h"
#include "Timer.h"
#include "TH1F.h"
#include "TH2F.h"
#include <fenv.h>
#include "TMVA/NeuralNet.icc"

Include dependency graph for NeuralNet.h:

This graph shows which files directly or indirectly include this file:

Classes
class	TMVA::DNN::Batch
	The Batch class encapsulates one mini-batch. More...
class	TMVA::DNN::ClassificationSettings
	Settings for classificationused to distinguish between different function signatures. More...
class	TMVA::DNN::Layer
	Layer defines the layout of a layer. More...
class	TMVA::DNN::LayerData
	LayerData holds the data of one layer. More...
class	TMVA::DNN::MeanVariance
class	TMVA::DNN::Net
	neural net More...
class	TMVA::DNN::Settings
	Settings for the training of the neural net. More...
class	TMVA::DNN::Steepest
	Steepest Gradient Descent algorithm (SGD) More...

Namespaces
namespace	TMVA
	create variable transformations
namespace	TMVA::DNN

Typedefs
typedef std::vector< char >	TMVA::DNN::DropContainer
typedef std::tuple< Settings &, Batch &, DropContainer & >	TMVA::DNN::pass_through_type

Enumerations
enum class	TMVA::DNN::EnumFunction {   TMVA::DNN::ZERO = '0' , TMVA::DNN::LINEAR = 'L' , TMVA::DNN::TANH = 'T' , TMVA::DNN::RELU = 'R' ,   TMVA::DNN::SYMMRELU = 'r' , TMVA::DNN::TANHSHIFT = 't' , TMVA::DNN::SIGMOID = 's' , TMVA::DNN::SOFTSIGN = 'S' ,   TMVA::DNN::GAUSS = 'G' , TMVA::DNN::GAUSSCOMPLEMENT = 'C' }
enum class	TMVA::DNN::EnumRegularization { TMVA::DNN::NONE , TMVA::DNN::L1 , TMVA::DNN::L2 , TMVA::DNN::L1MAX }
enum	TMVA::DNN::MinimizerType { TMVA::DNN::fSteepest }
	< list all the minimizer types More...
enum class	TMVA::DNN::ModeErrorFunction { TMVA::DNN::SUMOFSQUARES = 'S' , TMVA::DNN::CROSSENTROPY = 'C' , TMVA::DNN::CROSSENTROPY_MUTUALEXCLUSIVE = 'M' }
	error functions to be chosen from More...
enum class	TMVA::DNN::ModeOutput { TMVA::DNN::FETCH }
enum class	TMVA::DNN::ModeOutputValues : int { TMVA::DNN::DIRECT = 0x01 , TMVA::DNN::SIGMOID = 0x02 , TMVA::DNN::SOFTMAX = 0x04 , TMVA::DNN::BATCHNORMALIZATION = 0x08 }
enum class	TMVA::DNN::WeightInitializationStrategy { TMVA::DNN::XAVIER , TMVA::DNN::TEST , TMVA::DNN::LAYERSIZE , TMVA::DNN::XAVIERUNIFORM }
	weight initialization strategies to be chosen from More...

Functions
template<typename ItValue , typename ItFunction >
void	TMVA::DNN::applyFunctions (ItValue itValue, ItValue itValueEnd, ItFunction itFunction)
template<typename ItValue , typename ItFunction , typename ItInverseFunction , typename ItGradient >
void	TMVA::DNN::applyFunctions (ItValue itValue, ItValue itValueEnd, ItFunction itFunction, ItInverseFunction itInverseFunction, ItGradient itGradient)
template<typename ItSource , typename ItWeight , typename ItTarget >
void	TMVA::DNN::applyWeights (ItSource itSourceBegin, ItSource itSourceEnd, ItWeight itWeight, ItTarget itTargetBegin, ItTarget itTargetEnd)
template<typename ItSource , typename ItWeight , typename ItPrev >
void	TMVA::DNN::applyWeightsBackwards (ItSource itCurrBegin, ItSource itCurrEnd, ItWeight itWeight, ItPrev itPrevBegin, ItPrev itPrevEnd)
template<typename LAYERDATA >
void	TMVA::DNN::backward (LAYERDATA &prevLayerData, LAYERDATA &currLayerData)
	backward application of the weights (back-propagation of the error)
template<typename ItProbability , typename ItTruth , typename ItDelta , typename ItInvActFnc >
double	TMVA::DNN::crossEntropy (ItProbability itProbabilityBegin, ItProbability itProbabilityEnd, ItTruth itTruthBegin, ItTruth, ItDelta itDelta, ItDelta itDeltaEnd, ItInvActFnc, double patternWeight)
	cross entropy error function
template<typename LAYERDATA >
void	TMVA::DNN::forward (const LAYERDATA &prevLayerData, LAYERDATA &currLayerData)
	apply the weights (and functions) in forward direction of the DNN
double	TMVA::DNN::gaussDouble (double mean, double sigma)
template<typename T >
bool	TMVA::DNN::isFlagSet (T flag, T value)
ModeOutputValues	TMVA::DNN::operator& (ModeOutputValues lhs, ModeOutputValues rhs)
ModeOutputValues	TMVA::DNN::operator&= (ModeOutputValues &lhs, ModeOutputValues rhs)
ModeOutputValues	TMVA::DNN::operator| (ModeOutputValues lhs, ModeOutputValues rhs)
ModeOutputValues	TMVA::DNN::operator|= (ModeOutputValues &lhs, ModeOutputValues rhs)
int	TMVA::DNN::randomInt (int maxValue)
template<typename ItOutput , typename ItTruth , typename ItDelta , typename ItInvActFnc >
double	TMVA::DNN::softMaxCrossEntropy (ItOutput itProbabilityBegin, ItOutput itProbabilityEnd, ItTruth itTruthBegin, ItTruth, ItDelta itDelta, ItDelta itDeltaEnd, ItInvActFnc, double patternWeight)
	soft-max-cross-entropy error function (for mutual exclusive cross-entropy)
template<typename ItOutput , typename ItTruth , typename ItDelta , typename ItInvActFnc >
double	TMVA::DNN::sumOfSquares (ItOutput itOutputBegin, ItOutput itOutputEnd, ItTruth itTruthBegin, ItTruth itTruthEnd, ItDelta itDelta, ItDelta itDeltaEnd, ItInvActFnc itInvActFnc, double patternWeight)
double	TMVA::DNN::uniformDouble (double minValue, double maxValue)
template<typename LAYERDATA >
void	TMVA::DNN::update (const LAYERDATA &prevLayerData, LAYERDATA &currLayerData, double factorWeightDecay, EnumRegularization regularization)
	update the node values
template<typename ItSource , typename ItDelta , typename ItTargetGradient , typename ItGradient >
void	TMVA::DNN::update (ItSource itSource, ItSource itSourceEnd, ItDelta itTargetDeltaBegin, ItDelta itTargetDeltaEnd, ItTargetGradient itTargetGradientBegin, ItGradient itGradient)
	update the gradients
template<EnumRegularization Regularization, typename ItSource , typename ItDelta , typename ItTargetGradient , typename ItGradient , typename ItWeight >
void	TMVA::DNN::update (ItSource itSource, ItSource itSourceEnd, ItDelta itTargetDeltaBegin, ItDelta itTargetDeltaEnd, ItTargetGradient itTargetGradientBegin, ItGradient itGradient, ItWeight itWeight, double weightDecay)
	update the gradients, using regularization
template<typename ItWeight >
double	TMVA::DNN::weightDecay (double error, ItWeight itWeight, ItWeight itWeightEnd, double factorWeightDecay, EnumRegularization eRegularization)
	compute the weight decay for regularization (L1 or L2)

Detailed Description

Author

Peter Speckmayer

Version

1.0

LICENSE

net implementation

An implementation of a neural net for TMVA. This neural net uses multithreading

Definition in file NeuralNet.h.