#include <shark/Models/FFNet.h>
#include <shark/Models/Autoencoder.h>
#include <shark/Models/ImpulseNoiseModel.h>
#include <shark/Models/ConcatenatedModel.h>
#include <shark/ObjectiveFunctions/ErrorFunction.h>
#include <shark/ObjectiveFunctions/Loss/SquaredLoss.h>
#include <shark/ObjectiveFunctions/Loss/CrossEntropy.h>
#include <shark/ObjectiveFunctions/Loss/ZeroOneLoss.h>
#include <shark/ObjectiveFunctions/Regularizer.h>
#include <shark/Algorithms/GradientDescent/SteepestDescent.h>
#include <shark/Algorithms/GradientDescent/Rprop.h>

Typedefs
typedef Autoencoder< RectifierNeuron, LinearNeuron >	AutoencoderModel

typedef FFNet< RectifierNeuron, LinearNeuron >	Network

Functions
LabeledData< RealVector, unsigned int >	createProblem ()

template<class AutoencoderModel >
AutoencoderModel	trainAutoencoderModel (UnlabeledData< RealVector > const &data, std::size_t numHidden, double regularisation, double noiseStrength, std::size_t iterations)

Network	unsupervisedPreTraining (UnlabeledData< RealVector > const &data, std::size_t numHidden1, std::size_t numHidden2, std::size_t numOutputs, double regularisation, double noiseStrength, std::size_t iterations)

int	main ()

Typedef Documentation

§ AutoencoderModel

typedef Autoencoder<RectifierNeuron,LinearNeuron> AutoencoderModel

Definition at line 82 of file DeepNetworkTraining.cpp.

§ Network

typedef FFNet<RectifierNeuron,LinearNeuron> Network

Definition at line 83 of file DeepNetworkTraining.cpp.

Function Documentation

§ createProblem()

LabeledData<RealVector,unsigned int> createProblem ( )

Definition at line 20 of file DeepNetworkTraining.cpp.

References shark::coinToss(), shark::createLabeledDataFromRange(), and shark::blas::subrange().

Referenced by main().

§ main()

int main ( )

Definition at line 118 of file DeepNetworkTraining.cpp.

References createProblem(), shark::ZeroOneLoss< unsigned int, RealVector >::eval(), shark::IRpropPlusFull::init(), shark::LabeledData< InputT, LabelT >::inputs(), shark::LabeledData< InputT, LabelT >::labels(), shark::numberOfClasses(), shark::LabeledData< InputT, LabelT >::numberOfElements(), shark::ErrorFunction::numberOfVariables(), shark::ResultSet< SearchPointT, ResultT >::point, shark::FFNet< HiddenNeuron, OutputNeuron >::setParameterVector(), shark::ErrorFunction::setRegularizer(), shark::LabeledData< InputT, LabelT >::shuffle(), shark::AbstractSingleObjectiveOptimizer< PointType >::solution(), shark::splitAtElement(), shark::IRpropPlusFull::step(), unsupervisedPreTraining(), and shark::ResultSet< SearchPointT, ResultT >::value.

§ trainAutoencoderModel()

template<class AutoencoderModel >

AutoencoderModel trainAutoencoderModel	(	UnlabeledData< RealVector > const &	data,
		std::size_t	numHidden,
		double	regularisation,
		double	noiseStrength,
		std::size_t	iterations
	)

Definition at line 50 of file DeepNetworkTraining.cpp.

References shark::dataDimension(), shark::IRpropPlusFull::init(), shark::initRandomUniform(), shark::ConcatenatedModel< InputType, OutputType >::name(), noise, shark::ErrorFunction::numberOfVariables(), shark::ResultSet< SearchPointT, ResultT >::point, shark::ConcatenatedModel< InputType, OutputType >::setParameterVector(), shark::ErrorFunction::setRegularizer(), shark::Autoencoder< HiddenNeuron, OutputNeuron >::setStructure(), shark::AbstractSingleObjectiveOptimizer< PointType >::solution(), shark::IRpropPlusFull::step(), and shark::ResultSet< SearchPointT, ResultT >::value.

§ unsupervisedPreTraining()

Network unsupervisedPreTraining	(	UnlabeledData< RealVector > const &	data,
		std::size_t	numHidden1,
		std::size_t	numHidden2,
		std::size_t	numOutputs,
		double	regularisation,
		double	noiseStrength,
		std::size_t	iterations
	)

Definition at line 86 of file DeepNetworkTraining.cpp.

References shark::dataDimension(), shark::Autoencoder< HiddenNeuron, OutputNeuron >::encoderMatrix(), shark::Autoencoder< HiddenNeuron, OutputNeuron >::evalLayer(), shark::Autoencoder< HiddenNeuron, OutputNeuron >::hiddenBias(), shark::initRandomNormal(), and shark::FFNet< HiddenNeuron, OutputNeuron >::setStructure().