SHOGUN  4.2.0
 All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Friends Macros Modules Pages
List of all members | Public Member Functions | Public Attributes | Protected Member Functions | Protected Attributes
CAutoencoder Class Reference

Detailed Description

Represents a single layer neural autoencoder.

An autoencoder is a neural network that has three layers: an input layer, a hidden (encoding) layer, and a decoding layer. The network is trained to reconstruct its inputs, which forces the hidden layer to try to learn good representations of the inputs.

This class supports training normal autoencoders and denoising autoencoders [Vincent, 2008]. To use denoising autoencoders set noise_type and noise_parameter to specify the type and strength of the noise.

NOTE: LBFGS does not work properly with denoising autoencoders due to their stochastic nature. Use gradient descent instead.

Contractive autoencoders [Rifai, 2011] are also supported. To use them, call set_contraction_coefficient(). Denoising can also be used with contractive autoencoders through noise_type and noise_parameter.

Convolutional autoencoders [J Masci, 2011] are also supported. Simply build the autoencoder using CNeuralConvolutionalLayer objects.

NOTE: Contractive convolutional autoencoders are not supported.

Definition at line 86 of file Autoencoder.h.

Inheritance diagram for CAutoencoder:
[legend]

Public Member Functions

 CAutoencoder ()
 
 CAutoencoder (int32_t num_inputs, CNeuralLayer *hidden_layer, CNeuralLayer *decoding_layer=NULL, float64_t sigma=0.01)
 
 CAutoencoder (int32_t input_width, int32_t input_height, int32_t input_num_channels, CNeuralConvolutionalLayer *hidden_layer, CNeuralConvolutionalLayer *decoding_layer, float64_t sigma=0.01)
 
virtual bool train (CFeatures *data)
 
virtual CDenseFeatures
< float64_t > * 
transform (CDenseFeatures< float64_t > *data)
 
virtual CDenseFeatures
< float64_t > * 
reconstruct (CDenseFeatures< float64_t > *data)
 
virtual void set_contraction_coefficient (float64_t coeff)
 
virtual ~CAutoencoder ()
 
virtual const char * get_name () const
 
void set_noise_type (EAENoiseType noise_type)
 
EAENoiseType get_noise_type ()
 
void set_noise_parameter (float64_t noise_parameter)
 
float64_t get_noise_parameter ()
 
virtual void set_layers (CDynamicObjectArray *layers)
 
virtual void connect (int32_t i, int32_t j)
 
virtual void quick_connect ()
 
virtual void disconnect (int32_t i, int32_t j)
 
virtual void disconnect_all ()
 
virtual void initialize_neural_network (float64_t sigma=0.01f)
 
virtual CBinaryLabelsapply_binary (CFeatures *data)
 
virtual CRegressionLabelsapply_regression (CFeatures *data)
 
virtual CMulticlassLabelsapply_multiclass (CFeatures *data)
 
virtual void set_labels (CLabels *lab)
 
virtual EMachineType get_classifier_type ()
 
virtual EProblemType get_machine_problem_type () const
 
virtual float64_t check_gradients (float64_t approx_epsilon=1.0e-3, float64_t s=1.0e-9)
 
SGVector< float64_t > * get_layer_parameters (int32_t i)
 
int32_t get_num_parameters ()
 
SGVector< float64_tget_parameters ()
 
int32_t get_num_inputs ()
 
int32_t get_num_outputs ()
 
CDynamicObjectArrayget_layers ()
 
void set_optimization_method (ENNOptimizationMethod optimization_method)
 
ENNOptimizationMethod get_optimization_method () const
 
void set_l2_coefficient (float64_t l2_coefficient)
 
float64_t get_l2_coefficient () const
 
void set_l1_coefficient (float64_t l1_coefficient)
 
float64_t get_l1_coefficient () const
 
void set_dropout_hidden (float64_t dropout_hidden)
 
float64_t get_dropout_hidden () const
 
void set_dropout_input (float64_t dropout_input)
 
float64_t get_dropout_input () const
 
void set_max_norm (float64_t max_norm)
 
float64_t get_max_norm () const
 
void set_epsilon (float64_t epsilon)
 
float64_t get_epsilon () const
 
void set_max_num_epochs (int32_t max_num_epochs)
 
int32_t get_max_num_epochs () const
 
void set_gd_mini_batch_size (int32_t gd_mini_batch_size)
 
int32_t get_gd_mini_batch_size () const
 
void set_gd_learning_rate (float64_t gd_learning_rate)
 
float64_t get_gd_learning_rate () const
 
void set_gd_learning_rate_decay (float64_t gd_learning_rate_decay)
 
float64_t get_gd_learning_rate_decay () const
 
void set_gd_momentum (float64_t gd_momentum)
 
float64_t get_gd_momentum () const
 
void set_gd_error_damping_coeff (float64_t gd_error_damping_coeff)
 
float64_t get_gd_error_damping_coeff () const
 
virtual CLabelsapply (CFeatures *data=NULL)
 
virtual CStructuredLabelsapply_structured (CFeatures *data=NULL)
 
virtual CLatentLabelsapply_latent (CFeatures *data=NULL)
 
virtual CLabelsget_labels ()
 
void set_max_train_time (float64_t t)
 
float64_t get_max_train_time ()
 
void set_solver_type (ESolverType st)
 
ESolverType get_solver_type ()
 
virtual void set_store_model_features (bool store_model)
 
virtual bool train_locked (SGVector< index_t > indices)
 
virtual float64_t apply_one (int32_t i)
 
virtual CLabelsapply_locked (SGVector< index_t > indices)
 
virtual CBinaryLabelsapply_locked_binary (SGVector< index_t > indices)
 
virtual CRegressionLabelsapply_locked_regression (SGVector< index_t > indices)
 
virtual CMulticlassLabelsapply_locked_multiclass (SGVector< index_t > indices)
 
virtual CStructuredLabelsapply_locked_structured (SGVector< index_t > indices)
 
virtual CLatentLabelsapply_locked_latent (SGVector< index_t > indices)
 
virtual void data_lock (CLabels *labs, CFeatures *features)
 
virtual void post_lock (CLabels *labs, CFeatures *features)
 
virtual void data_unlock ()
 
virtual bool supports_locking () const
 
bool is_data_locked () const
 
virtual CSGObjectshallow_copy () const
 
virtual CSGObjectdeep_copy () const
 
virtual bool is_generic (EPrimitiveType *generic) const
 
template<class T >
void set_generic ()
 
template<>
void set_generic ()
 
template<>
void set_generic ()
 
template<>
void set_generic ()
 
template<>
void set_generic ()
 
template<>
void set_generic ()
 
template<>
void set_generic ()
 
template<>
void set_generic ()
 
template<>
void set_generic ()
 
template<>
void set_generic ()
 
template<>
void set_generic ()
 
template<>
void set_generic ()
 
template<>
void set_generic ()
 
template<>
void set_generic ()
 
template<>
void set_generic ()
 
template<>
void set_generic ()
 
void unset_generic ()
 
virtual void print_serializable (const char *prefix="")
 
virtual bool save_serializable (CSerializableFile *file, const char *prefix="")
 
virtual bool load_serializable (CSerializableFile *file, const char *prefix="")
 
void set_global_io (SGIO *io)
 
SGIOget_global_io ()
 
void set_global_parallel (Parallel *parallel)
 
Parallelget_global_parallel ()
 
void set_global_version (Version *version)
 
Versionget_global_version ()
 
SGStringList< char > get_modelsel_names ()
 
void print_modsel_params ()
 
char * get_modsel_param_descr (const char *param_name)
 
index_t get_modsel_param_index (const char *param_name)
 
void build_gradient_parameter_dictionary (CMap< TParameter *, CSGObject * > *dict)
 
virtual void update_parameter_hash ()
 
virtual bool parameter_hash_changed ()
 
virtual bool equals (CSGObject *other, float64_t accuracy=0.0, bool tolerant=false)
 
virtual CSGObjectclone ()
 

Public Attributes

SGIOio
 
Parallelparallel
 
Versionversion
 
Parameterm_parameters
 
Parameterm_model_selection_parameters
 
Parameterm_gradient_parameters
 
uint32_t m_hash
 

Protected Member Functions

virtual float64_t compute_error (SGMatrix< float64_t > targets)
 
virtual bool train_machine (CFeatures *data=NULL)
 
virtual bool train_gradient_descent (SGMatrix< float64_t > inputs, SGMatrix< float64_t > targets)
 
virtual bool train_lbfgs (SGMatrix< float64_t > inputs, SGMatrix< float64_t > targets)
 
virtual SGMatrix< float64_tforward_propagate (CFeatures *data, int32_t j=-1)
 
virtual SGMatrix< float64_tforward_propagate (SGMatrix< float64_t > inputs, int32_t j=-1)
 
virtual void set_batch_size (int32_t batch_size)
 
virtual float64_t compute_gradients (SGMatrix< float64_t > inputs, SGMatrix< float64_t > targets, SGVector< float64_t > gradients)
 
virtual float64_t compute_error (SGMatrix< float64_t > inputs, SGMatrix< float64_t > targets)
 
virtual bool is_label_valid (CLabels *lab) const
 
CNeuralLayerget_layer (int32_t i)
 
SGMatrix< float64_tfeatures_to_matrix (CFeatures *features)
 
SGMatrix< float64_tlabels_to_matrix (CLabels *labs)
 
virtual void store_model_features ()
 
virtual bool train_require_labels () const
 
virtual void load_serializable_pre () throw (ShogunException)
 
virtual void load_serializable_post () throw (ShogunException)
 
virtual void save_serializable_pre () throw (ShogunException)
 
virtual void save_serializable_post () throw (ShogunException)
 

Protected Attributes

float64_t m_contraction_coefficient
 
EAENoiseType m_noise_type
 
float64_t m_noise_parameter
 
int32_t m_num_inputs
 
int32_t m_num_layers
 
CDynamicObjectArraym_layers
 
SGMatrix< bool > m_adj_matrix
 
int32_t m_total_num_parameters
 
SGVector< float64_tm_params
 
SGVector< bool > m_param_regularizable
 
SGVector< int32_t > m_index_offsets
 
int32_t m_batch_size
 
bool m_is_training
 
ENNOptimizationMethod m_optimization_method
 
float64_t m_l2_coefficient
 
float64_t m_l1_coefficient
 
float64_t m_dropout_hidden
 
float64_t m_dropout_input
 
float64_t m_max_norm
 
float64_t m_epsilon
 
int32_t m_max_num_epochs
 
int32_t m_gd_mini_batch_size
 
float64_t m_gd_learning_rate
 
float64_t m_gd_learning_rate_decay
 
float64_t m_gd_momentum
 
float64_t m_gd_error_damping_coeff
 
float64_t m_max_train_time
 
CLabelsm_labels
 
ESolverType m_solver_type
 
bool m_store_model_features
 
bool m_data_locked
 

Constructor & Destructor Documentation

default constructor

Definition at line 43 of file Autoencoder.cpp.

CAutoencoder ( int32_t  num_inputs,
CNeuralLayer hidden_layer,
CNeuralLayer decoding_layer = NULL,
float64_t  sigma = 0.01 
)

Constructor

Parameters
num_inputsNumber of inputs
hidden_layerHidden layer. Can be any CNeuralLayer based object that supports being used as a hidden layer
decoding_layerDecoding layer. Must have the same number of neurons as num_inputs. Can be any CNeuralLayer based object that supports being used as an output layer. If NULL, a CNeuralLinearLayer is used.
sigmaStandard deviation of the gaussian used to initialize the parameters

Definition at line 48 of file Autoencoder.cpp.

CAutoencoder ( int32_t  input_width,
int32_t  input_height,
int32_t  input_num_channels,
CNeuralConvolutionalLayer hidden_layer,
CNeuralConvolutionalLayer decoding_layer,
float64_t  sigma = 0.01 
)

Constructor for convolutional autoencoders

Parameters
input_widthWidth of the input images
input_heightheight of the input images
input_num_channelsnumber of channels in the input images
hidden_layerHidden layer
decoding_layerDecoding layer. Should have the same dimensions as the inputs.
sigmaStandard deviation of the gaussian used to initialize the parameters

Definition at line 71 of file Autoencoder.cpp.

virtual ~CAutoencoder ( )
virtual

Definition at line 164 of file Autoencoder.h.

Member Function Documentation

CLabels * apply ( CFeatures data = NULL)
virtualinherited

apply machine to data if data is not specified apply to the current features

Parameters
data(test)data to be classified
Returns
classified labels

Definition at line 152 of file Machine.cpp.

CBinaryLabels * apply_binary ( CFeatures data)
virtualinherited

apply machine to data in means of binary classification problem

Reimplemented from CMachine.

Definition at line 158 of file NeuralNetwork.cpp.

CLatentLabels * apply_latent ( CFeatures data = NULL)
virtualinherited

apply machine to data in means of latent problem

Reimplemented in CLinearLatentMachine.

Definition at line 232 of file Machine.cpp.

CLabels * apply_locked ( SGVector< index_t indices)
virtualinherited

Applies a locked machine on a set of indices. Error if machine is not locked

Parameters
indicesindex vector (of locked features) that is predicted

Definition at line 187 of file Machine.cpp.

CBinaryLabels * apply_locked_binary ( SGVector< index_t indices)
virtualinherited

applies a locked machine on a set of indices for binary problems

Reimplemented in CKernelMachine.

Definition at line 238 of file Machine.cpp.

CLatentLabels * apply_locked_latent ( SGVector< index_t indices)
virtualinherited

applies a locked machine on a set of indices for latent problems

Definition at line 266 of file Machine.cpp.

CMulticlassLabels * apply_locked_multiclass ( SGVector< index_t indices)
virtualinherited

applies a locked machine on a set of indices for multiclass problems

Definition at line 252 of file Machine.cpp.

CRegressionLabels * apply_locked_regression ( SGVector< index_t indices)
virtualinherited

applies a locked machine on a set of indices for regression problems

Reimplemented in CKernelMachine.

Definition at line 245 of file Machine.cpp.

CStructuredLabels * apply_locked_structured ( SGVector< index_t indices)
virtualinherited

applies a locked machine on a set of indices for structured problems

Definition at line 259 of file Machine.cpp.

CMulticlassLabels * apply_multiclass ( CFeatures data)
virtualinherited

apply machine to data in means of multiclass classification problem

Reimplemented from CMachine.

Definition at line 199 of file NeuralNetwork.cpp.

virtual float64_t apply_one ( int32_t  i)
virtualinherited
CRegressionLabels * apply_regression ( CFeatures data)
virtualinherited

apply machine to data in means of regression problem

Reimplemented from CMachine.

Definition at line 187 of file NeuralNetwork.cpp.

CStructuredLabels * apply_structured ( CFeatures data = NULL)
virtualinherited

apply machine to data in means of SO classification problem

Reimplemented in CLinearStructuredOutputMachine.

Definition at line 226 of file Machine.cpp.

void build_gradient_parameter_dictionary ( CMap< TParameter *, CSGObject * > *  dict)
inherited

Builds a dictionary of all parameters in SGObject as well of those of SGObjects that are parameters of this object. Dictionary maps parameters to the objects that own them.

Parameters
dictdictionary of parameters to be built.

Definition at line 597 of file SGObject.cpp.

float64_t check_gradients ( float64_t  approx_epsilon = 1.0e-3,
float64_t  s = 1.0e-9 
)
virtualinherited

Checks if the gradients computed using backpropagation are correct by comparing them with gradients computed using numerical approximation. Used for testing purposes only.

Gradients are numerically approximated according to:

\[ c = max(\epsilon x, s) \]

\[ f'(x) = \frac{f(x + c)-f(x - c)}{2c} \]

Parameters
approx_epsilonConstant used during gradient approximation
sSome small value, used to prevent division by zero
Returns
Average difference between numerical gradients and backpropagation gradients

Definition at line 554 of file NeuralNetwork.cpp.

CSGObject * clone ( )
virtualinherited

Creates a clone of the current object. This is done via recursively traversing all parameters, which corresponds to a deep copy. Calling equals on the cloned object always returns true although none of the memory of both objects overlaps.

Returns
an identical copy of the given object, which is disjoint in memory. NULL if the clone fails. Note that the returned object is SG_REF'ed

Definition at line 714 of file SGObject.cpp.

float64_t compute_error ( SGMatrix< float64_t targets)
protectedvirtual

Computes the error between the output layer's activations and the given target activations.

Parameters
targetsdesired values for the network's output, matrix of size num_neurons_output_layer*batch_size

Reimplemented from CNeuralNetwork.

Reimplemented in CDeepAutoencoder.

Definition at line 154 of file Autoencoder.cpp.

float64_t compute_error ( SGMatrix< float64_t inputs,
SGMatrix< float64_t targets 
)
protectedvirtualinherited

Forward propagates the inputs and computes the error between the output layer's activations and the given target activations.

Parameters
inputsinputs to the network, a matrix of size m_num_inputs*m_batch_size
targetsdesired values for the network's output, matrix of size num_neurons_output_layer*batch_size

Definition at line 546 of file NeuralNetwork.cpp.

float64_t compute_gradients ( SGMatrix< float64_t inputs,
SGMatrix< float64_t targets,
SGVector< float64_t gradients 
)
protectedvirtualinherited

Applies backpropagation to compute the gradients of the error with repsect to every parameter in the network.

Parameters
inputsinputs to the network, a matrix of size m_num_inputs*m_batch_size
targetsdesired values for the output layer's activations. matrix of size m_layers[m_num_layers-1].get_num_neurons()*m_batch_size
gradientsarray to be filled with gradient values.
Returns
error between the targets and the activations of the last layer

Definition at line 467 of file NeuralNetwork.cpp.

void connect ( int32_t  i,
int32_t  j 
)
virtualinherited

Connects layer i as input to layer j. In order for forward and backpropagation to work correctly, i must be less that j

Definition at line 75 of file NeuralNetwork.cpp.

void data_lock ( CLabels labs,
CFeatures features 
)
virtualinherited

Locks the machine on given labels and data. After this call, only train_locked and apply_locked may be called

Only possible if supports_locking() returns true

Parameters
labslabels used for locking
featuresfeatures used for locking

Reimplemented in CKernelMachine.

Definition at line 112 of file Machine.cpp.

void data_unlock ( )
virtualinherited

Unlocks a locked machine and restores previous state

Reimplemented in CKernelMachine.

Definition at line 143 of file Machine.cpp.

CSGObject * deep_copy ( ) const
virtualinherited

A deep copy. All the instance variables will also be copied.

Definition at line 198 of file SGObject.cpp.

void disconnect ( int32_t  i,
int32_t  j 
)
virtualinherited

Disconnects layer i from layer j

Definition at line 88 of file NeuralNetwork.cpp.

void disconnect_all ( )
virtualinherited

Removes all connections in the network

Definition at line 93 of file NeuralNetwork.cpp.

bool equals ( CSGObject other,
float64_t  accuracy = 0.0,
bool  tolerant = false 
)
virtualinherited

Recursively compares the current SGObject to another one. Compares all registered numerical parameters, recursion upon complex (SGObject) parameters. Does not compare pointers!

May be overwritten but please do with care! Should not be necessary in most cases.

Parameters
otherobject to compare with
accuracyaccuracy to use for comparison (optional)
tolerantallows linient check on float equality (within accuracy)
Returns
true if all parameters were equal, false if not

Definition at line 618 of file SGObject.cpp.

SGMatrix< float64_t > features_to_matrix ( CFeatures features)
protectedinherited

Ensures the given features are suitable for use with the network and returns their feature matrix

Definition at line 614 of file NeuralNetwork.cpp.

SGMatrix< float64_t > forward_propagate ( CFeatures data,
int32_t  j = -1 
)
protectedvirtualinherited

Applies forward propagation, computes the activations of each layer up to layer j

Parameters
datainput features
jlayer index at which the propagation should stop. If -1, the propagation continues up to the last layer
Returns
activations of the last layer

Definition at line 439 of file NeuralNetwork.cpp.

SGMatrix< float64_t > forward_propagate ( SGMatrix< float64_t inputs,
int32_t  j = -1 
)
protectedvirtualinherited

Applies forward propagation, computes the activations of each layer up to layer j

Parameters
inputsinputs to the network, a matrix of size m_num_inputs*m_batch_size
jlayer index at which the propagation should stop. If -1, the propagation continues up to the last layer
Returns
activations of the last layer

Definition at line 446 of file NeuralNetwork.cpp.

virtual EMachineType get_classifier_type ( )
virtualinherited

get classifier type

Returns
classifier type CT_NEURALNETWORK

Reimplemented from CMachine.

Definition at line 188 of file NeuralNetwork.h.

float64_t get_dropout_hidden ( ) const
inherited

Returns dropout probability for hidden layers

Definition at line 292 of file NeuralNetwork.h.

float64_t get_dropout_input ( ) const
inherited

Returns dropout probability for input layers

Definition at line 312 of file NeuralNetwork.h.

float64_t get_epsilon ( ) const
inherited

Returns epsilon

Definition at line 346 of file NeuralNetwork.h.

float64_t get_gd_error_damping_coeff ( ) const
inherited

Definition at line 454 of file NeuralNetwork.h.

float64_t get_gd_learning_rate ( ) const
inherited

Returns gradient descent learning rate

Definition at line 393 of file NeuralNetwork.h.

float64_t get_gd_learning_rate_decay ( ) const
inherited

Returns gradient descent learning rate decay

Definition at line 410 of file NeuralNetwork.h.

int32_t get_gd_mini_batch_size ( ) const
inherited

Returns mini batch size

Definition at line 378 of file NeuralNetwork.h.

float64_t get_gd_momentum ( ) const
inherited

Returns gradient descent momentum multiplier

Definition at line 431 of file NeuralNetwork.h.

SGIO * get_global_io ( )
inherited

get the io object

Returns
io object

Definition at line 235 of file SGObject.cpp.

Parallel * get_global_parallel ( )
inherited

get the parallel object

Returns
parallel object

Definition at line 277 of file SGObject.cpp.

Version * get_global_version ( )
inherited

get the version object

Returns
version object

Definition at line 290 of file SGObject.cpp.

float64_t get_l1_coefficient ( ) const
inherited

Returns L1 coefficient

Definition at line 272 of file NeuralNetwork.h.

float64_t get_l2_coefficient ( ) const
inherited

Returns L2 coefficient

Definition at line 258 of file NeuralNetwork.h.

CLabels * get_labels ( )
virtualinherited

get labels

Returns
labels

Definition at line 76 of file Machine.cpp.

CNeuralLayer * get_layer ( int32_t  i)
protectedinherited

returns a pointer to layer i in the network

Definition at line 723 of file NeuralNetwork.cpp.

SGVector< float64_t > * get_layer_parameters ( int32_t  i)
inherited

returns a copy of a layer's parameters array

Parameters
iindex of the layer

Definition at line 712 of file NeuralNetwork.cpp.

CDynamicObjectArray * get_layers ( )
inherited

Returns an array holding the network's layers

Definition at line 744 of file NeuralNetwork.cpp.

EProblemType get_machine_problem_type ( ) const
virtualinherited

returns type of problem machine solves

Reimplemented from CMachine.

Definition at line 675 of file NeuralNetwork.cpp.

float64_t get_max_norm ( ) const
inherited

Returns maximum allowable L2 norm

Definition at line 328 of file NeuralNetwork.h.

int32_t get_max_num_epochs ( ) const
inherited

Returns maximum number of epochs

Definition at line 362 of file NeuralNetwork.h.

float64_t get_max_train_time ( )
inherited

get maximum training time

Returns
maximum training time

Definition at line 87 of file Machine.cpp.

SGStringList< char > get_modelsel_names ( )
inherited
Returns
vector of names of all parameters which are registered for model selection

Definition at line 498 of file SGObject.cpp.

char * get_modsel_param_descr ( const char *  param_name)
inherited

Returns description of a given parameter string, if it exists. SG_ERROR otherwise

Parameters
param_namename of the parameter
Returns
description of the parameter

Definition at line 522 of file SGObject.cpp.

index_t get_modsel_param_index ( const char *  param_name)
inherited

Returns index of model selection parameter with provided index

Parameters
param_namename of model selection parameter
Returns
index of model selection parameter with provided name, -1 if there is no such

Definition at line 535 of file SGObject.cpp.

virtual const char* get_name ( ) const
virtual

Returns the name of the SGSerializable instance. It MUST BE the CLASS NAME without the prefixed `C'.

Returns
name of the SGSerializable

Reimplemented from CNeuralNetwork.

Reimplemented in CDeepAutoencoder.

Definition at line 166 of file Autoencoder.h.

float64_t get_noise_parameter ( )

Returns noise parameter

Definition at line 201 of file Autoencoder.h.

EAENoiseType get_noise_type ( )

Returns noise type for denoising autoencoders

Definition at line 185 of file Autoencoder.h.

int32_t get_num_inputs ( )
inherited

returns the number of inputs the network takes

Definition at line 224 of file NeuralNetwork.h.

int32_t get_num_outputs ( )
inherited

returns the number of neurons in the output layer

Definition at line 739 of file NeuralNetwork.cpp.

int32_t get_num_parameters ( )
inherited

returns the totat number of parameters in the network

Definition at line 218 of file NeuralNetwork.h.

ENNOptimizationMethod get_optimization_method ( ) const
inherited

Returns optimization method

Definition at line 244 of file NeuralNetwork.h.

SGVector<float64_t> get_parameters ( )
inherited

return the network's parameter array

Definition at line 221 of file NeuralNetwork.h.

ESolverType get_solver_type ( )
inherited

get solver type

Returns
solver

Definition at line 102 of file Machine.cpp.

void initialize_neural_network ( float64_t  sigma = 0.01f)
virtualinherited

Initializes the network

Parameters
sigmastandard deviation of the gaussian used to randomly initialize the parameters

Definition at line 98 of file NeuralNetwork.cpp.

bool is_data_locked ( ) const
inherited
Returns
whether this machine is locked

Definition at line 296 of file Machine.h.

bool is_generic ( EPrimitiveType *  generic) const
virtualinherited

If the SGSerializable is a class template then TRUE will be returned and GENERIC is set to the type of the generic.

Parameters
genericset to the type of the generic if returning TRUE
Returns
TRUE if a class template.

Definition at line 296 of file SGObject.cpp.

bool is_label_valid ( CLabels lab) const
protectedvirtualinherited

check whether the labels is valid.

Subclasses can override this to implement their check of label types.

Parameters
labthe labels being checked, guaranteed to be non-NULL

Reimplemented from CMachine.

Definition at line 689 of file NeuralNetwork.cpp.

SGMatrix< float64_t > labels_to_matrix ( CLabels labs)
protectedinherited

converts the given labels into a matrix suitable for use with network

Returns
matrix of size get_num_outputs()*num_labels

Definition at line 630 of file NeuralNetwork.cpp.

bool load_serializable ( CSerializableFile file,
const char *  prefix = "" 
)
virtualinherited

Load this object from file. If it will fail (returning FALSE) then this object will contain inconsistent data and should not be used!

Parameters
filewhere to load from
prefixprefix for members
Returns
TRUE if done, otherwise FALSE

Definition at line 369 of file SGObject.cpp.

void load_serializable_post ( )
throw (ShogunException
)
protectedvirtualinherited

Can (optionally) be overridden to post-initialize some member variables which are not PARAMETER::ADD'ed. Make sure that at first the overridden method BASE_CLASS::LOAD_SERIALIZABLE_POST is called.

Exceptions
ShogunExceptionwill be thrown if an error occurs.

Reimplemented in CKernel, CWeightedDegreePositionStringKernel, CList, CAlphabet, CLinearHMM, CGaussianKernel, CInverseMultiQuadricKernel, CCircularKernel, and CExponentialKernel.

Definition at line 426 of file SGObject.cpp.

void load_serializable_pre ( )
throw (ShogunException
)
protectedvirtualinherited

Can (optionally) be overridden to pre-initialize some member variables which are not PARAMETER::ADD'ed. Make sure that at first the overridden method BASE_CLASS::LOAD_SERIALIZABLE_PRE is called.

Exceptions
ShogunExceptionwill be thrown if an error occurs.

Reimplemented in CDynamicArray< T >, CDynamicArray< float64_t >, CDynamicArray< float32_t >, CDynamicArray< int32_t >, CDynamicArray< char >, CDynamicArray< bool >, and CDynamicObjectArray.

Definition at line 421 of file SGObject.cpp.

bool parameter_hash_changed ( )
virtualinherited
Returns
whether parameter combination has changed since last update

Definition at line 262 of file SGObject.cpp.

virtual void post_lock ( CLabels labs,
CFeatures features 
)
virtualinherited

post lock

Definition at line 287 of file Machine.h.

void print_modsel_params ( )
inherited

prints all parameter registered for model selection and their type

Definition at line 474 of file SGObject.cpp.

void print_serializable ( const char *  prefix = "")
virtualinherited

prints registered parameters out

Parameters
prefixprefix for members

Definition at line 308 of file SGObject.cpp.

void quick_connect ( )
virtualinherited

Connects each layer to the layer after it. That is, connects layer i to as input to layer i+1 for all i.

Definition at line 81 of file NeuralNetwork.cpp.

CDenseFeatures< float64_t > * reconstruct ( CDenseFeatures< float64_t > *  data)
virtual

Reconstructs the input data

Parameters
dataInput features
Returns
Reconstructed features

Reimplemented in CDeepAutoencoder.

Definition at line 147 of file Autoencoder.cpp.

bool save_serializable ( CSerializableFile file,
const char *  prefix = "" 
)
virtualinherited

Save this object to file.

Parameters
filewhere to save the object; will be closed during returning if PREFIX is an empty string.
prefixprefix for members
Returns
TRUE if done, otherwise FALSE

Definition at line 314 of file SGObject.cpp.

void save_serializable_post ( )
throw (ShogunException
)
protectedvirtualinherited

Can (optionally) be overridden to post-initialize some member variables which are not PARAMETER::ADD'ed. Make sure that at first the overridden method BASE_CLASS::SAVE_SERIALIZABLE_POST is called.

Exceptions
ShogunExceptionwill be thrown if an error occurs.

Reimplemented in CKernel.

Definition at line 436 of file SGObject.cpp.

void save_serializable_pre ( )
throw (ShogunException
)
protectedvirtualinherited

Can (optionally) be overridden to pre-initialize some member variables which are not PARAMETER::ADD'ed. Make sure that at first the overridden method BASE_CLASS::SAVE_SERIALIZABLE_PRE is called.

Exceptions
ShogunExceptionwill be thrown if an error occurs.

Reimplemented in CKernel, CDynamicArray< T >, CDynamicArray< float64_t >, CDynamicArray< float32_t >, CDynamicArray< int32_t >, CDynamicArray< char >, CDynamicArray< bool >, and CDynamicObjectArray.

Definition at line 431 of file SGObject.cpp.

void set_batch_size ( int32_t  batch_size)
protectedvirtualinherited

Sets the batch size (the number of train/test cases) the network is expected to deal with. Allocates memory for the activations, local gradients, input gradients if necessary (if the batch size is different from it's previous value)

Parameters
batch_sizenumber of train/test cases the network is expected to deal with.

Definition at line 604 of file NeuralNetwork.cpp.

virtual void set_contraction_coefficient ( float64_t  coeff)
virtual

Sets the contraction coefficient

For contractive autoencoders [Rifai, 2011], a term:

\[ \frac{\lambda}{N} \sum_{k=0}^{N-1} \left \| J(x_k) \right \|^2_F \]

is added to the error, where \( \left \| J(x_k)) \right \|^2_F \) is the Frobenius norm of the Jacobian of the activations of the hidden layer with respect to its inputs, \( N \) is the batch size, and \( \lambda \) is the contraction coefficient.

Parameters
coeffContraction coefficient

Reimplemented in CDeepAutoencoder.

Definition at line 158 of file Autoencoder.h.

void set_dropout_hidden ( float64_t  dropout_hidden)
inherited

Sets the probabilty that a hidden layer neuron will be dropped out When using this, the recommended value is 0.5 default value 0.0 (no dropout)

For more details on dropout, see paper [Hinton, 2012]

Parameters
dropout_hiddendropout probability

Definition at line 286 of file NeuralNetwork.h.

void set_dropout_input ( float64_t  dropout_input)
inherited

Sets the probabilty that an input layer neuron will be dropped out When using this, a good value might be 0.2 default value 0.0 (no dropout)

For more details on dropout, see this paper [Hinton, 2012]

Parameters
dropout_inputdropout probability

Definition at line 306 of file NeuralNetwork.h.

void set_epsilon ( float64_t  epsilon)
inherited

Sets convergence criteria training stops when (E'- E)/E < epsilon where E is the error at the current iterations and E' is the error at the previous iteration default value is 1.0e-5

Parameters
epsilonconvergence criteria

Definition at line 340 of file NeuralNetwork.h.

void set_gd_error_damping_coeff ( float64_t  gd_error_damping_coeff)
inherited

Sets gradient descent error damping coefficient Used to damp the error fluctuations when stochastic gradient descent is used. damping is done according to: error_damped(i) = c*error(i) + (1-c)*error_damped(i-1) where c is the damping coefficient

If -1, the damping coefficient is automatically computed according to: c = 0.99*gd_mini_batch_size/training_set_size + 1e-2;

default value is -1

Parameters
gd_error_damping_coefferror damping coefficient

Definition at line 449 of file NeuralNetwork.h.

void set_gd_learning_rate ( float64_t  gd_learning_rate)
inherited

Sets gradient descent learning rate defualt value 0.1

Parameters
gd_learning_rategradient descent learning rate

Definition at line 387 of file NeuralNetwork.h.

void set_gd_learning_rate_decay ( float64_t  gd_learning_rate_decay)
inherited

Sets gradient descent learning rate decay learning rate is updated at each iteration i according to: alpha(i)=decay*alpha(i-1) default value is 1.0 (no decay)

Parameters
gd_learning_rate_decaygradient descent learning rate decay

Definition at line 404 of file NeuralNetwork.h.

void set_gd_mini_batch_size ( int32_t  gd_mini_batch_size)
inherited

Sets size of the mini-batch used during gradient descent training, if 0 full-batch training is performed default value is 0

Parameters
gd_mini_batch_sizemini batch size

Definition at line 372 of file NeuralNetwork.h.

void set_gd_momentum ( float64_t  gd_momentum)
inherited

Sets gradient descent momentum multiplier

default value is 0.9

For more details on momentum, see this paper [Sutskever, 2013]

Parameters
gd_momentumgradient descent momentum multiplier

Definition at line 425 of file NeuralNetwork.h.

void set_generic ( )
inherited

Definition at line 41 of file SGObject.cpp.

void set_generic ( )
inherited

Definition at line 46 of file SGObject.cpp.

void set_generic ( )
inherited

Definition at line 51 of file SGObject.cpp.

void set_generic ( )
inherited

Definition at line 56 of file SGObject.cpp.

void set_generic ( )
inherited

Definition at line 61 of file SGObject.cpp.

void set_generic ( )
inherited

Definition at line 66 of file SGObject.cpp.

void set_generic ( )
inherited

Definition at line 71 of file SGObject.cpp.

void set_generic ( )
inherited

Definition at line 76 of file SGObject.cpp.

void set_generic ( )
inherited

Definition at line 81 of file SGObject.cpp.

void set_generic ( )
inherited

Definition at line 86 of file SGObject.cpp.

void set_generic ( )
inherited

Definition at line 91 of file SGObject.cpp.

void set_generic ( )
inherited

Definition at line 96 of file SGObject.cpp.

void set_generic ( )
inherited

Definition at line 101 of file SGObject.cpp.

void set_generic ( )
inherited

Definition at line 106 of file SGObject.cpp.

void set_generic ( )
inherited

Definition at line 111 of file SGObject.cpp.

void set_generic ( )
inherited

set generic type to T

void set_global_io ( SGIO io)
inherited

set the io object

Parameters
ioio object to use

Definition at line 228 of file SGObject.cpp.

void set_global_parallel ( Parallel parallel)
inherited

set the parallel object

Parameters
parallelparallel object to use

Definition at line 241 of file SGObject.cpp.

void set_global_version ( Version version)
inherited

set the version object

Parameters
versionversion object to use

Definition at line 283 of file SGObject.cpp.

void set_l1_coefficient ( float64_t  l1_coefficient)
inherited

Sets L1 Regularization coeff default value is 0.0

Parameters
l1_coefficientl1_coefficient

Definition at line 266 of file NeuralNetwork.h.

void set_l2_coefficient ( float64_t  l2_coefficient)
inherited

Sets L2 Regularization coeff default value is 0.0

Parameters
l2_coefficientl2_coefficient

Definition at line 252 of file NeuralNetwork.h.

void set_labels ( CLabels lab)
virtualinherited

set labels

Parameters
lablabels

Reimplemented from CMachine.

Definition at line 696 of file NeuralNetwork.cpp.

void set_layers ( CDynamicObjectArray layers)
virtualinherited

Sets the layers of the network

Parameters
layersAn array of CNeuralLayer objects specifying the layers of the network. Must contain at least one input layer. The last layer in the array is treated as the output layer

Definition at line 55 of file NeuralNetwork.cpp.

void set_max_norm ( float64_t  max_norm)
inherited

Sets maximum allowable L2 norm for a neurons weights When using this, a good value might be 15 default value -1 (max-norm regularization disabled)

Parameters
max_normmaximum allowable L2 norm

Definition at line 322 of file NeuralNetwork.h.

void set_max_num_epochs ( int32_t  max_num_epochs)
inherited

Sets maximum number of iterations over the training set. If 0, training will continue until convergence. defualt value is 0

Parameters
max_num_epochsmaximum number of iterations over the training set

Definition at line 356 of file NeuralNetwork.h.

void set_max_train_time ( float64_t  t)
inherited

set maximum training time

Parameters
tmaximimum training time

Definition at line 82 of file Machine.cpp.

void set_noise_parameter ( float64_t  noise_parameter)

Sets noise parameter Controls the strength of the noise, depending on noise_type

Parameters
noise_parametercontrols the strength of noise

Definition at line 195 of file Autoencoder.h.

void set_noise_type ( EAENoiseType  noise_type)

Sets noise type for denoising autoencoders.

If set to AENT_DROPOUT, inputs are randomly set to zero during each iteration of training with probability noise_parameter.

If set to AENT_GAUSSIAN, gaussian noise with zero mean and noise_parameter standard deviation is added to the inputs.

Default value is AENT_NONE

Parameters
noise_typenoise type for denoising autoencoders

Definition at line 179 of file Autoencoder.h.

void set_optimization_method ( ENNOptimizationMethod  optimization_method)
inherited

Sets optimization method default is NNOM_LBFGS

Parameters
optimization_methodoptimiation method

Definition at line 238 of file NeuralNetwork.h.

void set_solver_type ( ESolverType  st)
inherited

set solver type

Parameters
stsolver type

Definition at line 97 of file Machine.cpp.

void set_store_model_features ( bool  store_model)
virtualinherited

Setter for store-model-features-after-training flag

Parameters
store_modelwhether model should be stored after training

Definition at line 107 of file Machine.cpp.

CSGObject * shallow_copy ( ) const
virtualinherited

A shallow copy. All the SGObject instance variables will be simply assigned and SG_REF-ed.

Reimplemented in CGaussianKernel.

Definition at line 192 of file SGObject.cpp.

virtual void store_model_features ( )
protectedvirtualinherited

Stores feature data of underlying model. After this method has been called, it is possible to change the machine's feature data and call apply(), which is then performed on the training feature data that is part of the machine's model.

Base method, has to be implemented in order to allow cross-validation and model selection.

NOT IMPLEMENTED! Has to be done in subclasses

Reimplemented in CKernelMachine, CKNN, CLinearMachine, CLinearMulticlassMachine, CKMeansBase, CTreeMachine< T >, CTreeMachine< ConditionalProbabilityTreeNodeData >, CTreeMachine< RelaxedTreeNodeData >, CTreeMachine< id3TreeNodeData >, CTreeMachine< VwConditionalProbabilityTreeNodeData >, CTreeMachine< CARTreeNodeData >, CTreeMachine< C45TreeNodeData >, CTreeMachine< CHAIDTreeNodeData >, CTreeMachine< NbodyTreeNodeData >, CGaussianProcessMachine, CHierarchical, CDistanceMachine, CKernelMulticlassMachine, and CLinearStructuredOutputMachine.

Definition at line 335 of file Machine.h.

virtual bool supports_locking ( ) const
virtualinherited
Returns
whether this machine supports locking

Reimplemented in CKernelMachine.

Definition at line 293 of file Machine.h.

bool train ( CFeatures data)
virtual

Trains the autoencoder

Parameters
dataTraining examples
Returns
True if training succeeded, false otherwise

Reimplemented from CMachine.

Definition at line 96 of file Autoencoder.cpp.

bool train_gradient_descent ( SGMatrix< float64_t inputs,
SGMatrix< float64_t targets 
)
protectedvirtualinherited

trains the network using gradient descent

Definition at line 261 of file NeuralNetwork.cpp.

bool train_lbfgs ( SGMatrix< float64_t inputs,
SGMatrix< float64_t targets 
)
protectedvirtualinherited

trains the network using L-BFGS

Definition at line 357 of file NeuralNetwork.cpp.

virtual bool train_locked ( SGVector< index_t indices)
virtualinherited

Trains a locked machine on a set of indices. Error if machine is not locked

NOT IMPLEMENTED

Parameters
indicesindex vector (of locked features) that is used for training
Returns
whether training was successful

Reimplemented in CKernelMachine.

Definition at line 239 of file Machine.h.

bool train_machine ( CFeatures data = NULL)
protectedvirtualinherited

trains the network

Reimplemented from CMachine.

Definition at line 229 of file NeuralNetwork.cpp.

virtual bool train_require_labels ( ) const
protectedvirtualinherited

returns whether machine require labels for training

Reimplemented in COnlineLinearMachine, CKMeansBase, CHierarchical, CLinearLatentMachine, CVwConditionalProbabilityTree, CConditionalProbabilityTree, and CLibSVMOneClass.

Definition at line 354 of file Machine.h.

CDenseFeatures< float64_t > * transform ( CDenseFeatures< float64_t > *  data)
virtual

Computes the activation of the hidden layer given the input data

Parameters
dataInput features
Returns
Transformed features

Reimplemented from CNeuralNetwork.

Reimplemented in CDeepAutoencoder.

Definition at line 140 of file Autoencoder.cpp.

void unset_generic ( )
inherited

unset generic type

this has to be called in classes specializing a template class

Definition at line 303 of file SGObject.cpp.

void update_parameter_hash ( )
virtualinherited

Updates the hash of current parameter combination

Definition at line 248 of file SGObject.cpp.

Member Data Documentation

SGIO* io
inherited

io

Definition at line 369 of file SGObject.h.

SGMatrix<bool> m_adj_matrix
protectedinherited

Describes the connections in the network: if there's a connection from layer i to layer j then m_adj_matrix(i,j) = 1.

Definition at line 596 of file NeuralNetwork.h.

int32_t m_batch_size
protectedinherited

number of train/test cases the network is expected to deal with. Default value is 1

Definition at line 618 of file NeuralNetwork.h.

float64_t m_contraction_coefficient
protected

For contractive autoencoders [Rifai, 2011], a term:

\[ \frac{\lambda}{N} \sum_{k=0}^{N-1} \left \| J(x_k) \right \|^2_F \]

is added to the error, where \( \left \| J(x_k)) \right \|^2_F \) is the Frobenius norm of the Jacobian of the activations of the hidden layer with respect to its inputs, \( N \) is the batch size, and \( \lambda \) is the contraction coefficient.

Default value is 0.0.

Definition at line 232 of file Autoencoder.h.

bool m_data_locked
protectedinherited

whether data is locked

Definition at line 370 of file Machine.h.

float64_t m_dropout_hidden
protectedinherited

Probabilty that a hidden layer neuron will be dropped out When using this, the recommended value is 0.5

default value 0.0 (no dropout)

For more details on dropout, see paper [Hinton, 2012]

Definition at line 642 of file NeuralNetwork.h.

float64_t m_dropout_input
protectedinherited

Probabilty that a input layer neuron will be dropped out When using this, a good value might be 0.2

default value 0.0 (no dropout)

For more details on dropout, see this paper [Hinton, 2012]

Definition at line 652 of file NeuralNetwork.h.

float64_t m_epsilon
protectedinherited

convergence criteria training stops when (E'- E)/E < epsilon where E is the error at the current iterations and E' is the error at the previous iteration default value is 1.0e-5

Definition at line 667 of file NeuralNetwork.h.

float64_t m_gd_error_damping_coeff
protectedinherited

Used to damp the error fluctuations when stochastic gradient descent is used. damping is done according to: error_damped(i) = c*error(i) + (1-c)*error_damped(i-1) where c is the damping coefficient

If -1, the damping coefficient is automatically computed according to: c = 0.99*gd_mini_batch_size/training_set_size + 1e-2;

default value is -1

Definition at line 711 of file NeuralNetwork.h.

float64_t m_gd_learning_rate
protectedinherited

gradient descent learning rate, defualt value 0.1

Definition at line 682 of file NeuralNetwork.h.

float64_t m_gd_learning_rate_decay
protectedinherited

gradient descent learning rate decay learning rate is updated at each iteration i according to: alpha(i)=decay*alpha(i-1) default value is 1.0 (no decay)

Definition at line 689 of file NeuralNetwork.h.

int32_t m_gd_mini_batch_size
protectedinherited

size of the mini-batch used during gradient descent training, if 0 full-batch training is performed default value is 0

Definition at line 679 of file NeuralNetwork.h.

float64_t m_gd_momentum
protectedinherited

gradient descent momentum multiplier

default value is 0.9

For more details on momentum, see this paper [Sutskever, 2013]

Definition at line 699 of file NeuralNetwork.h.

Parameter* m_gradient_parameters
inherited

parameters wrt which we can compute gradients

Definition at line 384 of file SGObject.h.

uint32_t m_hash
inherited

Hash of parameter values

Definition at line 387 of file SGObject.h.

SGVector<int32_t> m_index_offsets
protectedinherited

offsets specifying where each layer's parameters and parameter gradients are stored, i.e layer i's parameters are stored at m_params + m_index_offsets[i]

Definition at line 613 of file NeuralNetwork.h.

bool m_is_training
protectedinherited

True if the network is currently being trained initial value is false

Definition at line 623 of file NeuralNetwork.h.

float64_t m_l1_coefficient
protectedinherited

L1 Regularization coeff, default value is 0.0

Definition at line 632 of file NeuralNetwork.h.

float64_t m_l2_coefficient
protectedinherited

L2 Regularization coeff, default value is 0.0

Definition at line 629 of file NeuralNetwork.h.

CLabels* m_labels
protectedinherited

labels

Definition at line 361 of file Machine.h.

CDynamicObjectArray* m_layers
protectedinherited

network's layers

Definition at line 591 of file NeuralNetwork.h.

float64_t m_max_norm
protectedinherited

Maximum allowable L2 norm for a neurons weights When using this, a good value might be 15

default value -1 (max-norm regularization disabled)

Definition at line 659 of file NeuralNetwork.h.

int32_t m_max_num_epochs
protectedinherited

maximum number of iterations over the training set. If 0, training will continue until convergence. defualt value is 0

Definition at line 673 of file NeuralNetwork.h.

float64_t m_max_train_time
protectedinherited

maximum training time

Definition at line 358 of file Machine.h.

Parameter* m_model_selection_parameters
inherited

model selection parameters

Definition at line 381 of file SGObject.h.

float64_t m_noise_parameter
protected

Controls the strength of the noise, depending on noise_type

Definition at line 247 of file Autoencoder.h.

EAENoiseType m_noise_type
protected

Noise type for denoising autoencoders.

If set to AENT_DROPOUT, inputs are randomly set to zero during each iteration of training with probability noise_parameter.

If set to AENT_GAUSSIAN, gaussian noise with zero mean and noise_parameter standard deviation is added to the inputs.

Default value is AENT_NONE

Definition at line 244 of file Autoencoder.h.

int32_t m_num_inputs
protectedinherited

number of neurons in the input layer

Definition at line 585 of file NeuralNetwork.h.

int32_t m_num_layers
protectedinherited

number of layer

Definition at line 588 of file NeuralNetwork.h.

ENNOptimizationMethod m_optimization_method
protectedinherited

Optimization method, default is NNOM_LBFGS

Definition at line 626 of file NeuralNetwork.h.

SGVector<bool> m_param_regularizable
protectedinherited

Array that specifies which parameters are to be regularized. This is used to turn off regularization for bias parameters

Definition at line 607 of file NeuralNetwork.h.

Parameter* m_parameters
inherited

parameters

Definition at line 378 of file SGObject.h.

SGVector<float64_t> m_params
protectedinherited

array where all the parameters of the network are stored

Definition at line 602 of file NeuralNetwork.h.

ESolverType m_solver_type
protectedinherited

solver type

Definition at line 364 of file Machine.h.

bool m_store_model_features
protectedinherited

whether model features should be stored after training

Definition at line 367 of file Machine.h.

int32_t m_total_num_parameters
protectedinherited

total number of parameters in the network

Definition at line 599 of file NeuralNetwork.h.

Parallel* parallel
inherited

parallel

Definition at line 372 of file SGObject.h.

Version* version
inherited

version

Definition at line 375 of file SGObject.h.


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

SHOGUN Machine Learning Toolbox - Documentation