The implementation of the standard GAN module. More...

Public Member Functions
	GAN (Model generator, Model discriminator, InitializationRuleType &initializeRule, Noise &noiseFunction, const size_t noiseDim, const size_t batchSize, const size_t generatorUpdateStep, const size_t preTrainSize, const double multiplier, const double clippingParameter=0.01, const double lambda=10.0)
	Constructor for GAN class. More...

	GAN (const GAN &)
	Copy constructor. More...

	GAN (GAN &&)
	Move constructor. More...

const Model &	Discriminator () const
	Return the discriminator of the GAN. More...

Model &	Discriminator ()
	Modify the discriminator of the GAN. More...

template<typename Policy = PolicyType>
std::enable_if< std::is_same < Policy, StandardGAN >::value\|\|std::is_same < Policy, DCGAN >::value, double >::type	Evaluate (const arma::mat &parameters, const size_t i, const size_t batchSize)
	Evaluate function for the Standard GAN and DCGAN. More...

template<typename Policy = PolicyType>
std::enable_if< std::is_same < Policy, WGAN >::value, double >::type	Evaluate (const arma::mat &parameters, const size_t i, const size_t batchSize)
	Evaluate function for the WGAN. More...

template<typename Policy = PolicyType>
std::enable_if< std::is_same < Policy, WGANGP >::value, double >::type	Evaluate (const arma::mat &parameters, const size_t i, const size_t batchSize)
	Evaluate function for the WGAN-GP. More...

template<typename GradType , typename Policy = PolicyType>
std::enable_if< std::is_same < Policy, StandardGAN >::value\|\|std::is_same < Policy, DCGAN >::value, double >::type	EvaluateWithGradient (const arma::mat &parameters, const size_t i, GradType &gradient, const size_t batchSize)
	EvaluateWithGradient function for the Standard GAN and DCGAN. More...

template<typename GradType , typename Policy = PolicyType>
std::enable_if< std::is_same < Policy, WGAN >::value, double >::type	EvaluateWithGradient (const arma::mat &parameters, const size_t i, GradType &gradient, const size_t batchSize)
	EvaluateWithGradient function for the WGAN. More...

template<typename GradType , typename Policy = PolicyType>
std::enable_if< std::is_same < Policy, WGANGP >::value, double >::type	EvaluateWithGradient (const arma::mat &parameters, const size_t i, GradType &gradient, const size_t batchSize)
	EvaluateWithGradient function for the WGAN-GP. More...

void	Forward (const arma::mat &input)
	This function does a forward pass through the GAN network. More...

const Model &	Generator () const
	Return the generator of the GAN. More...

Model &	Generator ()
	Modify the generator of the GAN. More...

template<typename Policy = PolicyType>
std::enable_if< std::is_same < Policy, StandardGAN >::value\|\|std::is_same < Policy, DCGAN >::value, void > ::type	Gradient (const arma::mat &parameters, const size_t i, arma::mat &gradient, const size_t batchSize)
	Gradient function for Standard GAN and DCGAN. More...

template<typename Policy = PolicyType>
std::enable_if< std::is_same < Policy, WGAN >::value, void > ::type	Gradient (const arma::mat &parameters, const size_t i, arma::mat &gradient, const size_t batchSize)
	Gradient function for WGAN. More...

template<typename Policy = PolicyType>
std::enable_if< std::is_same < Policy, WGANGP >::value, void >::type	Gradient (const arma::mat &parameters, const size_t i, arma::mat &gradient, const size_t batchSize)
	Gradient function for WGAN-GP. More...

size_t	NumFunctions () const
	Return the number of separable functions (the number of predictor points). More...

const arma::mat &	Parameters () const
	Return the parameters of the network. More...

arma::mat &	Parameters ()
	Modify the parameters of the network. More...

void	Predict (arma::mat input, arma::mat &output)
	This function predicts the output of the network on the given input. More...

const arma::mat &	Predictors () const
	Get the matrix of data points (predictors). More...

arma::mat &	Predictors ()
	Modify the matrix of data points (predictors). More...

void	Reset ()

void	ResetData (arma::mat trainData)
	Initialize the generator, discriminator and weights of the model for training. More...

const arma::mat &	Responses () const
	Get the matrix of responses to the input data points. More...

arma::mat &	Responses ()
	Modify the matrix of responses to the input data points. More...

template<typename Archive >
void	serialize (Archive &ar, const unsigned int)
	Serialize the model. More...

void	Shuffle ()
	Shuffle the order of function visitation. More...

template<typename OptimizerType , typename... CallbackTypes>
double	Train (arma::mat trainData, OptimizerType &Optimizer, CallbackTypes &&...callbacks)
	Train function. More...

Detailed Description

template<typename Model, typename InitializationRuleType, typename Noise, typename PolicyType = StandardGAN>
class mlpack::ann::GAN< Model, InitializationRuleType, Noise, PolicyType >

The implementation of the standard GAN module.

Generative Adversarial Networks (GANs) are a class of artificial intelligence algorithms used in unsupervised machine learning, implemented by a system of two neural networks contesting with each other in a zero-sum game framework. This technique can generate photographs that look at least superficially authentic to human observers, having many realistic characteristics. GANs have been used in Text-to-Image Synthesis, Medical Drug Discovery, High Resolution Imagery Generation, Neural Machine Translation and so on.

For more information, see the following paper:

@article{Goodfellow14,
  author    = {Ian J. Goodfellow, Jean Pouget-Abadi, Mehdi Mirza, Bing Xu,
               David Warde-Farley, Sherjil Ozair, Aaron Courville and
               Yoshua Bengio},
  title     = {Generative Adversarial Nets},
  year      = {2014},
  url       = {http://arxiv.org/abs/1406.2661},
  eprint    = {1406.2661},
}

Template Parameters

Model	The class type of Generator and Discriminator.
InitializationRuleType	Type of Initializer.
Noise	The noise function to use.
PolicyType	The GAN variant to be used (GAN, DCGAN, WGAN or WGANGP).

Definition at line 63 of file gan.hpp.

Constructor & Destructor Documentation

GAN	(	Model	generator,
		Model	discriminator,
		InitializationRuleType &	initializeRule,
		Noise &	noiseFunction,
		const size_t	noiseDim,
		const size_t	batchSize,
		const size_t	generatorUpdateStep,
		const size_t	preTrainSize,
		const double	multiplier,
		const double	clippingParameter = `0.01`,
		const double	lambda = `10.0`
	)

Constructor for GAN class.

Parameters

generator	Generator network.
discriminator	Discriminator network.
initializeRule	Initialization rule to use for initializing parameters.
noiseFunction	Function to be used for generating noise.
noiseDim	Dimension of noise vector to be created.
batchSize	Batch size to be used for training.
generatorUpdateStep	Number of steps to train Discriminator before updating Generator.
preTrainSize	Number of pre-training steps of Discriminator.
multiplier	Ratio of learning rate of Discriminator to the Generator.
clippingParameter	Weight range for enforcing Lipschitz constraint.
lambda	Parameter for setting the gradient penalty.

GAN ( const GAN< Model, InitializationRuleType, Noise, PolicyType > & )

Copy constructor.

GAN ( GAN< Model, InitializationRuleType, Noise, PolicyType > && )

Move constructor.

Member Function Documentation

const Model& Discriminator ( ) const

inline

Return the discriminator of the GAN.

Definition at line 312 of file gan.hpp.

Model& Discriminator ( )

inline

Modify the discriminator of the GAN.

Definition at line 314 of file gan.hpp.

std::enable_if<std::is_same<Policy, StandardGAN>::value \|\| std::is_same<Policy, DCGAN>::value, double>::type Evaluate	(	const arma::mat &	parameters,
		const size_t	i,
		const size_t	batchSize
	)

Evaluate function for the Standard GAN and DCGAN.

This function gives the performance of the Standard GAN or DCGAN on the current input.

Parameters

parameters	The parameters of the network.
i	Index of the current input.
batchSize	Variable to store the present number of inputs.

std::enable_if<std::is_same<Policy, WGAN>::value, double>::type Evaluate	(	const arma::mat &	parameters,
		const size_t	i,
		const size_t	batchSize
	)

Evaluate function for the WGAN.

This function gives the performance of the WGAN on the current input.

Parameters

parameters	The parameters of the network.
i	Index of the current input.
batchSize	Variable to store the present number of inputs.

std::enable_if<std::is_same<Policy, WGANGP>::value, double>::type Evaluate	(	const arma::mat &	parameters,
		const size_t	i,
		const size_t	batchSize
	)

Evaluate function for the WGAN-GP.

This function gives the performance of the WGAN-GP on the current input.

Parameters

parameters	The parameters of the network.
i	Index of the current input.
batchSize	Variable to store the present number of inputs.

std::enable_if<std::is_same<Policy, StandardGAN>::value \|\| std::is_same<Policy, DCGAN>::value, double>::type EvaluateWithGradient	(	const arma::mat &	parameters,
		const size_t	i,
		GradType &	gradient,
		const size_t	batchSize
	)

EvaluateWithGradient function for the Standard GAN and DCGAN.

This function gives the performance of the Standard GAN or DCGAN on the current input, while updating Gradients.

Parameters

parameters	The parameters of the network.
i	Index of the current input.
gradient	Variable to store the present gradient.
batchSize	Variable to store the present number of inputs.

std::enable_if<std::is_same<Policy, WGAN>::value, double>::type EvaluateWithGradient	(	const arma::mat &	parameters,
		const size_t	i,
		GradType &	gradient,
		const size_t	batchSize
	)

EvaluateWithGradient function for the WGAN.

This function gives the performance of the WGAN on the current input, while updating Gradients.

Parameters

parameters	The parameters of the network.
i	Index of the current input.
gradient	Variable to store the present gradient.
batchSize	Variable to store the present number of inputs.

std::enable_if<std::is_same<Policy, WGANGP>::value, double>::type EvaluateWithGradient	(	const arma::mat &	parameters,
		const size_t	i,
		GradType &	gradient,
		const size_t	batchSize
	)

EvaluateWithGradient function for the WGAN-GP.

This function gives the performance of the WGAN-GP on the current input, while updating Gradients.

Parameters

parameters	The parameters of the network.
i	Index of the current input.
gradient	Variable to store the present gradient.
batchSize	Variable to store the present number of inputs.

void Forward ( const arma::mat & input )

This function does a forward pass through the GAN network.

Parameters

input Sampled noise.

const Model& Generator ( ) const

inline

Return the generator of the GAN.

Definition at line 308 of file gan.hpp.

Model& Generator ( )

inline

Modify the generator of the GAN.

Definition at line 310 of file gan.hpp.

std::enable_if<std::is_same<Policy, StandardGAN>::value \|\| std::is_same<Policy, DCGAN>::value, void>::type Gradient	(	const arma::mat &	parameters,
		const size_t	i,
		arma::mat &	gradient,
		const size_t	batchSize
	)

Gradient function for Standard GAN and DCGAN.

This function passes the gradient based on which network is being trained, i.e., Generator or Discriminator.

Parameters

parameters	present parameters of the network.
i	Index of the predictors.
gradient	Variable to store the present gradient.
batchSize	Variable to store the present number of inputs.

std::enable_if<std::is_same<Policy, WGAN>::value, void>::type Gradient	(	const arma::mat &	parameters,
		const size_t	i,
		arma::mat &	gradient,
		const size_t	batchSize
	)

Gradient function for WGAN.

This function passes the gradient based on which network is being trained, i.e., Generator or Discriminator.

Parameters

parameters	present parameters of the network.
i	Index of the predictors.
gradient	Variable to store the present gradient.
batchSize	Variable to store the present number of inputs.

std::enable_if<std::is_same<Policy, WGANGP>::value, void>::type Gradient	(	const arma::mat &	parameters,
		const size_t	i,
		arma::mat &	gradient,
		const size_t	batchSize
	)

Gradient function for WGAN-GP.

This function passes the gradient based on which network is being trained, i.e., Generator or Discriminator.

Parameters

parameters	present parameters of the network.
i	Index of the predictors.
gradient	Variable to store the present gradient.
batchSize	Variable to store the present number of inputs.

size_t NumFunctions ( ) const

inline

Return the number of separable functions (the number of predictor points).

Definition at line 317 of file gan.hpp.

const arma::mat& Parameters ( ) const

inline

Return the parameters of the network.

Definition at line 303 of file gan.hpp.

arma::mat& Parameters ( )

inline

Modify the parameters of the network.

Definition at line 305 of file gan.hpp.

void Predict	(	arma::mat	input,
		arma::mat &	output
	)

This function predicts the output of the network on the given input.

Parameters

input	The input of the Generator network.
output	Result of the Discriminator network.

const arma::mat& Predictors ( ) const

inline

Get the matrix of data points (predictors).

Definition at line 325 of file gan.hpp.

arma::mat& Predictors ( )

inline

Modify the matrix of data points (predictors).

Definition at line 327 of file gan.hpp.

void Reset ( )

void ResetData ( arma::mat trainData )

Initialize the generator, discriminator and weights of the model for training.

This function won't actually trigger training process.

Parameters

trainData The data points of real distribution.

const arma::mat& Responses ( ) const

inline

Get the matrix of responses to the input data points.

Definition at line 320 of file gan.hpp.

arma::mat& Responses ( )

inline

Modify the matrix of responses to the input data points.

Definition at line 322 of file gan.hpp.

void serialize	(	Archive &	ar,
		const unsigned	int
	)

Serialize the model.

void Shuffle ( )

Shuffle the order of function visitation.

This may be called by the optimizer.

double Train	(	arma::mat	trainData,
		OptimizerType &	Optimizer,
		CallbackTypes &&...	callbacks
	)

Train function.

Template Parameters

OptimizerType	Type of optimizer to use to train the model.
CallbackTypes	Types of Callback functions.

Parameters

trainData	The data points of real distribution.
Optimizer	Instantiated optimizer used to train the model.
callbacks	Callback function for ensmallen optimizer `OptimizerType`. See https://www.ensmallen.org/docs.html#callback-documentation.

Returns: The final objective of the trained model (NaN or Inf on error).

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

src/mlpack/methods/ann/gan/gan.hpp

Public Member Functions

Detailed Description

template<typename Model, typename InitializationRuleType, typename Noise, typename PolicyType = StandardGAN> class mlpack::ann::GAN< Model, InitializationRuleType, Noise, PolicyType >

Constructor & Destructor Documentation

Member Function Documentation

template<typename Model, typename InitializationRuleType, typename Noise, typename PolicyType = StandardGAN>
class mlpack::ann::GAN< Model, InitializationRuleType, Noise, PolicyType >