Classes
class	DenseAndSparseLinearBase
	Base class for implementationa of an objective that combines dense features and sparse features. More...

struct	DenseAndSparseMargin

class	GenericLinearClassifier
	This is a non-templated, runtime-polymorphic generic implementation of the linear classifier objective. More...

struct	GenericMarginClassifier
	A utility class template that, when instatiated with a `MarginFunction`, produces the corresponding linear classifier loss. More...

class	LinearClassifierBase
	Base class for objectives that use a linear classifier. More...

class	LinearClassifierImpBase
	Implementation helper for linear classifier derived classes. More...

struct	SquaredHingePhi

struct	HuberPhi

struct	LogisticPhi

class	Objective
	Class that models an optimization objective. More...

class	PointWiseRegularizer
	Base class for pointwise regularization functions. More...

class	Regularized_SquaredHingeSVC

struct	SquaredNormConfig

struct	HuberConfig

struct	ElasticConfig

class	SquaredNormRegularizer
	This class implements a squared norm (L2) regularizer. Thus `f(x) = 0.5 \|x\|^2`. More...

class	HuberRegularizer
	This class implements a huber regularizer. More...

class	ElasticNetRegularizer

Functions
std::unique_ptr< DenseAndSparseLinearBase >	make_sp_dense_squared_hinge (std::shared_ptr< const GenericFeatureMatrix > dense_features, real_t dense_reg_strength, std::shared_ptr< const GenericFeatureMatrix > sparse_features, real_t sparse_reg_strength)

std::unique_ptr< GenericLinearClassifier >	make_squared_hinge (std::shared_ptr< const GenericFeatureMatrix > X, std::unique_ptr< Objective > regularizer)

std::unique_ptr< GenericLinearClassifier >	make_logistic_loss (std::shared_ptr< const GenericFeatureMatrix > X, std::unique_ptr< Objective > regularizer)

std::unique_ptr< GenericLinearClassifier >	make_huber_hinge (std::shared_ptr< const GenericFeatureMatrix > X, std::unique_ptr< Objective > regularizer, real_t epsilon)

std::unique_ptr< Objective >	make_regularizer (const SquaredNormConfig &config)

std::unique_ptr< Objective >	make_regularizer (const HuberConfig &config)

std::unique_ptr< Objective >	make_regularizer (const ElasticConfig &config)

Detailed Description

Regularizers are implementations of Objective that depend only on the weight vector, but make no reference to external data. From the Objective interface, this is not visible, since the data is not part of the public interface.

Function Documentation

◆ make_huber_hinge()

std::unique_ptr< GenericLinearClassifier > dismec::objective::make_huber_hinge	(	std::shared_ptr< const GenericFeatureMatrix >	X,
		std::unique_ptr< Objective >	regularizer,
		real_t	epsilon
	)

Definition at line 185 of file generic_linear.cpp.

Referenced by dismec::make_loss().

◆ make_logistic_loss()

std::unique_ptr< GenericLinearClassifier > dismec::objective::make_logistic_loss	(	std::shared_ptr< const GenericFeatureMatrix >	X,
		std::unique_ptr< Objective >	regularizer
	)

Definition at line 180 of file generic_linear.cpp.

Referenced by dismec::make_loss().

◆ make_regularizer() [1/3]

std::unique_ptr< Objective > dismec::objective::make_regularizer ( const ElasticConfig & config )

Definition at line 134 of file regularizers_imp.cpp.

References dismec::objective::ElasticConfig::Epsilon, dismec::objective::ElasticConfig::IgnoreBias, dismec::objective::ElasticConfig::Interpolation, and dismec::objective::ElasticConfig::Strength.

◆ make_regularizer() [2/3]

std::unique_ptr< Objective > dismec::objective::make_regularizer ( const HuberConfig & config )

Definition at line 130 of file regularizers_imp.cpp.

References dismec::objective::HuberConfig::Epsilon, dismec::objective::HuberConfig::IgnoreBias, and dismec::objective::HuberConfig::Strength.

◆ make_regularizer() [3/3]

std::unique_ptr< Objective > dismec::objective::make_regularizer ( const SquaredNormConfig & config )

Definition at line 126 of file regularizers_imp.cpp.

References dismec::objective::SquaredNormConfig::IgnoreBias, and dismec::objective::SquaredNormConfig::Strength.

Referenced by dismec::init::create_ova_primal_initializer(), and dismec::DiSMECTraining::make_objective().

◆ make_sp_dense_squared_hinge()

std::unique_ptr< DenseAndSparseLinearBase > dismec::objective::make_sp_dense_squared_hinge	(	std::shared_ptr< const GenericFeatureMatrix >	dense_features,
		real_t	dense_reg_strength,
		std::shared_ptr< const GenericFeatureMatrix >	sparse_features,
		real_t	sparse_reg_strength
	)

Definition at line 247 of file dense_and_sparse.cpp.

Referenced by dismec::CascadeTraining::make_objective().

◆ make_squared_hinge()

std::unique_ptr< GenericLinearClassifier > dismec::objective::make_squared_hinge	(	std::shared_ptr< const GenericFeatureMatrix >	X,
		std::unique_ptr< Objective >	regularizer
	)

Definition at line 175 of file generic_linear.cpp.

Referenced by dismec::make_loss(), and TEST_CASE().

Classes

Functions