Super class of all kernelized (non-linear) SVM trainers. More...

#include <shark/Algorithms/Trainers/AbstractSvmTrainer.h>

Inheritance diagram for shark::AbstractSvmTrainer< InputType, LabelType, Model, Trainer >:

Public Types
typedef AbstractKernelFunction< InputType >	KernelType

Public Member Functions
	AbstractSvmTrainer (KernelType *kernel, double C, bool offset, bool unconstrained=false)

	AbstractSvmTrainer (KernelType *kernel, double negativeC, double positiveC, bool offset, bool unconstrained=false)

double	C () const
	Return the value of the regularization parameter C. More...

RealVector const &	regularizationParameters () const

RealVector &	regularizationParameters ()

KernelType *	kernel ()

const KernelType *	kernel () const

void	setKernel (KernelType *kernel)

bool	isUnconstrained () const

bool	trainOffset () const

double	CacheSize () const

void	setCacheSize (std::size_t size)

RealVector	parameterVector () const
	get the hyper-parameter vector More...

void	setParameterVector (RealVector const &newParameters)
	set the vector of hyper-parameters More...

size_t	numberOfParameters () const
	return the number of hyper-parameters More...

Public Member Functions inherited from shark::QpConfig
	QpConfig (bool precomputedFlag=false, bool sparsifyFlag=true)
	Constructor. More...

QpStoppingCondition &	stoppingCondition ()
	Read/write access to the stopping condition. More...

QpStoppingCondition const &	stoppingCondition () const
	Read access to the stopping condition. More...

QpSolutionProperties &	solutionProperties ()
	Access to the solution properties. More...

bool &	precomputeKernel ()
	Flag for using a precomputed kernel matrix. More...

bool const &	precomputeKernel () const
	Flag for using a precomputed kernel matrix. More...

bool &	sparsify ()
	Flag for sparsifying the model after training. More...

bool const &	sparsify () const
	Flag for sparsifying the model after training. More...

bool &	shrinking ()
	Flag for shrinking in the decomposition solver. More...

bool const &	shrinking () const
	Flag for shrinking in the decomposition solver. More...

bool &	s2do ()
	Flag for S2DO (instead of SMO) More...

bool const &	s2do () const
	Flag for S2DO (instead of SMO) More...

unsigned int &	verbosity ()
	Verbosity level of the solver. More...

unsigned int const &	verbosity () const
	Verbosity level of the solver. More...

unsigned long long const &	accessCount () const
	Number of kernel accesses. More...

void	setMinAccuracy (double a)

void	setMaxIterations (unsigned long long i)

void	setTargetValue (double v)

void	setMaxSeconds (double s)

Public Member Functions inherited from shark::IParameterizable
virtual	~IParameterizable ()

Protected Attributes
KernelType *	m_kernel

RealVector	m_regularizers
	Vector of regularization parameters. More...

bool	m_trainOffset

bool	m_unconstrained
	Is log(C) stored internally as a parameter instead of C? If yes, then we get rid of the constraint C > 0 on the level of the parameter interface. More...

std::size_t	m_cacheSize
	Number of values in the kernel cache. The size of the cache in bytes is the size of one entry (4 for float, 8 for double) times this number. More...

Protected Attributes inherited from shark::QpConfig
QpStoppingCondition	m_stoppingcondition
	conditions for when to stop the QP solver More...

QpSolutionProperties	m_solutionproperties
	properties of the approximate solution found by the solver More...

bool	m_precomputedKernelMatrix
	should the solver use a precomputed kernel matrix? More...

bool	m_sparsify
	should the trainer sparsify the model after training? More...

bool	m_shrinking
	should shrinking be used? More...

bool	m_s2do
	should S2DO be used instead of SMO? More...

unsigned int	m_verbosity
	verbosity level (currently unused) More...

unsigned long long	m_accessCount
	kernel access count More...

Detailed Description

template<class InputType, class LabelType, class Model = KernelClassifier<InputType>, class Trainer = AbstractTrainer< Model,LabelType>>
class shark::AbstractSvmTrainer< InputType, LabelType, Model, Trainer >

Super class of all kernelized (non-linear) SVM trainers.

: This class holds general information shared by most if not all SVM trainers. First of all, this includes the kernel and the regularization parameter. The class also manages meta-information of the training process, like the maximal size of the kernel cache, the stopping criterion, as well as information on the actual solution.

Definition at line 185 of file AbstractSvmTrainer.h.

Member Typedef Documentation

§ KernelType

template<class InputType, class LabelType, class Model = KernelClassifier<InputType>, class Trainer = AbstractTrainer< Model,LabelType>>

typedef AbstractKernelFunction<InputType> shark::AbstractSvmTrainer< InputType, LabelType, Model, Trainer >::KernelType

Definition at line 189 of file AbstractSvmTrainer.h.

Constructor & Destructor Documentation

§ AbstractSvmTrainer() [1/2]

template<class InputType, class LabelType, class Model = KernelClassifier<InputType>, class Trainer = AbstractTrainer< Model,LabelType>>

shark::AbstractSvmTrainer< InputType, LabelType, Model, Trainer >::AbstractSvmTrainer	(	KernelType *	kernel,
		double	C,
		bool	offset,
		bool	unconstrained = `false`
	)

inline

Constructor

Parameters

kernel	kernel function to use for training and prediction
C	regularization parameter - always the 'true' value of C, even when unconstrained is set
offset	train svm with offset - this is not supported for all SVM solvers.
unconstrained	when a C-value is given via setParameter, should it be piped through the exp-function before using it in the solver?

Definition at line 196 of file AbstractSvmTrainer.h.

§ AbstractSvmTrainer() [2/2]

template<class InputType, class LabelType, class Model = KernelClassifier<InputType>, class Trainer = AbstractTrainer< Model,LabelType>>

shark::AbstractSvmTrainer< InputType, LabelType, Model, Trainer >::AbstractSvmTrainer	(	KernelType *	kernel,
		double	negativeC,
		double	positiveC,
		bool	offset,
		bool	unconstrained = `false`
	)

inline

Constructor featuring two regularization parameters

Parameters

kernel	kernel function to use for training and prediction
negativeC	regularization parameter of the negative class (label 0)
positiveC	regularization parameter of the positive class (label 1)
offset	train svm with offset - this is not supported for all SVM solvers.
unconstrained	when a C-value is given via setParameter, should it be piped through the exp-function before using it in the solver?

Definition at line 210 of file AbstractSvmTrainer.h.

Member Function Documentation

§ C()

template<class InputType, class LabelType, class Model = KernelClassifier<InputType>, class Trainer = AbstractTrainer< Model,LabelType>>

double shark::AbstractSvmTrainer< InputType, LabelType, Model, Trainer >::C ( ) const

inline

Return the value of the regularization parameter C.

Definition at line 225 of file AbstractSvmTrainer.h.

§ CacheSize()

template<class InputType, class LabelType, class Model = KernelClassifier<InputType>, class Trainer = AbstractTrainer< Model,LabelType>>

double shark::AbstractSvmTrainer< InputType, LabelType, Model, Trainer >::CacheSize ( ) const

inline

Definition at line 254 of file AbstractSvmTrainer.h.

§ isUnconstrained()

template<class InputType, class LabelType, class Model = KernelClassifier<InputType>, class Trainer = AbstractTrainer< Model,LabelType>>

bool shark::AbstractSvmTrainer< InputType, LabelType, Model, Trainer >::isUnconstrained ( ) const

inline

Definition at line 248 of file AbstractSvmTrainer.h.

§ kernel() [1/2]

template<class InputType, class LabelType, class Model = KernelClassifier<InputType>, class Trainer = AbstractTrainer< Model,LabelType>>

KernelType* shark::AbstractSvmTrainer< InputType, LabelType, Model, Trainer >::kernel ( )

inline

Definition at line 241 of file AbstractSvmTrainer.h.

Referenced by shark::CSvmDerivative< InputType, CacheType >::CSvmDerivative().

§ kernel() [2/2]

template<class InputType, class LabelType, class Model = KernelClassifier<InputType>, class Trainer = AbstractTrainer< Model,LabelType>>

const KernelType* shark::AbstractSvmTrainer< InputType, LabelType, Model, Trainer >::kernel ( ) const

inline

Definition at line 243 of file AbstractSvmTrainer.h.

§ numberOfParameters()

template<class InputType, class LabelType, class Model = KernelClassifier<InputType>, class Trainer = AbstractTrainer< Model,LabelType>>

size_t shark::AbstractSvmTrainer< InputType, LabelType, Model, Trainer >::numberOfParameters ( ) const

inlinevirtual

return the number of hyper-parameters

Reimplemented from shark::IParameterizable.

Definition at line 282 of file AbstractSvmTrainer.h.

§ parameterVector()

template<class InputType, class LabelType, class Model = KernelClassifier<InputType>, class Trainer = AbstractTrainer< Model,LabelType>>

RealVector shark::AbstractSvmTrainer< InputType, LabelType, Model, Trainer >::parameterVector ( ) const

inlinevirtual

get the hyper-parameter vector

Reimplemented from shark::IParameterizable.

Definition at line 260 of file AbstractSvmTrainer.h.

§ regularizationParameters() [1/2]

template<class InputType, class LabelType, class Model = KernelClassifier<InputType>, class Trainer = AbstractTrainer< Model,LabelType>>

RealVector const& shark::AbstractSvmTrainer< InputType, LabelType, Model, Trainer >::regularizationParameters ( ) const

inline

Definition at line 231 of file AbstractSvmTrainer.h.

§ regularizationParameters() [2/2]

template<class InputType, class LabelType, class Model = KernelClassifier<InputType>, class Trainer = AbstractTrainer< Model,LabelType>>

RealVector& shark::AbstractSvmTrainer< InputType, LabelType, Model, Trainer >::regularizationParameters ( )

inline

Definition at line 236 of file AbstractSvmTrainer.h.

§ setCacheSize()

template<class InputType, class LabelType, class Model = KernelClassifier<InputType>, class Trainer = AbstractTrainer< Model,LabelType>>

void shark::AbstractSvmTrainer< InputType, LabelType, Model, Trainer >::setCacheSize ( std::size_t size )

inline

Definition at line 256 of file AbstractSvmTrainer.h.

§ setKernel()

template<class InputType, class LabelType, class Model = KernelClassifier<InputType>, class Trainer = AbstractTrainer< Model,LabelType>>

void shark::AbstractSvmTrainer< InputType, LabelType, Model, Trainer >::setKernel ( KernelType * kernel )

inline

Definition at line 245 of file AbstractSvmTrainer.h.

§ setParameterVector()

template<class InputType, class LabelType, class Model = KernelClassifier<InputType>, class Trainer = AbstractTrainer< Model,LabelType>>

void shark::AbstractSvmTrainer< InputType, LabelType, Model, Trainer >::setParameterVector ( RealVector const & newParameters )

inlinevirtual

set the vector of hyper-parameters

Reimplemented from shark::IParameterizable.

Definition at line 272 of file AbstractSvmTrainer.h.

§ trainOffset()

template<class InputType, class LabelType, class Model = KernelClassifier<InputType>, class Trainer = AbstractTrainer< Model,LabelType>>

bool shark::AbstractSvmTrainer< InputType, LabelType, Model, Trainer >::trainOffset ( ) const

inline

Definition at line 251 of file AbstractSvmTrainer.h.

Member Data Documentation

§ m_cacheSize

template<class InputType, class LabelType, class Model = KernelClassifier<InputType>, class Trainer = AbstractTrainer< Model,LabelType>>

std::size_t shark::AbstractSvmTrainer< InputType, LabelType, Model, Trainer >::m_cacheSize

protected

Number of values in the kernel cache. The size of the cache in bytes is the size of one entry (4 for float, 8 for double) times this number.

Definition at line 297 of file AbstractSvmTrainer.h.

§ m_kernel

template<class InputType, class LabelType, class Model = KernelClassifier<InputType>, class Trainer = AbstractTrainer< Model,LabelType>>

KernelType* shark::AbstractSvmTrainer< InputType, LabelType, Model, Trainer >::m_kernel

protected

Kernel object.

Definition at line 287 of file AbstractSvmTrainer.h.

§ m_regularizers

template<class InputType, class LabelType, class Model = KernelClassifier<InputType>, class Trainer = AbstractTrainer< Model,LabelType>>

RealVector shark::AbstractSvmTrainer< InputType, LabelType, Model, Trainer >::m_regularizers

protected

Vector of regularization parameters.

If the size of the vector is 1 there is only one regularization parameter for all classes, else there must be one for every class in the dataset. The exact meaning depends on the sub-class, but the value is always positive, and higher implies a less regular solution.

Definition at line 294 of file AbstractSvmTrainer.h.

§ m_trainOffset

template<class InputType, class LabelType, class Model = KernelClassifier<InputType>, class Trainer = AbstractTrainer< Model,LabelType>>

bool shark::AbstractSvmTrainer< InputType, LabelType, Model, Trainer >::m_trainOffset

protected

Definition at line 295 of file AbstractSvmTrainer.h.

§ m_unconstrained

template<class InputType, class LabelType, class Model = KernelClassifier<InputType>, class Trainer = AbstractTrainer< Model,LabelType>>

bool shark::AbstractSvmTrainer< InputType, LabelType, Model, Trainer >::m_unconstrained

protected

Is log(C) stored internally as a parameter instead of C? If yes, then we get rid of the constraint C > 0 on the level of the parameter interface.

Definition at line 296 of file AbstractSvmTrainer.h.

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

include/shark/Algorithms/Trainers/AbstractSvmTrainer.h

Public Types

Public Member Functions

Protected Attributes

Detailed Description

template<class InputType, class LabelType, class Model = KernelClassifier<InputType>, class Trainer = AbstractTrainer< Model,LabelType>> class shark::AbstractSvmTrainer< InputType, LabelType, Model, Trainer >

Member Typedef Documentation

§ KernelType

Constructor & Destructor Documentation

§ AbstractSvmTrainer() [1/2]

§ AbstractSvmTrainer() [2/2]

Member Function Documentation

§ C()

§ CacheSize()

§ isUnconstrained()

§ kernel() [1/2]

§ kernel() [2/2]

§ numberOfParameters()

§ parameterVector()

§ regularizationParameters() [1/2]

§ regularizationParameters() [2/2]

§ setCacheSize()

§ setKernel()

§ setParameterVector()

§ trainOffset()

Member Data Documentation

§ m_cacheSize

§ m_kernel

§ m_regularizers

§ m_trainOffset

§ m_unconstrained

template<class InputType, class LabelType, class Model = KernelClassifier<InputType>, class Trainer = AbstractTrainer< Model,LabelType>>
class shark::AbstractSvmTrainer< InputType, LabelType, Model, Trainer >