Super class of all support vector machine trainers. More...

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

Inheritance diagram for shark::QpConfig:

Public Member Functions
	QpConfig (bool precomputedFlag=false, bool sparsifyFlag=true)
	Constructor.

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

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

QpSolutionProperties &	solutionProperties ()
	Access to the solution properties.

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

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

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

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

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

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

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

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

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

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

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

void	setMinAccuracy (double a)

void	setMaxIterations (unsigned long long i)

void	setTargetValue (double v)

void	setMaxSeconds (double s)

Protected Attributes
QpStoppingCondition	m_stoppingcondition
	conditions for when to stop the QP solver

QpSolutionProperties	m_solutionproperties
	properties of the approximate solution found by the solver

bool	m_precomputedKernelMatrix
	should the solver use a precomputed kernel matrix?

bool	m_sparsify
	should the trainer sparsify the model after training?

bool	m_shrinking
	should shrinking be used?

bool	m_s2do
	should S2DO be used instead of SMO?

unsigned int	m_verbosity
	verbosity level (currently unused)

unsigned long long	m_accessCount
	kernel access count

Detailed Description

Super class of all support vector machine trainers.

: The QpConfig class holds two structures describing the stopping condition and the solution obtained by the underlying quadratic programming solvers. It provides a uniform interface for setting, e.g., the target solution accuracy and obtaining the accuracy of the actual solution.

Definition at line 71 of file AbstractSvmTrainer.h.

Constructor & Destructor Documentation

◆ QpConfig()

shark::QpConfig::QpConfig	(	bool	precomputedFlag = `false`,
		bool	sparsifyFlag = `true`
	)

inline

Constructor.

Definition at line 75 of file AbstractSvmTrainer.h.

Member Function Documentation

◆ accessCount()

unsigned long long const & shark::QpConfig::accessCount ( ) const

inline

Number of kernel accesses.

Definition at line 137 of file AbstractSvmTrainer.h.

References m_accessCount.

◆ precomputeKernel() [1/2]

bool & shark::QpConfig::precomputeKernel ( )

inline

Flag for using a precomputed kernel matrix.

Definition at line 97 of file AbstractSvmTrainer.h.

References m_precomputedKernelMatrix.

Referenced by shark::SquaredHingeCSvmTrainer< InputType, CacheType >::train(), shark::RankingSvmTrainer< InputType, CacheType >::train(), shark::EpsilonSvmTrainer< InputType, CacheType >::train(), and shark::OneClassSvmTrainer< InputType, CacheType >::train().

◆ precomputeKernel() [2/2]

bool const & shark::QpConfig::precomputeKernel ( ) const

inline

Flag for using a precomputed kernel matrix.

Definition at line 101 of file AbstractSvmTrainer.h.

References m_precomputedKernelMatrix.

◆ s2do() [1/2]

bool & shark::QpConfig::s2do ( )

inline

Flag for S2DO (instead of SMO)

Definition at line 121 of file AbstractSvmTrainer.h.

References m_s2do.

◆ s2do() [2/2]

bool const & shark::QpConfig::s2do ( ) const

inline

Flag for S2DO (instead of SMO)

Definition at line 125 of file AbstractSvmTrainer.h.

References m_s2do.

◆ setMaxIterations()

void shark::QpConfig::setMaxIterations ( unsigned long long i )

inline

Definition at line 143 of file AbstractSvmTrainer.h.

References m_stoppingcondition, and shark::QpStoppingCondition::maxIterations.

◆ setMaxSeconds()

void shark::QpConfig::setMaxSeconds ( double s )

inline

Definition at line 147 of file AbstractSvmTrainer.h.

References m_stoppingcondition, and shark::QpStoppingCondition::maxSeconds.

◆ setMinAccuracy()

void shark::QpConfig::setMinAccuracy ( double a )

inline

Definition at line 141 of file AbstractSvmTrainer.h.

References m_stoppingcondition, and shark::QpStoppingCondition::minAccuracy.

◆ setTargetValue()

void shark::QpConfig::setTargetValue ( double v )

inline

Definition at line 145 of file AbstractSvmTrainer.h.

References m_stoppingcondition, and shark::QpStoppingCondition::targetValue.

◆ shrinking() [1/2]

bool & shark::QpConfig::shrinking ( )

inline

Flag for shrinking in the decomposition solver.

Definition at line 113 of file AbstractSvmTrainer.h.

References m_shrinking.

◆ shrinking() [2/2]

bool const & shark::QpConfig::shrinking ( ) const

inline

Flag for shrinking in the decomposition solver.

Definition at line 117 of file AbstractSvmTrainer.h.

References m_shrinking.

◆ solutionProperties()

QpSolutionProperties & shark::QpConfig::solutionProperties ( )

inline

Access to the solution properties.

Definition at line 93 of file AbstractSvmTrainer.h.

References m_solutionproperties.

Referenced by main(), main(), shark::SquaredHingeLinearCSvmTrainer< InputType >::train(), and shark::OneClassSvmTrainer< InputType, CacheType >::trainSVM().

◆ sparsify() [1/2]

bool & shark::QpConfig::sparsify ( )

inline

Flag for sparsifying the model after training.

Definition at line 105 of file AbstractSvmTrainer.h.

References m_sparsify.

Referenced by shark::SvmLogisticInterpretation< InputType >::eval(), shark::SvmLogisticInterpretation< InputType >::evalDerivative(), main(), shark::CSvmTrainer< InputType, CacheType >::train(), shark::SquaredHingeCSvmTrainer< InputType, CacheType >::train(), shark::CSvmTrainer< InputType, CacheType >::train(), shark::EpsilonSvmTrainer< InputType, CacheType >::train(), and shark::OneClassSvmTrainer< InputType, CacheType >::train().

◆ sparsify() [2/2]

bool const & shark::QpConfig::sparsify ( ) const

inline

Flag for sparsifying the model after training.

Definition at line 109 of file AbstractSvmTrainer.h.

References m_sparsify.

◆ stoppingCondition() [1/2]

QpStoppingCondition & shark::QpConfig::stoppingCondition ( )

inline

Read/write access to the stopping condition.

Definition at line 85 of file AbstractSvmTrainer.h.

References m_stoppingcondition.

Referenced by shark::SvmLogisticInterpretation< InputType >::eval(), shark::SvmLogisticInterpretation< InputType >::evalDerivative(), main(), shark::SquaredHingeLinearCSvmTrainer< InputType >::train(), and shark::OneClassSvmTrainer< InputType, CacheType >::trainSVM().

◆ stoppingCondition() [2/2]

QpStoppingCondition const & shark::QpConfig::stoppingCondition ( ) const

inline

Read access to the stopping condition.

Definition at line 89 of file AbstractSvmTrainer.h.

References m_stoppingcondition.

◆ verbosity() [1/2]

unsigned int & shark::QpConfig::verbosity ( )

inline

Verbosity level of the solver.

Definition at line 129 of file AbstractSvmTrainer.h.

References m_verbosity.

Referenced by shark::SquaredHingeLinearCSvmTrainer< InputType >::train().

◆ verbosity() [2/2]

unsigned int const & shark::QpConfig::verbosity ( ) const

inline

Verbosity level of the solver.

Definition at line 133 of file AbstractSvmTrainer.h.

References m_verbosity.

Member Data Documentation

◆ m_accessCount

unsigned long long shark::QpConfig::m_accessCount

protected

kernel access count

Definition at line 165 of file AbstractSvmTrainer.h.

Referenced by accessCount(), shark::SquaredHingeCSvmTrainer< InputType, CacheType >::train(), and shark::OneClassSvmTrainer< InputType, CacheType >::trainSVM().

◆ m_precomputedKernelMatrix

bool shark::QpConfig::m_precomputedKernelMatrix

protected

should the solver use a precomputed kernel matrix?

Definition at line 155 of file AbstractSvmTrainer.h.

Referenced by precomputeKernel(), and precomputeKernel().

◆ m_s2do

bool shark::QpConfig::m_s2do

protected

should S2DO be used instead of SMO?

Definition at line 161 of file AbstractSvmTrainer.h.

Referenced by s2do(), and s2do().

◆ m_shrinking

bool shark::QpConfig::m_shrinking

protected

should shrinking be used?

Definition at line 159 of file AbstractSvmTrainer.h.

Referenced by shrinking(), shrinking(), and shark::OneClassSvmTrainer< InputType, CacheType >::trainSVM().

◆ m_solutionproperties

QpSolutionProperties shark::QpConfig::m_solutionproperties

protected

properties of the approximate solution found by the solver

Definition at line 153 of file AbstractSvmTrainer.h.

Referenced by solutionProperties().

◆ m_sparsify

bool shark::QpConfig::m_sparsify

protected

should the trainer sparsify the model after training?

Definition at line 157 of file AbstractSvmTrainer.h.

Referenced by sparsify(), and sparsify().

◆ m_stoppingcondition

QpStoppingCondition shark::QpConfig::m_stoppingcondition

protected

conditions for when to stop the QP solver

Definition at line 151 of file AbstractSvmTrainer.h.

Referenced by setMaxIterations(), setMaxSeconds(), setMinAccuracy(), setTargetValue(), stoppingCondition(), and stoppingCondition().

◆ m_verbosity

unsigned int shark::QpConfig::m_verbosity

protected

verbosity level (currently unused)

Definition at line 163 of file AbstractSvmTrainer.h.

Referenced by verbosity(), and verbosity().

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

include/shark/Algorithms/Trainers/AbstractSvmTrainer.h

Public Member Functions

Protected Attributes

Detailed Description

Constructor & Destructor Documentation

◆ QpConfig()

Member Function Documentation

◆ accessCount()

◆ precomputeKernel() [1/2]

◆ precomputeKernel() [2/2]

◆ s2do() [1/2]

◆ s2do() [2/2]

◆ setMaxIterations()

◆ setMaxSeconds()

◆ setMinAccuracy()

◆ setTargetValue()

◆ shrinking() [1/2]

◆ shrinking() [2/2]

◆ solutionProperties()

◆ sparsify() [1/2]

◆ sparsify() [2/2]

◆ stoppingCondition() [1/2]

◆ stoppingCondition() [2/2]

◆ verbosity() [1/2]

◆ verbosity() [2/2]

Member Data Documentation

◆ m_accessCount

◆ m_precomputedKernelMatrix

◆ m_s2do

◆ m_shrinking

◆ m_solutionproperties

◆ m_sparsify

◆ m_stoppingcondition

◆ m_verbosity