An optimizer that optimizes general objective functions. More...

#include <shark/Algorithms/AbstractOptimizer.h>

Inheritance diagram for shark::AbstractOptimizer< PointType, ResultT, SolutionTypeT >:

Public Types
enum	Feature { REQUIRES_VALUE = 1, REQUIRES_FIRST_DERIVATIVE = 2, REQUIRES_SECOND_DERIVATIVE = 4, CAN_SOLVE_CONSTRAINED = 8, REQUIRES_CLOSEST_FEASIBLE = 16 }
	Models features that the optimizer requires from the objective function. More...

typedef PointType	SearchPointType

typedef ResultT	ResultType

typedef SolutionTypeT	SolutionType

typedef AbstractObjectiveFunction< PointType, ResultType >	ObjectiveFunctionType

typedef TypedFlags< Feature >	Features

typedef TypedFeatureNotAvailableException< Feature >	FeatureNotAvailableException

Public Member Functions
const Features &	features () const

virtual void	updateFeatures ()

bool	requiresValue () const

bool	requiresFirstDerivative () const

bool	requiresSecondDerivative () const

bool	canSolveConstrained () const

bool	requiresClosestFeasible () const

virtual	~AbstractOptimizer ()

virtual std::size_t	numInitPoints () const =0
	Returns the number of points this method requires for initialisation. More...

virtual void	init (ObjectiveFunctionType const &function)
	Initialize the optimizer for the supplied objective function. More...

virtual void	init (ObjectiveFunctionType const &function, std::vector< SearchPointType > const &initPoints)=0
	Initialize the optimizer for the supplied objective function using a set of initialisation points. More...

virtual void	step (ObjectiveFunctionType const &function)=0
	Carry out one step of the optimizer for the supplied objective function. More...

virtual SolutionType const &	solution () const =0
	Accesses the best solution obtained so far. More...

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

virtual std::string	name () const
	returns the name of the object More...

Public Member Functions inherited from shark::ISerializable
virtual	~ISerializable ()
	Virtual d'tor. More...

virtual void	read (InArchive &archive)
	Read the component from the supplied archive. More...

virtual void	write (OutArchive &archive) const
	Write the component to the supplied archive. More...

void	load (InArchive &archive, unsigned int version)
	Versioned loading of components, calls read(...). More...

void	save (OutArchive &archive, unsigned int version) const
	Versioned storing of components, calls write(...). More...

	BOOST_SERIALIZATION_SPLIT_MEMBER ()

Protected Member Functions
void	checkFeatures (ObjectiveFunctionType const &objectiveFunction)
	Convenience function that checks whether the features of the supplied objective function match with the required features of the optimizer. More...

Protected Attributes
Features	m_features

Detailed Description

template<typename PointType, typename ResultT, typename SolutionTypeT>
class shark::AbstractOptimizer< PointType, ResultT, SolutionTypeT >

An optimizer that optimizes general objective functions.

After construction and configurationg the optimizer, init() is called with the objective function to be used. After that step() can be called until the required solution is found. The solution can be queried using solution(). The type of the solution depends on the optimisation problem at hand. It is allowed to add constrains on the features the objective function needs to offer

These are:

REQUIRES_VALUE: The function is evaluated to use the optimizer and the HAS_VALUE-flag must be set
REQUIRES_FIRST_DERIVATIVE: The first derivative needs to be evaluated and
HAS_FIRST_DERIVATIVE must be set
REQUIRES_SECOND_DERIVATIVE: The second derivative needs to be evaluated and
HAS_SECOND_DERIVATIVE must be set
CAN_SOLVE_CONSTRAINED: The optimizer can solve functions which are constrained and where the IS_CONSTRAINED_FEATURE is set.
REQUIRES_CLOSEST_FEASIBLE: If the function is constrained, it must offer a way to construct the closest feasible point and
CAN_PROVIDE_CLOSEST_FEASIBLE must be set

Also when init() is called as offered by the AbstractOptimizer interface, the function is required to have the CAN_PROPOSE_STARTING_POINT flag.

Template Parameters

PointType	The type of search space the optimizer works upon.
ResultT	The objective space the optimizer works upon.
SolutionTypeT	The type of the final solution.

Definition at line 68 of file AbstractOptimizer.h.

Member Typedef Documentation

◆ FeatureNotAvailableException

template<typename PointType, typename ResultT, typename SolutionTypeT>

typedef TypedFeatureNotAvailableException<Feature> shark::AbstractOptimizer< PointType, ResultT, SolutionTypeT >::FeatureNotAvailableException

Definition at line 85 of file AbstractOptimizer.h.

◆ Features

template<typename PointType, typename ResultT, typename SolutionTypeT>

typedef TypedFlags<Feature> shark::AbstractOptimizer< PointType, ResultT, SolutionTypeT >::Features

Definition at line 85 of file AbstractOptimizer.h.

◆ ObjectiveFunctionType

template<typename PointType, typename ResultT, typename SolutionTypeT>

typedef AbstractObjectiveFunction<PointType,ResultType> shark::AbstractOptimizer< PointType, ResultT, SolutionTypeT >::ObjectiveFunctionType

Definition at line 73 of file AbstractOptimizer.h.

◆ ResultType

template<typename PointType, typename ResultT, typename SolutionTypeT>

typedef ResultT shark::AbstractOptimizer< PointType, ResultT, SolutionTypeT >::ResultType

Definition at line 71 of file AbstractOptimizer.h.

◆ SearchPointType

template<typename PointType, typename ResultT, typename SolutionTypeT>

typedef PointType shark::AbstractOptimizer< PointType, ResultT, SolutionTypeT >::SearchPointType

Definition at line 70 of file AbstractOptimizer.h.

◆ SolutionType

template<typename PointType, typename ResultT, typename SolutionTypeT>

typedef SolutionTypeT shark::AbstractOptimizer< PointType, ResultT, SolutionTypeT >::SolutionType

Definition at line 72 of file AbstractOptimizer.h.

Member Enumeration Documentation

◆ Feature

template<typename PointType, typename ResultT, typename SolutionTypeT>

enum shark::AbstractOptimizer::Feature

Models features that the optimizer requires from the objective function.

See also: AbstractObjectiveFunction

Enumerator
REQUIRES_VALUE
REQUIRES_FIRST_DERIVATIVE
REQUIRES_SECOND_DERIVATIVE
CAN_SOLVE_CONSTRAINED
REQUIRES_CLOSEST_FEASIBLE

Definition at line 77 of file AbstractOptimizer.h.

Constructor & Destructor Documentation

◆ ~AbstractOptimizer()

template<typename PointType, typename ResultT, typename SolutionTypeT>

virtual shark::AbstractOptimizer< PointType, ResultT, SolutionTypeT >::~AbstractOptimizer ( )

inlinevirtual

Definition at line 104 of file AbstractOptimizer.h.

Member Function Documentation

◆ canSolveConstrained()

template<typename PointType, typename ResultT, typename SolutionTypeT>

bool shark::AbstractOptimizer< PointType, ResultT, SolutionTypeT >::canSolveConstrained ( ) const

inline

Definition at line 97 of file AbstractOptimizer.h.

Referenced by shark::AbstractOptimizer< RealVector, RealVector, std::vector< ResultSet< RealVector, RealVector > > >::checkFeatures().

◆ checkFeatures()

template<typename PointType, typename ResultT, typename SolutionTypeT>

void shark::AbstractOptimizer< PointType, ResultT, SolutionTypeT >::checkFeatures ( ObjectiveFunctionType const & objectiveFunction )

inlineprotected

Convenience function that checks whether the features of the supplied objective function match with the required features of the optimizer.

Parameters

[in] objectiveFunction The function to match with.

Exceptions

shark::Exception

Definition at line 152 of file AbstractOptimizer.h.

◆ features()

template<typename PointType, typename ResultT, typename SolutionTypeT>

const Features& shark::AbstractOptimizer< PointType, ResultT, SolutionTypeT >::features ( ) const

inline

Definition at line 85 of file AbstractOptimizer.h.

◆ init() [1/2]

template<typename PointType, typename ResultT, typename SolutionTypeT>

virtual void shark::AbstractOptimizer< PointType, ResultT, SolutionTypeT >::init ( ObjectiveFunctionType const & function )

inlinevirtual

Initialize the optimizer for the supplied objective function.

Be aware that function.init() has to be called before calling this function! This function will initialize the algorithm with a number of points proposed by the function. Note that this must fail if the function can not propose starting point(s).

Parameters

[in] function The objective function to initialize for.

Definition at line 119 of file AbstractOptimizer.h.

Referenced by shark::AbstractOptimizer< RealVector, RealVector, std::vector< ResultSet< RealVector, RealVector > > >::init().

◆ init() [2/2]

template<typename PointType, typename ResultT, typename SolutionTypeT>

virtual void shark::AbstractOptimizer< PointType, ResultT, SolutionTypeT >::init	(	ObjectiveFunctionType const &	function,
		std::vector< SearchPointType > const &	initPoints
	)

pure virtual

Initialize the optimizer for the supplied objective function using a set of initialisation points.

Most single objective algorithms only require a single point. However multi-objective algorithms need a set of initialisation points. The number of points required should be numInitPoints(). Otherwise the algorithm might use heuristics to generate additional points if needed.

Be aware that function.init() has to be called before calling this function!

Parameters

[in]	function	The objective function to initialize for.
[in]	initPoints	points used for initialisation. Should be at least numInitPoints().

Implemented in shark::RVEA, shark::IndicatorBasedRealCodedNSGAII< Indicator >, shark::IndicatorBasedRealCodedNSGAII< NSGA3Indicator >, shark::IndicatorBasedSteadyStateMOCMA< Indicator >, shark::SMSEMOA, shark::IndicatorBasedMOCMA< Indicator >, shark::AbstractSingleObjectiveOptimizer< PointType >, and shark::AbstractSingleObjectiveOptimizer< RealVector >.

◆ numInitPoints()

template<typename PointType, typename ResultT, typename SolutionTypeT>

virtual std::size_t shark::AbstractOptimizer< PointType, ResultT, SolutionTypeT >::numInitPoints ( ) const

pure virtual

Returns the number of points this method requires for initialisation.

The number of points supplied is allowed to be smaller, however in this case the optimizer will resort to techniques generating additional points if needed.

Implemented in shark::RVEA, shark::IndicatorBasedRealCodedNSGAII< Indicator >, shark::IndicatorBasedRealCodedNSGAII< NSGA3Indicator >, shark::MOEAD, shark::SMSEMOA, shark::IndicatorBasedMOCMA< Indicator >, shark::IndicatorBasedSteadyStateMOCMA< Indicator >, shark::AbstractSingleObjectiveOptimizer< PointType >, and shark::AbstractSingleObjectiveOptimizer< RealVector >.

Referenced by shark::AbstractOptimizer< RealVector, RealVector, std::vector< ResultSet< RealVector, RealVector > > >::init(), and shark::AbstractOptimizer< RealVector, RealVector, std::vector< ResultSet< RealVector, RealVector > > >::~AbstractOptimizer().

◆ requiresClosestFeasible()

template<typename PointType, typename ResultT, typename SolutionTypeT>

bool shark::AbstractOptimizer< PointType, ResultT, SolutionTypeT >::requiresClosestFeasible ( ) const

inline

Definition at line 100 of file AbstractOptimizer.h.

Referenced by shark::AbstractOptimizer< RealVector, RealVector, std::vector< ResultSet< RealVector, RealVector > > >::checkFeatures().

◆ requiresFirstDerivative()

template<typename PointType, typename ResultT, typename SolutionTypeT>

bool shark::AbstractOptimizer< PointType, ResultT, SolutionTypeT >::requiresFirstDerivative ( ) const

inline

Definition at line 91 of file AbstractOptimizer.h.

Referenced by shark::AbstractOptimizer< RealVector, RealVector, std::vector< ResultSet< RealVector, RealVector > > >::checkFeatures().

◆ requiresSecondDerivative()

template<typename PointType, typename ResultT, typename SolutionTypeT>

bool shark::AbstractOptimizer< PointType, ResultT, SolutionTypeT >::requiresSecondDerivative ( ) const

inline

Definition at line 94 of file AbstractOptimizer.h.

Referenced by shark::AbstractOptimizer< RealVector, RealVector, std::vector< ResultSet< RealVector, RealVector > > >::checkFeatures().

◆ requiresValue()

template<typename PointType, typename ResultT, typename SolutionTypeT>

bool shark::AbstractOptimizer< PointType, ResultT, SolutionTypeT >::requiresValue ( ) const

inline

Definition at line 87 of file AbstractOptimizer.h.

Referenced by shark::AbstractOptimizer< RealVector, RealVector, std::vector< ResultSet< RealVector, RealVector > > >::checkFeatures().

◆ solution()

template<typename PointType, typename ResultT, typename SolutionTypeT>

virtual SolutionType const& shark::AbstractOptimizer< PointType, ResultT, SolutionTypeT >::solution ( ) const

pure virtual

Accesses the best solution obtained so far.

Returns: An immutable reference to the best solution obtained so far.

Implemented in shark::AbstractSingleObjectiveOptimizer< PointType >, shark::AbstractSingleObjectiveOptimizer< RealVector >, shark::AbstractMultiObjectiveOptimizer< PointTypeT >, and shark::AbstractMultiObjectiveOptimizer< RealVector >.

Referenced by shark::AbstractOptimizer< RealVector, RealVector, std::vector< ResultSet< RealVector, RealVector > > >::init().

◆ step()

template<typename PointType, typename ResultT, typename SolutionTypeT>

virtual void shark::AbstractOptimizer< PointType, ResultT, SolutionTypeT >::step ( ObjectiveFunctionType const & function )

pure virtual

Carry out one step of the optimizer for the supplied objective function.

Parameters

[in] function The objective function to initialize for.

Implemented in shark::PointSearch, shark::IRpropMinus, shark::NestedGridSearch, shark::IRpropPlusFull, shark::IRpropPlus, shark::LMCMA, shark::GridSearch, shark::RpropPlus, shark::RVEA, shark::IndicatorBasedRealCodedNSGAII< Indicator >, shark::IndicatorBasedRealCodedNSGAII< NSGA3Indicator >, shark::SMSEMOA, shark::CrossEntropyMethod, shark::IndicatorBasedSteadyStateMOCMA< Indicator >, shark::IndicatorBasedMOCMA< Indicator >, shark::VDCMA, shark::MOEAD, shark::Adam, shark::RpropMinus, shark::CMSA, shark::SimplexDownhill, shark::CMA, shark::TrustRegionNewton, shark::SteepestDescent, shark::ElitistCMA, and shark::AbstractLineSearchOptimizer.

Referenced by shark::AbstractOptimizer< RealVector, RealVector, std::vector< ResultSet< RealVector, RealVector > > >::init(), and shark::OptimizationTrainer< Model, LabelTypeT >::train().

◆ updateFeatures()

template<typename PointType, typename ResultT, typename SolutionTypeT>

virtual void shark::AbstractOptimizer< PointType, ResultT, SolutionTypeT >::updateFeatures ( )

inlinevirtual

Definition at line 85 of file AbstractOptimizer.h.

Member Data Documentation

◆ m_features

template<typename PointType, typename ResultT, typename SolutionTypeT>

Features shark::AbstractOptimizer< PointType, ResultT, SolutionTypeT >::m_features

protected

Definition at line 85 of file AbstractOptimizer.h.

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

include/shark/Algorithms/AbstractOptimizer.h

Public Types

Public Member Functions

Protected Member Functions

Protected Attributes

Detailed Description

template<typename PointType, typename ResultT, typename SolutionTypeT> class shark::AbstractOptimizer< PointType, ResultT, SolutionTypeT >

Member Typedef Documentation

◆ FeatureNotAvailableException

◆ Features

◆ ObjectiveFunctionType

◆ ResultType

◆ SearchPointType

◆ SolutionType

Member Enumeration Documentation

◆ Feature

Constructor & Destructor Documentation

◆ ~AbstractOptimizer()

Member Function Documentation

◆ canSolveConstrained()

◆ checkFeatures()

◆ features()

◆ init() [1/2]

◆ init() [2/2]

◆ numInitPoints()

◆ requiresClosestFeasible()

◆ requiresFirstDerivative()

◆ requiresSecondDerivative()

◆ requiresValue()

◆ solution()

◆ step()

◆ updateFeatures()

Member Data Documentation

◆ m_features

template<typename PointType, typename ResultT, typename SolutionTypeT>
class shark::AbstractOptimizer< PointType, ResultT, SolutionTypeT >