shark::NegativeGaussianProcessEvidence< InputType, OutputType, LabelType > Class Template Reference

Evidence for model selection of a regularization network/Gaussian process. More...

#include <shark/ObjectiveFunctions/NegativeGaussianProcessEvidence.h>

+ Inheritance diagram for shark::NegativeGaussianProcessEvidence< InputType, OutputType, LabelType >:

Public Types

typedef LabeledData< InputType, LabelType > DatasetType
 
typedef AbstractKernelFunction< InputTypeKernelType
 
- Public Types inherited from shark::AbstractObjectiveFunction< RealVector, double >
enum  Feature
 List of features that are supported by an implementation. More...
 
typedef RealVector SearchPointType
 
typedef double ResultType
 
typedef boost::mpl::if_< std::is_arithmetic< double >, SearchPointType, RealMatrix >::type FirstOrderDerivative
 
typedef TypedFlags< FeatureFeatures
 This statement declares the member m_features. See Core/Flags.h for details. More...
 
typedef TypedFeatureNotAvailableException< FeatureFeatureNotAvailableException
 

Public Member Functions

 NegativeGaussianProcessEvidence (DatasetType const &dataset, KernelType *kernel, bool unconstrained=false)
 
std::string name () const
 From INameable: return the class name. More...
 
std::size_t numberOfVariables () const
 Accesses the number of variables. More...
 
double eval (const RealVector &parameters) const
 
double evalDerivative (const RealVector &parameters, FirstOrderDerivative &derivative) const
 
void setThreshold (double d)
 set threshold value for truncating partial derivatives More...
 
void setThresholds (RealVector &c)
 set threshold values for truncating partial derivatives More...
 
- Public Member Functions inherited from shark::AbstractObjectiveFunction< RealVector, double >
const Featuresfeatures () const
 
virtual void updateFeatures ()
 
bool hasValue () const
 returns whether this function can calculate it's function value More...
 
bool hasFirstDerivative () const
 returns whether this function can calculate the first derivative More...
 
bool hasSecondDerivative () const
 returns whether this function can calculate the second derivative More...
 
bool canProposeStartingPoint () const
 returns whether this function can propose a starting point. More...
 
bool isConstrained () const
 returns whether this function can return More...
 
bool hasConstraintHandler () const
 returns whether this function can return More...
 
bool canProvideClosestFeasible () const
 Returns whether this function can calculate thee closest feasible to an infeasible point. More...
 
bool isThreadSafe () const
 Returns true, when the function can be usd in parallel threads. More...
 
bool isNoisy () const
 Returns true, when the function can be usd in parallel threads. More...
 
 AbstractObjectiveFunction ()
 Default ctor. More...
 
virtual ~AbstractObjectiveFunction ()
 Virtual destructor. More...
 
virtual void init ()
 
void setRng (random::rng_type *rng)
 Sets the Rng used by the objective function. More...
 
virtual bool hasScalableDimensionality () const
 
virtual void setNumberOfVariables (std::size_t numberOfVariables)
 Adjusts the number of variables if the function is scalable. More...
 
virtual std::size_t numberOfObjectives () const
 
virtual bool hasScalableObjectives () const
 
virtual void setNumberOfObjectives (std::size_t numberOfObjectives)
 Adjusts the number of objectives if the function is scalable. More...
 
std::size_t evaluationCounter () const
 Accesses the evaluation counter of the function. More...
 
AbstractConstraintHandler< SearchPointType > const & getConstraintHandler () const
 Returns the constraint handler of the function if it has one. More...
 
virtual bool isFeasible (const SearchPointType &input) const
 Tests whether a point in SearchSpace is feasible, e.g., whether the constraints are fulfilled. More...
 
virtual void closestFeasible (SearchPointType &input) const
 If supported, the supplied point is repaired such that it satisfies all of the function's constraints. More...
 
virtual SearchPointType proposeStartingPoint () const
 Proposes a starting point in the feasible search space of the function. More...
 
ResultType operator() (SearchPointType const &input) const
 Evaluates the function. Useful together with STL-Algorithms like std::transform. More...
 
virtual ResultType evalDerivative (SearchPointType const &input, SecondOrderDerivative &derivative) const
 Evaluates the objective function and calculates its gradient. More...
 
- Public Member Functions inherited from shark::INameable
virtual ~INameable ()
 

Additional Inherited Members

- Protected Member Functions inherited from shark::AbstractObjectiveFunction< RealVector, double >
void announceConstraintHandler (AbstractConstraintHandler< SearchPointType > const *handler)
 helper function which is called to announce the presence of an constraint handler. More...
 
- Protected Attributes inherited from shark::AbstractObjectiveFunction< RealVector, double >
Features m_features
 
std::size_t m_evaluationCounter
 Evaluation counter, default value: 0. More...
 
AbstractConstraintHandler< SearchPointType > const * m_constraintHandler
 
random::rng_type * mep_rng
 

Detailed Description

template<class InputType = RealVector, class OutputType = RealVector, class LabelType = RealVector>
class shark::NegativeGaussianProcessEvidence< InputType, OutputType, LabelType >

Evidence for model selection of a regularization network/Gaussian process.

Let \(M\) denote the (kernel Gram) covariance matrix and \(t\) the corresponding label vector. For the evidence we have:

\[ E = 1/2 \cdot [ -\log(\det(M)) - t^T M^{-1} t - N \log(2 \pi)] \]

The evidence is also known as marginal (log)likelihood. For details, please see:

C.E. Rasmussen & C.K.I. Williams, Gaussian Processes for Machine Learning, section 5.4, MIT Press, 2006

C.M. Bishop, Pattern Recognition and Machine Learning, section 6.4.3, Springer, 2006

The regularization parameter can be encoded in different ways. The exponential encoding is the proper choice for unconstraint optimization. Be careful not to mix up different encodings between trainer and evidence.

Definition at line 66 of file NegativeGaussianProcessEvidence.h.

Member Typedef Documentation

◆ DatasetType

template<class InputType = RealVector, class OutputType = RealVector, class LabelType = RealVector>
typedef LabeledData<InputType,LabelType> shark::NegativeGaussianProcessEvidence< InputType, OutputType, LabelType >::DatasetType

Definition at line 69 of file NegativeGaussianProcessEvidence.h.

◆ KernelType

template<class InputType = RealVector, class OutputType = RealVector, class LabelType = RealVector>
typedef AbstractKernelFunction<InputType> shark::NegativeGaussianProcessEvidence< InputType, OutputType, LabelType >::KernelType

Definition at line 70 of file NegativeGaussianProcessEvidence.h.

Constructor & Destructor Documentation

◆ NegativeGaussianProcessEvidence()

template<class InputType = RealVector, class OutputType = RealVector, class LabelType = RealVector>
shark::NegativeGaussianProcessEvidence< InputType, OutputType, LabelType >::NegativeGaussianProcessEvidence ( DatasetType const &  dataset,
KernelType kernel,
bool  unconstrained = false 
)
inline
Parameters
datasettraining data for the Gaussian process
kernelpointer to external kernel function
unconstrainedexponential encoding of regularization parameter for unconstraint optimization

Definition at line 75 of file NegativeGaussianProcessEvidence.h.

References shark::AbstractObjectiveFunction< RealVector, double >::HAS_FIRST_DERIVATIVE, shark::AbstractKernelFunction< InputTypeT >::hasFirstParameterDerivative(), shark::AbstractObjectiveFunction< RealVector, double >::m_features, and shark::NegativeGaussianProcessEvidence< InputType, OutputType, LabelType >::setThreshold().

Member Function Documentation

◆ eval()

template<class InputType = RealVector, class OutputType = RealVector, class LabelType = RealVector>
double shark::NegativeGaussianProcessEvidence< InputType, OutputType, LabelType >::eval ( const RealVector &  parameters) const
inlinevirtual

Let \(M\) denote the (kernel Gram) covariance matrix and \(t\) the label vector. For the evidence we have:

\[ E= 1/2 \cdot [ -\log(\det(M)) - t^T M^{-1} t - N \log(2 \pi) ] \]

Reimplemented from shark::AbstractObjectiveFunction< RealVector, double >.

Definition at line 97 of file NegativeGaussianProcessEvidence.h.

References shark::AbstractObjectiveFunction< RealVector, double >::m_evaluationCounter, shark::LabeledData< InputT, LabelT >::numberOfElements(), shark::IParameterizable< VectorType >::numberOfParameters(), shark::IParameterizable< VectorType >::setParameterVector(), and SHARK_ASSERT.

◆ evalDerivative()

template<class InputType = RealVector, class OutputType = RealVector, class LabelType = RealVector>
double shark::NegativeGaussianProcessEvidence< InputType, OutputType, LabelType >::evalDerivative ( const RealVector &  parameters,
FirstOrderDerivative derivative 
) const
inlinevirtual

Let \(M\) denote the regularized (kernel Gram) covariance matrix. For the evidence we have:

\[ E = 1/2 \cdot [ -\log(\det(M)) - t^T M^{-1} t - N \log(2 \pi) ] \]

For a kernel parameter \(p\) and \(C = \beta^{-1}\) we get the derivatives:

\[ dE/dC = 1/2 \cdot [ -tr(M^{-1}) + (M^{-1} t)^2 ] \]

\[ dE/dp = 1/2 \cdot [ -tr(M^{-1} dM/dp) + t^T (M^{-1} dM/dp M^{-1}) t ] \]

Reimplemented from shark::AbstractObjectiveFunction< RealVector, double >.

Definition at line 143 of file NegativeGaussianProcessEvidence.h.

References shark::AbstractObjectiveFunction< RealVector, double >::m_evaluationCounter, shark::LabeledData< InputT, LabelT >::numberOfElements(), shark::IParameterizable< VectorType >::numberOfParameters(), shark::IParameterizable< VectorType >::setParameterVector(), and SHARK_ASSERT.

◆ name()

template<class InputType = RealVector, class OutputType = RealVector, class LabelType = RealVector>
std::string shark::NegativeGaussianProcessEvidence< InputType, OutputType, LabelType >::name ( ) const
inlinevirtual

From INameable: return the class name.

Reimplemented from shark::INameable.

Definition at line 88 of file NegativeGaussianProcessEvidence.h.

◆ numberOfVariables()

template<class InputType = RealVector, class OutputType = RealVector, class LabelType = RealVector>
std::size_t shark::NegativeGaussianProcessEvidence< InputType, OutputType, LabelType >::numberOfVariables ( ) const
inlinevirtual

◆ setThreshold()

template<class InputType = RealVector, class OutputType = RealVector, class LabelType = RealVector>
void shark::NegativeGaussianProcessEvidence< InputType, OutputType, LabelType >::setThreshold ( double  d)
inline

◆ setThresholds()

template<class InputType = RealVector, class OutputType = RealVector, class LabelType = RealVector>
void shark::NegativeGaussianProcessEvidence< InputType, OutputType, LabelType >::setThresholds ( RealVector &  c)
inline

set threshold values for truncating partial derivatives

Definition at line 223 of file NegativeGaussianProcessEvidence.h.

References SHARK_ASSERT.


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