URANIE::Modeler::TGPLooCostFunction Class Reference

Description of the class TGPLooCostFunction. More...

#include <TGPLooCostFunction.h>

Inheritance diagram for URANIE::Modeler::TGPLooCostFunction:

Collaboration diagram for URANIE::Modeler::TGPLooCostFunction:

Public Member Functions
Constructor and Destructor
	TGPLooCostFunction (TGPBuilder *gpb)
	Standard constructor.
virtual	~TGPLooCostFunction ()
	Default destructor.
virtual URANIE::Relauncher::TEval *	rebuild (int rank, Bool_t chef=kTRUE)
virtual Bool_t	unbuild ()
Estimate
virtual Int_t	eval (Double_t *theta, Double_t *res, int=0)
	Evaluate the cost function.
virtual Int_t	eval (vector< URANIE::DataServer::UEntry * > *, vector< URANIE::DataServer::UEntry * > *, int=0)
Public Member Functions inherited from URANIE::Modeler::TGPCostFunction
	TGPCostFunction (TGPBuilder *gpb)
	Standard constructor.
virtual	~TGPCostFunction ()
	Default destructor.
TGPBuilder *	getBuilder ()
	Return a pointer to the GP Builder.
Double_t *	getCovarianceMatrix ()
	Return a pointer to the GP Builder covariance matrix.
Double_t *	getCorrelationMatrix ()
	Return a pointer to the GP Builder correlation matrix.
Double_t *	getObservationVector ()
	Return a pointer to the observations outputs.
Int_t	getNObs ()
	Return the number of observations.
void	setBuilder (TGPBuilder *gpb)
void	setLog ()
void	unsetLog ()
void	changeLog ()
Bool_t	getLog ()
virtual void	printLog (Option_t *option="")

Private Attributes
Double_t *	_BetaPrior
	pointer to the a priori means of the deterministic trend parameters (size: _n * _p)
Double_t *	_QPrior
	pointer to the a priori covariance matrix of the deterministic trend parameters (size: _n * _n)

Additional Inherited Members
Public Attributes inherited from URANIE::Modeler::TGPCostFunction
Bool_t	_blog
	Boolean to decide if the log information is shown or not.
TGPBuilder *	gpBuilder
	pointer to a gaussian process builder object.
Protected Attributes inherited from URANIE::Modeler::TGPCostFunction
Double_t *	_H
	pointer to the deterministic trend matrix (size: _n * _p)
Double_t *	_K
	pointer to the covariance matrix (size: _n * _n)
Double_t *	_C
	pointer to the correlation matrix (size: _n * _n)
Double_t *	_yObs
	pointer to the vector of observations outputs (size: _n)
Int_t	_n
	number of observations
Int_t	_p
	number of deterministic trend coefficients

Detailed Description

Description of the class TGPLooCostFunction.

This class defines a Leave One Out cost function used to select the optimal hyper-parameters of a Gaussian Process. It is a wrapper to the gpLoo0, gpLooh and gpLoohBayes cost functions provided by the "gaussian process Library" (gpLib), by Jean-Marc Martinez.

Constructor & Destructor Documentation

TGPLooCostFunction()

URANIE::Modeler::TGPLooCostFunction::TGPLooCostFunction

(

TGPBuilder *

gpb

)

Standard constructor.

A TGPLooCostFunction object need to be associated with a TGPBuilder object which will provide the covariance matrix and the observations.

Parameters

gpb	(TGPBuilder*): pointer to a Gaussian Process Builder object.

Referenced by ClassImp().

~TGPLooCostFunction()

virtual URANIE::Modeler::TGPLooCostFunction::~TGPLooCostFunction ( )

virtual

Default destructor.

Referenced by ClassImp().

Member Function Documentation

eval() [1/2]

virtual Int_t URANIE::Modeler::TGPLooCostFunction::eval ( Double_t *  theta, Double_t *  res, int  = 0  )

virtual

Evaluate the cost function.

This function provides compatibility with the Relauncher module. It takes two arrays of double and returns an integer: 1 if all went well, 0 otherwise.

Parameters

theta	(Double_t*): a single set of parameters to optimise.
res	(Double_t*): output of the cost function.

Warning

If a measurement error is taken into account and its variance is provided by the user, the last element of the array "theta" is the GP variance. If the measurement error variance exists but is unknown, the last element of "theta" is the value of the parameter alpha.

Implements URANIE::Modeler::TGPCostFunction.

Referenced by ClassImp().

eval() [2/2]

virtual Int_t URANIE::Modeler::TGPLooCostFunction::eval ( vector< URANIE::DataServer::UEntry * > *  , vector< URANIE::DataServer::UEntry * > *  , int  = 0  )

virtual

rebuild()

virtual URANIE::Relauncher::TEval * URANIE::Modeler::TGPLooCostFunction::rebuild ( int  rank, Bool_t  chef = kTRUE  )

virtual

Referenced by ClassImp().

unbuild()

virtual Bool_t URANIE::Modeler::TGPLooCostFunction::unbuild ( )

virtual

Referenced by ClassImp().

Member Data Documentation

_BetaPrior

Double_t* URANIE::Modeler::TGPLooCostFunction::_BetaPrior

private

pointer to the a priori means of the deterministic trend parameters (size: _n * _p)

Referenced by ClassImp().

_QPrior

Double_t* URANIE::Modeler::TGPLooCostFunction::_QPrior

private

pointer to the a priori covariance matrix of the deterministic trend parameters (size: _n * _n)

Referenced by ClassImp().