Global chi-square likelihood with no explicit nuisance parameters. More...

#include <ChiSquaredLikelihood.h>

Inheritance diagram for ChiSquaredLikelihood:

[legend]

Collaboration diagram for ChiSquaredLikelihood:

[legend]

Public Member Functions
	ChiSquaredLikelihood (const ModelFn &model, std::shared_ptr< LikelihoodContext > ctx, std::size_t p_dim, RealMatrix covariance_inv)
	Constructs a chi-square likelihood.

double	nll (const std::vector< double > &theta) const override
	Evaluates the chi-square negative log-likelihood.

std::size_t	dim () const override
	Returns the total optimization dimension.

RealMatrix	observable_curvature (const std::vector< double > &r) const override
	Returns the observable-space curvature matrix.

RealMatrix	nuisance_curvature (const std::vector< double > &eta) const override
	Returns the nuisance-space curvature matrix.

Public Member Functions inherited from BaseLikelihood
	BaseLikelihood (const ModelFn &model, std::shared_ptr< LikelihoodContext > ctx, size_t p_dim)
	Constructs a likelihood from a model and statistical context.

std::vector< fit_app::ParameterDefinition >	get_param_defs () const override
	Returns all parameter definitions in likelihood-vector order.

void	enable_debug_trace (std::size_t max_evals=25)
	Enables debug tracing for likelihood evaluations.

void	disable_debug_trace ()
	Disables debug tracing of likelihood evaluations.

std::size_t	p_dimension () const override
	Returns the dimension of the fitted-parameter block.

std::size_t	eta_dimension () const override
	Returns the dimension of the nuisance-parameter block.

std::vector< double >	central_p () const override
	Returns the central values of the fitted parameters.

std::vector< double >	central_eta () const override
	Returns the central values of the nuisance parameters.

std::vector< double >	predict (const std::vector< double > &p, const std::vector< double > &eta) const override

std::vector< double >	residuals (const std::vector< double > &p, const std::vector< double > &eta) const override

double	nll_from_split (const std::vector< double > &p, const std::vector< double > &eta) const override

Public Member Functions inherited from ILikelihood
virtual	~ILikelihood ()=default
	Virtual destructor for safe polymorphic deletion.

Additional Inherited Members
Protected Attributes inherited from BaseLikelihood
std::shared_ptr< LikelihoodContext >	ctx
	Shared statistical context used by the likelihood.

ModelFn	model
	Model function evaluated by the likelihood.

std::size_t	p_dim
	Dimension of the fitted-parameter block.

bool	debug_trace_enabled_ = false
	Whether evaluation debug tracing is enabled.

std::size_t	debug_trace_max_evals_ = 0
	Maximum number of traced evaluations.

std::size_t	debug_eval_count_ = 0
	Number of evaluations already traced.

bool	debug_have_ref_theta_ = false
	Whether the debug reference parameter vector is set.

bool	debug_have_ref_res_ = false
	Whether the debug reference residual vector is set.

std::vector< double >	debug_ref_theta_
	First traced parameter vector used as debug reference.

std::vector< double >	debug_ref_res_
	First traced residual vector used as debug reference.

Detailed Description

Global chi-square likelihood with no explicit nuisance parameters.

The optimization vector is interpreted as the parameter vector $p$ and the nuisance vector $\eta$ is always empty. The negative log-likelihood is evaluated as

$\mathrm{NLL}(p) = \frac{1}{2} \left(f(p, \emptyset) - O_{\mathrm{exp}}\right)^\top C^{-1} \left(f(p, \emptyset) - O_{\mathrm{exp}}\right),$

where $f(p, \emptyset)$ is the model prediction, $O_{\mathrm{exp}}$ is the vector of experimental observations and $C^{-1}$ is the supplied inverse covariance matrix.

Note: This class derives from BaseLikelihood for API compatibility, but overrides the likelihood evaluation and reports a zero-dimensional nuisance space.

Definition at line 48 of file ChiSquaredLikelihood.h.

Constructor & Destructor Documentation

◆ ChiSquaredLikelihood()

ChiSquaredLikelihood::ChiSquaredLikelihood	(	const ModelFn &	model,
		std::shared_ptr< LikelihoodContext >	ctx,
		std::size_t	p_dim,
		RealMatrix	covariance_inv
	)

Constructs a chi-square likelihood.

Parameters

model	Model function mapping $(p, \eta)$ to observable predictions. For this likelihood, `eta` is expected to be empty.
ctx	Shared likelihood context containing observations and parameter definitions.
p_dim	Number of parameters of interest.
covariance_inv	Inverse covariance matrix used in the quadratic form.

Definition at line 5 of file ChiSquaredLikelihood.cpp.

Member Function Documentation

◆ dim()

std::size_t ChiSquaredLikelihood::dim ( ) const

overridevirtual

Returns the total optimization dimension.

For this class, the dimension is equal to the number of parameters of interest because no nuisance parameters are explicitly represented.

Returns: Parameter-space dimension.

Reimplemented from BaseLikelihood.

Definition at line 63 of file ChiSquaredLikelihood.cpp.

◆ nll()

double ChiSquaredLikelihood::nll ( const std::vector< double > & theta ) const

overridevirtual

Evaluates the chi-square negative log-likelihood.

Parameters

theta Parameter vector. Its size must be exactly p_dim.

Returns: Finite NLL value when the model and covariance are valid; a large penalty value is returned when non-finite numerical values are encountered.

Exceptions

std::invalid_argument if theta does not have dimension p_dim.

Reimplemented from BaseLikelihood.

Definition at line 15 of file ChiSquaredLikelihood.cpp.

◆ nuisance_curvature()

RealMatrix ChiSquaredLikelihood::nuisance_curvature ( const std::vector< double > & eta ) const

overridevirtual

Returns the nuisance-space curvature matrix.

Parameters

eta	Nuisance vector. Must be empty for this likelihood.

Returns: An empty $0 \times 0$ matrix.

Exceptions

std::invalid_argument if eta is not empty.

Reimplemented from BaseLikelihood.

Definition at line 73 of file ChiSquaredLikelihood.cpp.

◆ observable_curvature()

RealMatrix ChiSquaredLikelihood::observable_curvature ( const std::vector< double > & r ) const

overridevirtual

Returns the observable-space curvature matrix.

Parameters

r	Residual vector. The argument is accepted for interface compatibility and is not used because the curvature is constant.

Returns: The inverse covariance matrix supplied at construction.

Reimplemented from BaseLikelihood.

Definition at line 67 of file ChiSquaredLikelihood.cpp.

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

/home/runner/work/Hyperiso/Hyperiso/Hyperiso/Hyperiso/core/src/Statistic/domain/Likelihood/ChiSquaredLikelihood.h
/home/runner/work/Hyperiso/Hyperiso/Hyperiso/Hyperiso/core/src/Statistic/domain/Likelihood/ChiSquaredLikelihood.cpp

Public Member Functions

Additional Inherited Members

Detailed Description

Constructor & Destructor Documentation

◆ ChiSquaredLikelihood()

Member Function Documentation

◆ dim()

◆ nll()

◆ nuisance_curvature()

◆ observable_curvature()