sgpp::datadriven::NegativeLogLikelihood Class Reference

Metric that quantifies the likelihood of a dataset given the density function. More...

#include <NegativeLogLikelihood.hpp>

Inheritance diagram for sgpp::datadriven::NegativeLogLikelihood:

Public Member Functions
Metric *	clone () const override
	Standard clone method.
double	measure (const DataVector &predictedValues, const DataVector &trueValues, const ModelFittingBase &model, Dataset &testDataset) const override
	Quantify the NLL of the predicted values (i.e.
double	measureLowerIsBetter (const DataVector &predictedValues, const DataVector &trueValues, const ModelFittingBase &model, Dataset &testDataset) const override
	Quantify the NLL of the predicted values (i.e.
Public Member Functions inherited from sgpp::datadriven::Metric
	Metric ()=default
	Default constructor.
	Metric (const Metric &rhs)=default
	Copy constructor.
	Metric (Metric &&rhs)=default
	Move constructor.
Metric &	operator= (const Metric &rhs)=default
	Copy assign operator.
Metric &	operator= (Metric &&rhs)=default
	Move assign operator.
virtual	~Metric ()=default
	virtual destructor.

Detailed Description

Metric that quantifies the likelihood of a dataset given the density function.

The smaller the negative log likelihood the better the fit

Member Function Documentation

clone()

Metric * sgpp::datadriven::NegativeLogLikelihood::clone ( ) const

overridevirtual

Standard clone method.

Returns

the cloned metric instance

Implements sgpp::datadriven::Metric.

measure()

double sgpp::datadriven::NegativeLogLikelihood::measure ( const DataVector &  predictedValues, const DataVector &  trueValues, const ModelFittingBase &  model, Dataset &  testDataset  ) const

overridevirtual

Quantify the NLL of the predicted values (i.e.

adding the logs of the predicted values and ignorign the true values). Note that probabilities <= 0 are simply ignored (the model can provide those)

Parameters

predictedValues	probabilites calculated by the model for testing data
trueValues	ignored
model	reference to the model
testDataset	dataset with test data

Returns

the negative log likelihood of the predicted probabilities

Implements sgpp::datadriven::Metric.

References sgpp::base::DataVector::get().

Referenced by measureLowerIsBetter().

measureLowerIsBetter()

double sgpp::datadriven::NegativeLogLikelihood::measureLowerIsBetter ( const DataVector &  predictedValues, const DataVector &  trueValues, const ModelFittingBase &  model, Dataset &  testDataset  ) const

overridevirtual

Quantify the NLL of the predicted values (i.e.

adding the logs of the predicted values and ignorign the true values). Note that probabilities <= 0 are simply ignored (the model can provide those)

Parameters

predictedValues	probabilites calculated by the model for testing data
trueValues	ignored
model	reference to the model
testDataset	dataset with test data

Returns

the negative log likelihood of the predicted probabilities

Implements sgpp::datadriven::Metric.

References measure().

---

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

• datadriven/src/sgpp/datadriven/datamining/modules/scoring/NegativeLogLikelihood.hpp
• datadriven/src/sgpp/datadriven/datamining/modules/scoring/NegativeLogLikelihood.cpp