LightGBM::ObjectiveFunction Class Reference

The interface of Objective Function. More...

#include <objective_function.h>

Inheritance diagram for LightGBM::ObjectiveFunction:

Public Member Functions
virtual	~ObjectiveFunction ()
	virtual destructor
virtual void	Init (const Metadata &metadata, data_size_t num_data)=0
	Initialize.
virtual void	GetGradients (const double *score, score_t *gradients, score_t *hessians) const =0
	calculating first order derivative of loss function
virtual const char *	GetName () const =0
virtual bool	IsConstantHessian () const
virtual bool	IsRenewTreeOutput () const
virtual double	RenewTreeOutput (double ori_output, const double *, const data_size_t *, const data_size_t *, data_size_t) const
virtual double	RenewTreeOutput (double ori_output, double, const data_size_t *, const data_size_t *, data_size_t) const
virtual double	BoostFromScore (int) const
virtual bool	ClassNeedTrain (int) const
virtual bool	SkipEmptyClass () const
virtual int	NumModelPerIteration () const
virtual int	NumPredictOneRow () const
virtual bool	NeedAccuratePrediction () const
	The prediction should be accurate or not. True will disable early stopping for prediction.
virtual void	ConvertOutput (const double *input, double *output) const
virtual std::string	ToString () const =0
ObjectiveFunction &	operator= (const ObjectiveFunction &)=delete
	Disable copy.
	ObjectiveFunction (const ObjectiveFunction &)=delete
	Disable copy.

Static Public Member Functions
static LIGHTGBM_EXPORT ObjectiveFunction *	CreateObjectiveFunction (const std::string &type, const Config &config)
	Create object of objective function.
static LIGHTGBM_EXPORT ObjectiveFunction *	CreateObjectiveFunction (const std::string &str)
	Load objective function from string object.

Detailed Description

The interface of Objective Function.

Member Function Documentation

CreateObjectiveFunction()

ObjectiveFunction * LightGBM::ObjectiveFunction::CreateObjectiveFunction ( const std::string &  type, const Config &  config  )

static

Create object of objective function.

Parameters

type	Specific type of objective function
config	Config for objective function

GetGradients()

virtual void LightGBM::ObjectiveFunction::GetGradients ( const double *  score, score_t *  gradients, score_t *  hessians  ) const

pure virtual

calculating first order derivative of loss function

Parameters

score

prediction score in this round \gradients Output gradients \hessians Output hessians

Implemented in LightGBM::BinaryLogloss, LightGBM::MulticlassSoftmax, LightGBM::MulticlassOVA, LightGBM::LambdarankNDCG, LightGBM::RegressionL2loss, LightGBM::RegressionL1loss, LightGBM::RegressionHuberLoss, LightGBM::RegressionFairLoss, LightGBM::RegressionPoissonLoss, LightGBM::RegressionQuantileloss, LightGBM::RegressionMAPELOSS, LightGBM::RegressionGammaLoss, LightGBM::RegressionTweedieLoss, LightGBM::CrossEntropy, and LightGBM::CrossEntropyLambda.

Init()

virtual void LightGBM::ObjectiveFunction::Init ( const Metadata &  metadata, data_size_t  num_data  )

pure virtual

Initialize.

Parameters

metadata	Label data
num_data	Number of data

Implemented in LightGBM::BinaryLogloss, LightGBM::MulticlassSoftmax, LightGBM::MulticlassOVA, LightGBM::LambdarankNDCG, LightGBM::RegressionL2loss, LightGBM::RegressionPoissonLoss, LightGBM::RegressionMAPELOSS, LightGBM::CrossEntropy, and LightGBM::CrossEntropyLambda.

NeedAccuratePrediction()

virtual bool LightGBM::ObjectiveFunction::NeedAccuratePrediction ( ) const

inlinevirtual

The prediction should be accurate or not. True will disable early stopping for prediction.

Reimplemented in LightGBM::BinaryLogloss, LightGBM::MulticlassSoftmax, LightGBM::MulticlassOVA, and LightGBM::LambdarankNDCG.

---

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

• External/LightGBM_2.2.3/LightGBM-2.2.3/include/LightGBM/objective_function.h
• External/LightGBM_2.2.3/LightGBM-2.2.3/src/objective/objective_function.cpp