Base class for pair-wise learning to rank. More...

Inheritance diagram for xgboost::obj::LambdaRankObj< Loss, Cache >:

Public Member Functions
void	Configure (Args const &args) override
	Configure the objective with the specified parameters.

void	SaveConfig (Json *p_out) const override
	Save configuration to JSON object.

void	LoadConfig (Json const &in) override
	Load configuration from JSON object.

ObjInfo	Task () const override
	Return task of this objective.

bst_target_t	Targets (MetaInfo const &info) const override
	Return number of targets for input matrix.

const char *	RankEvalMetric (StringView metric) const

void	GetGradient (HostDeviceVector< float > const &predt, MetaInfo const &info, std::int32_t iter, HostDeviceVector< GradientPair > *out_gpair) override

Public Member Functions inherited from xgboost::ObjFunction
	~ObjFunction () override=default
	virtual destructor

virtual void	GetGradient (const HostDeviceVector< bst_float > &preds, const MetaInfo &info, int iteration, HostDeviceVector< GradientPair > *out_gpair)=0
	Get gradient over each of predictions, given existing information.

virtual const char *	DefaultEvalMetric () const =0

virtual Json	DefaultMetricConfig () const
	Return the configuration for the default metric.

virtual void	PredTransform (HostDeviceVector< bst_float > *) const
	transform prediction values, this is only called when Prediction is called

virtual void	EvalTransform (HostDeviceVector< bst_float > *io_preds)
	transform prediction values, this is only called when Eval is called, usually it redirect to PredTransform

virtual bst_float	ProbToMargin (bst_float base_score) const
	transform probability value back to margin this is used to transform user-set base_score back to margin used by gradient boosting

virtual void	InitEstimation (MetaInfo const &info, linalg::Tensor< float, 1 > *base_score) const
	Make initialize estimation of prediction.

virtual void	UpdateTreeLeaf (HostDeviceVector< bst_node_t > const &, MetaInfo const &, float, HostDeviceVector< float > const &, std::int32_t, RegTree *) const
	Update the leaf values after a tree is built.

Protected Member Functions
std::shared_ptr< Cache >	GetCache () const

linalg::VectorView< double >	GroupLoss (bst_group_t g, linalg::Vector< double > *v) const

template<bool unbiased, typename Delta >
void	CalcLambdaForGroup (std::int32_t iter, common::Span< float const > g_predt, linalg::VectorView< float const > g_label, float w, common::Span< std::size_t const > g_rank, bst_group_t g, Delta delta, common::Span< GradientPair > g_gpair)

Protected Attributes
linalg::Vector< double >	li_

linalg::Vector< double >	lj_

linalg::Vector< double >	ti_plus_

linalg::Vector< double >	tj_minus_

linalg::Vector< double >	li_full_

linalg::Vector< double >	lj_full_

ltr::LambdaRankParam	param_

std::shared_ptr< ltr::RankingCache >	p_cache_

Protected Attributes inherited from xgboost::ObjFunction
Context const *	ctx_

Additional Inherited Members
Static Public Member Functions inherited from xgboost::ObjFunction
static constexpr float	DefaultBaseScore ()

static ObjFunction *	Create (const std::string &name, Context const *ctx)
	Create an objective function according to name.

Detailed Description

template<typename Loss, typename Cache>
class xgboost::obj::LambdaRankObj< Loss, Cache >

Base class for pair-wise learning to rank.

See From RankNet to LambdaRank to LambdaMART: An Overview for a description of the algorithm.

In addition to ranking, this also implements Unbiased LambdaMART: An Unbiased Pairwise Learning-to-Rank Algorithm.

Member Function Documentation

◆ Configure()

template<typename Loss , typename Cache >

void xgboost::obj::LambdaRankObj< Loss, Cache >::Configure ( Args const & args )

inlineoverridevirtual

Configure the objective with the specified parameters.

Parameters

args	arguments to the objective function.

Implements xgboost::ObjFunction.

◆ LoadConfig()

template<typename Loss , typename Cache >

void xgboost::obj::LambdaRankObj< Loss, Cache >::LoadConfig ( Json const & in )

inlineoverridevirtual

Load configuration from JSON object.

Parameters

in	JSON object containing the configuration

Implements xgboost::Configurable.

◆ SaveConfig()

template<typename Loss , typename Cache >

void xgboost::obj::LambdaRankObj< Loss, Cache >::SaveConfig ( Json * out ) const

inlineoverridevirtual

Save configuration to JSON object.

Parameters

out	pointer to output JSON object

Implements xgboost::Configurable.

◆ Targets()

template<typename Loss , typename Cache >

bst_target_t xgboost::obj::LambdaRankObj< Loss, Cache >::Targets ( MetaInfo const & info ) const

inlineoverridevirtual

Return number of targets for input matrix.

Right now XGBoost supports only multi-target regression.

Reimplemented from xgboost::ObjFunction.

◆ Task()

template<typename Loss , typename Cache >

ObjInfo xgboost::obj::LambdaRankObj< Loss, Cache >::Task ( ) const

inlineoverridevirtual

Return task of this objective.

Implements xgboost::ObjFunction.

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

External/xgboost/src/objective/lambdarank_obj.cc

Public Member Functions

Protected Member Functions

Protected Attributes

Additional Inherited Members

Detailed Description

Member Function Documentation

◆ Configure()

◆ LoadConfig()

◆ SaveConfig()

◆ Targets()

◆ Task()