xgboost::Metric Class Reference

interface of evaluation metric used to evaluate model performance. This has nothing to do with training, but merely act as evaluation purpose. More...

#include <metric.h>

Inheritance diagram for xgboost::Metric:

Public Member Functions
virtual void	Configure (const std::vector< std::pair< std::string, std::string > > &)
	Configure the Metric with the specified parameters.
void	LoadConfig (Json const &) override
	Load configuration from JSON object By default, metric has no internal configuration; override this function to maintain internal configuration.
void	SaveConfig (Json *p_out) const override
	Save configuration to JSON object By default, metric has no internal configuration; override this function to maintain internal configuration.
virtual double	Evaluate (HostDeviceVector< bst_float > const &preds, std::shared_ptr< DMatrix > p_fmat)=0
	Evaluate a metric with DMatrix as input.
virtual const char *	Name () const =0
	~Metric () override=default
	virtual destructor

Static Public Member Functions
static Metric *	Create (const std::string &name, Context const *ctx)
	create a metric according to name.

Protected Attributes
Context const *	ctx_ {nullptr}

Detailed Description

interface of evaluation metric used to evaluate model performance. This has nothing to do with training, but merely act as evaluation purpose.

Member Function Documentation

Configure()

virtual void xgboost::Metric::Configure ( const std::vector< std::pair< std::string, std::string > > &  )

inlinevirtual

Configure the Metric with the specified parameters.

Parameters

args	arguments to the objective function.

Reimplemented in xgboost::metric::EvalRankWithCache< Cache >, xgboost::metric::EvalRankWithCache< ltr::MAPCache >, xgboost::metric::EvalRankWithCache< ltr::NDCGCache >, and xgboost::metric::EvalRankWithCache< ltr::PreCache >.

Create()

Metric * xgboost::Metric::Create ( const std::string &  name, Context const *  ctx  )

static

create a metric according to name.

Parameters

name	name of the metric. name can be in form metric[@]param and the name will be matched in the registry.
ctx	A global context

Returns

the created metric.

Evaluate()

virtual double xgboost::Metric::Evaluate ( HostDeviceVector< bst_float > const &  preds, std::shared_ptr< DMatrix >  p_fmat  )

pure virtual

Evaluate a metric with DMatrix as input.

Parameters

preds	Prediction
p_fmat	DMatrix that contains related information like labels.

Implemented in xgboost::metric::EvalRankWithCache< Cache >, xgboost::metric::EvalRankWithCache< ltr::MAPCache >, xgboost::metric::EvalRankWithCache< ltr::NDCGCache >, xgboost::metric::EvalRankWithCache< ltr::PreCache >, and xgboost::MetricNoCache.

LoadConfig()

void xgboost::Metric::LoadConfig ( Json const &  )

inlineoverridevirtual

Load configuration from JSON object By default, metric has no internal configuration; override this function to maintain internal configuration.

Parameters

in	JSON object containing the configuration

Implements xgboost::Configurable.

Name()

virtual const char * xgboost::Metric::Name ( ) const

pure virtual

Returns

name of metric

Implemented in xgboost::metric::EvalROCAUC, xgboost::metric::EvalPRAUC, xgboost::metric::EvalAMS, xgboost::metric::EvalRank, xgboost::metric::EvalCox, xgboost::metric::EvalRankWithCache< Cache >, xgboost::metric::EvalRankWithCache< ltr::MAPCache >, xgboost::metric::EvalRankWithCache< ltr::NDCGCache >, and xgboost::metric::EvalRankWithCache< ltr::PreCache >.

SaveConfig()

void xgboost::Metric::SaveConfig ( Json *  p_out ) const

inlineoverridevirtual

Save configuration to JSON object By default, metric has no internal configuration; override this function to maintain internal configuration.

Parameters

out	pointer to output JSON object

Implements xgboost::Configurable.

---

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

• External/xgboost/include/xgboost/metric.h
• External/xgboost/src/metric/metric.cc