Used for learning a tree by single machine. More...

#include <serial_tree_learner.h>

Inheritance diagram for LightGBM::SerialTreeLearner:

Public Member Functions
	SerialTreeLearner (const Config *config)

void	Init (const Dataset *train_data, bool is_constant_hessian) override
	Initialize tree learner with training dataset.

void	ResetTrainingData (const Dataset *train_data) override

void	ResetConfig (const Config *config) override
	Reset tree configs.

Tree *	Train (const score_t gradients, const score_t hessians, bool is_constant_hessian, Json &forced_split_json) override
	training tree model on dataset

Tree *	FitByExistingTree (const Tree old_tree, const score_t gradients, const score_t *hessians) const override
	use a existing tree to fit the new gradients and hessians.

Tree *	FitByExistingTree (const Tree old_tree, const std::vector< int > &leaf_pred, const score_t gradients, const score_t *hessians) override

void	SetBaggingData (const data_size_t *used_indices, data_size_t num_data) override
	Set bagging data.

void	AddPredictionToScore (const Tree tree, double out_score) const override
	Using last trained tree to predict score then adding to out_score;.

void	RenewTreeOutput (Tree tree, const ObjectiveFunction obj, const double prediction, data_size_t total_num_data, const data_size_t bag_indices, data_size_t bag_cnt) const override

void	RenewTreeOutput (Tree tree, const ObjectiveFunction obj, double prediction, data_size_t total_num_data, const data_size_t *bag_indices, data_size_t bag_cnt) const override

Public Member Functions inherited from LightGBM::TreeLearner
virtual	~TreeLearner ()
	virtual destructor

TreeLearner &	operator= (const TreeLearner &)=delete
	Disable copy.

	TreeLearner (const TreeLearner &)=delete
	Disable copy.

Protected Member Functions
virtual void	BeforeTrain ()
	Some initial works before training.

virtual bool	BeforeFindBestSplit (const Tree *tree, int left_leaf, int right_leaf)
	Some initial works before FindBestSplit.

virtual void	FindBestSplits ()

virtual void	ConstructHistograms (const std::vector< int8_t > &is_feature_used, bool use_subtract)

virtual void	FindBestSplitsFromHistograms (const std::vector< int8_t > &is_feature_used, bool use_subtract)

virtual void	Split (Tree tree, int best_leaf, int left_leaf, int *right_leaf)
	Partition tree and data according best split.

virtual int32_t	ForceSplits (Tree tree, Json &forced_split_json, int left_leaf, int right_leaf, int cur_depth, bool *aborted_last_force_split)

virtual data_size_t	GetGlobalDataCountInLeaf (int leaf_idx) const
	Get the number of data in a leaf.

Protected Attributes
data_size_t	num_data_
	number of data

int	num_features_
	number of features

const Dataset *	train_data_
	training data

const score_t *	gradients_
	gradients of current iteration

const score_t *	hessians_
	hessians of current iteration

std::unique_ptr< DataPartition >	data_partition_
	training data partition on leaves

Random	random_
	used for generate used features

std::vector< int8_t >	is_feature_used_
	used for sub feature training, is_feature_used_[i] = false means don't used feature i

FeatureHistogram *	parent_leaf_histogram_array_
	pointer to histograms array of parent of current leaves

FeatureHistogram *	smaller_leaf_histogram_array_
	pointer to histograms array of smaller leaf

FeatureHistogram *	larger_leaf_histogram_array_
	pointer to histograms array of larger leaf

std::vector< SplitInfo >	best_split_per_leaf_
	store best split points for all leaves

std::unique_ptr< LeafSplits >	smaller_leaf_splits_
	stores best thresholds for all feature for smaller leaf

std::unique_ptr< LeafSplits >	larger_leaf_splits_
	stores best thresholds for all feature for larger leaf

std::vector< int >	valid_feature_indices_

std::vector< score_t >	ordered_gradients_
	gradients of current iteration, ordered for cache optimized

std::vector< score_t >	ordered_hessians_
	hessians of current iteration, ordered for cache optimized

std::vector< std::unique_ptr< OrderedBin > >	ordered_bins_
	Store ordered bin.

bool	has_ordered_bin_ = false
	True if has ordered bin.

std::vector< char >	is_data_in_leaf_
	is_data_in_leaf_[i] != 0 means i-th data is marked

HistogramPool	histogram_pool_
	used to cache historical histogram to speed up

const Config *	config_
	config of tree learner

int	num_threads_

std::vector< int >	ordered_bin_indices_

bool	is_constant_hessian_

Additional Inherited Members
Static Public Member Functions inherited from LightGBM::TreeLearner
static TreeLearner *	CreateTreeLearner (const std::string &learner_type, const std::string &device_type, const Config *config)
	Create object of tree learner.

Detailed Description

Used for learning a tree by single machine.

Member Function Documentation

◆ AddPredictionToScore()

void LightGBM::SerialTreeLearner::AddPredictionToScore	(	const Tree *	tree,
		double *	out_score
	)		const

inlineoverridevirtual

Using last trained tree to predict score then adding to out_score;.

Parameters

out_score output score

Implements LightGBM::TreeLearner.

◆ FitByExistingTree() [1/2]

Tree * LightGBM::SerialTreeLearner::FitByExistingTree	(	const Tree *	old_tree,
		const score_t *	gradients,
		const score_t *	hessians
	)		const

overridevirtual

use a existing tree to fit the new gradients and hessians.

Implements LightGBM::TreeLearner.

◆ FitByExistingTree() [2/2]

Tree * LightGBM::SerialTreeLearner::FitByExistingTree	(	const Tree *	old_tree,
		const std::vector< int > &	leaf_pred,
		const score_t *	gradients,
		const score_t *	hessians
	)

overridevirtual

Implements LightGBM::TreeLearner.

◆ GetGlobalDataCountInLeaf()

data_size_t LightGBM::SerialTreeLearner::GetGlobalDataCountInLeaf ( int leaf_idx ) const

inlineprotectedvirtual

Get the number of data in a leaf.

Parameters

leaf_idx The index of leaf

Returns: The number of data in the leaf_idx leaf

◆ Init()

void LightGBM::SerialTreeLearner::Init	(	const Dataset *	train_data,
		bool	is_constant_hessian
	)

overridevirtual

Initialize tree learner with training dataset.

Parameters

train_data	The used training data
is_constant_hessian	True if all hessians share the same value

Implements LightGBM::TreeLearner.

◆ RenewTreeOutput() [1/2]

void LightGBM::SerialTreeLearner::RenewTreeOutput	(	Tree *	tree,
		const ObjectiveFunction *	obj,
		const double *	prediction,
		data_size_t	total_num_data,
		const data_size_t *	bag_indices,
		data_size_t	bag_cnt
	)		const

overridevirtual

Implements LightGBM::TreeLearner.

◆ RenewTreeOutput() [2/2]

void LightGBM::SerialTreeLearner::RenewTreeOutput	(	Tree *	tree,
		const ObjectiveFunction *	obj,
		double	prediction,
		data_size_t	total_num_data,
		const data_size_t *	bag_indices,
		data_size_t	bag_cnt
	)		const

overridevirtual

Implements LightGBM::TreeLearner.

◆ ResetConfig()

void LightGBM::SerialTreeLearner::ResetConfig ( const Config * config )

overridevirtual

Reset tree configs.

Parameters

config config of tree

Implements LightGBM::TreeLearner.

◆ ResetTrainingData()

void LightGBM::SerialTreeLearner::ResetTrainingData ( const Dataset * train_data )

overridevirtual

Implements LightGBM::TreeLearner.

◆ SetBaggingData()

void LightGBM::SerialTreeLearner::SetBaggingData	(	const data_size_t *	used_indices,
		data_size_t	num_data
	)

inlineoverridevirtual

Set bagging data.

Parameters

used_indices	Used data indices
num_data	Number of used data

Implements LightGBM::TreeLearner.

◆ Split()

void LightGBM::SerialTreeLearner::Split	(	Tree *	tree,
		int	best_leaf,
		int *	left_leaf,
		int *	right_leaf
	)

protectedvirtual

Partition tree and data according best split.

Parameters

tree	Current tree, will be splitted on this function.
best_leaf	The index of leaf that will be splitted.
left_leaf	The index of left leaf after splitted.
right_leaf	The index of right leaf after splitted.

◆ Train()

Tree * LightGBM::SerialTreeLearner::Train	(	const score_t *	gradients,
		const score_t *	hessians,
		bool	is_constant_hessian,
		Json &	forced_split_json
	)

overridevirtual

training tree model on dataset

Parameters

gradients	The first order gradients
hessians	The second order gradients
is_constant_hessian	True if all hessians share the same value

Returns: A trained tree

Implements LightGBM::TreeLearner.

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

External/LightGBM_2.2.3/LightGBM-2.2.3/src/treelearner/serial_tree_learner.h
External/LightGBM_2.2.3/LightGBM-2.2.3/src/treelearner/serial_tree_learner.cpp

Public Member Functions

Protected Member Functions

Protected Attributes

Additional Inherited Members

Detailed Description

Member Function Documentation

◆ AddPredictionToScore()

◆ FitByExistingTree() [1/2]

◆ FitByExistingTree() [2/2]

◆ GetGlobalDataCountInLeaf()

◆ Init()

◆ RenewTreeOutput() [1/2]

◆ RenewTreeOutput() [2/2]

◆ ResetConfig()

◆ ResetTrainingData()

◆ SetBaggingData()

◆ Split()

◆ Train()