CNum::Model::Tree::TreeBoosterNode Class Reference

A node used in a TreeBooster used for gather and storing information about the decision making process. More...

#include <TreeBoosterNode.h>

Public Member Functions
	TreeBoosterNode (TreeBoosterNode *left=nullptr, TreeBoosterNode *right=nullptr)
	Overloaded default constructor.
virtual	~TreeBoosterNode ()
std::string	to_json_string ()
	Save TreeBoosterNode data for an entire TreeBooster in a JSON formatted string.

Static Public Member Functions
static Split	find_best_split_hist (const CNum::DataStructs::Matrix< int > &X, std::shared_ptr< CNum::Data::Shelf[]> shelves, const double *g, const double *h, bool histogram_cache, const arena_view_t &hist_view, DataPartition &partition, double weight_decay=0.0, double reg_lambda=1.0, double gamma=0)
	Find the best split at a tree node with the histogram method (maximizing gain).
static Split	find_best_split_greedy (CNum::DataStructs::Matrix< double > &X, const double *g, const double *h, double weight_decay=0.0, double reg_lambda=1.0, double gamma=0)
	Find the best split at a tree node with the exact greedy method proposed in Chen & Guestrin's XGBoost (minimizing loss).

Public Attributes
Split	_split
double	_value
TreeBoosterNode *	_left
TreeBoosterNode *	_right

Friends
class	TreeBooster

Detailed Description

A node used in a TreeBooster used for gather and storing information about the decision making process.

Constructor & Destructor Documentation

TreeBoosterNode()

CNum::Model::Tree::TreeBoosterNode::TreeBoosterNode	(	TreeBoosterNode *	left = nullptr,
		TreeBoosterNode *	right = nullptr )

Overloaded default constructor.

Parameters

left	The left child of the tree
right	The right child of the tree

~TreeBoosterNode()

virtual CNum::Model::Tree::TreeBoosterNode::~TreeBoosterNode ( )

virtual

Member Function Documentation

find_best_split_greedy()

Split CNum::Model::Tree::TreeBoosterNode::find_best_split_greedy ( CNum::DataStructs::Matrix< double > & X, const double * g, const double * h, double weight_decay = 0.0, double reg_lambda = 1.0, double gamma = 0 )

static

Find the best split at a tree node with the exact greedy method proposed in Chen & Guestrin's XGBoost (minimizing loss).

Available next release

Parameters

X	The dataset
g	The gradient array
h	The hessian array
weight_decay	A parameter used in deteriming whether or not a splits effect is significant enough to take
reg_lambda	Reg Lambda; A regularization parameter
gamma	Gamma; A regularization parameter

Returns

The best split

find_best_split_hist()

Split CNum::Model::Tree::TreeBoosterNode::find_best_split_hist ( const CNum::DataStructs::Matrix< int > & X, std::shared_ptr< CNum::Data::Shelf[]> shelves, const double * g, const double * h, bool histogram_cache, const arena_view_t & hist_view, DataPartition & partition, double weight_decay = 0.0, double reg_lambda = 1.0, double gamma = 0 )

static

Find the best split at a tree node with the histogram method (maximizing gain).

Parameters

X	The dataset
shelves	The bins and values associated with their boundaries
g	The gradient array
h	The hessian array
histogram_cache	Whether or not the histograms have already been built
hist_view	The view of the histograms
partition	The partition of the node's slice of the dataset
weight_decay	A parameter used in deteriming whether or not a splits effect is significant enough to take
reg_lambda	Reg Lambda; A regularization parameter
gamma	Gamma; A regularization parameter

Returns

The best split

to_json_string()

std::string CNum::Model::Tree::TreeBoosterNode::to_json_string

(

)

Save TreeBoosterNode data for an entire TreeBooster in a JSON formatted string.

Returns

The JSON string

Friends And Related Symbol Documentation

TreeBooster

friend class TreeBooster

friend

Member Data Documentation

_left

TreeBoosterNode* CNum::Model::Tree::TreeBoosterNode::_left

_right

TreeBoosterNode* CNum::Model::Tree::TreeBoosterNode::_right

_split

Split CNum::Model::Tree::TreeBoosterNode::_split

_value

double CNum::Model::Tree::TreeBoosterNode::_value

---

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

• include/CNum/Model/Tree/TreeBoosterNode.h