TreeNode

Full name Microsoft.VisualBasic.MachineLearning.XGBoost.train.TreeNode Assembly Microsoft.VisualBasic.MachineLearning.XGBoost Members 27

01 Syntax

Microsoft.VisualBasic.MachineLearning.XGBoost.train.TreeNode

02 Methods

Name	Overloads	Summary
clean_up	1	release memory
Grad_add	1
Hess_add	1
num_sample_add	1
Grad_setter	1
Hess_setter	1
update_best_split	1

03 Properties

Name	Overloads	Summary
leafValue	1	叶节点的输出值（leaf score）。该值等于 -G / (H + lambda)，对于回归任务是加性空间下的叶值，对于分类任务是 log-odds 空间下的叶值。该属性将内部的 leaf_score 字段只读地暴露给下游的可解释性分析（例如 TreeSHAP），而不会改变任何训练/预测行为。

04 Fields

Name	Overloads	Summary
index	1
depth	1
feature_dim	1
is_leaf	1
num_sample	1
Grad	1
Hess	1
G_left	1
H_left	1
nan_go_to	1
Grad_missing	1
Hess_missing	1
split_feature	1
split_threshold	1
split_left_child_catvalue	1
nan_child	1
left_child	1
right_child	1
cat_feature_col_value_GH	1

05 Members

method clean_up #

clean_up

release memory

property leafValue #

leafValue

叶节点的输出值（leaf score）。该值等于 -G / (H + lambda)，对于回归任务是加性空间下的叶值，对于分类任务是 log-odds 空间下的叶值。

该属性将内部的 leaf_score 字段只读地暴露给下游的可解释性分析（例如 TreeSHAP），而不会改变任何训练/预测行为。

field index #

index

field depth #

depth

field feature_dim #

feature_dim

field is_leaf #

is_leaf

field num_sample #

num_sample

field Grad #

Grad

field Hess #

Hess

field G_left #

G_left

field H_left #

H_left

field nan_go_to #

nan_go_to

field Grad_missing #

Grad_missing

field Hess_missing #

Hess_missing

field split_feature #

split_feature

field split_threshold #

split_threshold

field split_left_child_catvalue #

split_left_child_catvalue

field nan_child #

nan_child

field left_child #

left_child

field right_child #

right_child

field cat_feature_col_value_GH #

cat_feature_col_value_GH

method Grad_add #

Grad_add(Double)

method Hess_add #

Hess_add(Double)

method num_sample_add #

num_sample_add(Double)

method Grad_setter #

Grad_setter(Double)

method Hess_setter #

Hess_setter(Double)

method update_best_split #

update_best_split(Int32, Double, Double, Double)