koho::BestSplitter Class Reference

Splitter to find the best split for a node. More...

#include <decision_tree.h>

Collaboration diagram for koho::BestSplitter:

Public Member Functions
	BestSplitter (OutputsIdx_t n_outputs, ClassesIdx_t *n_classes, ClassesIdx_t n_classes_max, FeaturesIdx_t n_features, SamplesIdx_t n_samples, ClassWeights_t *class_weight, FeaturesIdx_t max_features, unsigned long max_thresholds, RandomState const &random_state)
	Create and initialize a new best splitter. More...
void	init_node (Classes_t *y, SamplesIdx_t start, SamplesIdx_t end)
	Initialize node and calculate weighted histograms for all outputs and impurity for the node. More...
void	split_feature (Features_t *X, Classes_t *y, std::vector< SamplesIdx_t > &s, FeaturesIdx_t feature, int &NA, Features_t &threshold, SamplesIdx_t &pos, double &improvement)
	Find the best split and partition (actually sorted) samples for a given feature. More...
void	split_feature_extreme_random (Features_t *X, Classes_t *y, std::vector< SamplesIdx_t > &s, FeaturesIdx_t feature, int &NA, Features_t &threshold, SamplesIdx_t &pos, double &improvement)
void	split_node (Features_t *X, Classes_t *y, FeaturesIdx_t &feature, int &NA, Features_t &threshold, SamplesIdx_t &pos, double &improvement)
	Find the best split and partition (actually sorted) samples. More...

Public Attributes
GiniCriterion	criterion

Protected Attributes
FeaturesIdx_t	n_features
SamplesIdx_t	n_samples
FeaturesIdx_t	max_features
unsigned long	max_thresholds
RandomState	random_state
std::vector< SamplesIdx_t >	samples
SamplesIdx_t	start
SamplesIdx_t	end

Detailed Description

Splitter to find the best split for a node.

Constructor & Destructor Documentation

BestSplitter()

koho::BestSplitter::BestSplitter	(	OutputsIdx_t	n_outputs,
		ClassesIdx_t *	n_classes,
		ClassesIdx_t	n_classes_max,
		FeaturesIdx_t	n_features,
		SamplesIdx_t	n_samples,
		ClassWeights_t *	class_weight,
		FeaturesIdx_t	max_features,
		unsigned long	max_thresholds,
		RandomState const &	random_state
	)

Create and initialize a new best splitter.

Member Function Documentation

init_node()

void koho::BestSplitter::init_node	(	Classes_t *	y,
		SamplesIdx_t	start,
		SamplesIdx_t	end
	)

Initialize node and calculate weighted histograms for all outputs and impurity for the node.

split_feature()

void koho::BestSplitter::split_feature	(	Features_t *	X,
		Classes_t *	y,
		std::vector< SamplesIdx_t > &	s,
		FeaturesIdx_t	feature,
		int &	NA,
		Features_t &	threshold,
		SamplesIdx_t &	pos,
		double &	improvement
	)

Find the best split and partition (actually sorted) samples for a given feature.

split_feature_extreme_random()

void koho::BestSplitter::split_feature_extreme_random	(	Features_t *	X,
		Classes_t *	y,
		std::vector< SamplesIdx_t > &	s,
		FeaturesIdx_t	feature,
		int &	NA,
		Features_t &	threshold,
		SamplesIdx_t &	pos,
		double &	improvement
	)

Find a split and partition samples for a given feature using the Extreme Random Tree formulation for the threshold.

split_node()

void koho::BestSplitter::split_node	(	Features_t *	X,
		Classes_t *	y,
		FeaturesIdx_t &	feature,
		int &	NA,
		Features_t &	threshold,
		SamplesIdx_t &	pos,
		double &	improvement
	)

Find the best split and partition (actually sorted) samples.

Find the split (feature, threshold) on samples[start:end] and partition samples[start:end] into samples[start:pos] and samples[pos:end] according to split.

Assuming: init_node()

Member Data Documentation

criterion

GiniCriterion koho::BestSplitter::criterion

end

SamplesIdx_t koho::BestSplitter::end

protected

max_features

FeaturesIdx_t koho::BestSplitter::max_features

protected

max_thresholds

unsigned long koho::BestSplitter::max_thresholds

protected

n_features

FeaturesIdx_t koho::BestSplitter::n_features

protected

n_samples

SamplesIdx_t koho::BestSplitter::n_samples

protected

random_state

RandomState koho::BestSplitter::random_state

protected

samples

std::vector<SamplesIdx_t> koho::BestSplitter::samples

protected

start

SamplesIdx_t koho::BestSplitter::start

protected

---

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

• decision_tree.h
• decision_tree.cpp