BART Class Reference

Bayesian Additive Regression Trees. More...

#include <BART.h>

Public Member Functions
void	learn (const vector< float > &x, const vector< float > &y)
	learning on x vector which represents matrix.
void	predict (const vector< float > &x, int nSamples, vector< float > &scores) const
	prediction on x vector which represents matrix
	BART (int ntrees, int iterations, int burn_cnt, int restart_cnt, bart_params &tree_pr)
	a simple default ctor
void	operator= (const BART &other)
	a simple assignment operator to shallow copy all BART model with all trees.
	~BART ()
	a dctor to free all tree memory

Data Fields
int	ntrees
	The nubmer of trees/restarts.
int	iter_count
	the number of steps to call next_gen_tree for each tree
int	burn_count
	the burn count
int	restart_count
	number of restarts
bart_params	tree_params
	additional tree parameters

Detailed Description

Bayesian Additive Regression Trees.

A Monte-Carlo-Markov-Chain process to create trees based on bayesian assumtions
to reach maximum likelihood of tree based on the observations with Metropolis_Hastings algorithm.
It's very usefull for casual inference:
In 2016 at Atlantic Casual Infernce Confernce - 1st place in data competition

Member Function Documentation

learn()

void BART::learn	(	const vector< float > &	x,
		const vector< float > &	y
	)

learning on x vector which represents matrix.

y is the labels

Parameters

x	a vector which represnts matrix. the data is ordered by observations. we first see first observation all features and than second obseravtion all features...
y	labels vector for each observation in x

operator=()

void BART::operator= ( const BART &  other )

inline

a simple assignment operator to shallow copy all BART model with all trees.

not allocating new memory for trees. pointing to same objects

predict()

void BART::predict	(	const vector< float > &	x,
		int	nSamples,
		vector< float > &	scores
	)		const

prediction on x vector which represents matrix

Parameters

x	a vector which represnts matrix. the data is ordered by observations. we first see first observation all features and than second obseravtion all features...
nSamples	the number of samples in x

Returns

scores the result scores for each observation

---

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

• Internal/MedAlgo/MedAlgo/BART.h
• Internal/MedAlgo/MedAlgo/BART.cpp