ppforest2::Forest Struct Reference

A projection pursuit random forest. More...

#include <Forest.hpp>

Public Member Functions
	Forest ()
	Forest (TrainingSpec::Ptr training_spec)
void	accept (Model::Visitor &visitor) const override
	Accept a model visitor (double dispatch).
void	add_tree (std::unique_ptr< BootstrapTree > tree)
	Add a tree to the forest.
double	oob_error (types::FeatureMatrix const &x, types::ResponseVector const &y) const
	Out-of-bag error rate.
types::ResponseVector	oob_predict (types::FeatureMatrix const &x) const
	Out-of-bag predictions by majority vote.
bool	operator!= (Forest const &other) const
bool	operator== (Forest const &other) const
types::ResponseVector	predict (types::FeatureMatrix const &data) const override
	Predict a matrix of observations.
types::FeatureMatrix	predict (types::FeatureMatrix const &data, Proportions) const override
	Predict vote proportions for a matrix of observations.
types::Response	predict (types::FeatureVector const &data) const override
	Predict a single observation.
Public Member Functions inherited from ppforest2::Model
virtual	~Model ()=default

Static Public Member Functions
static Forest	train (TrainingSpec const &training_spec, types::FeatureMatrix const &x, types::ResponseVector const &y)
	Train a random forest.
Static Public Member Functions inherited from ppforest2::Model
static Ptr	train (TrainingSpec const &spec, types::FeatureMatrix const &x, types::ResponseVector const &y)
	Train a model from a training specification.

Public Attributes
std::vector< BootstrapTree::Ptr >	trees
Public Attributes inherited from ppforest2::Model
bool	degenerate = false
	Whether the model contains degenerate nodes/splits.
TrainingSpec::Ptr	training_spec
	Training specification used to build this model.

Additional Inherited Members
Public Types inherited from ppforest2::Model
using	Ptr = std::shared_ptr<Model>

Detailed Description

A projection pursuit random forest.

An ensemble of BootstrapTree instances, each trained on a bootstrap sample. Predictions are made by majority vote. Out-of-bag estimation and vote-proportion predictions are supported.

TrainingSpec spec(pp::pda(0.0), dr::uniform(3), sr::mean_of_means(), 500, 0);
Forest forest = Forest::train(spec, x, y);
 
types::ResponseVector preds = forest.predict(x_test);
double oob = forest.oob_error(x, y);
 
// Vote proportions — (n × G) matrix, rows sum to 1.
types::FeatureMatrix probs = forest.predict(x_test, Proportions{});

Constructor & Destructor Documentation

Forest() [1/2]

ppforest2::Forest::Forest

(

)

Forest() [2/2]

ppforest2::Forest::Forest ( TrainingSpec::Ptr training_spec )

explicit

Member Function Documentation

accept()

void ppforest2::Forest::accept ( Model::Visitor & visitor ) const

overridevirtual

Accept a model visitor (double dispatch).

Implements ppforest2::Model.

add_tree()

void ppforest2::Forest::add_tree

(

std::unique_ptr< BootstrapTree >

tree

)

Add a tree to the forest.

Parameters

tree	Tree to add (ownership transferred).

oob_error()

double ppforest2::Forest::oob_error	(	types::FeatureMatrix const &	x,
		types::ResponseVector const &	y ) const

Out-of-bag error rate.

For each observation, predicts using majority vote of only the trees where it was out-of-bag, then computes the overall misclassification rate.

Parameters

x	Training feature matrix (n × p).
y	Training response vector (n).

Returns

Error rate in [0, 1], or −1 if no observation has any OOB tree.

oob_predict()

types::ResponseVector ppforest2::Forest::oob_predict

(

types::FeatureMatrix const &

x

)

const

Out-of-bag predictions by majority vote.

For each observation, predicts using majority vote of only the trees where it was out-of-bag. Observations with no OOB tree receive a sentinel value (−1) and are excluded when computing oob_error().

Parameters

x	Training feature matrix (n × p).

Returns

OOB predictions (n); −1 where no OOB tree exists.

operator!=()

bool ppforest2::Forest::operator!=

(

Forest const &

other

)

const

operator==()

bool ppforest2::Forest::operator==

(

Forest const &

other

)

const

predict() [1/3]

types::ResponseVector ppforest2::Forest::predict ( types::FeatureMatrix const & data ) const

overridevirtual

Predict a matrix of observations.

Parameters

data	Feature matrix (n × p).

Returns

Predictions (n).

Implements ppforest2::Model.

predict() [2/3]

types::FeatureMatrix ppforest2::Forest::predict ( types::FeatureMatrix const & data, Proportions  ) const

overridevirtual

Predict vote proportions for a matrix of observations.

For each observation, counts votes from every tree and returns the proportion of trees that voted for each group. The number of groups G is derived from the root node of the first tree.

Parameters

data	Feature matrix (n × p).

Returns

Vote proportions matrix (n × G), rows sum to 1.0.

Implements ppforest2::Model.

predict() [3/3]

types::Response ppforest2::Forest::predict ( types::FeatureVector const & data ) const

overridevirtual

Predict a single observation.

Parameters

data	Feature vector (p).

Returns

Prediction.

Implements ppforest2::Model.

train()

static Forest ppforest2::Forest::train ( TrainingSpec const & training_spec, types::FeatureMatrix const & x, types::ResponseVector const & y )

static

Train a random forest.

Forest-level parameters (size, seed, threads, max_retries) are read from the training specification.

Parameters

training_spec	Training specification.
x	Feature matrix (n × p).
y	Response vector (n).

Returns

Trained forest.

Member Data Documentation

trees

std::vector<BootstrapTree::Ptr> ppforest2::Forest::trees

---

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

• core/src/models/Forest.hpp