ppforest2::Tree Class Reference

Abstract base class for projection pursuit decision trees. More...

#include <Tree.hpp>

Public Types
using	FeatureMatrix = types::FeatureMatrix
using	FeatureVector = types::FeatureVector
using	GroupPartition = stats::GroupPartition
using	Outcome = types::Outcome
using	OutcomeVector = types::OutcomeVector
using	Ptr = std::unique_ptr<Tree>
using	RNG = stats::RNG
using	Root = TreeNode::Ptr
Public Types inherited from ppforest2::Model
using	Ptr = std::shared_ptr<Model>

Public Member Functions
void	accept (Model::Visitor &visitor) const override=0
	Accept a model visitor (mode-specific dispatch).
bool	operator!= (Tree const &other) const
bool	operator== (Tree const &other) const
virtual types::OutcomeVector	predict (types::FeatureMatrix const &x) const
	Predict a matrix of observations.
types::Outcome	predict (types::FeatureVector const &x) const override
	Predict a single observation.
Public Member Functions inherited from ppforest2::Model
virtual	~Model ()=default

Static Public Member Functions
static Ptr	train (TrainingSpec const &spec, types::FeatureMatrix &x, types::OutcomeVector &y)
	Train a tree from a response vector.
static Ptr	train (TrainingSpec const &spec, types::FeatureMatrix &x, types::OutcomeVector &y, stats::RNG &rng)
	Train a tree from a response vector.
Static Public Member Functions inherited from ppforest2::Model
static void	check_train_inputs (types::FeatureMatrix const &x, types::OutcomeVector const &y)
	Validate common training inputs (y non-empty, matching x rows).
static Ptr	train (TrainingSpec const &spec, types::FeatureMatrix &x, types::OutcomeVector &y)
	Train a model from a training specification.

Public Attributes
Root	root
	Root node of the tree.
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.

Protected Member Functions
	Tree (TreeNode::Ptr root, TrainingSpec::Ptr spec)

Static Protected Member Functions
static Root	build_root (TrainingSpec const &spec, FeatureMatrix &x, OutcomeVector &y, GroupPartition const &y_part, RNG &rng)
	Build the root node of a tree.

Detailed Description

Abstract base class for projection pursuit decision trees.

Each internal node projects data onto a linear combination of features and splits on the projected value. Leaf values depend on the mode — group labels for classification, mean response for regression — implemented in the concrete subclasses ClassificationTree and RegressionTree.

Construct via the static Tree::train factory, which dispatches to the correct concrete type based on training_spec.mode.

auto tree = Tree::train(spec, x, y);  // classification or regression
Outcome label = tree->predict(x.row(0));
OutcomeVector preds = tree->predict(x);

Member Typedef Documentation

FeatureMatrix

using ppforest2::Tree::FeatureMatrix = types::FeatureMatrix

FeatureVector

using ppforest2::Tree::FeatureVector = types::FeatureVector

GroupPartition

using ppforest2::Tree::GroupPartition = stats::GroupPartition

Outcome

using ppforest2::Tree::Outcome = types::Outcome

OutcomeVector

using ppforest2::Tree::OutcomeVector = types::OutcomeVector

Ptr

using ppforest2::Tree::Ptr = std::unique_ptr<Tree>

RNG

using ppforest2::Tree::RNG = stats::RNG

Root

using ppforest2::Tree::Root = TreeNode::Ptr

Constructor & Destructor Documentation

Tree()

ppforest2::Tree::Tree ( TreeNode::Ptr root, TrainingSpec::Ptr spec )

inlineprotected

Member Function Documentation

accept()

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

overridepure virtual

Accept a model visitor (mode-specific dispatch).

Implements ppforest2::Model.

Implemented in ppforest2::ClassificationTree, and ppforest2::RegressionTree.

build_root()

static Root ppforest2::Tree::build_root ( TrainingSpec const & spec, FeatureMatrix & x, OutcomeVector & y, GroupPartition const & y_part, RNG & rng )

staticprotected

Build the root node of a tree.

Iteratively grows the tree from the given group partition. Shared implementation used by ClassificationTree::train and RegressionTree::train.

x and y are mutable: regression's ByCutpoint grouping strategy reorders rows in place. Classification doesn't mutate either — the reference is still non-const to keep the signature uniform across modes and avoid the "aliased mutable vs immutable view" smell of the prior pointer-based design.

Parameters

spec	Training specification (strategies + mode).
x	Feature matrix. Mutated by regression's in-place row reorder; untouched by classification.
y	Response vector. Same mutation contract as x.
y_part	Initial group partition for the root node.
rng	Random number generator (tree-local).

Returns

Root TreeNode of the constructed tree.

operator!=()

bool ppforest2::Tree::operator!=

(

Tree const &

other

)

const

operator==()

bool ppforest2::Tree::operator==

(

Tree const &

other

)

const

predict() [1/2]

virtual types::OutcomeVector ppforest2::Model::predict ( types::FeatureMatrix const & x ) const

inlinevirtual

Predict a matrix of observations.

Default implementation iterates rows and dispatches to the single-row predict. Subclasses may override to vectorize.

Parameters

x	Feature matrix (n × p).

Returns

Predicted group labels (n).

Reimplemented from ppforest2::Model.

predict() [2/2]

types::Outcome ppforest2::Tree::predict ( types::FeatureVector const & x ) const

overridevirtual

Predict a single observation.

Walks the tree and returns the leaf value. Same implementation for both modes — the leaf value is produced by the mode-specific leaf strategy during training.

Implements ppforest2::Model.

train() [1/2]

static Ptr ppforest2::Tree::train ( TrainingSpec const & spec, types::FeatureMatrix & x, types::OutcomeVector & y )

static

Train a tree from a response vector.

Dispatches to ClassificationTree::train or RegressionTree::train based on training_spec.mode. Creates an RNG from the spec's seed.

x and y are taken by mutable reference because some strategies — notably ByCutpoint for regression — permute rows in place during training. Classification training does not mutate them, so the classification path pays no cost for this signature. The alternative (const-correct public API + defensive copy at the regression dispatch) would force a full-matrix copy per single-tree regression call, which is a real cost the library shouldn't absorb when the natural callers (R bindings, CLI) discard the data right after training anyway. Callers who need to preserve the original row order must copy before calling.

Parameters

spec	Training specification (strategies + mode).
x	Feature matrix (n × p). May be permuted during regression training.
y	Response vector (n) — integer labels for classification, continuous response for regression. May be permuted during regression training.

Returns

Trained tree as a polymorphic pointer.

train() [2/2]

static Ptr ppforest2::Tree::train ( TrainingSpec const & spec, types::FeatureMatrix & x, types::OutcomeVector & y, stats::RNG & rng )

static

Train a tree from a response vector.

Dispatches to ClassificationTree::train or RegressionTree::train based on training_spec.mode. Creates an RNG from the spec's seed.

x and y are taken by mutable reference because some strategies — notably ByCutpoint for regression — permute rows in place during training. Classification training does not mutate them, so the classification path pays no cost for this signature. The alternative (const-correct public API + defensive copy at the regression dispatch) would force a full-matrix copy per single-tree regression call, which is a real cost the library shouldn't absorb when the natural callers (R bindings, CLI) discard the data right after training anyway. Callers who need to preserve the original row order must copy before calling.

Parameters

spec	Training specification (strategies + mode).
x	Feature matrix (n × p). May be permuted during regression training.
y	Response vector (n) — integer labels for classification, continuous response for regression. May be permuted during regression training.

Returns

Trained tree as a polymorphic pointer.

Member Data Documentation

root

Root ppforest2::Tree::root

Root node of the tree.

---

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

• core/src/models/Tree.hpp