ppforest2 Namespace Reference

Namespaces
namespace	cli
	Command-line interface: argument parsing, subcommands, and benchmark/evaluation orchestration.
namespace	dr
	Dimensionality reduction strategies for variable selection.
namespace	io
namespace	math
	Numeric comparison utilities.
namespace	pp
	Projection pursuit strategies.
namespace	serialization
	JSON serialization and deserialization for ppforest2 models.
namespace	sr
	Split rule strategies for computing decision thresholds.
namespace	stats
	Statistical infrastructure for training and evaluation.
namespace	sys
	System-level utilities (process memory measurement).
namespace	types
	Core numeric type aliases for the ppforest2 library.
namespace	utils
	Utility functions for std::map manipulation.
namespace	viz

Classes
struct	BootstrapTree
struct	Forest
	A projection pursuit random forest. More...
struct	Model
	Abstract base class for predictive models (trees and forests). More...
struct	Proportions
	Tag type for requesting vote-proportion predictions. More...
struct	TrainingSpec
	Training configuration for projection pursuit trees and forests. More...
struct	Tree
	A single projection pursuit decision tree. More...
struct	TreeCondition
	Internal split node in a projection pursuit tree. More...
struct	TreeNode
	Abstract base class for nodes in a projection pursuit tree. More...
struct	TreeResponse
	Leaf node in a projection pursuit tree. More...
class	UserError
	Exception for user-facing input validation errors. More...
struct	VariableImportance
	Grouped result of the three variable importance measures. More...
struct	VIVisitor
	Visitor that accumulates per-variable contributions for VI2 and VI3. More...

Typedefs
using	Threshold = types::Feature
	Scalar threshold type for split decisions.

Functions
void	user_error (bool condition, char const *message)
	Throw a UserError if the condition is false.
void	user_error (bool condition, std::string const &message)
	Throw a UserError if the condition is false.
void	validate_json_keys (nlohmann::json const &j, std::string const &context, std::initializer_list< std::string > allowed)
	Validate that a JSON object contains only expected keys.
types::FeatureVector	variable_importance_permuted (Forest const &forest, types::FeatureMatrix const &x, types::ResponseVector const &y, int seed=0)
	VI1 — Permuted importance.
types::FeatureVector	variable_importance_projections (Forest const &forest, int n_vars, types::FeatureVector const *scale=nullptr)
	VI2 — Projections importance (forest).
types::FeatureVector	variable_importance_projections (Tree const &tree, int n_vars, types::FeatureVector const *scale=nullptr)
	VI2 — Projections importance (single tree).
types::FeatureVector	variable_importance_weighted_projections (Forest const &forest, types::FeatureMatrix const &x, types::ResponseVector const &y, types::FeatureVector const *scale=nullptr)
	VI3 — Weighted projections importance.

Typedef Documentation

Threshold

using ppforest2::Threshold = types::Feature

Scalar threshold type for split decisions.

Function Documentation

user_error() [1/2]

void ppforest2::user_error	(	bool	condition,
		char const *	message )

Throw a UserError if the condition is false.

Parameters

condition	Condition that must hold for valid input.
message	Actionable error message for the user.

user_error() [2/2]

void ppforest2::user_error	(	bool	condition,
		std::string const &	message )

Throw a UserError if the condition is false.

Parameters

condition	Condition that must hold for valid input.
message	Actionable error message for the user.

validate_json_keys()

void ppforest2::validate_json_keys ( nlohmann::json const & j, std::string const & context, std::initializer_list< std::string > allowed )

inline

Validate that a JSON object contains only expected keys.

Parameters

j	The JSON object to check.
context	Human-readable strategy description (e.g. "PDA").
allowed	Set of allowed key names.

Exceptions

std::runtime_error

if an unknown key is found.

variable_importance_permuted()

types::FeatureVector ppforest2::variable_importance_permuted	(	Forest const &	forest,
		types::FeatureMatrix const &	x,
		types::ResponseVector const &	y,
		int	seed = 0 )

VI1 — Permuted importance.

For each tree, measures the drop in OOB accuracy when each variable is randomly permuted among the OOB observations. Results are averaged over all trees.

Parameters

forest	Trained random forest.
x	Training feature matrix (n × p).
y	Training response vector (n).
seed	RNG seed for the permutations.

Returns

FeatureVector of size p with per-variable importance.

variable_importance_projections() [1/2]

types::FeatureVector ppforest2::variable_importance_projections	(	Forest const &	forest,
		int	n_vars,
		types::FeatureVector const *	scale = nullptr )

VI2 — Projections importance (forest).

At every split node s with G_s groups, accumulates |a_j| / G_s for each variable j. When scale is provided each |a_j| is first multiplied by σ_j so that coefficients are comparable across variables with different units. Results are averaged over all trees.

Parameters

forest	Trained random forest.
n_vars	Number of predictor variables (p).
scale	Optional per-variable σ vector (size p).

Returns

FeatureVector of size p with per-variable importance.

variable_importance_projections() [2/2]

types::FeatureVector ppforest2::variable_importance_projections	(	Tree const &	tree,
		int	n_vars,
		types::FeatureVector const *	scale = nullptr )

VI2 — Projections importance (single tree).

Same as the forest overload but for a single tree (no averaging).

Parameters

tree	Trained tree.
n_vars	Number of predictor variables (p).
scale	Optional per-variable σ vector (size p).

Returns

FeatureVector of size p with per-variable importance.

variable_importance_weighted_projections()

types::FeatureVector ppforest2::variable_importance_weighted_projections	(	Forest const &	forest,
		types::FeatureMatrix const &	x,
		types::ResponseVector const &	y,
		types::FeatureVector const *	scale = nullptr )

VI3 — Weighted projections importance.

Each tree's contribution is weighted by (1 − e_k), where e_k is its OOB error rate, and each split node contributes I_s × |a_j|. The result is normalised by B × (G − 1).

Parameters

forest	Trained random forest.
x	Training feature matrix (n × p).
y	Training response vector (n).
scale	Optional per-variable σ vector (size p).

Returns

FeatureVector of size p with per-variable importance.