cmp::statistics::KFold Class Reference

K-Fold cross-validation partition generator. More...

#include <statistics.h>

Classes
class	KFoldIterator
	Iterator class to traverse folds for K-Fold cross-validation. More...

Public Member Functions
	KFold (Eigen::Index n_splits, Eigen::Index n_obs, bool shuffle=false, unsigned int random_state=42)
	Constructor for the KFold partitioner.
KFoldIterator	begin () const
	Gets the iterator pointing to the first fold.
KFoldIterator	end () const
	Gets the iterator pointing to the end of the folds.
std::pair< Eigen::VectorXs, Eigen::VectorXs >	operator() (Eigen::Index split) const
	Directly accesses the train/test indices for a specific split without iterating.
size_t	nFolds () const
	Gets the total number of splits/folds \( K \).

Private Attributes
Eigen::Index	nSplits_
	Number of folds/splits \( K \).
Eigen::Index	nObs_
	Total number of observations \( N \) in the dataset.
bool	shuffle_
	Flag indicating if the dataset indices should be shuffled before splitting.
unsigned int	rngState_
	Seed for the random number generator.
std::default_random_engine	rng_
	Random number engine used for shuffling.
Eigen::VectorXs	indices_
	Internal array storing the (potentially shuffled) sequence of indices.

Detailed Description

K-Fold cross-validation partition generator.

Mathematical Formulation

Partitions a dataset \(\mathcal{D}\) of size \(N\) into \(K\) mutually exclusive, approximately equal-sized subsets \(\{F_1, \dots, F_K\}\) such that:

\[ \bigcup_{k=1}^K F_k = \mathcal{D}, \quad F_i \cap F_j = \emptyset \quad \forall i \neq j \]

During the \(k\)-th fold iteration, the model is trained on \(\mathcal{D} \setminus F_k\) and validated on \(F_k\).

• ### Implementation Algorithm

1. Computes the base size of each fold: \(S = \lfloor N/K \rfloor\).
2. Assigns indices to splits. The \(k\)-th split covers index range \([k \cdot S, (k+1) \cdot S)\), and the last fold covers any remainder.
3. Returns the respective train and test index vectors.

Constructor & Destructor Documentation

KFold()

cmp::statistics::KFold::KFold ( Eigen::Index  n_splits, Eigen::Index  n_obs, bool  shuffle = false, unsigned int  random_state = 42  )

inline

Constructor for the KFold partitioner.

Parameters

n_splits	Number of folds \( K \). Must be at least 2 and at most n_obs.
n_obs	Total number of observations \( N \) in the dataset. Must be at least 2.
shuffle	Whether to shuffle the data before splitting into folds.
random_state	Seed for the random number generator (used if shuffle is true).

Exceptions

std::invalid_argument

if the input parameters violate minimum constraints.

Member Function Documentation

begin()

KFoldIterator cmp::statistics::KFold::begin ( ) const

inline

Gets the iterator pointing to the first fold.

Returns

A KFoldIterator initialized to split 0.

end()

KFoldIterator cmp::statistics::KFold::end ( ) const

inline

Gets the iterator pointing to the end of the folds.

Returns

A KFoldIterator initialized to split K.

nFolds()

size_t cmp::statistics::KFold::nFolds ( ) const

inline

Gets the total number of splits/folds \( K \).

Returns

The number of folds.

operator()()

std::pair< Eigen::VectorXs, Eigen::VectorXs > cmp::statistics::KFold::operator() ( Eigen::Index  split ) const

inline

Directly accesses the train/test indices for a specific split without iterating.

Parameters

split

The specific split index \( k \) to extract.

Returns

A pair of training and test index vectors.

Exceptions

std::out_of_range

if the requested split is invalid.

Member Data Documentation

indices_

Eigen::VectorXs cmp::statistics::KFold::indices_

private

Internal array storing the (potentially shuffled) sequence of indices.

nObs_

Eigen::Index cmp::statistics::KFold::nObs_

private

Total number of observations \( N \) in the dataset.

nSplits_

Eigen::Index cmp::statistics::KFold::nSplits_

private

Number of folds/splits \( K \).

rng_

std::default_random_engine cmp::statistics::KFold::rng_

private

Random number engine used for shuffling.

rngState_

unsigned int cmp::statistics::KFold::rngState_

private

Seed for the random number generator.

shuffle_

bool cmp::statistics::KFold::shuffle_

private

Flag indicating if the dataset indices should be shuffled before splitting.

---

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

• include/statistics.h