CrossValidation Class Reference

Lightweight K-fold / LOO cross-validation utilities and 1-D grid search. More...

#include <OpenMS/ML/CROSSVALIDATION/CrossValidation.h>

Public Types
enum class	CandidateTieBreak { PreferLarger , PreferSmaller , PreferAny }
	Tie-breaking preference for equal (within tolerance) CV scores. More...

Static Public Member Functions
static std::vector< std::vector< Size > >	makeKFolds (Size n, Size K)
	Build K folds for indices [0, n). More...
template<typename CandIter , typename TrainEval , typename ScoreFn >
static std::pair< typename std::iterator_traits< CandIter >::value_type, double >	gridSearch1D (CandIter cbegin, CandIter cend, const std::vector< std::vector< Size >> &folds, TrainEval train_eval, ScoreFn score, double tie_tol=1e-12, CandidateTieBreak tie_break=CandidateTieBreak::PreferLarger)
	One-dimensional grid search with external cross-validation evaluation. More...

Detailed Description

Lightweight K-fold / LOO cross-validation utilities and 1-D grid search.

Provides:

• makeKFolds(n, K): deterministic round-robin fold assignment (LOO if K==n)
• gridSearch1D(...): evaluate a 1-D candidate grid via CV and pick the best

Tie-breaking uses a tiny absolute tolerance and (optionally) prefers larger candidates (useful for smoother/regularized models).

References:

• Stone, M. (1974) Cross-Validatory Choice and Assessment of Statistical Predictions. J. Roy. Stat. Soc. B, 36(2):111–147.

See also

GridSearch