model_selection.hpp File Reference

Model selection functions. More...

#include "statcpp/basic_statistics.hpp"
#include "statcpp/dispersion_spread.hpp"
#include "statcpp/linear_regression.hpp"
#include "statcpp/random_engine.hpp"
#include <algorithm>
#include <cmath>
#include <cstddef>
#include <iterator>
#include <limits>
#include <numeric>
#include <stdexcept>
#include <utility>
#include <vector>

Include dependency graph for model_selection.hpp:

This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Classes
struct	statcpp::cv_result
	Structure to store cross-validation results. More...
struct	statcpp::regularized_regression_result
	Structure to store regularized regression results. More...

Namespaces
namespace	statcpp
namespace	statcpp::detail
	Internal helper functions.

Functions
double	statcpp::aic (double log_likelihood, std::size_t k)
	Calculate AIC (Akaike Information Criterion)
double	statcpp::aic_linear (const simple_regression_result &model, std::size_t n)
	Calculate AIC from simple regression model.
double	statcpp::aic_linear (const multiple_regression_result &model, std::size_t n)
	Calculate AIC from multiple regression model.
double	statcpp::aicc (double log_likelihood, std::size_t n, std::size_t k)
	Calculate AICc (corrected AIC)
double	statcpp::bic (double log_likelihood, std::size_t n, std::size_t k)
	Calculate BIC (Bayesian Information Criterion)
double	statcpp::bic_linear (const simple_regression_result &model, std::size_t n)
	Calculate BIC from simple regression model.
double	statcpp::bic_linear (const multiple_regression_result &model, std::size_t n)
	Calculate BIC from multiple regression model.
template<typename IteratorX , typename IteratorY >
double	statcpp::press_statistic (IteratorX x_first, IteratorX x_last, IteratorY y_first, IteratorY y_last, const simple_regression_result &model)
	Calculate PRESS statistic.
std::vector< std::vector< std::size_t > >	statcpp::create_cv_folds (std::size_t n, std::size_t k, bool shuffle=true)
	Generate indices for k-fold cross-validation.
cv_result	statcpp::cross_validate_linear (const std::vector< std::vector< double > > &X, const std::vector< double > &y, std::size_t k=5)
	Perform k-fold cross-validation for multiple regression model.
cv_result	statcpp::loocv_linear (const std::vector< std::vector< double > > &X, const std::vector< double > &y)
	Perform leave-one-out cross-validation.
void	statcpp::detail::standardize_features (const std::vector< std::vector< double > > &X, std::vector< std::vector< double > > &X_scaled, std::vector< double > &X_mean, std::vector< double > &X_std, std::size_t n, std::size_t p)
	特徴量の標準化 (平均0, 標準偏差1)
std::vector< double >	statcpp::detail::rescale_coefficients (const std::vector< double > &beta, const std::vector< double > &X_mean, const std::vector< double > &X_std, double y_mean, std::size_t p, bool standardize)
	標準化済み係数を元のスケールに逆変換する
regularized_regression_result	statcpp::ridge_regression (const std::vector< std::vector< double > > &X, const std::vector< double > &y, double lambda, bool standardize=true, std::size_t max_iter=1000, double tol=1e-6)
	Perform Ridge regression (L2 regularization)
regularized_regression_result	statcpp::lasso_regression (const std::vector< std::vector< double > > &X, const std::vector< double > &y, double lambda, bool standardize=true, std::size_t max_iter=1000, double tol=1e-6)
	Perform Lasso regression (L1 regularization)
regularized_regression_result	statcpp::elastic_net_regression (const std::vector< std::vector< double > > &X, const std::vector< double > &y, double lambda, double alpha=0.5, bool standardize=true, std::size_t max_iter=1000, double tol=1e-6)
	Perform Elastic Net regression (L1 + L2 regularization)
std::pair< double, std::vector< double > >	statcpp::cv_ridge (const std::vector< std::vector< double > > &X, const std::vector< double > &y, const std::vector< double > &lambda_grid, std::size_t k=5)
	Select optimal lambda for Ridge regression using cross-validation.
std::pair< double, std::vector< double > >	statcpp::cv_lasso (const std::vector< std::vector< double > > &X, const std::vector< double > &y, const std::vector< double > &lambda_grid, std::size_t k=5)
	Select optimal lambda for Lasso regression using cross-validation.
std::vector< double >	statcpp::generate_lambda_grid (const std::vector< std::vector< double > > &X, const std::vector< double > &y, std::size_t n_lambda=100, double lambda_min_ratio=0.0001)
	Automatically generate lambda grid for regularized regression.

Detailed Description

Model selection functions.

Provides model selection and evaluation metrics including AIC, BIC, cross-validation, and regularized regression.

Definition in file model_selection.hpp.