parameters_wrapper.hpp File Reference

Contains methods to check and wrap R-inputs into PPCKO-coherent ones. More...

#include <vector>
#include <algorithm>
#include <numeric>
#include <iostream>
#include <utility>
#include <string>
#include <stdexcept>
#include "traits_ko.hpp"
#include "mesh.hpp"

Go to the source code of this file.

Enumerations
enum	REM_NAN { NR = 0 , MR = 1 , ZR = 2 }
	The available strategy for removing non-dummy NaNs. More...

Functions
std::string	wrap_string_CV_to_be_printed (const std::string &id_cv)
	Creating a string to print on the screen which PPCKO version is being used.
void	check_threshold_ppc (const double &threshold_ppc)
	Check if 'threshold_ppc' input is between 0 and 1. Eventually, raises and error.
void	check_alpha (const double &alpha)
	Check if 'alpha' input is greater than 0. Eventually, raises and error.
void	check_k (const int &k, const int &max_k)
	Check if 'k' input is an integer between 0 and the number of available evaluations of the functional object. Eventually, raises and error.
void	check_solver (bool solver_ex, const std::string &id_cv, int k)
	Check if, if using 'gep_solver', the number of PPCs is not retrieved through explanatory power criterion. Eventually, raises and error.
std::vector< double >	wrap_alpha_vec (Rcpp::Nullable< Rcpp::NumericVector > alpha_vec)
	Wrapping the R-vector representing the regularization parameter input space into a coherent C++ object, checking parameters consistency, eventually throwing an error, eventually sorting them in increasing order.
std::vector< int >	wrap_k_vec (Rcpp::Nullable< Rcpp::IntegerVector > k_vec, int k_max)
	Wrapping the R-vector representing the number of PPCs input space into a coherent C++ object, checking parameters consistency, eventually throwing an error, eventually sorting them in increasing order.
std::vector< double >	wrap_disc_ev (Rcpp::Nullable< Rcpp::NumericVector > disc_ev, double a, double b, int dim)
	Wrapping the points over which the discrete evaluations of the functional object are available. Check consistency of domain extremes and passed points, eventualy throwing an error.
std::pair< int, int >	wrap_sizes_set_CV (Rcpp::Nullable< int > min_size_ts, Rcpp::Nullable< int > max_size_ts, int number_time_instants)
	Wrapping the minimum and maximum dimension of the training set, checking their consitency, eventually raising an error.
int	wrap_num_thread (Rcpp::Nullable< int > num_threads)
	Wrapping the number of threads for OMP.
REM_NAN	wrap_id_rem_nans (Rcpp::Nullable< std::string > id_rem_nan)
	Wrapping the strategy for handling non-dummy NaNs.

Detailed Description

Contains methods to check and wrap R-inputs into PPCKO-coherent ones.

Author

Andrea Enrico Franzoni

Enumeration Type Documentation

REM_NAN

enum REM_NAN

The available strategy for removing non-dummy NaNs.

Enumerator
NR	Not replacing NaN: not to be used by the user, necessary for handling dummy NaNs.
MR	Replacing nans with mean (could change the mean of the distribution)
ZR	Replacing nans with 0s (could change the sd of the distribution)

Function Documentation

check_alpha()

void check_alpha ( const double & alpha )

inline

Check if 'alpha' input is greater than 0. Eventually, raises and error.

Parameters

alpha

regularization parameter

check_k()

void check_k ( const int & k, const int & max_k )

inline

Check if 'k' input is an integer between 0 and the number of available evaluations of the functional object. Eventually, raises and error.

Parameters

k	number of PPCs passed as parameter
max_k	the maximum value for 'k' (number of available evaluations of the functional object)

check_solver()

void check_solver ( bool solver_ex, const std::string & id_cv, int k )

inline

Check if, if using 'gep_solver', the number of PPCs is not retrieved through explanatory power criterion. Eventually, raises and error.

Parameters

solver_ex	'true' if using ex_solver
id_cv	which PPCKO version is used
k	the input parameter k

check_threshold_ppc()

void check_threshold_ppc ( const double & threshold_ppc )

inline

Check if 'threshold_ppc' input is between 0 and 1. Eventually, raises and error.

Parameters

threshold_ppc

requested explanatory power for the predictor

wrap_alpha_vec()

std::vector< double > wrap_alpha_vec ( Rcpp::Nullable< Rcpp::NumericVector > alpha_vec )

inline

Wrapping the R-vector representing the regularization parameter input space into a coherent C++ object, checking parameters consistency, eventually throwing an error, eventually sorting them in increasing order.

Parameters

alpha_vec

Rcpp::Nullable<Rcpp::NumericVector>

Returns

the R-vector is shifted to a std::vector<double>

Note

If the input is 'NULL': a vector in logarithmic scale, from 10 to the power of -10 up to the power of 11 is generated

wrap_disc_ev()

std::vector< double > wrap_disc_ev ( Rcpp::Nullable< Rcpp::NumericVector > disc_ev, double a, double b, int dim )

inline

Wrapping the points over which the discrete evaluations of the functional object are available. Check consistency of domain extremes and passed points, eventualy throwing an error.

Parameters

disc_ev	Rcpp::Nullable<Rcpp::NumericVector> containing the domain points
a	left domain extreme
b	right domain extreme
dim	the number of discrete evaluations

Returns

an std::vector<double>

Note

If the 'disc_ev' is 'NULL': an equally spaced grid from 'a' to 'b' is generated. For surfaces, the two dimensions are wrapped separately

wrap_id_rem_nans()

REM_NAN wrap_id_rem_nans ( Rcpp::Nullable< std::string > id_rem_nan )

inline

Wrapping the strategy for handling non-dummy NaNs.

Parameters

id_rem_nan

string indicating the straegy for removing non-dummy NaNs

Returns

the correpsonding value of 'REM_NAN' (default: 'MR')

wrap_k_vec()

std::vector< int > wrap_k_vec ( Rcpp::Nullable< Rcpp::IntegerVector > k_vec, int k_max )

inline

Wrapping the R-vector representing the number of PPCs input space into a coherent C++ object, checking parameters consistency, eventually throwing an error, eventually sorting them in increasing order.

Parameters

k_vec	Rcpp::Nullable<Rcpp::IntegerVector>
k_max	the maximum number possible of retained PPCs

Returns

the R-vector is shifted to a std::vector<int>

Note

If the input is 'NULL': a vector from 1 to 'k_max' is generated

wrap_num_thread()

int wrap_num_thread ( Rcpp::Nullable< int > num_threads )

inline

Wrapping the number of threads for OMP.

Parameters

num_threads

indicates how many threads to be used by multi-threading directives.

Returns

the number of threads

if omp is not included: will return 1. If not, a number going from 1 up to the maximum cores available by the machine used (default, or if the input is smaller than 1 or bigger than the maximum number of available cores)

Note

omp requested

wrap_sizes_set_CV()

std::pair< int, int > wrap_sizes_set_CV ( Rcpp::Nullable< int > min_size_ts, Rcpp::Nullable< int > max_size_ts, int number_time_instants )

inline

Wrapping the minimum and maximum dimension of the training set, checking their consitency, eventually raising an error.

Parameters

min_size_ts	minimum dimension of the training set
max_size_ts	maximum dimension of the training set
number_time_instants	number of total time instants available

Returns

a pair containig minimum and maximum dimension training set

Note

• 'min_dim_ts' has to be at least 2, but less than 'max_size_ts'. Default value: ceil of 'number_time_instants'/2
• 'max_size_ts' has to be greater than 'min_dim_ts' but less than number_time_instants. Default value: 'number_time_instants' (to be consistent with Eigen, look at the reference to better understand this choice)

See also

train_validation_set_strategy()

wrap_string_CV_to_be_printed()

std::string wrap_string_CV_to_be_printed ( const std::string & id_cv )

inline

Creating a string to print on the screen which PPCKO version is being used.

Parameters

id_cv

string indicating the version used

Returns

a string to be printed