R/RcppExports.R

In orthoDr: Semi-Parametric Dimension Reduction Models Using Orthogonality Constrained Optimization

# Generated by using Rcpp::compileAttributes() -> do not edit by hand
# Generator token: 10BE3573-1514-4C36-9D1C-5A225CD40393

#' @title General solver \code{C++} function
#' @name gen_solver
#' @description A general purpose optimization solver with orthogonality constraint. For details, please see the original \code{MATLAB} code by Wen and Yin (2013). This is an internal function and should not be called directly.
#' @keywords internal
#' @param B A matrix of the parameters \code{B}, the columns are subject to the orthogonality constraint
#' @param f A function that calculates the objective function value. The first argument should be \code{B}. Returns a single value.
#' @param g A function that calculates the gradient. The first argument should be \code{B}. Returns a matrix with the same dimension as \code{B}. If not specified, then the numerical approximation is used.
#' @param env Environment passed to the Rcpp function for evaluating \code{f} and \code{g}
#' @param useg If true, the gradient is calculated using \code{g} function, otherwise numerically approximated
#' @param rho (don't change) Parameter for control the linear approximation in line search
#' @param eta (don't change) Factor for decreasing the step size in the backtracking line search
#' @param gamma (don't change) Parameter for updating C by Zhang and Hager (2004)
#' @param tau (don't change) Step size for updating
#' @param epsilon (don't change) Parameter for approximating numerical gradient, if \code{g} is not given.
#' @param btol (don't change) The \code{$B$} parameter tolerance level
#' @param ftol (don't change) Functional value tolerance level
#' @param gtol (don't change) Gradient tolerance level
#' @param maxitr Maximum number of iterations
#' @param verbose Should information be displayed
#' 
#' @references Wen, Z., & Yin, W. (2013). A feasible method for optimization with orthogonality constraints. 
#' Mathematical Programming, 142(1), 397-434.
#' DOI: \doi{10.1007/s10107-012-0584-1}
#' 
#' @references Zhang, H., & Hager, W. W. (2004). A nonmonotone line search technique and its application to unconstrained optimization. 
#' SIAM journal on Optimization, 14(4), 1043-1056.
#' DOI: \doi{10.1137/S1052623403428208}
#' 
#' @examples
#' # This function should be called internally. When having all objects pre-computed, one can call
#' # gen_solver(B, f, g, env, useg, rho, eta, gamma, tau, epsilon, btol, ftol, gtol, maxitr, verbose)
#' # to solve for the parameters B.
gen_solver <- function(B, f, g, env, useg, rho, eta, gamma, tau, epsilon, btol, ftol, gtol, maxitr, verbose) {
    .Call(`_orthoDr_gen_solver`, B, f, g, env, useg, rho, eta, gamma, tau, epsilon, btol, ftol, gtol, maxitr, verbose)
}

#' @title pdose_direct_solver
#' @name pdose_direct_solver
#' @description The direct learning optimization function for personalized dose finding.
#' @keywords internal
#' @param B A matrix of the parameters \code{B}, the columns are subject to the orthogonality constraint
#' @param X The covariate matrix
#' @param A observed dose levels
#' @param a_dist A kernel distance matrix for the observed dose and girds of the dose levels
#' @param a_seq A grid of dose levels
#' @param R The perosnalzied medicine reward
#' @param lambda The penalty for the GCV for the kernel ridge regression
#' @param bw A Kernel bandwidth, assuming each variable have unit variance
#' @param rho (don't change) Parameter for control the linear approximation in line search
#' @param eta (don't change) Factor for decreasing the step size in the backtracking line search
#' @param gamma (don't change) Parameter for updating C by Zhang and Hager (2004)
#' @param tau (don't change) Step size for updating
#' @param epsilon (don't change) Parameter for approximating numerical gradient
#' @param btol (don't change) The \code{$B$} parameter tolerance level
#' @param ftol (don't change) Estimation equation 2-norm tolerance level
#' @param gtol (don't change) Gradient tolerance level
#' @param maxitr Maximum number of iterations
#' @param verbose Should information be displayed
#' @return The optimizer \code{B} for the esitmating equation.
#' 
#' @references Zhou, W., Zhu, R., & Zeng, D. (2021). A parsimonious personalized dose-finding model via dimension reduction. 
#' Biometrika, 108(3), 643-659.
#' DOI: \doi{10.1093/biomet/asaa087}
#' 
pdose_direct_solver <- function(B, X, A, a_dist, a_seq, R, lambda, bw, rho, eta, gamma, tau, epsilon, btol, ftol, gtol, maxitr, verbose, ncore) {
    .Call(`_orthoDr_pdose_direct_solver`, B, X, A, a_dist, a_seq, R, lambda, bw, rho, eta, gamma, tau, epsilon, btol, ftol, gtol, maxitr, verbose, ncore)
}

#' @title The prediction function for the personalized direct learning dose model
#' @name dosepred
#' @description Predict the fitted dose from the direct learning dose model
#' @keywords internal
#' @param B A matrix of the parameters \code{B}, the columns are subject to the orthogonality constraint
#' @param X The covariate matrix
#' @param X_test The test covariate matrix
#' @param bw A Kernel bandwidth, assuming each variable have unit variance
#' @param w The kernel ridge regression coefficient
#' @return The predicted dose
dosepred <- function(B, X, X_test, bw, W) {
    .Call(`_orthoDr_dosepred`, B, X, X_test, bw, W)
}

#' @title pdose_semi_solver
#' @name pdose_semi_solver
#' @description The pseudo direct learning optimization function for personalized dose finding with dimension reduction.
#' @keywords internal
#' @param B A matrix of the parameters \code{B}, the columns are subject to the orthogonality constraint
#' @param X The covariate matrix
#' @param R The perosnalzied medicine reward
#' @param A observed dose levels
#' @param a_dist A kernel distance matrix for the observed dose and girds of the dose levels
#' @param a_seq A grid of dose levels
#' @param lambda The penalty for the GCV for the kernel ridge regression
#' @param bw A Kernel bandwidth, assuming each variable have unit variance
#' @param rho (don't change) Parameter for control the linear approximation in line search
#' @param eta (don't change) Factor for decreasing the step size in the backtracking line search
#' @param gamma (don't change) Parameter for updating C by Zhang and Hager (2004)
#' @param tau (don't change) Step size for updating
#' @param epsilon (don't change) Parameter for approximating numerical gradient
#' @param btol (don't change) The \code{$B$} parameter tolerance level
#' @param ftol (don't change) Estimation equation 2-norm tolerance level
#' @param gtol (don't change) Gradient tolerance level
#' @param maxitr Maximum number of iterations
#' @param verbose Should information be displayed
#' @return The optimizer \code{B} for the esitmating equation.
#' 
#' @references Zhou, W., Zhu, R., & Zeng, D. (2021). A parsimonious personalized dose-finding model via dimension reduction. 
#' Biometrika, 108(3), 643-659.
#' DOI: \doi{10.1093/biomet/asaa087}
#' 
pdose_semi_solver <- function(B, X, R, A, a_dist, a_seq, lambda, bw, rho, eta, gamma, tau, epsilon, btol, ftol, gtol, maxitr, verbose, ncore) {
    .Call(`_orthoDr_pdose_semi_solver`, B, X, R, A, a_dist, a_seq, lambda, bw, rho, eta, gamma, tau, epsilon, btol, ftol, gtol, maxitr, verbose, ncore)
}

#' @title local_f
#' @name local_f
#' @description local method f value function
#' @keywords internal
local_f <- function(B, X, Y, bw, ncore) {
    .Call(`_orthoDr_local_f`, B, X, Y, bw, ncore)
}

#' @title local semi regression solver \code{C++} function
#' @name local_solver
#' @description Sovling the local semiparametric estimating equations. This is an internal function and should not be called directly.
#' @keywords internal
#' @param B A matrix of the parameters \code{B}, the columns are subject to the orthogonality constraint
#' @param X A matrix of the parameters \code{X}
#' @param Y A matrix of the parameters \code{Y}
#' @param bw Kernel bandwidth for X
#' @param rho (don't change) Parameter for control the linear approximation in line search
#' @param eta (don't change) Factor for decreasing the step size in the backtracking line search
#' @param gamma (don't change) Parameter for updating C by Zhang and Hager (2004)
#' @param tau (don't change) Step size for updating
#' @param epsilon (don't change) Parameter for apprximating numerical gradient, if \code{g} is not given.
#' @param btol (don't change) The \code{$B$} parameter tolerance level
#' @param ftol (don't change) Functional value tolerance level
#' @param gtol (don't change) Gradient tolerance level
#' @param maxitr Maximum number of iterations
#' @param verbose Should information be displayed
#' 
#' @references Ma, Y., & Zhu, L. (2013). Efficient estimation in sufficient dimension reduction. 
#' Annals of statistics, 41(1), 250.
#' DOI: \doi{10.1214/12-AOS1072}
#'
local_solver <- function(B, X, Y, bw, rho, eta, gamma, tau, epsilon, btol, ftol, gtol, maxitr, verbose, ncore) {
    .Call(`_orthoDr_local_solver`, B, X, Y, bw, rho, eta, gamma, tau, epsilon, btol, ftol, gtol, maxitr, verbose, ncore)
}

#' @title phd_init
#' @name phd_init
#' @description phd initial value function
#' @keywords internal
phd_init <- function(B, X, Y, bw, ncore) {
    .Call(`_orthoDr_phd_init`, B, X, Y, bw, ncore)
}

#' @title semi-phd solver \code{C++} function
#' @name phd_solver
#' @description Sovling the semi-phd estimating equations. This is an internal function and should not be called directly.
#' @keywords internal
#' @param B A matrix of the parameters \code{B}, the columns are subject to the orthogonality constraint
#' @param X A matrix of the parameters \code{X}
#' @param Y A matrix of the parameters \code{Y}
#' @param bw Kernel bandwidth for X
#' @param rho (don't change) Parameter for control the linear approximation in line search
#' @param eta (don't change) Factor for decreasing the step size in the backtracking line search
#' @param gamma (don't change) Parameter for updating C by Zhang and Hager (2004)
#' @param tau (don't change) Step size for updating
#' @param epsilon (don't change) Parameter for apprximating numerical gradient, if \code{g} is not given.
#' @param btol (don't change) The \code{$B$} parameter tolerance level
#' @param ftol (don't change) Functional value tolerance level
#' @param gtol (don't change) Gradient tolerance level
#' @param maxitr Maximum number of iterations
#' @param verbose Should information be displayed
#' 
#' @references Ma, Y., & Zhu, L. (2012). A semiparametric approach to dimension reduction. 
#' Journal of the American Statistical Association, 107(497), 168-179.
#' DOI: \doi{10.1080/01621459.2011.646925}
#'
phd_solver <- function(B, X, Y, bw, rho, eta, gamma, tau, epsilon, btol, ftol, gtol, maxitr, verbose, ncore) {
    .Call(`_orthoDr_phd_solver`, B, X, Y, bw, rho, eta, gamma, tau, epsilon, btol, ftol, gtol, maxitr, verbose, ncore)
}

#' @title save_init
#' @name save_init
#' @description save initial value function
#' @keywords internal
save_init <- function(B, X, Y, bw, ncore) {
    .Call(`_orthoDr_save_init`, B, X, Y, bw, ncore)
}

#' @title semi-save solver \code{C++} function
#' @name save_solver
#' @description Sovling the semi-save estimating equations. This is an internal function and should not be called directly.
#' @keywords internal
#' @param B A matrix of the parameters \code{B}, the columns are subject to the orthogonality constraint
#' @param X A matrix of the parameters \code{X}
#' @param Y A matrix of the parameters \code{Y}
#' @param bw Kernel bandwidth for X
#' @param rho (don't change) Parameter for control the linear approximation in line search
#' @param eta (don't change) Factor for decreasing the step size in the backtracking line search
#' @param gamma (don't change) Parameter for updating C by Zhang and Hager (2004)
#' @param tau (don't change) Step size for updating
#' @param epsilon (don't change) Parameter for apprximating numerical gradient, if \code{g} is not given.
#' @param btol (don't change) The \code{$B$} parameter tolerance level
#' @param ftol (don't change) Functional value tolerance level
#' @param gtol (don't change) Gradient tolerance level
#' @param maxitr Maximum number of iterations
#' @param verbose Should information be displayed
#' 
#' @references Ma, Y., & Zhu, L. (2012). A semiparametric approach to dimension reduction. 
#' Journal of the American Statistical Association, 107(497), 168-179.
#' DOI: \doi{10.1080/01621459.2011.646925}
#'
save_solver <- function(B, X, Y, bw, rho, eta, gamma, tau, epsilon, btol, ftol, gtol, maxitr, verbose, ncore) {
    .Call(`_orthoDr_save_solver`, B, X, Y, bw, rho, eta, gamma, tau, epsilon, btol, ftol, gtol, maxitr, verbose, ncore)
}

#' @title seff_init
#' @name seff_init
#' @description semiparametric efficient method initial value function
#' @keywords internal
seff_init <- function(B, X, Y, bw, ncore) {
    .Call(`_orthoDr_seff_init`, B, X, Y, bw, ncore)
}

#' @title Eff semi regression solver \code{C++} function
#' @name seff_solver
#' @description Sovling the semiparametric efficient estimating equations. This is an internal function and should not be called directly.
#' @keywords internal
#' @param B A matrix of the parameters \code{B}, the columns are subject to the orthogonality constraint
#' @param X A matrix of the parameters \code{X}
#' @param Y A matrix of the parameters \code{Y}
#' @param bw Kernel bandwidth for X
#' @param rho (don't change) Parameter for control the linear approximation in line search
#' @param eta (don't change) Factor for decreasing the step size in the backtracking line search
#' @param gamma (don't change) Parameter for updating C by Zhang and Hager (2004)
#' @param tau (don't change) Step size for updating
#' @param epsilon (don't change) Parameter for apprximating numerical gradient, if \code{g} is not given.
#' @param btol (don't change) The \code{$B$} parameter tolerance level
#' @param ftol (don't change) Functional value tolerance level
#' @param gtol (don't change) Gradient tolerance level
#' @param maxitr Maximum number of iterations
#' @param verbose Should information be displayed
#' 
#' @references Ma, Y., & Zhu, L. (2013). Efficient estimation in sufficient dimension reduction. 
#' Annals of statistics, 41(1), 250.
#' DOI: \doi{10.1214/12-AOS1072}
#'
seff_solver <- function(B, X, Y, bw, rho, eta, gamma, tau, epsilon, btol, ftol, gtol, maxitr, verbose, ncore) {
    .Call(`_orthoDr_seff_solver`, B, X, Y, bw, rho, eta, gamma, tau, epsilon, btol, ftol, gtol, maxitr, verbose, ncore)
}

#' @title sir_init
#' @name sir_init
#' @description sir initial value function
#' @keywords internal
sir_init <- function(B, X, Y, bw, ncore) {
    .Call(`_orthoDr_sir_init`, B, X, Y, bw, ncore)
}

#' @title semi-sir solver \code{C++} function
#' @name sir_solver
#' @description Sovling the semi-sir estimating equations. This is an internal function and should not be called directly.
#' @keywords internal
#' @param B A matrix of the parameters \code{B}, the columns are subject to the orthogonality constraint
#' @param X A matrix of the parameters \code{X}
#' @param Y A matrix of the parameters \code{Y}
#' @param bw Kernel bandwidth for X
#' @param rho (don't change) Parameter for control the linear approximation in line search
#' @param eta (don't change) Factor for decreasing the step size in the backtracking line search
#' @param gamma (don't change) Parameter for updating C by Zhang and Hager (2004)
#' @param tau (don't change) Step size for updating
#' @param epsilon (don't change) Parameter for apprximating numerical gradient, if \code{g} is not given.
#' @param btol (don't change) The \code{$B$} parameter tolerance level
#' @param ftol (don't change) Functional value tolerance level
#' @param gtol (don't change) Gradient tolerance level
#' @param maxitr Maximum number of iterations
#' @param verbose Should information be displayed
#' 
#' @references Ma, Y., & Zhu, L. (2012). A semiparametric approach to dimension reduction. 
#' Journal of the American Statistical Association, 107(497), 168-179.
#' DOI: \doi{10.1080/01621459.2011.646925}
#'
sir_solver <- function(B, X, Y, bw, rho, eta, gamma, tau, epsilon, btol, ftol, gtol, maxitr, verbose, ncore) {
    .Call(`_orthoDr_sir_solver`, B, X, Y, bw, rho, eta, gamma, tau, epsilon, btol, ftol, gtol, maxitr, verbose, ncore)
}

#' @title surv_dm_solver \code{C++} function
#' @name surv_dm_solver
#' @description The main optimization function for survival dimensional reduction, the IR-Semi method. This is an internal function and should not be called directly.
#' @keywords internal
#' @param B A matrix of the parameters \code{B}, the columns are subject to the orthogonality constraint
#' @param X The covariate matrix (This matrix is ordered by the order of Y for faster computation)
#' @param Phit Phit as defined in Sun et al. (2017)
#' @param Fail_Ind The locations of the failure subjects
#' @param bw Kernel bandwidth for X
#' @param bw_optim whether to optimize the bandwidth
#' @param rho (don't change) Parameter for control the linear approximation in line search
#' @param eta (don't change) Factor for decreasing the step size in the backtracking line search
#' @param gamma (don't change) Parameter for updating C by Zhang and Hager (2004)
#' @param tau (don't change) Step size for updating
#' @param epsilon (don't change) Parameter for approximating numerical gradient
#' @param btol (don't change) The \code{$B$} parameter tolerance level
#' @param ftol (don't change) Estimation equation 2-norm tolerance level
#' @param gtol (don't change) Gradient tolerance level
#' @param maxitr Maximum number of iterations
#' @param verbose Should information be displayed
#' @param ncore The number of cores for parallel computing
#' @return The optimizer \code{B} for the estimating equation.
#' 
#' @references Sun, Q., Zhu, R., Wang, T., & Zeng, D. (2019). Counting process-based dimension reduction methods for censored outcomes. 
#' Biometrika, 106(1), 181-196.
#' DOI: \doi{10.1093/biomet/asy064}
#' 
#' @examples
#' # This function should be called internally. When having all objects pre-computed, one can call
#' # surv_solver(B, X, Phit, Fail.Ind,
#' #             rho, eta, gamma, tau, epsilon, btol, ftol, gtol, maxitr, verbose)
#' # to solve for the parameters B.
#' 
surv_dm_solver <- function(B, X, Phit, Fail_Ind, bw, rho, eta, gamma, tau, epsilon, btol, ftol, gtol, maxitr, verbose, ncore) {
    .Call(`_orthoDr_surv_dm_solver`, B, X, Phit, Fail_Ind, bw, rho, eta, gamma, tau, epsilon, btol, ftol, gtol, maxitr, verbose, ncore)
}

#' @title surv_dn_solver \code{C++} function
#' @name surv_dn_solver
#' @description The main optimization function for survival dimensional reduction, the IR-CP method. This is an internal function and should not be called directly.
#' @keywords internal
#' @param B A matrix of the parameters \code{B}, the columns are subject to the orthogonality constraint
#' @param X The covariate matrix (This matrix is ordered by the order of Y for faster computation)
#' @param Phit Phit as defined in Sun et al. (2017)
#' @param Fail_Ind The locations of the failure subjects
#' @param bw Kernel bandwidth for X
#' @param rho (don't change) Parameter for control the linear approximation in line search
#' @param eta (don't change) Factor for decreasing the step size in the backtracking line search
#' @param gamma (don't change) Parameter for updating C by Zhang and Hager (2004)
#' @param tau (don't change) Step size for updating
#' @param epsilon (don't change) Parameter for approximating numerical gradient
#' @param btol (don't change) The \code{$B$} parameter tolerance level
#' @param ftol (don't change) Estimation equation 2-norm tolerance level
#' @param gtol (don't change) Gradient tolerance level
#' @param maxitr Maximum number of iterations
#' @param verbose Should information be displayed
#' @param ncore The number of cores for parallel computing
#' @return The optimizer \code{B} for the estimating equation.
#' 
#' @references Sun, Q., Zhu, R., Wang, T., & Zeng, D. (2019). Counting process-based dimension reduction methods for censored outcomes. 
#' Biometrika, 106(1), 181-196.
#' DOI: \doi{10.1093/biomet/asy064}
#' 
#' @examples
#' # This function should be called internally. When having all objects pre-computed, one can call
#' # surv_solver(B, X, Phit, Fail.Ind,
#' #             rho, eta, gamma, tau, epsilon, btol, ftol, gtol, maxitr, verbose)
#' # to solve for the parameters B.
#'
surv_dn_solver <- function(B, X, Phit, Fail_Ind, bw, rho, eta, gamma, tau, epsilon, btol, ftol, gtol, maxitr, verbose, ncore) {
    .Call(`_orthoDr_surv_dn_solver`, B, X, Phit, Fail_Ind, bw, rho, eta, gamma, tau, epsilon, btol, ftol, gtol, maxitr, verbose, ncore)
}

#' @title surv_forward_solver \code{C++} function
#' @name surv_forward_solver
#' @description The main optimization function for survival dimensional reduction, the forward method. This is an internal function and should not be called directly.
#' @keywords internal
#' @param B A matrix of the parameters \code{B}, the columns are subject to the orthogonality constraint
#' @param X The covariate matrix (This matrix is ordered by the order of Y for faster computation)
#' @param Phit Phit as defined in Sun et al. (2017)
#' @param Fail_Ind The locations of the failure subjects
#' @param bw Kernel bandwidth for X
#' @param rho (don't change) Parameter for control the linear approximation in line search
#' @param eta (don't change) Factor for decreasing the step size in the backtracking line search
#' @param gamma (don't change) Parameter for updating C by Zhang and Hager (2004)
#' @param tau (don't change) Step size for updating
#' @param epsilon (don't change) Parameter for approximating numerical gradient
#' @param btol (don't change) The \code{$B$} parameter tolerance level
#' @param ftol (don't change) Estimation equation 2-norm tolerance level
#' @param gtol (don't change) Gradient tolerance level
#' @param maxitr Maximum number of iterations
#' @param verbose Should information be displayed
#' @param ncore The number of cores for parallel computing
#' @return The optimizer \code{B} for the esitmating equation.
#' 
#' @references 
#' Sun, Q., Zhu, R., Wang, T., & Zeng, D. (2019). Counting process-based dimension reduction methods for censored outcomes. 
#' Biometrika, 106(1), 181-196.
#' DOI: \doi{10.1093/biomet/asy064}
#' 
#' @examples
#' # This function should be called internally. When having all objects pre-computed, one can call
#' # surv_solver(B, X, Phit, Fail.Ind,
#' #             rho, eta, gamma, tau, epsilon, btol, ftol, gtol, maxitr, verbose)
#' # to solve for the parameters B.
#'
surv_forward_solver <- function(B, X, Fail_Ind, bw, rho, eta, gamma, tau, epsilon, btol, ftol, gtol, maxitr, verbose, ncore) {
    .Call(`_orthoDr_surv_forward_solver`, B, X, Fail_Ind, bw, rho, eta, gamma, tau, epsilon, btol, ftol, gtol, maxitr, verbose, ncore)
}

#' @title KernelDist_cross
#' @name KernelDist_cross
#' @description Calculate the kernel distance between testing data and training data
#' @keywords internal
#' @param TestX testing data
#' @param X training data
KernelDist_cross <- function(TestX, X) {
    .Call(`_orthoDr_KernelDist_cross`, TestX, X)
}