R/soft_threshold.R

In Tenenhaus/RGCCA: Regularized and Sparse Generalized Canonical Correlation Analysis for Multiblock Data

#' The function soft_threshold() soft-thresholds a vector such that the L1-norm
#' constraint is satisfied.
#' @param x A numeric vector.
#' @param sumabs A numeric constraint on x's L1 norm.
#'
#' @return Returns a vector resulting from the soft thresholding of \eqn{x}
#' given sumabs.
#' @noRd
soft_threshold <- function(x, sumabs = 1) {
  proj <- proj_l1_l2(x, sumabs)
  if (proj$l2_sat) {
    x_proj <- soft(x, proj$lambda)
    return(x_proj / norm(x_proj, type = "2"))
  } else {
    return(proj$sol)
  }
}