R/RandProject.R

In btaschler/mcap: Model-Based Clustering in High Dimensions via Adaptive Projections

RandProject <- function(xx, q, method = 'gaussian'){
  #' Random projection of input matrix X
  #'
  #' Perform a random projection (Gauss, Achlioptas or Li) of the input matrix X.
  #' Based on the RandPro package. 
  #' 
  #' @seealso \code{\link[RandPro]{form_matrix}}
  #' 
  #' @param xx The data matrix (n x p).
  #' @param q The target dimension. 
  #' @param method Projection method using either a Gaussian (\code{"gaussian"}), 
  #'               a sparse (\code{"achlioptas"}) or very sparse (\code{"li"}) projection 
  #'               matrix. Default: \code{"gaussian"}.
  #' 
  #' @return Projected matrix (n x q).
  #' @examples 
  #'   ## Project to 3 dimensions with Gaussian projection matrix
  #'   RandProject(xx=matrix(rnorm(500),50,10), q=3)
  #'   
  #'   ## Achlioptas (sparse)
  #'   RandProject(xx=matrix(rnorm(500),50,10), q=2, method='achlioptas')
  #'   
  #'   ## Li (very sparse)
  #'   RandProject(xx=matrix(rnorm(5000),50,100), q=10, method='li')
  #' @export
  
  p <- ncol(xx)
  rr <- RandPro::form_matrix(rows = p, 
                             cols = q, 
                             JLT = FALSE, 
                             eps = 0.1, 
                             projection = method)

  return(xx %*% rr)
}