R/mmd_l_multi.R

In AnthonyEbert/EasyMMD: Allows users to compute the Maximum Mean Discrepancy (MMD) between two samples.

#' Compute the permuted Jack-knife MMD_linear
#' @param x_obs Numeric Vector
#' @param x_sim Numeric Vector
#' @param k Integer Number of permutations
#' @param var Numeric Kernel size
#' @param agg Logical Should output be aggregated by mean?
#' @export
MMD_l_multi <- function(x_obs, x_sim, k, var = 1, agg = TRUE){
  output <- rep(NA, k)
  n <- length(x_obs)
  stopifnot(n == length(x_sim))

  for(i in 1:k){
    output[i] <- MMD_l(sample(x_obs, n), sample(x_sim, n), var = var)
  }

  if(agg){
    output <- mean(output)
  }

  return(output)
}