several modifs - pkg looks better (but untested)

[valse.git] / pkg / R / computeGridLambda.R

#' computeGridLambda
#'
#' Construct the data-driven grid for the regularization parameters used for the Lasso estimator
#'
#' @param phiInit value for phi
#' @param rhoInit	value for rho
#' @param piInit	value for pi
#' @param gamInit value for gamma
#' @param X matrix of covariates (of size n*p)
#' @param Y matrix of responses (of size n*m)
#' @param gamma power of weights in the penalty
#' @param mini minimum number of iterations in EM algorithm
#' @param maxi maximum number of iterations in EM algorithm
#' @param tau threshold to stop EM algorithm
#'
#' @return the grid of regularization parameters
#'
#' @export
computeGridLambda = function(phiInit, rhoInit, piInit, gamInit, X, Y,
	gamma, mini, maxi, tau)
{
	n = nrow(X)
	p = dim(phiInit)[1]
	m = dim(phiInit)[2]
	k = dim(phiInit)[3]

	# TODO: explain why gamma=1 instad of just 'gamma'?
  list_EMG = EMGLLF(phiInit, rhoInit, piInit, gamInit, mini, maxi,
		gamma=1, lamba=0, X, Y, tau)
	grid = array(0, dim=c(p,m,k))
	for (i in 1:p)
	{
		for (j in 1:m)
			grid[i,j,] = abs(list_EMG$S[i,j,]) / (n*list_EMG$pi^gamma)
	}
	grid = unique(grid)
	grid = grid[grid <= 1]
	grid
}