CCC.R

   1 #' constructionModelesLassoMLE
   2 #'
   3 #' TODO: description
   4 #'
   5 #' @param ...
   6 #'
   7 #' @return ...
   8 #'
   9 #' export
  10 constructionModelesLassoMLE = function(phiInit, rhoInit, piInit, gamInit, mini, maxi,
  11     gamma, X, Y, seuil, tau, selected, ncores=3, verbose=FALSE)
  12 {
  13     if (ncores > 1)
  14     {
  15         cl = parallel::makeCluster(ncores)
  16         parallel::clusterExport( cl, envir=environment(),
  17             varlist=c("phiInit","rhoInit","gamInit","mini","maxi","gamma","X","Y","seuil",
  18             "tau","selected","ncores","verbose") )
  19     }
  20
  21     # Individual model computation
  22     computeAtLambda <- function(lambda)
  23     {
  24         if (ncores > 1)
  25             require("valse") #// nodes start with an ampty environment
  26
  27         if (verbose)
  28             print(paste("Computations for lambda=",lambda))
  29
  30         n = dim(X)[1]
  31         p = dim(phiInit)[1]
  32         m = dim(phiInit)[2]
  33         k = dim(phiInit)[3]
  34
  35         sel.lambda = selected[[lambda]]
  36 #       col.sel = which(colSums(sel.lambda)!=0) #if boolean matrix
  37         col.sel <- which( sapply(sel.lambda,length) > 0 ) #if list of selected vars
  38
  39         if (length(col.sel) == 0)
  40             return (NULL)
  41
  42         # lambda == 0 because we compute the EMV: no penalization here
  43         res = EMGLLF(phiInit[col.sel,,],rhoInit,piInit,gamInit,mini,maxi,gamma,0,
  44             X[,col.sel],Y,tau)
  45
  46         # Eval dimension from the result + selected
  47         phiLambda2 = res_EM$phi
  48         rhoLambda = res_EM$rho
  49         piLambda = res_EM$pi
  50         phiLambda = array(0, dim = c(p,m,k))
  51         for (j in seq_along(col.sel))
  52             phiLambda[col.sel[j],,] = phiLambda2[j,,]
  53
  54         dimension = 0
  55         for (j in 1:p)
  56         {
  57             b = setdiff(1:m, sel.lambda[,j])
  58             if (length(b) > 0)
  59                 phiLambda[j,b,] = 0.0
  60             dimension = dimension + sum(sel.lambda[,j]!=0)
  61         }
  62
  63         # on veut calculer la vraisemblance avec toutes nos estimations
  64         densite = vector("double",n)
  65         for (r in 1:k)
  66         {
  67             delta = Y%*%rhoLambda[,,r] - (X[, col.sel]%*%phiLambda[col.sel,,r])
  68             densite = densite + piLambda[r] *
  69                 det(rhoLambda[,,r])/(sqrt(2*base::pi))^m * exp(-tcrossprod(delta)/2.0)
  70         }
  71         llhLambda = c( sum(log(densite)), (dimension+m+1)*k-1 )
  72         list("phi"= phiLambda, "rho"= rhoLambda, "pi"= piLambda, "llh" = llhLambda)
  73     }
  74
  75     #Pour chaque lambda de la grille, on calcule les coefficients
  76     out =
  77         if (ncores > 1)
  78             parLapply(cl, glambda, computeAtLambda)
  79         else
  80             lapply(glambda, computeAtLambda)
  81
  82     if (ncores > 1)
  83         parallel::stopCluster(cl)
  84
  85     out
  86 }