R/initSmallEM.R

   1 library(MASS) #generalized inverse of matrix Monroe-Penrose
   2
   3 vec_bin = function(X,r){
   4   Z = c()
   5   indice = c()
   6   j=1
   7   for(i in 1:length(X)){
   8     if(X[i] == r){
   9       Z[i] = 1
  10       indice[j] = i
  11       j=j+1
  12     }
  13     else{
  14       Z[i] = 0
  15     }
  16   }
  17   return(list(Z,indice))
  18 }
  19
  20 initSmallEM = function(k,X,Y,tau){
  21   n = nrow(Y)
  22   m = ncol(Y)
  23   p = ncol(X)
  24
  25   betaInit1 = array(0, dim=c(p,m,k,20))
  26   sigmaInit1 = array(0, dim = c(m,m,k,20))
  27   phiInit1 = array(0, dim = c(p,m,k,20))
  28   rhoInit1 = array(0, dim = c(m,m,k,20))
  29   piInit1 = matrix(0,20,k)
  30   gamInit1 = array(0, dim=c(n,k,20))
  31   LLFinit1 = list()
  32
  33
  34   for(repet in 1:20){
  35     clusters = hclust(dist(y)) #default distance : euclidean
  36     clusterCut = cutree(clusters,k)
  37     Zinit1[,repet] = clusterCut #retourne les indices (à quel cluster indiv_i appartient) d'un clustering hierarchique (nb de cluster = k)
  38
  39     for(r in 1:k){
  40       Z = Zinit1[,repet]
  41       Z_bin = vec_bin(Z,r)
  42       Z_vec = Z_bin[[1]] #vecteur 0 et 1 aux endroits où Z==r
  43       Z_indice = Z_bin[[2]] #renvoit les indices où Z==r
  44
  45       betaInit1[,,r,repet] = ginv(t(x[Z_indice,])%*%x[Z_indice,])%*%t(x[Z_indice,])%*%y[Z_indice,]
  46       sigmaInit1[,,r,repet] = diag(m)
  47       phiInit1[,,r,repet] = betaInit1[,,r,repet]/sigmaInit1[,,r,repet]
  48       rhoInit1[,,r,repet] = solve(sigmaInit1[,,r,repet])
  49       piInit1[repet,r] = sum(Z_vec)/n
  50     }
  51
  52     for(i in 1:n){
  53       for(r in 1:k){
  54         dotProduct = (y[i,]%*%rhoInit1[,,r,repet]-x[i,]%*%phiInit1[,,r,repet]) %*% (y[i,]%*%rhoInit1[,,r,repet]-x[i,]%*%phiInit1[,,r,repet])
  55         Gam[i,r] = piInit1[repet,r]*det(rhoInit1[,,r,repet])*exp(-0.5*dotProduct)
  56       }
  57       sumGamI = sum(gam[i,])
  58       gamInit1[i,,repet]= Gam[i,] / sumGamI
  59     }
  60
  61     miniInit = 10
  62     maxiInit = 11
  63
  64     new_EMG = EMGLLF(phiInit1[,,,repet],rhoInit1[,,,repet],piInit1[repet,],gamInit1[,,repet],miniInit,maxiInit,1,0,x,y,tau)
  65     ##.C("EMGLLF", phiInit = phiInit, rhoInit = rhoInit, ...)
  66     LLFEessai = new_EMG[[4]]
  67     LLFinit1[[repet]] = LLFEessai[[length(LLFEessai)]]
  68   }
  69
  70   b = which.max(LLFinit1)
  71
  72   phiInit = phiInit1[,,,b]
  73   rhoInit = rhoInit1[,,,b]
  74   piInit = piInit1[b,]
  75   gamInit = gamInit1[,,b]
  76
  77   return(list(phiInit, rhoInit, piInit, gamInit))
  78 }
  79
  80