pkg/R/main.R

   1 #' runValse
   2 #'
   3 #' Main function
   4 #'
   5 #' @param X matrix of covariates (of size n*p)
   6 #' @param Y matrix of responses (of size n*m)
   7 #' @param procedure among 'LassoMLE' or 'LassoRank'
   8 #' @param selecMod method to select a model among 'DDSE', 'DJump', 'BIC' or 'AIC'
   9 #' @param gamma integer for the power in the penaly, by default = 1
  10 #' @param mini integer, minimum number of iterations in the EM algorithm, by default = 10
  11 #' @param maxi integer, maximum number of iterations in the EM algorithm, by default = 100
  12 #' @param eps real, threshold to say the EM algorithm converges, by default = 1e-4
  13 #' @param kmin integer, minimum number of clusters, by default = 2
  14 #' @param kmax integer, maximum number of clusters, by default = 10
  15 #' @param rank.min integer, minimum rank in the low rank procedure, by default = 1
  16 #' @param rank.max integer, maximum rank in the low rank procedure, by default = 5
  17 #' @param ncores_outer Number of cores for the outer loop on k
  18 #' @param ncores_inner Number of cores for the inner loop on lambda
  19 #' @param thresh real, threshold to say a variable is relevant, by default = 1e-8
  20 #' @param grid_lambda, a vector with regularization parameters if known, by default numeric(0)
  21 #' @param size_coll_mod (Maximum) size of a collection of models, by default 50
  22 #' @param fast TRUE to use compiled C code, FALSE for R code only
  23 #' @param verbose TRUE to show some execution traces
  24 #' @param plot TRUE to plot the selected models after run
  25 #'
  26 #' @return
  27 #' The selected model (except if 'DDSE' or 'DJump' is used to select a model and the collection of models
  28 #' has less than 11 models, the function returns the collection as it can not select one - in that case,
  29 #' it is adviced to use 'AIC' or 'BIC' to select a model)
  30 #'
  31 #' @examples
  32 #' n = 50; m = 10; p = 5
  33 #' beta = array(0, dim=c(p,m,2))
  34 #' beta[,,1] = 1
  35 #' beta[,,2] = 2
  36 #' data = generateXY(n, c(0.4,0.6), rep(0,p), beta, diag(0.5, p), diag(0.5, m))
  37 #' X = data$X
  38 #' Y = data$Y
  39 #' res = runValse(X, Y)
  40 #' X <- matrix(runif(100), nrow=50)
  41 #' Y <- matrix(runif(100), nrow=50)
  42 #' res = runValse(X, Y)
  43 #'
  44 #' @export
  45 runValse <- function(X, Y, procedure = "LassoMLE", selecMod = "DDSE", gamma = 1, mini = 10,
  46   maxi = 50, eps = 1e-04, kmin = 2, kmax = 3, rank.min = 1, rank.max = 5, ncores_outer = 1,
  47   ncores_inner = 1, thresh = 1e-08, grid_lambda = numeric(0), size_coll_mod = 50,
  48   fast = TRUE, verbose = FALSE, plot = TRUE)
  49 {
  50   n <- nrow(X)
  51   p <- ncol(X)
  52   m <- ncol(Y)
  53
  54   if (verbose)
  55     print("main loop: over all k and all lambda")
  56
  57   if (ncores_outer > 1) {
  58     cl <- parallel::makeCluster(ncores_outer, outfile = "")
  59     parallel::clusterExport(cl = cl, envir = environment(), varlist = c("X",
  60       "Y", "procedure", "selecMod", "gamma", "mini", "maxi", "eps", "kmin",
  61       "kmax", "rank.min", "rank.max", "ncores_outer", "ncores_inner", "thresh",
  62       "size_coll_mod", "verbose", "p", "m"))
  63   }
  64
  65   # Compute models with k components
  66   computeModels <- function(k)
  67   {
  68     if (ncores_outer > 1)
  69       require("valse") #nodes start with an empty environment
  70
  71     if (verbose)
  72       print(paste("Parameters initialization for k =", k))
  73     # smallEM initializes parameters by k-means and regression model in each
  74     # component, doing this 20 times, and keeping the values maximizing the
  75     # likelihood after 10 iterations of the EM algorithm.
  76     P <- initSmallEM(k, X, Y, fast)
  77     if (length(grid_lambda) == 0)
  78     {
  79       grid_lambda <- computeGridLambda(P$phiInit, P$rhoInit, P$piInit, P$gamInit,
  80                                        X, Y, gamma, mini, maxi, eps, fast)
  81     }
  82     if (length(grid_lambda) > size_coll_mod)
  83       grid_lambda <- grid_lambda[seq(1, length(grid_lambda), length.out = size_coll_mod)]
  84
  85     if (verbose)
  86       print("Compute relevant parameters")
  87     # select variables according to each regularization parameter from the grid:
  88     # S$selected corresponding to selected variables
  89     S <- selectVariables(P$phiInit, P$rhoInit, P$piInit, P$gamInit, mini, maxi,
  90       gamma, grid_lambda, X, Y, thresh, eps, ncores_inner, fast)
  91
  92     if (procedure == "LassoMLE") {
  93       if (verbose)
  94         print("run the procedure Lasso-MLE")
  95       # compute parameter estimations, with the Maximum Likelihood Estimator,
  96       # restricted on selected variables.
  97       models <- constructionModelesLassoMLE(P$phiInit, P$rhoInit, P$piInit,
  98         P$gamInit, mini, maxi, gamma, X, Y, eps, S, ncores_inner, fast, verbose)
  99     } else {
 100       if (verbose)
 101         print("run the procedure Lasso-Rank")
 102       # compute parameter estimations, with the Low Rank Estimator, restricted on
 103       # selected variables.
 104       models <- constructionModelesLassoRank(S, k, mini, maxi, X, Y, eps, rank.min,
 105         rank.max, ncores_inner, fast, verbose)
 106     }
 107     # warning! Some models are NULL after running selectVariables
 108     models <- models[sapply(models, function(cell) !is.null(cell))]
 109     models
 110   }
 111
 112   # List (index k) of lists (index lambda) of models
 113   models_list <-
 114     if (ncores_outer > 1) {
 115       parLapply(cl, kmin:kmax, computeModels)
 116     } else {
 117       lapply(kmin:kmax, computeModels)
 118     }
 119   if (ncores_outer > 1)
 120     parallel::stopCluster(cl)
 121
 122   if (!requireNamespace("capushe", quietly = TRUE))
 123   {
 124     warning("'capushe' not available: returning all models")
 125     return(models_list)
 126   }
 127
 128   # Get summary 'tableauRecap' from models
 129   tableauRecap <- do.call(rbind, lapply(seq_along(models_list), function(i)
 130   {
 131     models <- models_list[[i]]
 132     # For a collection of models (same k, several lambda):
 133     LLH <- sapply(models, function(model) model$llh[1])
 134     k <- length(models[[1]]$pi)
 135     sumPen <- sapply(models, function(model) k * (dim(model$rho)[1] + sum(model$phi[,
 136       , 1] != 0) + 1) - 1)
 137     data.frame(model = paste(i, ".", seq_along(models), sep = ""), pen = sumPen/n,
 138       complexity = sumPen, contrast = -LLH)
 139   }))
 140   tableauRecap <- tableauRecap[which(tableauRecap[, 4] != Inf), ]
 141   if (verbose)
 142     print(tableauRecap)
 143
 144   if (nrow(tableauRecap) > 10) {
 145     modSel <- capushe::capushe(tableauRecap, n)
 146     indModSel <- if (selecMod == "DDSE")
 147     {
 148       as.numeric(modSel@DDSE@model)
 149     } else if (selecMod == "Djump")
 150     {
 151       as.numeric(modSel@Djump@model)
 152     } else if (selecMod == "BIC")
 153     {
 154       modSel@BIC_capushe$model
 155     } else if (selecMod == "AIC")
 156     {
 157       modSel@AIC_capushe$model
 158     }
 159     listMod <- as.integer(unlist(strsplit(as.character(indModSel), "[.]")))
 160     modelSel <- models_list[[listMod[1]]][[listMod[2]]]
 161     modelSel$models <- tableauRecap
 162
 163     if (plot)
 164       print(plot_valse(X, Y, modelSel))
 165     return(modelSel)
 166   }
 167   tableauRecap
 168 }