| 1 | #' valse |
| 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 rang.min integer, minimum rank in the low rank procedure, by default = 1 |
| 16 | #' @param rang.max integer, maximum rank in the |
| 17 | #' |
| 18 | #' @return a list with estimators of parameters |
| 19 | #' |
| 20 | #' @examples |
| 21 | #' #TODO: a few examples |
| 22 | #' @export |
| 23 | valse = function(X,Y,procedure = 'LassoMLE',selecMod = 'DDSE',gamma = 1,mini = 10, |
| 24 | maxi = 50,eps = 1e-4,kmin = 2,kmax = 2, |
| 25 | rang.min = 1,rang.max = 10, ncores_k=1, ncores_lambda=3, verbose=FALSE) |
| 26 | { |
| 27 | p = dim(X)[2] |
| 28 | m = dim(Y)[2] |
| 29 | n = dim(X)[1] |
| 30 | |
| 31 | tableauRecap = list() |
| 32 | if (verbose) |
| 33 | print("main loop: over all k and all lambda") |
| 34 | |
| 35 | if (ncores_k > 1) |
| 36 | { |
| 37 | cl = parallel::makeCluster(ncores_k) |
| 38 | parallel::clusterExport( cl=cl, envir=environment(), varlist=c("X","Y","procedure", |
| 39 | "selecMod","gamma","mini","maxi","eps","kmin","kmax","rang.min","rang.max", |
| 40 | "ncores_k","ncores_lambda","verbose","p","m","k","tableauRecap") ) |
| 41 | } |
| 42 | |
| 43 | # Compute model with k components |
| 44 | computeModel <- function(k) |
| 45 | { |
| 46 | if (ncores_k > 1) |
| 47 | require("valse") #nodes start with an empty environment |
| 48 | |
| 49 | if (verbose) |
| 50 | print(paste("Parameters initialization for k =",k)) |
| 51 | #smallEM initializes parameters by k-means and regression model in each component, |
| 52 | #doing this 20 times, and keeping the values maximizing the likelihood after 10 |
| 53 | #iterations of the EM algorithm. |
| 54 | P = initSmallEM(k, X, Y) |
| 55 | grid_lambda <- computeGridLambda(P$phiInit, P$rhoInit, P$piInit, P$gamInit, X, Y, |
| 56 | gamma, mini, maxi, eps) |
| 57 | |
| 58 | # TODO: 100 = magic number |
| 59 | if (length(grid_lambda)>100) |
| 60 | grid_lambda = grid_lambda[seq(1, length(grid_lambda), length.out = 100)] |
| 61 | |
| 62 | if (verbose) |
| 63 | print("Compute relevant parameters") |
| 64 | #select variables according to each regularization parameter |
| 65 | #from the grid: A1 corresponding to selected variables, and |
| 66 | #A2 corresponding to unselected variables. |
| 67 | S = selectVariables(P$phiInit,P$rhoInit,P$piInit,P$gamInit,mini,maxi,gamma, |
| 68 | grid_lambda,X,Y,1e-8,eps,ncores_lambda) |
| 69 | |
| 70 | if (procedure == 'LassoMLE') |
| 71 | { |
| 72 | if (verbose) |
| 73 | print('run the procedure Lasso-MLE') |
| 74 | #compute parameter estimations, with the Maximum Likelihood |
| 75 | #Estimator, restricted on selected variables. |
| 76 | model = constructionModelesLassoMLE(phiInit, rhoInit, piInit, gamInit, mini, |
| 77 | maxi, gamma, X, Y, thresh, eps, S$selected) |
| 78 | llh = matrix(ncol = 2) |
| 79 | for (l in seq_along(model[[k]])) |
| 80 | llh = rbind(llh, model[[k]][[l]]$llh) |
| 81 | LLH = llh[-1,1] |
| 82 | D = llh[-1,2] |
| 83 | } |
| 84 | else |
| 85 | { |
| 86 | if (verbose) |
| 87 | print('run the procedure Lasso-Rank') |
| 88 | #compute parameter estimations, with the Low Rank |
| 89 | #Estimator, restricted on selected variables. |
| 90 | model = constructionModelesLassoRank(S$Pi, S$Rho, mini, maxi, X, Y, eps, A1, |
| 91 | rank.min, rank.max) |
| 92 | |
| 93 | ################################################ |
| 94 | ### Regarder la SUITE |
| 95 | phi = runProcedure2()$phi |
| 96 | Phi2 = Phi |
| 97 | if (dim(Phi2)[1] == 0) |
| 98 | Phi[, , 1:k,] <- phi |
| 99 | else |
| 100 | { |
| 101 | Phi <- array(0, dim = c(p, m, kmax, dim(Phi2)[4] + dim(phi)[4])) |
| 102 | Phi[, , 1:(dim(Phi2)[3]), 1:(dim(Phi2)[4])] <<- Phi2 |
| 103 | Phi[, , 1:k,-(1:(dim(Phi2)[4]))] <<- phi |
| 104 | } |
| 105 | } |
| 106 | tableauRecap[[k]] = matrix(c(LLH, D, rep(k, length(model[[k]])), 1:length(model[[k]])), ncol = 4)) |
| 107 | } |
| 108 | |
| 109 | model <- |
| 110 | if (ncores_k > 1) |
| 111 | parLapply(cl, kmin:kmax, computeModel) |
| 112 | else |
| 113 | lapply(kmin:kmax, computeModel) |
| 114 | if (ncores_k > 1) |
| 115 | parallel::stopCluster(cl) |
| 116 | |
| 117 | if (verbose) |
| 118 | print('Model selection') |
| 119 | tableauRecap = do.call( rbind, tableaurecap ) #stack list cells into a matrix |
| 120 | tableauRecap = tableauRecap[rowSums(tableauRecap[, 2:4])!=0,] |
| 121 | tableauRecap = tableauRecap[(tableauRecap[,1])!=Inf,] |
| 122 | data = cbind(1:dim(tableauRecap)[1], tableauRecap[,2], tableauRecap[,2], tableauRecap[,1]) |
| 123 | |
| 124 | require(capushe) |
| 125 | modSel = capushe(data, n) |
| 126 | indModSel <- |
| 127 | if (selecMod == 'DDSE') |
| 128 | as.numeric(modSel@DDSE@model) |
| 129 | else if (selecMod == 'Djump') |
| 130 | as.numeric(modSel@Djump@model) |
| 131 | else if (selecMod == 'BIC') |
| 132 | modSel@BIC_capushe$model |
| 133 | else if (selecMod == 'AIC') |
| 134 | modSel@AIC_capushe$model |
| 135 | model[[tableauRecap[indModSel,3]]][[tableauRecap[indModSel,4]]] |
| 136 | } |