Commit | Line | Data |
---|---|---|
ffdf9447 | 1 | #' selectVariables |
bb551124 | 2 | #' |
7064275b | 3 | #' It is a function which construct, for a given lambda, the sets of relevant variables. |
e01c9b1f | 4 | #' |
5 | #' @param phiInit an initial estimator for phi (size: p*m*k) | |
6 | #' @param rhoInit an initial estimator for rho (size: m*m*k) | |
ffdf9447 | 7 | #' @param piInit\tan initial estimator for pi (size : k) |
e01c9b1f | 8 | #' @param gamInit an initial estimator for gamma |
ffdf9447 BA |
9 | #' @param mini\t\tminimum number of iterations in EM algorithm |
10 | #' @param maxi\t\tmaximum number of iterations in EM algorithm | |
11 | #' @param gamma\t power in the penalty | |
e01c9b1f | 12 | #' @param glambda grid of regularization parameters |
ffdf9447 BA |
13 | #' @param X\t\t\t matrix of regressors |
14 | #' @param Y\t\t\t matrix of responses | |
43d76c49 | 15 | #' @param thresh real, threshold to say a variable is relevant, by default = 1e-8 |
ffdf9447 | 16 | #' @param eps\t\t threshold to say that EM algorithm has converged |
4cc632c9 | 17 | #' @param ncores Number or cores for parallel execution (1 to disable) |
e01c9b1f | 18 | #' |
7064275b | 19 | #' @return a list of outputs, for each lambda in grid: selected,Rho,Pi |
cad71b2c BA |
20 | #' |
21 | #' @examples TODO | |
e01c9b1f | 22 | #' |
cad71b2c | 23 | #' @export |
bb551124 | 24 | #' |
ffdf9447 BA |
25 | selectVariables <- function(phiInit, rhoInit, piInit, gamInit, mini, maxi, gamma, |
26 | glambda, X, Y, thresh = 1e-08, eps, ncores = 3, fast = TRUE) | |
27 | { | |
fb6e49cb | 28 | if (ncores > 1) |
29 | { | |
ffdf9447 BA |
30 | cl <- parallel::makeCluster(ncores, outfile = "") |
31 | parallel::clusterExport(cl = cl, varlist = c("phiInit", "rhoInit", "gamInit", | |
32 | "mini", "maxi", "glambda", "X", "Y", "thresh", "eps"), envir = environment()) | |
fb6e49cb | 33 | } |
34 | ||
35 | # Computation for a fixed lambda | |
36 | computeCoefs <- function(lambda) | |
37 | { | |
ffdf9447 BA |
38 | params <- EMGLLF(phiInit, rhoInit, piInit, gamInit, mini, maxi, gamma, lambda, |
39 | X, Y, eps, fast) | |
fb6e49cb | 40 | |
ffdf9447 BA |
41 | p <- dim(phiInit)[1] |
42 | m <- dim(phiInit)[2] | |
fb6e49cb | 43 | |
ffdf9447 BA |
44 | # selectedVariables: list where element j contains vector of selected variables |
45 | # in [1,m] | |
46 | selectedVariables <- lapply(1:p, function(j) | |
47 | { | |
48 | # from boolean matrix mxk of selected variables obtain the corresponding boolean | |
49 | # m-vector, and finally return the corresponding indices | |
50 | seq_len(m)[apply(abs(params$phi[j, , ]) > thresh, 1, any)] | |
fb6e49cb | 51 | }) |
52 | ||
ffdf9447 | 53 | list(selected = selectedVariables, Rho = params$rho, Pi = params$pi) |
fb6e49cb | 54 | } |
55 | ||
56 | # For each lambda in the grid, we compute the coefficients | |
ffdf9447 BA |
57 | out <- if (ncores > 1) |
58 | parLapply(cl, glambda, computeCoefs) else lapply(glambda, computeCoefs) | |
59 | if (ncores > 1) | |
fb6e49cb | 60 | parallel::stopCluster(cl) |
ffdf9447 BA |
61 | # Suppress models which are computed twice En fait, ca ca fait la comparaison de |
62 | # tous les parametres On veut juste supprimer ceux qui ont les memes variables | |
63 | # sélectionnées sha1_array <- lapply(out, digest::sha1) out[ | |
64 | # duplicated(sha1_array) ] | |
65 | selec <- lapply(out, function(model) model$selected) | |
66 | ind_dup <- duplicated(selec) | |
67 | ind_uniq <- which(!ind_dup) | |
68 | out2 <- list() | |
69 | for (l in 1:length(ind_uniq)) | |
70 | { | |
71 | out2[[l]] <- out[[ind_uniq[l]]] | |
fb6e49cb | 72 | } |
73 | out2 | |
09ab3c16 | 74 | } |