template for CV in RCC

R functions

miss_gen = function(Y, percent = 0.2){
  N = dim(Y)[1]
  P = dim(Y)[2]
  colindex = sample(P,(P*percent))
  rowindex = sample(N,(N*percent))
  missindex = list(colindex = colindex, rowindex = rowindex)
  return(missindex)
}

# methods
flashG_wrapper = function(Y, missindex, rank_max,type = "constant"){
  # delete the value in the missing position
  Y_miss = Y
  Y_miss[missindex$rowindex, missindex$colindex] = NA
  data = flashr::flash_set_data(Y_miss)
  g_flash = flashr::flash_add_greedy(data,Kmax=rank_max,ash_param=list(method = "fdr"),var_type = type)
  misshat_flash = g_flash$EL %*% t(g_flash$EF)
  rmse = sqrt(mean((Y[missindex$rowindex, missindex$colindex] - misshat_flash[missindex$rowindex, missindex$colindex])^2))
  return(rmse)
}

flashB_wrapper = function(Y, missindex, rank_max,type = "constant"){
  # delete the value in the missing position
  Y_miss = Y
  Y_miss[missindex$rowindex, missindex$colindex] = NA
  data = flashr::flash_set_data(Y_miss)
  g_flash = flashr::flash_add_greedy(data,Kmax=rank_max,ash_param=list(method = "fdr"),var_type = type)
  b_flash = flash_backfit(data,g_flash)
  misshat_flash = b_flash$EL %*% t(b_flash$EF)
  rmse = sqrt(mean((Y[missindex$rowindex, missindex$colindex] - misshat_flash[missindex$rowindex, missindex$colindex])^2))
  return(rmse)
}

softImpute_wrapper = function(Y, missindex, rank_max){
  # delete the value in the missing position
  Y_miss = Y
  Y_miss[missindex$rowindex, missindex$colindex] = NA
  gsoft = softImpute::softImpute(Y_miss, rank.max = rank_max, lambda = 30)
  misshat_soft =  gsoft$u %*% diag(gsoft$d) %*% t(gsoft$v)
  rmse = sqrt(mean((Y[missindex$rowindex, missindex$colindex] - misshat_soft[missindex$rowindex, missindex$colindex])^2))
  return(rmse)
}

PMD_wrapper = function(Y, missindex, rank_max){
  # delete the value in the missing position
  Y_miss = Y
  Y_miss[missindex$rowindex, missindex$colindex] = NA
  out = PMA::PMD(Y_miss, K = rank_max)
  misshat_PMD =  out$u %*% diag(out$d) %*% t(out$v)
  rmse = sqrt(mean((Y[missindex$rowindex, missindex$colindex] - misshat_PMD[missindex$rowindex, missindex$colindex])^2))
  return(rmse)
}

library(R.matlab)
# call the R functions
source("~/HG/flash/data/CVmissflashr/template/Rfunction.R")
# read the data
# you need to change the path for different data
Y_data = R.matlab::readMat("~/HG/flash/data/CVmissflashr/Breastcancer/example.mat")
Y = Y_data$Y
N = dim(Y)[1]
P = dim(Y)[2]
Y = Y - rowMeans(Y) %*% t(rep(1,P))
result = c()
missindex = miss_gen(Y, percent = 0.1)
# creat a folder for the output and matlab in the folder we make
if(file.exists("NBSFAout")){
  unlink("NBSFAout", recursive= T)
}
system("mkdir NBSFAout")
# write the missing index
writeMat("./NBSFAout/missindex.mat", missindex = missindex)

flashG_rmse = flashG_wrapper(Y, missindex, rank_max = 50)
pmd_rmse  = PMD_wrapper(Y, missindex, rank_max= 50 )
soft_rmse = softImpute_wrapper(Y, missindex, rank_max = 50)
# then call the matlab estiamtion 
# this for the folder of the data
system("matlab -nosplash -nodesktop -r \"addpath(\'~/HG/flash/data/CVmissflashr/template\'); run(\'missvalue.m\');exit;\" ")
g_nsfa = readMat("./NBSFAout/NSFAresult.mat")
nsfa_rmse = as.vector(g_nsfa$rmse.NSFA)

result = c(flashG_rmse, nsfa_rmse,pmd_rmse,soft_rmse)

saveRDS(result, "./NBSFAout/output.rds")

matlab code

addpath('~/HG/flash/data/missingvalue/methods/NBSF/nsfa-master/');
addpath('~/HG/flash/data/missingvalue/methods/NBSF/nsfa-master/utils/');

% you need to change the path and the data file (not centered)
load ../example.mat;
Ycentered=Y-repmat(mean(Y,2),1,size(Y,2));
settings=defaultsettings();
[settings.D,settings.N]=size(Ycentered);
settings.iterations=100;
% now the missing is decide by the this is the missing index.(or you can upload the missing matrix)

file_path = fullfile(pwd,'NBSFAout/missindex.mat');
load(file_path);
mvmask = ones(settings.D,settings.N);
mvmask(missindex.rowindex,missindex.colindex) = 0;

initialsample=init_nsfa(settings);
[finalsample,resultstable]=nsfa(Ycentered,mvmask,initialsample,settings);
Y_hat = (finalsample.G *  finalsample.X);

sse_NSFA = sum(sum((Ycentered(missindex.rowindex,missindex.colindex) - Y_hat(missindex.rowindex,missindex.colindex)).^2));
mse_NSFA = sse_NSFA/(length(missindex.rowindex)*length(missindex.colindex))
rmse_NSFA = sqrt(mse_NSFA)

save_path = fullfile(pwd,'NBSFAout/NSFAresult');
save(save_path,'rmse_NSFA');