R/utils.R

   1 get_testIndices <- function(n, CV, v, shuffle_inds) {
   2   if (CV$type == "vfold") {
   3     # Slice indices (optionnally shuffled)
   4     first_index = round((v-1) * n / CV$V) + 1
   5     last_index = round(v * n / CV$V)
   6     test_indices = first_index:last_index
   7     if (!is.null(shuffle_inds))
   8       test_indices <- shuffle_inds[test_indices]
   9   }
  10   else
  11     # Monte-Carlo cross-validation
  12     test_indices = sample(n, round(n * CV$test_size))
  13   test_indices
  14 }
  15
  16 splitTrainTest <- function(data, target, testIdx) {
  17   dataTrain <- data[-testIdx,]
  18   targetTrain <- target[-testIdx]
  19   dataTest <- data[testIdx,]
  20   targetTest <- target[testIdx]
  21   # [HACK] R will cast 1-dim matrices into vectors:
  22   if (!is.matrix(dataTrain) && !is.data.frame(dataTrain))
  23     dataTrain <- as.matrix(dataTrain)
  24   if (!is.matrix(dataTest) && !is.data.frame(dataTest))
  25     dataTest <- as.matrix(dataTest)
  26   list(dataTrain=dataTrain, targetTrain=targetTrain,
  27        dataTest=dataTest, targetTest=targetTest)
  28 }