add embedded ensemble feature selection

DESCRIPTION

@@ -78,6 +78,7 @@ Collate:
     'assertions.R'
     'auto_fselector.R'
     'bibentries.R'
+    'embedded_ensemble_fselect.R'
     'ensemble_fselect.R'
     'extract_inner_fselect_archives.R'
     'extract_inner_fselect_results.R'

NAMESPACE

@@ -36,6 +36,7 @@ export(auto_fselector)
 export(callback_batch_fselect)
 export(clbk)
 export(clbks)
+export(embedded_ensemble_fselect)
 export(ensemble_fselect)
 export(extract_inner_fselect_archives)
 export(extract_inner_fselect_results)

R/embedded_ensemble_fselect.R

@@ -0,0 +1,128 @@
+#' @title Embedded Ensemble Feature Selection
+#'
+#' @include CallbackBatchFSelect.R
+#'
+#' @description
+#' Ensemble feature selection using multiple learners.
+#' The ensemble feature selection method is designed to identify the most informative features from a given dataset by leveraging multiple machine learning models and resampling techniques.
+#' Returns an [EnsembleFSResult].
+#'
+#' @details
+#' The method begins by applying an initial resampling technique specified by the user, to create **multiple subsamples** from the original dataset.
+#' This resampling process helps in generating diverse subsets of data for robust feature selection.
+#'
+#' For each subsample generated in the previous step, the method applies learners
+#' that support **embedded feature selection**.
+#' These learners are then scored on their ability to predict on the resampled
+#' test sets, storing the selected features during training, for each
+#' combination of subsample and learner.
+#'
+#' Results are stored in an [EnsembleFSResult].
+#'
+#' @param learners (list of [mlr3::Learner])\cr
+#'  The learners to be used for feature selection.
+#'  All learners must have the `selected_features` property, i.e. implement
+#'  embedded feature selection (e.g. regularized models).
+#' @param init_resampling ([mlr3::Resampling])\cr
+#'  The initial resampling strategy of the data, from which each train set
+#'  will be passed on to the learners and each test set will be used for
+#'  prediction.
+#'  Can only be [mlr3::ResamplingSubsampling] or [mlr3::ResamplingBootstrap].
+#' @param measure ([mlr3::Measure])\cr
+#'  The measure used to score each learner on the test sets generated by
+#'  `init_resampling`.
+#'  If `NULL`, default measure is used.
+#' @param store_benchmark_result (`logical(1)`)\cr
+#'  Whether to store the benchmark result in [EnsembleFSResult] or not.
+#'
+#' @template param_task
+#'
+#' @returns an [EnsembleFSResult] object.
+#'
+#' @source
+#' `r format_bib("meinshausen2010", "hedou2024")`
+#' @export
+#' @examples
+#' \donttest{
+#'   eefsr = embedded_ensemble_fselect(
+#'     task = tsk("sonar"),
+#'     learners = lrns(c("classif.rpart", "classif.featureless")),
+#'     init_resampling = rsmp("subsampling", repeats = 5),
+#'     measure = msr("classif.ce")
+#'   )
+#'   eefsr
+#' }
+embedded_ensemble_fselect = function(
+    task,
+    learners,
+    init_resampling,
+    measure,
+    store_benchmark_result = TRUE
+  ) {
+  # TOCHECK: DO WE NEED callbacks?
+  assert_task(task)
+  # check all learners support `selected_features()`, ie embedded feature selection
+  assert_learners(as_learners(learners), task = task, properties = "selected_features")
+  assert_resampling(init_resampling)
+  assert_choice(class(init_resampling)[1], choices = c("ResamplingBootstrap", "ResamplingSubsampling"))
+  assert_flag(store_benchmark_result)
+
+  init_resampling$instantiate(task)
+  grid = map_dtr(seq(init_resampling$iters), function(i) {
+    # create task and resampling for each outer iteration
+    task_train = task$clone()$filter(init_resampling$train_set(i))
+    pred_task  = task$clone()$filter(init_resampling$test_set(i))
+    # TOCHECK: better way to keep the test sets?
+    resampling = rsmp("insample")$instantiate(task_train)
+
+    data.table(
+      resampling_iteration = i,
+      learner_id = map_chr(learners, "id"),
+      learner = learners,
+      task = list(task_train),
+      pred_task = list(pred_task),
+      resampling = list(resampling)
+    )
+  })
+
+  design = grid[, c("learner", "task", "resampling"), with = FALSE]
+  print(design)
+
+  bmr = benchmark(design, store_models = TRUE)
+
+  trained_learners = bmr$score()$learner
+
+  # extract selected features
+  features = map(trained_learners, function(learner) {
+    learner$selected_features()
+  })
+
+  # extract n_features
+  n_features = map_int(features, length)
+
+  # TOCHECK: better way to predict on the kept test sets? if measure needs
+  # extract scores from the init_resampling's test set
+  scores = map_dbl(seq(length(trained_learners)), function(i) {
+    learner = trained_learners[[i]]
+    pred_task = grid$pred_task[[i]] # TOCHECK: 1-1 with the trained_learners?
+    p = learner$predict(pred_task)
+    p$score(measure) # TOCHECK: what if train_task is needed, can we do this with a resampling scheme?
+  })
+
+  set(grid, j = "features", value = features)
+  set(grid, j = "n_features", value = n_features)
+  set(grid, j = measure$id, value = scores)
+  set(grid, j = "learner", value = NULL)
+  # remove R6 objects
+  set(grid, j = "task", value = NULL)
+  set(grid, j = "pred_task", value = NULL)
+  set(grid, j = "resampling", value = NULL)
+
+  EnsembleFSResult$new(
+    result = grid,
+    features = task$feature_names,
+    benchmark_result = if (store_benchmark_result) bmr,
+    measure_id = measure$id,
+    minimize = measure$minimize
+  )
+}