Merge pull request #725 from mlr-org/ppl_bagging_with_replacement

add `replace` argument to ppl("bagging")

Author: mb706

NEWS.md

@@ -1,8 +1,9 @@
 # mlr3pipelines 0.5.0-9000
 
-* Feature: The `$add_pipeop()` method got an argument `clone` (old behaviour `TRUE` by default)
-* Bugfix: `PipeOpFeatureUnion` in some rare cases dropped variables called `"x"`
-* Compatibility with upcoming paradox release
+* `pipeline_bagging()` gets the `replace` argument (old behaviour `FALSE` by default).
+* Feature: The `$add_pipeop()` method got an argument `clone` (old behaviour `TRUE` by default).
+* Bugfix: `PipeOpFeatureUnion` in some rare cases dropped variables called `"x"`.
+* Compatibility with upcoming paradox release.
 
 # mlr3pipelines 0.5.0-2
 

R/pipeline_bagging.R

@@ -28,6 +28,9 @@
 #'   predictions respectively.
 #'   If `NULL` (default), no averager is added to the end of the graph.
 #'   Note that setting `collect_multipliciy = TRUE` during construction of the averager is required.
+#' @param replace `logical(1)` \cr
+#'   Whether to sample with replacement.
+#'   Default `FALSE`.
 #' @return [`Graph`]
 #' @export
 #' @examples
@@ -36,9 +39,14 @@
 #' lrn_po = po("learner", lrn("regr.rpart"))
 #' task = mlr_tasks$get("boston_housing")
 #' gr = pipeline_bagging(lrn_po, 3, averager = po("regravg", collect_multiplicity = TRUE))
-#' resample(task, GraphLearner$new(gr), rsmp("holdout"))
+#' resample(task, GraphLearner$new(gr), rsmp("holdout"))$aggregate()
+#'
+#' # The original bagging method uses boosting by sampling with replacement.
+#' gr = ppl("bagging", lrn_po, frac = 1, replace = TRUE,
+#'   averager = po("regravg", collect_multiplicity = TRUE))
+#' resample(task, GraphLearner$new(gr), rsmp("holdout"))$aggregate()
 #' }
-pipeline_bagging = function(graph, iterations = 10, frac = 0.7, averager = NULL) {
+pipeline_bagging = function(graph, iterations = 10, frac = 0.7, averager = NULL, replace = FALSE) {
   g = as_graph(graph)
   assert_count(iterations)
   assert_number(frac, lower = 0, upper = 1)
@@ -50,7 +58,7 @@ pipeline_bagging = function(graph, iterations = 10, frac = 0.7, averager = NULL)
   }
 
   po("replicate", param_vals = list(reps = iterations)) %>>!%
-    po("subsample", param_vals = list(frac = frac)) %>>!%
+    po("subsample", param_vals = list(frac = frac, replace = replace)) %>>!%
     g %>>!%
     averager
 }

man/mlr_graphs_bagging.Rd

@@ -39,3 +39,36 @@ test_that("Bagging Pipeline", {
   expect_true(all(map_lgl(predict_out, function(x) "PredictionClassif" %in% class(x))))
 })
 
+test_that("Bagging with replacement", {
+  tsk = tsk("iris")
+  lrn = lrn("classif.rpart")
+  p = ppl("bagging", graph = po(lrn), replace = TRUE, averager = po("classifavg", collect_multiplicity = TRUE))
+  expect_graph(p)
+  res = resample(tsk, GraphLearner$new(p), rsmp("holdout"))
+  expect_resample_result(res)
+
+  tsk$filter(1:140)
+  expect_equal(anyDuplicated(tsk$data()), 0)  # make sure no duplicates
+
+  p = ppl("bagging", iterations = 2, frac = 1,
+    graph = lrn("classif.debug", save_tasks = TRUE),
+    replace = TRUE, averager = po("classifavg", collect_multiplicity = TRUE)
+  )
+  p$train(tsk)
+
+  expect_true(anyDuplicated(p$pipeops$classif.debug$state[[1]]$model$task_train$data()) != 0)
+
+  getOrigId = function(data) {
+    tsk$data()[, origline := .I][data, on = colnames(tsk$data()), origline]
+  }
+  orig_id_1 = getOrigId(p$pipeops$classif.debug$state[[1]]$model$task_train$data())
+  orig_id_2 = getOrigId(p$pipeops$classif.debug$state[[2]]$model$task_train$data())
+
+  expect_equal(length(orig_id_1), 140)
+  expect_equal(length(orig_id_2), 140)
+  # if we sampled the same values twice, the all.equal() would just give TRUE
+  expect_string(all.equal(orig_id_1, orig_id_2))
+
+  expect_true(length(unique(orig_id_1)) < 140)
+  expect_true(length(unique(orig_id_2)) < 140)
+})