feat: features can be always included with the always_include colum…

…n role (#89)

NEWS.md

@@ -1,5 +1,6 @@
 # mlr3fselect (development version)
 
+* feat: Features can be always included with the `always_include` column role.
 * fix: Add `$phash()` method to `AutoFSelector`.
 * fix: Include `FSelector` in hash of `AutoFSelector`.
 * refactor: Change default batch size of `FSelectorRandomSearch` to 10.

R/ObjectiveFSelect.R

@@ -82,7 +82,9 @@ ObjectiveFSelect = R6Class("ObjectiveFSelect",
  tasks = map(private$.xss, function(x) {
  state = self$task$feature_names[unlist(x)]
  task = self$task$clone()
- task$select(state)
+ always_included = task$col_roles$always_included
+ task$set_col_roles(always_included, "feature")
+ task$select(c(state, always_included))
  task
  })
 

R/zzz.R

@@ -16,6 +16,10 @@
  x = utils::getFromNamespace("bbotk_reflections", ns = "bbotk")
  x$optimizer_properties = c(x$optimizer_properties, "requires_model")
 
+ x = utils::getFromNamespace("mlr_reflections", ns = "mlr3")
+ x$task_col_roles$classif = c(x$task_col_roles$classif, "always_included")
+ x$task_col_roles$regr = c(x$task_col_roles$regr, "always_included")
+
  # callbacks
  x = utils::getFromNamespace("mlr_callbacks", ns = "mlr3misc")
  x$add("mlr3fselect.backup", load_callback_backup)

tests/testthat/test_FSelectInstanceSingleCrit.R

@@ -66,3 +66,55 @@ test_that("result$features works", {
  inst$assign_result(xdt, y)
  expect_character(inst$result_feature_set)
 })
+
+test_that("always include variable works", {
+ task = tsk("pima")
+ task$set_col_roles("glucose", "always_included")
+
+ learner = lrn("classif.rpart")
+ resampling = rsmp("cv", folds = 3)
+
+ instance = fselect(
+ fselector = fs("random_search", batch_size = 100),
+ task = task,
+ learner = learner,
+ resampling = resampling,
+ measure = msr("classif.ce"),
+ terminator = trm("evals", n_evals = 100),
+ store_models = TRUE
+ )
+
+ data = as.data.table(instance$archive)
+
+ expect_names(instance$archive$cols_x, disjunct.from = "gloucose")
+ expect_names(names(instance$archive$data), disjunct.from = "gloucose")
+ walk(data$resample_result, function(rr) {
+ expect_names(names(rr$learners[[1]]$state$data_prototype), must.include = "glucose")
+ })
+})
+
+test_that("always include variables works", {
+ task = tsk("pima")
+ task$set_col_roles(c("glucose", "age"), "always_included")
+
+ learner = lrn("classif.rpart")
+ resampling = rsmp("cv", folds = 3)
+
+ instance = fselect(
+ fselector = fs("random_search", batch_size = 100),
+ task = task,
+ learner = learner,
+ resampling = resampling,
+ measure = msr("classif.ce"),
+ terminator = trm("evals", n_evals = 100),
+ store_models = TRUE
+ )
+
+ data = as.data.table(instance$archive)
+
+ expect_names(instance$archive$cols_x, disjunct.from = c("glucose", "age"))
+ expect_names(names(instance$archive$data), disjunct.from = c("glucose", "age"))
+ walk(data$resample_result, function(rr) {
+ expect_names(names(rr$learners[[1]]$state$data_prototype), must.include = c("glucose", "age"))
+ })
+})