update tests

mlr-org · Nov 18, 2024 · 0e2f93f · 0e2f93f
1 parent efc0155
commit 0e2f93f
Showing 1 changed file with 70 additions and 71 deletions.
diff --git a/tests/testthat/test_ensemble_fselect.R b/tests/testthat/test_ensemble_fselect.R
@@ -1,4 +1,4 @@
-test_that("ensemble feature selection works", {
+test_that("efs works", {
   task = tsk("sonar")
   with_seed(42, {
     efsr = ensemble_fselect(
@@ -7,6 +7,7 @@ test_that("ensemble feature selection works", {
       learners = lrns(c("classif.rpart", "classif.featureless")),
       init_resampling = rsmp("subsampling", repeats = 2),
       inner_resampling = rsmp("cv", folds = 3),
+      inner_measure = msr("classif.ce"),
       measure = msr("classif.ce"),
       terminator = trm("evals", n_evals = 5)
     )
@@ -17,8 +18,10 @@ test_that("ensemble feature selection works", {
   expect_list(efsr$result$features, any.missing = FALSE, len = 4)
   expect_vector(efsr$result$n_features, size = 4)
   expect_vector(efsr$result$classif.ce, size = 4)
+  expect_vector(efsr$result$classif.ce_inner, size = 4)
   expect_benchmark_result(efsr$benchmark_result)
   expect_equal(efsr$measure, "classif.ce")
+  expect_true(efsr$minimize) # classification error
   expect_equal(efsr$n_learners, 2)
   expect_equal(efsr$n_resamples, 2)
 
@@ -27,7 +30,7 @@ test_that("ensemble feature selection works", {
   expect_error(efsr$stability(stability_args = list(20)), "have names")
   stability = efsr$stability(stability_measure = "jaccard", global = FALSE)
   expect_numeric(stability, len = 2)
-  expect_equal(names(stability), c("classif.rpart", "classif.featureless"))
+  expect_equal(names(stability), c("classif.rpart.fselector", "classif.featureless.fselector"))
 
   # default feature ranking
   feature_ranking = efsr$feature_ranking()
@@ -36,7 +39,7 @@ test_that("ensemble feature selection works", {
 
   # pareto_front
   pf = efsr$pareto_front()
-  expect_data_table(pf, nrows = 3)
+  expect_data_table(pf, nrows = 2)
   expect_equal(names(pf), c("n_features", "classif.ce"))
   pf_pred = efsr$pareto_front(type = "estimated")
   expect_data_table(pf_pred, nrows = max(efsr$result$n_features))
@@ -52,70 +55,33 @@ test_that("ensemble feature selection works", {
 
   # data.table conversion
   tab = as.data.table(efsr)
-  expect_equal(names(tab), c("resampling_iteration", "learner_id", "features", "n_features", "classif.ce", "task", "learner", "resampling"))
+  expect_equal(names(tab), c("learner_id", "resampling_iteration", "classif.ce",
+                             "features", "n_features", "classif.ce_inner",
+                             "task", "learner", "resampling"))
+  # scores on train and test sets are different (even though same measure used)
+  assert_true(all(tab$classif.ce != tab$classif.ce_inner))
+
+  # change to use inner measure
+  expect_error(efsr$set_active_measure(measure_id = "XYZ"), regexp = "Must be element")
+  expect_error(efsr$set_active_measure(measure_id = "classif.ce_inner"), regexp = "is missing")
+  efsr$set_active_measure(measure_id = "classif.ce_inner", minimize = TRUE)
+  expect_equal(efsr$measure, "classif.ce_inner")
+  expect_true(efsr$minimize) # classification error
+  pf2 = efsr$pareto_front()
+  expect_data_table(pf2, nrows = 3) # pareto front has changed
 })
 
-test_that("ensemble feature selection works without benchmark result", {
+test_that("efs works with rfe", {
   task = tsk("sonar")
   with_seed(42, {
     efsr = ensemble_fselect(
-      fselector = fs("random_search"),
+      fselector = fs("rfe", subset_sizes = c(60, 20, 10, 5)),
       task = task,
       learners = lrns(c("classif.rpart", "classif.featureless")),
       init_resampling = rsmp("subsampling", repeats = 2),
       inner_resampling = rsmp("cv", folds = 3),
-      measure = msr("classif.ce"),
-      terminator = trm("evals", n_evals = 3),
-      store_benchmark_result = FALSE
-    )
-  })
-
-  expect_character(efsr$man)
-  expect_data_table(efsr$result, nrows = 4)
-  expect_list(efsr$result$features, any.missing = FALSE, len = 4)
-  expect_vector(efsr$result$n_features, size = 4)
-  expect_vector(efsr$result$classif.ce, size = 4)
-  expect_null(efsr$benchmark_result)
-  expect_equal(efsr$measure, "classif.ce")
-  expect_equal(efsr$n_learners, 2)
-  expect_equal(efsr$n_resamples, 2)
-
-  # stability
-  expect_number(efsr$stability(stability_measure = "jaccard"))
-  stability = efsr$stability(stability_measure = "jaccard", global = FALSE)
-  expect_numeric(stability, len = 2)
-  expect_equal(names(stability), c("classif.rpart", "classif.featureless"))
-
-  # default feature ranking
-  feature_ranking = efsr$feature_ranking()
-  expect_data_table(feature_ranking, nrows = length(task$feature_names))
-  expect_equal(names(feature_ranking), c("feature", "score", "norm_score", "borda_score"))
-
-  # pareto_front
-  pf = efsr$pareto_front()
-  expect_data_table(pf, nrows = 3)
-  expect_equal(names(pf), c("n_features", "classif.ce"))
-
-  # knee_points
-  kps = efsr$knee_points()
-  expect_data_table(kps, nrows = 1)
-  expect_equal(names(kps), c("n_features", "classif.ce"))
-
-  # data.table conversion
-  tab = as.data.table(efsr)
-  expect_equal(names(tab), c("resampling_iteration", "learner_id", "features", "n_features", "classif.ce"))
-})
-
-test_that("ensemble feature selection works with rfe", {
-  task = tsk("sonar")
-  with_seed(42, {
-    efsr = ensemble_fselect(
-      fselector = fs("rfe", n_features = 2, feature_fraction = 0.8),
-      task = task,
-      learners = lrns(c("classif.rpart", "classif.featureless")),
-      init_resampling = rsmp("subsampling", repeats = 2),
-      inner_resampling = rsmp("cv", folds = 3),
-      measure = msr("classif.ce"),
+      inner_measure = msr("classif.ce"),
+      measure = msr("classif.acc"),
       terminator = trm("none")
     )
   })
@@ -124,18 +90,25 @@ test_that("ensemble feature selection works with rfe", {
   expect_data_table(efsr$result, nrows = 4)
   expect_list(efsr$result$features, any.missing = FALSE, len = 4)
   expect_vector(efsr$result$n_features, size = 4)
-  expect_vector(efsr$result$classif.ce, size = 4)
+  expect_vector(efsr$result$classif.ce_inner, size = 4)
+  expect_vector(efsr$result$classif.acc, size = 4)
   expect_list(efsr$result$importance, any.missing = FALSE, len = 4)
   expect_benchmark_result(efsr$benchmark_result)
-  expect_equal(efsr$measure, "classif.ce")
+  expect_equal(efsr$measure, "classif.acc")
+  expect_false(efsr$minimize) # accuracy
   expect_equal(efsr$n_learners, 2)
   expect_equal(efsr$n_resamples, 2)
 
+  # change active measure
+  efsr$set_active_measure(measure_id = "classif.ce_inner", minimize = TRUE)
+  expect_equal(efsr$measure, "classif.ce_inner")
+  expect_true(efsr$minimize) # classification error
+
   # stability
   expect_number(efsr$stability(stability_measure = "jaccard"))
   stability = efsr$stability(stability_measure = "jaccard", global = FALSE)
   expect_numeric(stability, len = 2)
-  expect_equal(names(stability), c("classif.rpart", "classif.featureless"))
+  expect_equal(names(stability), c("classif.rpart.fselector", "classif.featureless.fselector"))
 
   # default feature ranking
   feature_ranking = efsr$feature_ranking()
@@ -145,25 +118,28 @@ test_that("ensemble feature selection works with rfe", {
   # pareto_front
   pf = efsr$pareto_front()
   expect_data_table(pf, nrows = 4)
-  expect_equal(names(pf), c("n_features", "classif.ce"))
+  expect_equal(names(pf), c("n_features", "classif.ce_inner"))
   pf_pred = efsr$pareto_front(type = "estimated")
   expect_data_table(pf_pred, nrows = max(efsr$result$n_features))
-  expect_equal(names(pf_pred), c("n_features", "classif.ce"))
+  expect_equal(names(pf_pred), c("n_features", "classif.ce_inner"))
 
   # knee_points
   kps = efsr$knee_points(type = "estimated")
   expect_data_table(kps, nrows = 1)
-  expect_equal(names(kps), c("n_features", "classif.ce"))
+  expect_equal(names(kps), c("n_features", "classif.ce_inner"))
 
   # data.table conversion
   tab = as.data.table(efsr)
-  expect_names(names(tab), identical.to = c("resampling_iteration", "learner_id", "features", "n_features", "classif.ce", "importance", "task", "learner", "resampling"))
+  expect_equal(names(tab), c("learner_id", "resampling_iteration", "classif.acc",
+                             "features", "n_features", "classif.ce_inner",
+                             "importance", "task", "learner", "resampling"))
 })
 
 test_that("EnsembleFSResult initialization", {
   result = data.table(a = 1, b = 3)
   expect_error(EnsembleFSResult$new(result = result, features = LETTERS, measure_id = "a"), "missing elements")
 
+  errors = c(0.13, 0.24, 0.16, 0.11, 0.25, 0.18, 0.15, 0.1, 0.16)
   result = data.table(
     resampling_iteration = c(1, 1, 1, 2, 2, 2, 3, 3, 3),
     learner_id = rep(c("classif.xgboost", "classif.rpart", "classif.ranger"), 3),
@@ -178,7 +154,8 @@ test_that("EnsembleFSResult initialization", {
       c("V2"),
       c("V4", "V12"),
       c("V6", "V15", "V19", "V7")),
-    classif.ce = c(0.13, 0.24, 0.16, 0.11, 0.25, 0.18, 0.15, 0.1, 0.16)
+    classif.ce = errors,
+    classif.acc = 1 - errors
   )
 
   # a feature set includes "V20" which is not included in input "features"
@@ -189,27 +166,49 @@ test_that("EnsembleFSResult initialization", {
   expect_class(efsr, "EnsembleFSResult")
   expect_equal(efsr$n_learners, 3)
   expect_equal(efsr$n_resamples, 3)
+  expect_equal(efsr$measure, "classif.ce")
+  expect_true(efsr$minimize)
   tab = as.data.table(efsr)
   expect_data_table(tab)
   expect_equal(names(tab), c("resampling_iteration", "learner_id", "n_features",
-                             "features", "classif.ce"))
+                             "features", "classif.ce", "classif.acc"))
+
+  # non-existent `inner_measure_id`
+  expect_error(EnsembleFSResult$new(
+    result = result, features = paste0("V", 1:20), measure_id = "classif.ce", inner_measure_id = "a"
+  ), "missing")
+  # `use_inner_measure = TRUE`, no `inner_measure_id` though
+  expect_error(EnsembleFSResult$new(
+    result = result, features = paste0("V", 1:20), use_inner_measure = TRUE, measure_id = "classif.ce"
+  ), "Must be of type")
+  # `use_inner_measure = TRUE`, and `inner_measure_id` is
+  efsr2 = EnsembleFSResult$new(
+    result = result, features = paste0("V", 1:20), use_inner_measure = TRUE,
+    measure_id = "classif.ce", inner_measure_id = "classif.acc",
+    minimize = FALSE # now this applies to the inner measure (accuracy)
+  )
+  expect_class(efsr2, "EnsembleFSResult")
+  expect_equal(efsr2$n_learners, 3)
+  expect_equal(efsr2$n_resamples, 3)
+  expect_equal(efsr2$measure, "classif.acc")
+  expect_false(efsr2$minimize)
 })
 
 test_that("different callbacks can be set", {
   callback_test = callback_batch_fselect("mlr3fselect.test",
     on_eval_before_archive = function(callback, context) {
-      context$aggregated_performance[, callback_active := context$instance$objective$learner$id == "classif.rpart"]
+      context$aggregated_performance[, callback_active := context$instance$objective$learner$id == "classif.rpart.fselector"]
     }
   )
 
   efsr = ensemble_fselect(
-    # 4-5 evaluations on sonar
-    fselector = fs("rfe", n_features = 25, feature_fraction = 0.8),
+    fselector = fs("rfe", subset_sizes = c(60, 20, 10, 5)),
     task = tsk("sonar"),
     learners = lrns(c("classif.rpart", "classif.featureless")),
     init_resampling = rsmp("subsampling", repeats = 2),
     inner_resampling = rsmp("cv", folds = 3),
-    measure = msr("classif.ce"),
+    inner_measure = msr("classif.ce"),
+    measure = msr("classif.acc"),
     terminator = trm("none"),
     callbacks = list(list(callback_test), list())
   )