Feat/inner valid (#294)

* anticipate mlr3 and mlr3pipelines changes * update workflows * typo in workflow * renaming * fix workflow * ... * wip * ... * ... * fix: add patterns to global variables --------- Co-authored-by: be-marc <[email protected]>
mlr-org · Jun 20, 2024 · c566afb · c566afb
1 parent e9905f4
commit c566afb
Show file tree

Hide file tree

Showing 28 changed files with 406 additions and 137 deletions.
diff --git a/.github/workflows/test-task-1.yml b/.github/workflows/test-task-1.yml
@@ -0,0 +1,43 @@
+# r cmd check workflow of the mlr3 ecosystem v0.1.0
+# https://github.com/mlr-org/actions
+on:
+  workflow_dispatch:
+  push:
+    branches:
+      - main
+  pull_request:
+    branches:
+      - main
+
+name: mlr3 & mlr3pipelines change
+
+jobs:
+  r-cmd-check:
+    runs-on: ${{ matrix.config.os }}
+
+    name: ${{ matrix.config.os }} (${{ matrix.config.r }})
+
+    env:
+      GITHUB_PAT: ${{ secrets.GITHUB_TOKEN }}
+
+    strategy:
+      fail-fast: false
+      matrix:
+        config:
+          - {os: ubuntu-latest,   r: 'release'}
+
+    steps:
+      - uses: actions/checkout@v3
+
+      - name: mlr3
+        run: 'echo -e "Remotes:\n    mlr-org/mlr3@feat/train-predict,\n    mlr-org/mlr3pipelines$fixt/uses_test_task" >> DESCRIPTION'
+
+      - uses: r-lib/actions/setup-r@v2
+        with:
+          r-version: ${{ matrix.config.r }}
+
+      - uses: r-lib/actions/setup-r-dependencies@v2
+        with:
+          extra-packages: any::rcmdcheck
+          needs: check
+      - uses: r-lib/actions/check-r-package@v2
diff --git a/.github/workflows/test-task-2.yml b/.github/workflows/test-task-2.yml
@@ -0,0 +1,43 @@
+# r cmd check workflow of the mlr3 ecosystem v0.1.0
+# https://github.com/mlr-org/actions
+on:
+  workflow_dispatch:
+  push:
+    branches:
+      - main
+  pull_request:
+    branches:
+      - main
+
+name: mlr3 & mlr3pipelines change
+
+jobs:
+  r-cmd-check:
+    runs-on: ${{ matrix.config.os }}
+
+    name: ${{ matrix.config.os }} (${{ matrix.config.r }})
+
+    env:
+      GITHUB_PAT: ${{ secrets.GITHUB_TOKEN }}
+
+    strategy:
+      fail-fast: false
+      matrix:
+        config:
+          - {os: ubuntu-latest,   r: 'release'}
+
+    steps:
+      - uses: actions/checkout@v3
+
+      - name: mlr3
+        run: 'echo -e "Remotes:\n    mlr-org/mlr3@feat/train-predict,\n    mlr-org/mlr3pipelines$feat/test-rows" >> DESCRIPTION'
+
+      - uses: r-lib/actions/setup-r@v2
+        with:
+          r-version: ${{ matrix.config.r }}
+
+      - uses: r-lib/actions/setup-r-dependencies@v2
+        with:
+          extra-packages: any::rcmdcheck
+          needs: check
+      - uses: r-lib/actions/check-r-package@v2
diff --git a/DESCRIPTION b/DESCRIPTION
@@ -28,7 +28,7 @@ Imports:
     checkmate,
     data.table,
     mlr3misc (>= 0.9.4),
-    paradox,
+    paradox (>= 1.0.0),
     R6
 Suggests:
     DiceKriging,
@@ -45,6 +45,8 @@ Suggests:
     rmarkdown,
     testthat (>= 3.0.0),
     xgboost (>= 1.6.0)
+Remotes:
+    mlr-org/mlr3@feat/inner_valid
 Config/testthat/edition: 3
 Encoding: UTF-8
 NeedsCompilation: no

diff --git a/NAMESPACE b/NAMESPACE
@@ -34,6 +34,7 @@ import(paradox)
 importFrom(R6,R6Class)
 importFrom(mlr3,LearnerClassif)
 importFrom(mlr3,LearnerRegr)
+importFrom(mlr3,assert_validate)
 importFrom(mlr3,mlr_learners)
 importFrom(stats,predict)
 importFrom(stats,reformulate)

diff --git a/NEWS.md b/NEWS.md
@@ -1,5 +1,8 @@
 # mlr3learners (development version)
 
+* feat: `LearnerClassifXgboost` and `LearnerRegrXgboost` now support internal tuning and validation.
+  This now also works in conjunction with `mlr3pipelines`.
+
 # mlr3learners 0.6.0
 
 * Adaption to new paradox version 1.0.0.

diff --git a/R/LearnerClassifXgboost.R b/R/LearnerClassifXgboost.R
@@ -32,13 +32,9 @@
 #'
 #' @section Early stopping:
 #' Early stopping can be used to find the optimal number of boosting rounds.
-#' The `early_stopping_set` parameter controls which set is used to monitor the performance.
-#' Set `early_stopping_set = "test"` to monitor the performance of the model on the test set while training.
-#' The test set for early stopping can be set with the `"test"` row role in the [mlr3::Task].
-#' Additionally, the range must be set in which the performance must increase with `early_stopping_rounds` and the maximum number of boosting rounds with `nrounds`.
-#' While resampling, the test set is automatically applied from the [mlr3::Resampling].
-#' Not that using the test set for early stopping can potentially bias the performance scores.
-#' See the section on early stopping in the examples.
+#' The `early_stopping` parameter controls which set is used to monitor the performance.
+#' Set `early_stopping_rounds` to an integer vaulue to monitor the performance of the model on the validation set while training.
+#' For information on how to configure the valdiation set, see the *Validation* section of [`mlr3::Learner`].
 #'
 #' @templateVar id classif.xgboost
 #' @template learner
@@ -55,19 +51,24 @@
 #' # Train learner with early stopping on spam data set
 #' task = tsk("spam")
 #'
-#' # Split task into training and test set
+#' # Split task into training and validation data
 #' split = partition(task, ratio = 0.8)
-#' task$set_row_roles(split$test, "test")
+#' task$divide(split$test)
+#'
+#' task
 #'
 #' # Set early stopping parameter
 #' learner = lrn("classif.xgboost",
 #'   nrounds = 100,
 #'   early_stopping_rounds = 10,
-#'   early_stopping_set = "test"
+#'   validate = "internal_valid"
 #' )
 #'
 #' # Train learner with early stopping
 #' learner$train(task)
+#'
+#' learner$internal_tuned_values
+#' learner$internal_valid_scores
 #' }
 LearnerClassifXgboost = R6Class("LearnerClassifXgboost",
   inherit = LearnerClassif,
@@ -77,6 +78,16 @@ LearnerClassifXgboost = R6Class("LearnerClassifXgboost",
     #' @description
     #' Creates a new instance of this [R6][R6::R6Class] class.
     initialize = function() {
+
+      p_nrounds = p_int(1L,
+        tags = c("train", "hotstart", "internal_tuning"),
+        aggr = crate(function(x) as.integer(ceiling(mean(unlist(x)))), .parent = topenv()),
+        in_tune_fn = crate(function(domain, param_vals) {
+          assert_true(!is.null(param_vals$early_stopping), .var.name = "early stopping rounds is set")
+          assert_integerish(domain$upper, len = 1L, any.missing = FALSE) }, .parent = topenv()),
+        disable_in_tune = list(early_stopping_rounds = NULL)
+      )
+
       ps = ps(
         alpha                       = p_dbl(0, default = 0, tags = "train"),
         approxcontrib               = p_lgl(default = FALSE, tags = "predict"),
@@ -89,7 +100,6 @@ LearnerClassifXgboost = R6Class("LearnerClassifXgboost",
         device                      = p_uty(default = "cpu", tags = "train"),
         disable_default_eval_metric = p_lgl(default = FALSE, tags = "train"),
         early_stopping_rounds       = p_int(1L, default = NULL, special_vals = list(NULL), tags = "train"),
-        early_stopping_set          = p_fct(c("none", "train", "test"), default = "none", tags = "train"),
         eta                         = p_dbl(0, 1, default = 0.3, tags = c("train", "control")),
         eval_metric                 = p_uty(tags = "train"),
         feature_selector            = p_fct(c("cyclic", "shuffle", "random", "greedy", "thrifty"), default = "cyclic", tags = "train", depends = quote(booster == "gblinear")),
@@ -108,8 +118,8 @@ LearnerClassifXgboost = R6Class("LearnerClassifXgboost",
         min_child_weight            = p_dbl(0, default = 1, tags = c("train", "control")),
         missing                     = p_dbl(default = NA, tags = c("train", "predict"), special_vals = list(NA, NA_real_, NULL)),
         monotone_constraints        = p_uty(default = 0, tags = c("train", "control"), custom_check = crate(function(x) { checkmate::check_integerish(x, lower = -1, upper = 1, any.missing = FALSE) })), # nolint
+        nrounds                     = p_nrounds,
         normalize_type              = p_fct(c("tree", "forest"), default = "tree", tags = "train", depends = quote(booster == "dart")),
-        nrounds                     = p_int(1L, tags = c("train", "hotstart")),
         nthread                     = p_int(1L, default = 1L, tags = c("train", "control", "threads")),
         ntreelimit                  = p_int(1L, default = NULL, special_vals = list(NULL), tags = "predict"),
         num_parallel_tree           = p_int(1L, default = 1L, tags = c("train", "control")),
@@ -141,18 +151,17 @@ LearnerClassifXgboost = R6Class("LearnerClassifXgboost",
         verbose                     = p_int(0L, 2L, default = 1L, tags = "train"),
         watchlist                   = p_uty(default = NULL, tags = "train"),
         xgb_model                   = p_uty(default = NULL, tags = "train")
-
       )
 
       # custom defaults
-      ps$values = list(nrounds = 1L, nthread = 1L, verbose = 0L, early_stopping_set = "none")
+      ps$values = list(nrounds = 1L, nthread = 1L, verbose = 0L)
 
       super$initialize(
         id = "classif.xgboost",
         predict_types = c("response", "prob"),
         param_set = ps,
         feature_types = c("logical", "integer", "numeric"),
-        properties = c("weights", "missings", "twoclass", "multiclass", "importance", "hotstart_forward"),
+        properties = c("weights", "missings", "twoclass", "multiclass", "importance", "hotstart_forward", "internal_tuning", "validation"),
         packages = c("mlr3learners", "xgboost"),
         label = "Extreme Gradient Boosting",
         man = "mlr3learners::mlr_learners_classif.xgboost"
@@ -174,8 +183,30 @@ LearnerClassifXgboost = R6Class("LearnerClassifXgboost",
       set_names(imp$Gain, imp$Feature)
     }
   ),
-
+  active = list(
+    #' @field internal_valid_scores
+    #' The last observation of the validation scores for all metrics.
+    #' Extracted from `model$evaluation_log`
+    internal_valid_scores = function() {
+      self$state$internal_valid_scores
+    },
+    #' @field internal_tuned_values
+    #' Returns the early stopped iterations if `early_stopping_rounds` was set during training.
+    internal_tuned_values = function() {
+      self$state$internal_tuned_values
+    },
+    #' @field validate
+    #' How to construct the internal validation data. This parameter can be either `NULL`,
+    #' a ratio, `"test"`, or `"predefined"`.
+    validate = function(rhs) {
+      if (!missing(rhs)) {
+        private$.validate = assert_validate(rhs)
+      }
+      private$.validate
+    }
+  ),
   private = list(
+    .validate = NULL,
     .train = function(task) {
 
       pv = self$param_set$get_values(tags = "train")
@@ -217,19 +248,18 @@ LearnerClassifXgboost = R6Class("LearnerClassifXgboost",
         xgboost::setinfo(data, "weight", task$weights$weight)
       }
 
-      if (pv$early_stopping_set != "none") {
-        pv$watchlist = c(pv$watchlist, list(train = data))
-      }
-
       # the last element in the watchlist is used as the early stopping set
 
-      if (pv$early_stopping_set == "test" && !is.null(task$row_roles$test)) {
-        test_data = task$data(rows = task$row_roles$test, cols = task$feature_names)
-        test_label = nlvls - as.integer(task$truth(rows = task$row_roles$test))
+      internal_valid_task = task$internal_valid_task
+      if (!is.null(pv$early_stopping_rounds) && is.null(internal_valid_task)) {
+        stopf("Learner (%s): Configure field 'validate' to enable early stopping.", self$id)
+      }
+      if (!is.null(internal_valid_task)) {
+        test_data = internal_valid_task$data(cols = internal_valid_task$feature_names)
+        test_label = nlvls - as.integer(internal_valid_task$truth(rows = internal_valid_task$row_roles$test))
         test_data = xgboost::xgb.DMatrix(data = as_numeric_matrix(test_data), label = test_label)
         pv$watchlist = c(pv$watchlist, list(test = test_data))
       }
-      pv$early_stopping_set = NULL
 
       invoke(xgboost::xgb.train, data = data, .args = pv)
     },
@@ -282,7 +312,6 @@ LearnerClassifXgboost = R6Class("LearnerClassifXgboost",
       if (!is.null(pars_train$early_stopping_rounds)) {
         stop("The parameter `early_stopping_rounds` is set. Early stopping and hotstarting are incompatible.")
       }
-      pars$early_stopping_set = NULL
 
       # Calculate additional boosting iterations
       # niter in model and nrounds in ps should be equal after train and continue
@@ -295,10 +324,29 @@ LearnerClassifXgboost = R6Class("LearnerClassifXgboost",
       data = xgboost::xgb.DMatrix(data = as_numeric_matrix(data), label = label)
 
       invoke(xgboost::xgb.train, data = data, xgb_model = model, .args = pars)
+    },
+
+    .extract_internal_tuned_values = function() {
+      if (is.null(self$state$param_vals$early_stopping_rounds)) {
+        return(named_list())
+      }
+      list(nrounds = self$model$niter)
+    },
+
+    .extract_internal_valid_scores = function() {
+      if (is.null(self$model$evaluation_log)) {
+        return(named_list())
+      }
+      as.list(self$model$evaluation_log[
+        get(".N"),
+        set_names(get(".SD"), gsub("^test_", "", colnames(get(".SD",)))),
+        .SDcols = patterns("^test_")
+      ])
     }
   )
 )
 
+
 #' @export
 default_values.LearnerClassifXgboost = function(x, search_space, task, ...) { # nolint
   special_defaults = list(