feat: regression learners and estimate memory functions (#18)

* ... * feat: add regression single learner * ... * docs: ids * export internal measure * docs * fix * ... * ...
mlr-org · Sep 14, 2024 · 6afb23d · 6afb23d
1 parent eed0eeb
commit 6afb23d
Show file tree

Hide file tree

Showing 65 changed files with 2,241 additions and 352 deletions.
diff --git a/DESCRIPTION b/DESCRIPTION
@@ -44,11 +44,9 @@ Suggests:
  xgboost
 Remotes:
  catboost/catboost/catboost/R-package,
- mlr-org/mlr3@learner_size,
- mlr-org/mlr3extralearners@mlr3automl,
- mlr-org/mlr3learners@mlr3automl,
- mlr-org/mlr3mbo@adbo,
- mlr-org/mlr3pipelines
+ mlr-org/mlr3,
+ mlr-org/mlr3extralearners,
+ mlr-org/mlr3mbo@adbo
 Config/testthat/edition: 3
 Config/testthat/parallel: false
 Encoding: UTF-8
@@ -67,7 +65,17 @@ Collate:
  'LearnerClassifAutoSVM.R'
  'LearnerClassifAutoXgboost.R'
  'LearnerRegrAuto.R'
- 'helper.R'
+ 'LearnerRegrAutoCatboost.R'
+ 'LearnerRegrAutoGlmnet.R'
+ 'LearnerRegrAutoKKNN.R'
+ 'LearnerRegrAutoLightGBM.R'
+ 'LearnerRegrAutoNnet.R'
+ 'LearnerRegrAutoRanger.R'
+ 'LearnerRegrAutoSVM.R'
+ 'LearnerRegrAutoXgboost.R'
  'build_graph.R'
+ 'estimate_memory.R'
+ 'helper.R'
+ 'internal_measure.R'
  'train_auto.R'
  'zzz.R'
diff --git a/NAMESPACE b/NAMESPACE
@@ -1,9 +1,36 @@
 # Generated by roxygen2: do not edit by hand
 
+S3method(estimate_memory,Learner)
+S3method(estimate_memory,LearnerClassifCatboost)
+S3method(estimate_memory,LearnerClassifLightGBM)
+S3method(estimate_memory,LearnerClassifRanger)
+S3method(estimate_memory,LearnerClassifXgboost)
+S3method(estimate_memory,LearnerRegrCatboost)
+S3method(estimate_memory,LearnerRegrLightGBM)
+S3method(estimate_memory,LearnerRegrRanger)
+S3method(estimate_memory,LearnerRegrXgboost)
 export(LearnerClassifAuto)
+export(LearnerClassifAutoCatboost)
+export(LearnerClassifAutoGlmnet)
+export(LearnerClassifAutoKKNN)
+export(LearnerClassifAutoLightGBM)
+export(LearnerClassifAutoNnet)
+export(LearnerClassifAutoRanger)
 export(LearnerClassifAutoSVM)
 export(LearnerClassifAutoXgboost)
 export(LearnerRegrAuto)
+export(LearnerRegrAutoCatboost)
+export(LearnerRegrAutoGlmnet)
+export(LearnerRegrAutoKKNN)
+export(LearnerRegrAutoLightGBM)
+export(LearnerRegrAutoNnet)
+export(LearnerRegrAutoRanger)
+export(LearnerRegrAutoSVM)
+export(LearnerRegrAutoXgboost)
+export(catboost_internal_measure)
+export(estimate_memory)
+export(lightgbm_internal_measure)
+export(xgboost_internal_measure)
 import(R6)
 import(checkmate)
 import(data.table)

diff --git a/R/LearnerClassifAuto.R b/R/LearnerClassifAuto.R
@@ -4,6 +4,7 @@
 #' Classification auto learner.
 #'
 #' @template param_id
+#' @template param_learner_ids
 #'
 #' @export
 LearnerClassifAuto = R6Class("LearnerClassifAuto",
@@ -21,47 +22,52 @@ LearnerClassifAuto = R6Class("LearnerClassifAuto",
 
  #' @description
  #' Creates a new instance of this [R6][R6::R6Class] class.
- initialize = function(id = "classif.auto") {
+ initialize = function(
+ id = "classif.auto",
+ learner_ids = c("glmnet", "kknn", "lda", "nnet", "ranger", "svm", "xgboost", "catboost", "extra_trees", "lightgbm")
+ ) {
+ assert_subset(learner_ids, c("glmnet", "kknn", "lda", "nnet", "ranger", "svm", "xgboost", "catboost", "extra_trees", "lightgbm"))
+ if (all(learner_ids %in% c("lda", "extra_trees"))) {
+ stop("Learner 'lda' and 'extra_trees' must be combined with other learners")
+ }
+
+ private$.learner_ids = learner_ids
+ self$tuning_space = tuning_space[private$.learner_ids]
+
  param_set = ps(
- # learner
- learner_ids = p_uty(default = c("glmnet", "kknn", "lda", "nnet", "ranger", "svm", "xgboost", "catboost", "extra_trees", "lightgbm"),
- custom_check = function(x) {
- if (all(x %in% c("lda", "extra_trees"))) {
- return("Learner 'lda' and 'extra_trees' must be combined with other learners")
- }
- check_subset(x, c("glmnet", "kknn", "lda", "nnet", "ranger", "svm", "xgboost", "catboost", "extra_trees", "lightgbm"))
- }),
- learner_timeout = p_int(lower = 1L, default = 900L),
- xgboost_eval_metric = p_uty(),
- catboost_eval_metric = p_uty(),
- lightgbm_eval_metric = p_uty(),
+ learner_timeout = p_int(lower = 1L, default = 900L, tags = c("train", "super")),
+ # internal eval metric
+ xgboost_eval_metric = p_uty(tags = c("train", "xgboost")),
+ catboost_eval_metric = p_uty(tags = c("train", "catboost")),
+ lightgbm_eval_metric = p_uty(tags = c("train", "lightgbm")),
  # system
- max_nthread = p_int(lower = 1L, default = 1L),
- max_memory = p_int(lower = 1L, default = 32000L),
+ max_nthread = p_int(lower = 1L, default = 1L, tags = c("train", "catboost", "lightgbm", "ranger", "xgboost")),
+ max_memory = p_int(lower = 1L, default = 32000L, tags = c("train", "catboost", "lightgbm", "ranger", "xgboost")),
  # large data
- large_data_size = p_int(lower = 1L, default = 1e6),
- large_data_learner_ids = p_uty(),
- large_data_nthread = p_int(lower = 1L, default = 4L),
+ large_data_size = p_int(lower = 1L, default = 1e6, tags = c("train", "super")),
+ large_data_learner_ids = p_uty(tags = c("train", "super")),
+ large_data_nthread = p_int(lower = 1L, default = 4L, tags = c("train", "catboost", "lightgbm", "ranger", "xgboost")),
  # small data
- small_data_size = p_int(lower = 1L, default = 5000L),
- small_data_resampling = p_uty(),
- max_cardinality = p_int(lower = 1L, default = 100L),
- extra_trees_max_cardinality = p_int(lower = 1L, default = 40L),
+ small_data_size = p_int(lower = 1L, default = 5000L, tags = c("train", "super")),
+ small_data_resampling = p_uty(tags = c("train", "super")),
+ # cardinality
+ max_cardinality = p_int(lower = 1L, default = 100L, tags = c("train", "super")),
+ extra_trees_max_cardinality = p_int(lower = 1L, default = 40L, tags = c("train", "extra_trees")),
  # tuner
- resampling = p_uty(),
- terminator = p_uty(),
- measure = p_uty(),
- lhs_size = p_int(lower = 1L, default = 4L),
- callbacks = p_uty(),
- store_benchmark_result = p_lgl(default = FALSE))
+ resampling = p_uty(tags = c("train", "super")),
+ terminator = p_uty(tags = c("train", "super")),
+ measure = p_uty(tags = c("train", "super")),
+ lhs_size = p_int(lower = 1L, default = 4L, tags = c("train", "super")),
+ callbacks = p_uty(tags = c("train", "super")),
+ store_benchmark_result = p_lgl(default = FALSE, tags = c("train", "super")),
+ store_models = p_lgl(default = FALSE, tags = c("train", "super")))
 
  param_set$set_values(
- learner_ids = c("glmnet", "kknn", "lda", "nnet", "ranger", "svm", "xgboost", "catboost", "extra_trees", "lightgbm"),
  learner_timeout = 900L,
  max_nthread = 1L,
  max_memory = 32000L,
  large_data_size = 1e6L,
- large_data_learner_ids = c("lda", "ranger", "xgboost", "catboost", "extra_trees", "lightgbm"),
+ large_data_learner_ids = intersect(c("lda", "ranger", "xgboost", "catboost", "extra_trees", "lightgbm"), private$.learner_ids),
  large_data_nthread = 4L,
  small_data_size = 5000L,
  small_data_resampling = rsmp("cv", folds = 10L),
@@ -71,13 +77,19 @@ LearnerClassifAuto = R6Class("LearnerClassifAuto",
  terminator = trm("run_time", secs = 14400L),
  measure = msr("classif.ce"),
  lhs_size = 4L,
- store_benchmark_result = FALSE)
+ store_benchmark_result = FALSE,
+ store_models = FALSE)
+
+ # subset to relevant parameters for selected learners
+ param_set = param_set$subset(ids = unique(param_set$ids(any_tags = c("super", learner_ids))))
+
+ self$graph = build_graph(private$.learner_ids, "classif")
 
  super$initialize(
  id = id,
  task_type = "classif",
  param_set = param_set,
- packages = c("mlr3tuning", "mlr3learners", "mlr3pipelines", "mlr3mbo", "mlr3automl", "xgboost", "catboost", "lightgbm", "ranger", "nnet", "kknn", "glmnet", "MASS", "e1071"),
+ packages = c("mlr3tuning", "mlr3learners", "mlr3pipelines", "mlr3mbo", "mlr3automl", self$graph$packages),
  feature_types = c("logical", "integer", "numeric", "character", "factor"),
  predict_types = c("response", "prob"),
  properties = c("missings", "weights", "twoclass", "multiclass"),
@@ -86,8 +98,10 @@ LearnerClassifAuto = R6Class("LearnerClassifAuto",
  ),
 
  private = list(
+ .learner_ids = NULL,
+
  .train = function(task) {
- train_auto(self, task, task_type = "classif")
+ train_auto(self, private, task)
  },
 
  .predict = function(task) {

diff --git a/R/LearnerClassifAutoCatboost.R b/R/LearnerClassifAutoCatboost.R
@@ -13,26 +13,7 @@ LearnerClassifAutoCatboost = R6Class("LearnerClassifAutoCatboost",
  #' @description
  #' Creates a new instance of this [R6][R6::R6Class] class.
  initialize = function(id = "classif.auto_catboost") {
- super$initialize(id = id)
-
- # reduce parameter set to the relevant parameters
- private$.param_set = private$.param_set$subset(
- c("learner_ids",
- "learner_timeout",
- "catboost_eval_metric",
- "small_data_size",
- "small_data_resampling",
- "max_cardinality",
- "resampling",
- "terminator",
- "measure",
- "lhs_size",
- "callbacks",
- "store_benchmark_result")
- )
-
- self$param_set$set_values(learner_ids = "catboost")
- self$packages = c("mlr3tuning", "mlr3learners", "mlr3pipelines", "mlr3mbo", "mlr3automl", "catboost")
+ super$initialize(id = id, learner_ids = "catboost")
  }
  )
 )

diff --git a/R/LearnerClassifAutoGlmnet.R b/R/LearnerClassifAutoGlmnet.R
@@ -13,25 +13,7 @@ LearnerClassifAutoGlmnet = R6Class("LearnerClassifAutoGlmnet",
  #' @description
  #' Creates a new instance of this [R6][R6::R6Class] class.
  initialize = function(id = "classif.auto_glmnet") {
- super$initialize(id = id)
-
- # reduce parameter set to the relevant parameters
- private$.param_set = private$.param_set$subset(
- c("learner_ids",
- "learner_timeout",
- "small_data_size",
- "small_data_resampling",
- "max_cardinality",
- "resampling",
- "terminator",
- "measure",
- "lhs_size",
- "callbacks",
- "store_benchmark_result")
- )
-
- self$param_set$set_values(learner_ids = "glmnet")
- self$packages = c("mlr3tuning", "mlr3learners", "mlr3pipelines", "mlr3mbo", "mlr3automl", "glmnet")
+ super$initialize(id = id, learner_ids = "glmnet")
  }
  )
 )

diff --git a/R/LearnerClassifAutoKKNN.R b/R/LearnerClassifAutoKKNN.R
@@ -13,25 +13,7 @@ LearnerClassifAutoKKNN = R6Class("LearnerClassifAutoKKNN",
  #' @description
  #' Creates a new instance of this [R6][R6::R6Class] class.
  initialize = function(id = "classif.auto_kknn") {
- super$initialize(id = id)
-
- # reduce parameter set to the relevant parameters
- private$.param_set = private$.param_set$subset(
- c("learner_ids",
- "learner_timeout",
- "small_data_size",
- "small_data_resampling",
- "max_cardinality",
- "resampling",
- "terminator",
- "measure",
- "lhs_size",
- "callbacks",
- "store_benchmark_result")
- )
-
- self$param_set$set_values(learner_ids = "kknn")
- self$packages = c("mlr3tuning", "mlr3learners", "mlr3pipelines", "mlr3mbo", "mlr3automl", "kknn")
+ super$initialize(id = id, learner_ids = "kknn")
  }
  )
 )

diff --git a/R/LearnerClassifAutoLightGBM.R b/R/LearnerClassifAutoLightGBM.R
@@ -13,26 +13,7 @@ LearnerClassifAutoLightGBM = R6Class("LearnerClassifAutoLightGBM",
  #' @description
  #' Creates a new instance of this [R6][R6::R6Class] class.
  initialize = function(id = "classif.auto_lightgbm") {
- super$initialize(id = id)
-
- # reduce parameter set to the relevant parameters
- private$.param_set = private$.param_set$subset(
- c("learner_ids",
- "learner_timeout",
- "lightgbm_eval_metric",
- "small_data_size",
- "small_data_resampling",
- "max_cardinality",
- "resampling",
- "terminator",
- "measure",
- "lhs_size",
- "callbacks",
- "store_benchmark_result")
- )
-
- self$param_set$set_values(learner_ids = "lightgbm")
- self$packages = c("mlr3tuning", "mlr3learners", "mlr3pipelines", "mlr3mbo", "mlr3automl", "lightgbm")
+ super$initialize(id = id, learner_ids = "lightgbm")
  }
  )
 )

diff --git a/R/LearnerClassifAutoNnet.R b/R/LearnerClassifAutoNnet.R
@@ -13,25 +13,7 @@ LearnerClassifAutoNnet = R6Class("LearnerClassifAutoNnet",
  #' @description
  #' Creates a new instance of this [R6][R6::R6Class] class.
  initialize = function(id = "classif.auto_nnet") {
- super$initialize(id = id)
-
- # reduce parameter set to the relevant parameters
- private$.param_set = private$.param_set$subset(
- c("learner_ids",
- "learner_timeout",
- "small_data_size",
- "small_data_resampling",
- "max_cardinality",
- "resampling",
- "terminator",
- "measure",
- "lhs_size",
- "callbacks",
- "store_benchmark_result")
- )
-
- self$param_set$set_values(learner_ids = "nnet")
- self$packages = c("mlr3tuning", "mlr3learners", "mlr3pipelines", "mlr3mbo", "mlr3automl", "nnet")
+ super$initialize(id = id, learner_ids = "nnet")
  }
  )
 )

diff --git a/R/LearnerClassifAutoRanger.R b/R/LearnerClassifAutoRanger.R
@@ -13,25 +13,7 @@ LearnerClassifAutoRanger = R6Class("LearnerClassifAutoRanger",
  #' @description
  #' Creates a new instance of this [R6][R6::R6Class] class.
  initialize = function(id = "classif.auto_ranger") {
- super$initialize(id = id)
-
- # reduce parameter set to the relevant parameters
- private$.param_set = private$.param_set$subset(
- c("learner_ids",
- "learner_timeout",
- "small_data_size",
- "small_data_resampling",
- "max_cardinality",
- "resampling",
- "terminator",
- "measure",
- "lhs_size",
- "callbacks",
- "store_benchmark_result")
- )
-
- self$param_set$set_values(learner_ids = "ranger")
- self$packages = c("mlr3tuning", "mlr3learners", "mlr3pipelines", "mlr3mbo", "mlr3automl", "ranger")
+ super$initialize(id = id, learner_ids = "ranger")
  }
  )
 )

diff --git a/R/LearnerClassifAutoSVM.R b/R/LearnerClassifAutoSVM.R
@@ -13,25 +13,7 @@ LearnerClassifAutoSVM = R6Class("LearnerClassifAutoSVM",
  #' @description
  #' Creates a new instance of this [R6][R6::R6Class] class.
  initialize = function(id = "classif.auto_svm") {
- super$initialize(id = id)
-
- # reduce parameter set to the relevant parameters
- private$.param_set = private$.param_set$subset(
- c("learner_ids",
- "learner_timeout",
- "small_data_size",
- "small_data_resampling",
- "max_cardinality",
- "resampling",
- "terminator",
- "measure",
- "lhs_size",
- "callbacks",
- "store_benchmark_result")
- )
-
- self$param_set$set_values(learner_ids = "svm")
- self$packages = c("mlr3tuning", "mlr3learners", "mlr3pipelines", "mlr3mbo", "mlr3automl", "e1071")
+ super$initialize(id = id, learner_ids = "svm")
  }
  )
 )