Merge pull request #2792 from mlr-org/cran-2.18.0

CRAN release v2.18.0

Author: pat-s

Diff for: .Rbuildignore

@@ -44,3 +44,4 @@ revdep/*
 ^\.github$
 ^clang-.*
 ^gfortran.*
+^cran-comments\.md$

Diff for: NEWS.md

@@ -5,41 +5,16 @@
 
 # mlr 2.18.0
 
-- Internal changes only.
-
-
-# mlr 2.17.1.9006
-
 - Many praznik filters are now also able to deal with regression tasks (#2790, @bommert)
 - `praznik_MRMR`: Remove handling of survival tasks (#2790, @bommert)
-
-
-# mlr 2.17.1.9005
-
 - xgboost: update `objective` default from `reg:linear` (deprecated) to `reg:squarederror`
 - issue a warning if `blocking` was set in the Task but `blocking.cv` was not set within `makeResampleDesc() (#2788)
-
-# mlr 2.17.1.9003
-
 - Fix order of learners in `generateLearningCurveData()` (#2768)
-
-
-# mlr 2.17.1.9002
-
  - `getFeatureImportance()`: Account for feature importance weight of linear xgboost models
-
-
-# mlr 2.17.1.9001
-
 - Fix learner note for learner glmnet (the default of param `s` did not match the learner note) (#2747)
-- Remove dep {hrbrthemes} used in `createSpatialResamplingPlots()`. The package caused issues on R-devel. In addition users should set custom themes by themselves.
+- Remove dependency {hrbrthemes} used in `createSpatialResamplingPlots()`. The package caused issues on R-devel. In addition users should set custom themes by themselves.
 - Explicitly return value in `getNestedTuneResultsOptPathDf()` (#2754)
 
-
-# mlr 2.17.1.9000
-
-- Internal changes only
-
 # mlr 2.17.1
 
 ## Learners - bugfixes
@@ -235,7 +210,7 @@ PR: #2638 (@pfistl)
 ## filters - new
 
 - Ensemble features are now supported. These filters combine multiple single filters to create a final ranking based on certain statistical operations. All new filters are listed in a dedicated section "ensemble filters" in the [tutorial](https://mlr.mlr-org.com/articles/tutorial/filter_methods.html).
-  Tuning of simple features is not supported yet because of a [missing feature](https://github.com/berndbischl/ParamHelpers/pull/206) in _ParamHelpers_. (@pat-s, #2456)
+  Tuning of simple features is not supported yet because of a [missing feature](https://github.com/mlr-org/ParamHelpers/pull/206) in _ParamHelpers_. (@pat-s, #2456)
 
 # mlr 2.14.0
 

Diff for: R/CostSensWeightedPairsWrapper.R

@@ -23,21 +23,23 @@
 #' One-versus-one Binary Classification.
 #' In: Proceedings of the Sixth Asian Conference on Machine Learning.
 #' JMLR Workshop and Conference Proceedings, vol 39, pp. 371-386. JMLR W&CP (2014).
-#' <http://www.jmlr.org/proceedings/papers/v39/lin14.pdf>
+#' <https://www.jmlr.org/proceedings/papers/v39/lin14.pdf>
 #' @family costsens
 #' @aliases CostSensWeightedPairsWrapper CostSensWeightedPairsModel
 makeCostSensWeightedPairsWrapper = function(learner) {
   learner = checkLearner(learner, "classif", props = "weights")
   learner = setPredictType(learner, "response")
   id = stri_paste("costsens", learner$id, sep = ".")
-  makeHomogeneousEnsemble(id, "costsens", learner, package = learner$package,
+  makeHomogeneousEnsemble(id, "costsens", learner,
+    package = learner$package,
     learner.subclass = "CostSensWeightedPairsWrapper", model.subclass = "CostSensWeightedPairsModel")
 }
 
 #' @export
 trainLearner.CostSensWeightedPairsWrapper = function(.learner, .task, .subset = NULL, ...) {
 
   # note that no hyperpars can be in ..., they would refer to the wrapper
+
   .task = subsetTask(.task, subset = .subset)
   costs = getTaskCosts(.task)
   td = getTaskDesc(.task)
@@ -57,7 +59,8 @@ trainLearner.CostSensWeightedPairsWrapper = function(.learner, .task, .subset =
         models[[counter]] = y[1]
       } else {
         feats$..y.. = y
-        task = makeClassifTask(data = feats, target = "..y..",
+        task = makeClassifTask(
+          data = feats, target = "..y..",
           check.data = FALSE, fixup.data = "quiet")
         w = abs(costs[, a1] - costs[, a2])
         models[[counter]] = train(.learner$next.learner, task, weights = w)

Diff for: R/MulticlassWrapper.R

@@ -11,7 +11,7 @@
 #' are generated is defined by an error-correcting-output-code (ECOC) code book.
 #' This also allows the simple and well-known one-vs-one and one-vs-rest
 #' approaches. Decoding is currently done via Hamming decoding, see
-#' e.g. here <http://jmlr.org/papers/volume11/escalera10a/escalera10a.pdf>.
+#' e.g. here <https://jmlr.org/papers/volume11/escalera10a/escalera10a.pdf>.
 #'
 #' Currently, the approach always operates on the discrete predicted labels
 #' of the binary base models (instead of their probabilities) and the created
@@ -41,7 +41,8 @@ makeMulticlassWrapper = function(learner, mcw.method = "onevsrest") {
   pv = list(mcw.method = mcw.method)
   id = stri_paste(learner$id, "multiclass", sep = ".")
 
-  x = makeHomogeneousEnsemble(id = id, type = "classif", next.learner = learner,
+  x = makeHomogeneousEnsemble(
+    id = id, type = "classif", next.learner = learner,
     package = learner$package, par.set = ps, par.vals = pv,
     learner.subclass = "MulticlassWrapper", model.subclass = "MulticlassModel")
   x = setPredictType(x, predict.type = "response")
@@ -58,7 +59,8 @@ trainLearner.MulticlassWrapper = function(.learner, .task, .subset = NULL, .weig
   args = list(x = x, learner = .learner, task = .task, weights = .weights)
   parallelLibrary("mlr", master = FALSE, level = "mlr.ensemble", show.info = FALSE)
   exportMlrOptions(level = "mlr.ensemble")
-  models = parallelMap(i = seq_along(x$row.inds), doMulticlassTrainIteration,
+  models = parallelMap(
+    i = seq_along(x$row.inds), doMulticlassTrainIteration,
     more.args = args, level = "mlr.ensemble")
   m = makeHomChainModel(.learner, models)
   m$cm = cm