Merge pull request #102 from mlr-org/fix_catboost

Fix catboost, closes #100

.github/workflows/rcmdcheck.yml

@@ -62,10 +62,10 @@ jobs:
           pak::pkg_install("rcmdcheck")
         shell: Rscript {0}
 
-      - name: Install CatBoost
+      - name: Install catboost
         run: |
           install.packages("remotes")
-          remotes::install_url('https://github.com/catboost/catboost/releases/download/v0.24.1/catboost-R-Linux-0.24.1.tgz', INSTALL_opts = c("--no-multiarch"))
+          remotes::install_url('https://github.com/catboost/catboost/releases/download/v0.26.1/catboost-R-Linux-0.26.1.tgz', INSTALL_opts = c("--no-multiarch"))
         shell: Rscript {0}
 
       - name: Install Python

.gitignore

@@ -125,3 +125,4 @@ docs/
 
 .ccache/
 .vscode/settings.json
+catboost_info/

DESCRIPTION

@@ -1,6 +1,6 @@
 Package: mlr3extralearners
 Title: Extra Learners For mlr3
-Version: 0.5.2
+Version: 0.5.3
 Authors@R:
     c(person(given = "Raphael",
              family = "Sonabend",

NAMESPACE

@@ -80,6 +80,7 @@ export(LearnerSurvPenalized)
 export(LearnerSurvRandomForestSRC)
 export(LearnerSurvSVM)
 export(create_learner)
+export(install_catboost)
 export(install_learners)
 export(list_mlr3learners)
 export(lrn)

NEWS.md

@@ -1,3 +1,9 @@
+# mlr3extralearners 0.5.3
+
+* Fixed bugs in catboost for classification
+* Removed factor feature types from catboost
+* Added `install_catboost` to make installation from catboost simpler
+
 # mlr3extralearners 0.5.2
 
 * Fixed learner tests

R/fn_install_catboost.R

@@ -0,0 +1,50 @@
+#' Helper function to install catboost
+#' @title Install catboost
+#' @param version `(character(1))`
+#' Version to install, if `NULL` installs latest
+#' @param os `(character(1))` \cr
+#' Operating system to install on, if `NULL` automatically detected
+#' @param install_required `(logical(1))` \cr
+#' If `TRUE` (default) then installs required packages: {curl}, {jsonlite},
+#' {devtools}
+#' @param INSTALL_opts `(character())` \cr
+#' Passed to [devtools::install_url]
+#' @param ... `ANY` \cr
+#' Other arguments passed to [devtools::install_url]
+#' @export
+install_catboost <- function(version = NULL, os = NULL,
+                             install_required = TRUE,
+                             INSTALL_opts = c("--no-multiarch",
+                                              "--no-test-load"), ...) {
+
+  if (is.null(version)) {
+
+    if (!requireNamespace("jsonlite", quietly = TRUE) && install_required) {
+      utils::install.packages("jsonlite", repos = "https://cloud.r-project.org")
+    }
+
+    if (!requireNamespace("curl", quietly = TRUE) && install_required) {
+      utils::install.packages("curl", repos = "https://cloud.r-project.org")
+    }
+
+    version <-  jsonlite::fromJSON(
+        "https://api.github.com/repos/catboost/catboost/releases"
+    )$tag_name[1]
+  }
+
+  version <- gsub("v", "", version)
+
+  if (is.null(os)) {
+    os <- as.character(Sys.info()["sysname"])
+  }
+
+  url <- sprintf(
+    "https://github.com/catboost/catboost/releases/download/v%s/catboost-R-%s-%s.tgz",
+    version, os, version)
+
+  if (!requireNamespace("devtools", quietly = TRUE) && install_required) {
+    utils::install.packages("devtools", repos = "https://cloud.r-project.org")
+  }
+
+  devtools::install_url(url, INSTALL_opts = INSTALL_opts, ...)
+}

R/learner_catboost_classif_catboost.R

@@ -6,6 +6,10 @@
 #' @templateVar id classif.catboost
 #' @templateVar caller catboost.train
 #'
+#' @section Installation:
+#' The easiest way to install catboost is with the helper function
+#' [install_catboost].
+#'
 #' @section Custom mlr3 defaults:
 #' - `logging_level`:
 #'   - Actual default: "Verbose"
@@ -25,11 +29,11 @@
 #'   - Reason for change: consistent with other mlr3 learners
 #'
 #' @references
-#' CatBoost: unbiased boosting with categorical features.
+#' catboost: unbiased boosting with categorical features.
 #' Liudmila Prokhorenkova, Gleb Guse, Aleksandr Vorobev, Anna Veronika Dorogush and Andrey Gulin.
 #' 2017. https://arxiv.org/abs/1706.09516.
 #'
-#' CatBoost: gradient boosting with categorical features support.
+#' catboost: gradient boosting with categorical features support.
 #' Anna Veronika Dorogush, Vasily Ershov and Andrey Gulin.
 #' 2018. https://arxiv.org/abs/1810.11363.
 #'
@@ -155,7 +159,7 @@ LearnerClassifCatboost = R6Class("LearnerClassifCatboost",
       super$initialize(
         id = "classif.catboost",
         packages = "catboost",
-        feature_types = c("logical", "integer", "numeric", "factor", "ordered"),
+        feature_types = c("numeric", "factor", "ordered"),
         predict_types = c("response", "prob"),
         param_set = ps,
         properties = c(
@@ -182,18 +186,14 @@ LearnerClassifCatboost = R6Class("LearnerClassifCatboost",
 
   private = list(
     .train = function(task) {
-      # integer/logical features must be converted to numerics explicitly
 
-      data = task$data(cols = task$feature_names)
-      to_numerics = task$feature_types$id[task$feature_types$type %in%
-        c("integer", "logical")]
-      if (length(to_numerics)) {
-        data[, (to_numerics) := lapply(.SD, as.numeric), .SDcols = to_numerics]
+      if (packageVersion('catboost') < '0.21') {
+        stop('catboost v0.21 or greater is required, update with install_catboost')
       }
 
       # target is encoded as integer values from 0
       # if binary, the positive class is 1
-      is_binary = (length(task$class_names) == 2L)
+      is_binary = length(task$class_names) == 2L
       label = if (is_binary) {
         ifelse(task$data(cols = task$target_names)[[1L]] == task$positive,
           yes = 1L,
@@ -204,7 +204,7 @@ LearnerClassifCatboost = R6Class("LearnerClassifCatboost",
 
       # data must be a dataframe
       learn_pool = mlr3misc::invoke(catboost::catboost.load_pool,
-        data = data.table::setDF(data),
+        data = task$data(cols = task$feature_names),
         label = label,
         weight = task$weights$weight,
         thread_count = self$param_set$values$thread_count)
@@ -219,29 +219,16 @@ LearnerClassifCatboost = R6Class("LearnerClassifCatboost",
       pars$loss_function_twoclass = NULL
       pars$loss_function_multiclass = NULL
 
-      mlr3misc::invoke(catboost::catboost.train,
-        learn_pool = learn_pool,
-        test_pool = NULL,
-        params = pars)
+      catboost::catboost.train(learn_pool, NULL, pars)
     },
 
     .predict = function(task) {
-      # integer/logical features must be converted to numerics explicitly
-
-      data = task$data(cols = task$feature_names)
-      to_numerics = task$feature_types$id[task$feature_types$type %in%
-        c("integer", "logical")]
-      if (length(to_numerics)) {
-        data[, (to_numerics) := lapply(.SD, as.numeric), .SDcols = to_numerics]
-      }
 
-      # target was encoded as integer values based on the train_task
-      # to later revert this, again use the train_task
-      is_binary = (length(self$state$train_task$class_names) == 2L)
+      is_binary = (length(task$class_names) == 2L)
 
       # data must be a dataframe
       pool = mlr3misc::invoke(catboost::catboost.load_pool,
-        data = data.table::setDF(data),
+        data = task$data(cols = task$feature_names),
         thread_count = self$param_set$values$thread_count)
 
       prediction_type = if (self$predict_type == "response") {
@@ -257,20 +244,20 @@ LearnerClassifCatboost = R6Class("LearnerClassifCatboost",
 
       if (self$predict_type == "response") {
         response = if (is_binary) {
-          ifelse(preds == 1L,
-            yes = self$state$train_task$positive,
-            no = setdiff(
-              self$state$train_task$class_names,
-              self$state$train_task$positive))
+          ifelse(preds == 1L, yes = task$positive, no = task$negative)
         } else {
-          self$state$train_task$class_names[preds + 1L]
+          task$class_names[preds + 1L]
         }
-        list(response = response)
+        list(response = as.character(unname(response)))
       } else {
+
         if (is_binary && is.null(dim(preds))) {
           preds = matrix(c(preds, 1 - preds), ncol = 2L, nrow = length(preds))
+          colnames(preds) = c(task$positive, task$negative)
+        } else {
+          colnames(preds) = self$state$train_task$class_names
         }
-        colnames(preds) = self$state$train_task$class_names
+
         list(prob = preds)
       }
     }

R/learner_catboost_regr_catboost.R

@@ -6,6 +6,10 @@
 #' @templateVar id regr.catboost
 #' @templateVar caller catboost.train
 #'
+#' @section Installation:
+#' The easiest way to install catboost is with the helper function
+#' [install_catboost].
+#'
 #' @section Custom mlr3 defaults:
 #' - `logging_level`:
 #'   - Actual default: "Verbose"
@@ -25,11 +29,11 @@
 #'   - Reason for change: consistent with other mlr3 learners
 #'
 #' @references
-#' CatBoost: unbiased boosting with categorical features.
+#' catboost: unbiased boosting with categorical features.
 #' Liudmila Prokhorenkova, Gleb Guse, Aleksandr Vorobev, Anna Veronika Dorogush and Andrey Gulin.
 #' 2017. https://arxiv.org/abs/1706.09516.
 #'
-#' CatBoost: gradient boosting with categorical features support.
+#' catboost: gradient boosting with categorical features support.
 #' Anna Veronika Dorogush, Vasily Ershov and Andrey Gulin.
 #' 2018. https://arxiv.org/abs/1810.11363.
 #'
@@ -149,7 +153,7 @@ LearnerRegrCatboost = R6Class("LearnerRegrCatboost",
       super$initialize(
         id = "regr.catboost",
         packages = "catboost",
-        feature_types = c("logical", "integer", "numeric", "factor", "ordered"),
+        feature_types = c("numeric", "factor", "ordered"),
         predict_types = "response",
         param_set = ps,
         properties = c(
@@ -176,17 +180,16 @@ LearnerRegrCatboost = R6Class("LearnerRegrCatboost",
 
   private = list(
     .train = function(task) {
-      # integer/logical features must be converted to numerics explicitly
-      data = task$data(cols = task$feature_names)
-      to_numerics = task$feature_types$id[task$feature_types$type %in%
-        c("integer", "logical")]
-      if (length(to_numerics)) {
-        data[, (to_numerics) := lapply(.SD, as.numeric), .SDcols = to_numerics]
+
+      if (packageVersion('catboost') < '0.21') {
+        stop('catboost v0.21 or greater is required, update with install_catboost')
       }
 
+      self$state$feature_names = task$feature_names
+
       # data must be a dataframe
       learn_pool = mlr3misc::invoke(catboost::catboost.load_pool,
-        data = data.table::setDF(data),
+        data = task$data(cols = task$feature_names),
         label = task$data(cols = task$target_names)[[1L]],
         weight = task$weights$weight,
         thread_count = self$param_set$values$thread_count)
@@ -198,18 +201,9 @@ LearnerRegrCatboost = R6Class("LearnerRegrCatboost",
     },
 
     .predict = function(task) {
-      # integer/logical features must be converted to numerics explicitly
 
-      data = task$data(cols = task$feature_names)
-      to_numerics = task$feature_types$id[task$feature_types$type %in%
-        c("integer", "logical")]
-      if (length(to_numerics)) {
-        data[, (to_numerics) := lapply(.SD, as.numeric), .SDcols = to_numerics]
-      }
-
-      # data must be a dataframe
       pool = mlr3misc::invoke(catboost::catboost.load_pool,
-        data = data.table::setDF(data),
+        data = task$data(cols = self$state$feature_names),
         thread_count = self$param_set$values$thread_count)
 
       preds = mlr3misc::invoke(catboost::catboost.predict,

inst/paramtest/test_paramtest_catboost_classif_catboost.R

@@ -1,6 +1,7 @@
 library(mlr3extralearners)
-remotes::install_url('https://github.com/catboost/catboost/releases/download/v0.24.1/catboost-R-Linux-0.24.1.tgz', # nolint
-                      INSTALL_opts = c("--no-multiarch"))
+if (!requireNamespace("catboost", quietly = TRUE)) {
+  install_catboost("0.26.1")
+}
 
 test_that("classif.catboost_catboost.train", {
   learner = lrn("classif.catboost")

inst/paramtest/test_paramtest_catboost_regr_catboost.R

@@ -1,6 +1,7 @@
 library(mlr3extralearners)
-remotes::install_url('https://github.com/catboost/catboost/releases/download/v0.24.1/catboost-R-Linux-0.24.1.tgz', # nolint
-                      INSTALL_opts = c("--no-multiarch"))
+if (!requireNamespace("catboost", quietly = TRUE)) {
+  install_catboost("0.26.1")
+}
 
 
 test_that("regr.catboost_catboost.train", {