mlr-org · awinterstetter · Jan 15, 2026 · Jan 15, 2026 · Jan 15, 2026 · Jan 15, 2026
diff --git a/NEWS.md b/NEWS.md
@@ -9,6 +9,12 @@
   - `LearnerRegrPcr`
   - `LearnerRegrPlsr`
   - `LearnerRegrLaGP`
+  - `LearnerRegrFrbs`
+  - `LearnerRegrBgp`
+  - `LearnerRegrBgpllm`
+  - `LearnerRegrBtgp`
+  - `LearnerRegrBtlm`
+
 
 ## Breaking Changes
 

diff --git a/R/learner_RRF_regr_RRF.R b/R/learner_RRF_regr_RRF.R
@@ -94,7 +94,7 @@ LearnerRegrRRF = R6Class("LearnerRegrRRF",
     .train = function(task) {
       pars = self$param_set$get_values(tags = "train")
 
-      mlr3misc::invoke(RRF::RRF,
+      invoke(RRF::RRF,
         formula = task$formula(),
         data = task$data(),
         .args = pars)
@@ -104,7 +104,7 @@ LearnerRegrRRF = R6Class("LearnerRegrRRF",
       pars = self$param_set$get_values(tags = "predict")
       newdata = ordered_features(task, self)
 
-      pred = mlr3misc::invoke(predict,
+      pred = invoke(predict,
         object = self$model,
         newdata = newdata,
         type = self$predict_type,

diff --git a/R/learner_frbs_regr_frbs.R b/R/learner_frbs_regr_frbs.R
@@ -82,7 +82,7 @@ LearnerRegrFrbs = R6Class("LearnerRegrFrbs",
 
     .predict = function(task) {
       newdata = as.matrix(ordered_features(task, self), drop = FALSE)
-      pred = mlr3misc::invoke(predict, object = self$model, newdata = newdata)
+      pred = invoke(predict, object = self$model, newdata = newdata)
 
       if (is.matrix(pred) || is.data.frame(pred)) {
         pred = pred[, 1L]

diff --git a/R/learner_pls_regr_plsr.R b/R/learner_pls_regr_plsr.R
@@ -56,7 +56,7 @@ LearnerRegrPlsr = R6Class("LearnerRegrPlsr",
         quote = c("left", "right"),
         env = environment())
 
-      mlr3misc::invoke(pls::plsr,
+      invoke(pls::plsr,
         formula = formula,
         data = task$data(),
         .args = pars)
@@ -66,7 +66,7 @@ LearnerRegrPlsr = R6Class("LearnerRegrPlsr",
       pars = self$param_set$get_values(tags = "predict")
       newdata = ordered_features(task, self)
 
-      pred = mlr3misc::invoke(predict, self$model,
+      pred = invoke(predict, self$model,
         newdata = newdata,
         type = "response",
         # specifying comps forces that pred is a matrix

diff --git a/R/learner_tgp_regr_bgp.R b/R/learner_tgp_regr_bgp.R
@@ -9,6 +9,7 @@
 #'
 #' @section Initial parameter values:
 #' * `verb` is initialized to `0` to silence printing.
+#' * `pred.n` is initialized to `FALSE` to skip prediction during training.
 #'
 #' @templateVar id regr.bgp
 #' @template learner
@@ -26,27 +27,27 @@ LearnerRegrBgp = R6Class("LearnerRegrBgp",
     #' Creates a new instance of this [R6][R6::R6Class] class.
     initialize = function() {
       param_set = ps(
+        meanfn = p_fct(default = "linear", levels = c("constant", "linear"), tags = "train"),
         bprior = p_fct(default = "bflat", levels = c("b0", "b0not", "bflat", "bmle", "bmznot", "bmzt"), tags = "train"),
-        BTE = p_uty(default = c(1000L, 4000L, 2L), tags = c("train", "predict"), custom_check = mlr3misc::crate({function(x) {
+        corr   = p_fct(default = "expsep", levels = c("exp", "expsep", "matern", "sim"), tags = "train"),
+        BTE    = p_uty(default = c(1000L, 4000L, 2L), tags = c("train", "predict"), custom_check = mlr3misc::crate({function(x) {
           if (!checkmate::test_integerish(x, len = 3, lower = 0)) {
             return("`BTE` must be an integerish vector of length 3 with non-negative entries")
           }
           TRUE
         }})),
-        corr   = p_fct(default = "expsep", levels = c("exp", "expsep", "matern", "sim"), tags = "train"),
-        Ds2x   = p_lgl(default = FALSE, tags = c("train", "predict")),
-        improv = p_lgl(default = FALSE, tags = c("train", "predict")),
+        R      = p_int(default = 1L, lower = 1L, tags = c("train", "predict")),
+        m0r1   = p_lgl(default = TRUE, tags = "train"),
         itemps = p_uty(default = NULL, tags = "train"),
+        pred.n = p_lgl(init = FALSE, tags = c("train", "predict")),
         krige  = p_lgl(default = TRUE, tags = c("train", "predict")),
-        m0r1   = p_lgl(default = TRUE, tags = "train"),
-        MAP    = p_lgl(default = TRUE, tags = "predict"),
-        meanfn = p_fct(default = "linear", levels = c("constant", "linear"), tags = "train"),
+        zcov   = p_lgl(default = FALSE, tags = c("train", "predict")),
+        Ds2x   = p_lgl(default = FALSE, tags = c("train", "predict")),
+        improv = p_lgl(default = FALSE, tags = c("train", "predict")),
         nu     = p_dbl(default = 1.5, tags = "train", depends = quote(corr == "matern")),
-        R      = p_int(default = 1L, lower = 1L, tags = c("train", "predict")),
-        sens.p = p_uty(default = NULL, tags = c("train", "predict")),
         trace  = p_lgl(default = FALSE, tags = c("train", "predict")),
         verb   = p_int(init= 0L, lower = 0L, upper = 4L, tags = c("train", "predict")),
-        zcov   = p_lgl(default = FALSE, tags = c("train", "predict"))
+        MAP    = p_lgl(default = TRUE, tags = "predict")
       )
 
       super$initialize(
@@ -65,7 +66,7 @@ LearnerRegrBgp = R6Class("LearnerRegrBgp",
       pars = self$param_set$get_values(tags = "train")
       data = as_numeric_matrix(task$data(cols = task$feature_names))
 
-      mlr3misc::invoke(tgp::bgp,
+      invoke(tgp::bgp,
         X = data,
         Z = task$truth(),
         .args = pars
@@ -76,7 +77,11 @@ LearnerRegrBgp = R6Class("LearnerRegrBgp",
       pars = self$param_set$get_values(tags = "predict")
       newdata = as_numeric_matrix(ordered_features(task, self))
 
-      pred = mlr3misc::invoke(predict,
+      if (self$predict_type == "se") {
+        pars$krige = TRUE
+      }
+
+      pred = invoke(predict,
         object = self$model,
         XX = newdata,
         .args = pars
@@ -85,9 +90,6 @@ LearnerRegrBgp = R6Class("LearnerRegrBgp",
       if (self$predict_type == "response") {
         list(response = pred$ZZ.km)
       } else {
-        if (is.null(pred$ZZ.ks2)) {
-          stop("Standard errors requested but `ZZ.ks2` was not returned; try setting `krige = TRUE`.")
-        }
         list(response = pred$ZZ.km, se = sqrt(pred$ZZ.ks2))
       }
     }

diff --git a/R/learner_tgp_regr_bgpllm.R b/R/learner_tgp_regr_bgpllm.R
@@ -9,6 +9,7 @@
 #'
 #' @section Initial parameter values:
 #' * `verb` is initialized to `0` to silence printing.
+#' * `pred.n` is initialized to `FALSE` to skip prediction during training.
 #'
 #' @templateVar id regr.bgpllm
 #' @template learner
@@ -26,16 +27,10 @@ LearnerRegrBgpllm = R6Class("LearnerRegrBgpllm",
     #' Creates a new instance of this [R6][R6::R6Class] class.
     initialize = function() {
       param_set = ps(
+        meanfn = p_fct(default = "linear", levels = c("constant", "linear"), tags = "train"),
         bprior = p_fct(default = "bflat", levels = c("b0", "b0not", "bflat", "bmle", "bmznot", "bmzt"), tags = "train"),
-        BTE = p_uty(default = c(2000L, 4000L, 2L), tags = c("train", "predict"), custom_check = mlr3misc::crate({function(x) {
-          if (!checkmate::test_integerish(x, len = 3, lower = 0)) {
-            return("`BTE` must be an integerish vector of length 3 with non-negative entries")
-          }
-          TRUE
-        }})),
         corr   = p_fct(default = "expsep", levels = c("exp", "expsep", "matern", "sim"), tags = "train"),
-        Ds2x   = p_lgl(default = FALSE, tags = c("train", "predict")),
-        gamma = p_uty(default = c(10, 0.2, 0.7), tags = "train", custom_check = mlr3misc::crate({function(x) {
+        gamma  = p_uty(default = c(10, 0.2, 0.7), tags = "train", custom_check = mlr3misc::crate({function(x) {
           if (!checkmate::test_numeric(x, len = 3, lower = 0)) {
             return("`gamma` must be a numeric vector of length 3 with non-negative entries")
           }
@@ -44,18 +39,24 @@ LearnerRegrBgpllm = R6Class("LearnerRegrBgpllm",
           }
           TRUE
         }})),
-        improv = p_lgl(default = FALSE, tags = c("train", "predict")),
+        BTE    = p_uty(default = c(2000L, 4000L, 2L), tags = c("train", "predict"), custom_check = mlr3misc::crate({function(x) {
+          if (!checkmate::test_integerish(x, len = 3, lower = 0)) {
+            return("`BTE` must be an integerish vector of length 3 with non-negative entries")
+          }
+          TRUE
+        }})),
+        R      = p_int(default = 1L, lower = 1L, tags = c("train", "predict")),
+        m0r1   = p_lgl(default = TRUE, tags = "train"),
         itemps = p_uty(default = NULL, tags = "train"),
+        pred.n = p_lgl(init = FALSE, tags = c("train", "predict")),
         krige  = p_lgl(default = TRUE, tags = c("train", "predict")),
-        m0r1   = p_lgl(default = TRUE, tags = "train"),
-        MAP    = p_lgl(default = TRUE, tags = "predict"),
-        meanfn = p_fct(default = "linear", levels = c("constant", "linear"), tags = "train"),
+        zcov   = p_lgl(default = FALSE, tags = c("train", "predict")),
+        Ds2x   = p_lgl(default = FALSE, tags = c("train", "predict")),
+        improv = p_lgl(default = FALSE, tags = c("train", "predict")),
         nu     = p_dbl(default = 1.5, tags = "train", depends = quote(corr == "matern")),
-        R      = p_int(default = 1L, lower = 1L, tags = c("train", "predict")),
-        sens.p = p_uty(default = NULL, tags = c("train", "predict")),
         trace  = p_lgl(default = FALSE, tags = c("train", "predict")),
         verb   = p_int(init = 0L, lower = 0L, upper = 4L, tags = c("train", "predict")),
-        zcov   = p_lgl(default = FALSE, tags = c("train", "predict"))
+        MAP    = p_lgl(default = TRUE, tags = "predict")
       )
 
       super$initialize(
@@ -76,7 +77,7 @@ LearnerRegrBgpllm = R6Class("LearnerRegrBgpllm",
       x = as_numeric_matrix(task$data(cols = task$feature_names))
       y = task$truth()
 
-      mlr3misc::invoke(
+      invoke(
         tgp::bgpllm,
         X = x,
         Z = y,
@@ -88,7 +89,11 @@ LearnerRegrBgpllm = R6Class("LearnerRegrBgpllm",
 
       newdata = as_numeric_matrix(ordered_features(task, self))
 
-      pred = mlr3misc::invoke(
+      if (self$predict_type == "se") {
+        pars$krige = TRUE
+      }
+
+      pred = invoke(
         predict,
         self$model,
         XX = newdata,
@@ -98,9 +103,6 @@ LearnerRegrBgpllm = R6Class("LearnerRegrBgpllm",
       if (self$predict_type == "response") {
         list(response = pred$ZZ.km)
       } else {
-        if (is.null(pred$ZZ.ks2)) {
-          stop("Standard errors requested but `ZZ.ks2` was not returned; try setting `krige = TRUE`.")
-        }
         list(response = pred$ZZ.km, se = sqrt(pred$ZZ.ks2))
       }
     }

diff --git a/R/learner_tgp_regr_btgp.R b/R/learner_tgp_regr_btgp.R
@@ -6,9 +6,14 @@
 #' Bayesian treed Gaussian process regression model.
 #' Calls [tgp::btgp()] from \CRANpkg{tgp}.
 #' For the predicted mean `ZZ.km` and for the predicted variance `ZZ.ks2` are chosen.
+#' 
+#' Factor features are one-hot encoded with reference encoding before fitting.
+#' If factors are present, `basemax` is set to the number of non-factor features
+#' so that tree proposals account for the numeric part of the design.
 #'
 #' @section Initial parameter values:
 #' * `verb` is initialized to `0` to silence printing.
+#' * `pred.n` is initialized to `FALSE` to skip prediction during training.
 #'
 #' @templateVar id regr.btgp
 #' @template learner
@@ -26,36 +31,36 @@ LearnerRegrBtgp = R6Class("LearnerRegrBtgp",
     #' Creates a new instance of this [R6][R6::R6Class] class.
     initialize = function() {
       param_set = ps(
+        meanfn  = p_fct(default = "linear", levels = c("constant", "linear"), tags = "train"),
         bprior  = p_fct(default = "bflat", levels = c("b0", "b0not", "bflat", "bmle", "bmznot", "bmzt"), tags = "train"),
+        corr    = p_fct(default = "expsep", levels = c("exp", "expsep", "matern", "sim"), tags = "train"),
+        tree    = p_uty(default = c(0.5, 2), tags = "train", custom_check = mlr3misc::crate({function(x) {
+          if (checkmate::test_numeric(x, len = 2, any.missing = FALSE)) {
+            if (x[1] >= 0 && x[1] <= 1 && x[2] >= 0) {
+              return(TRUE)
+            }
+          }
+          "`tree` must be numeric length 2 with first element in [0, 1] and second >= 0"
+        }})),
         BTE     = p_uty(default = c(2000L, 7000L, 2L), tags = c("train", "predict"), custom_check = mlr3misc::crate({function(x) {
           if (!checkmate::test_integerish(x, len = 3, lower = 0)) {
             return("`BTE` must be an integerish vector of length 3 with non-negative entries")
           }
           TRUE
         }})),
-        corr    = p_fct(default = "expsep", levels = c("exp", "expsep", "matern", "sim"), tags = "train"),
-        Ds2x    = p_lgl(default = FALSE, tags = c("train", "predict")),
-        improv  = p_lgl(default = FALSE, tags = c("train", "predict")),
+        R       = p_int(default = 1L, lower = 1L, tags = c("train", "predict")),
+        m0r1    = p_lgl(default = TRUE, tags = "train"),
+        linburn = p_lgl(default = FALSE, tags = "train"),
         itemps  = p_uty(default = NULL, tags = "train"),
+        pred.n  = p_lgl(init = FALSE, tags = c("train", "predict")),
         krige   = p_lgl(default = TRUE, tags = c("train", "predict")),
-        linburn = p_lgl(default = FALSE, tags = "train"),
-        m0r1    = p_lgl(default = TRUE, tags = "train"),
-        MAP     = p_lgl(default = TRUE, tags = "predict"),
-        meanfn  = p_fct(default = "linear", levels = c("constant", "linear"), tags = "train"),
+        zcov    = p_lgl(default = FALSE, tags = c("train", "predict")),
+        Ds2x    = p_lgl(default = FALSE, tags = c("train", "predict")),
+        improv  = p_lgl(default = FALSE, tags = c("train", "predict")),
         nu      = p_dbl(default = 1.5, tags = "train", depends = quote(corr == "matern")),
-        R       = p_int(default = 1L, lower = 1L, tags = c("train", "predict")),
-        sens.p  = p_uty(default = NULL, tags = c("train", "predict")),
         trace   = p_lgl(default = FALSE, tags = c("train", "predict")),
-        tree    = p_uty(default = c(0.5, 2), tags = "train", custom_check = mlr3misc::crate({function(x) {
-          if (checkmate::test_numeric(x, len = 2, any.missing = FALSE)) {
-            if (x[1] >= 0 && x[1] <= 1 && x[2] >= 0) {
-              return(TRUE)
-            }
-          }
-          "`tree` must be numeric length 2 with first element in [0, 1] and second >= 0"
-        }})),
         verb    = p_int(init = 0L, lower = 0L, upper = 4L, tags = c("train", "predict")),
-        zcov    = p_lgl(default = FALSE, tags = c("train", "predict"))
+        MAP     = p_lgl(default = TRUE, tags = "predict")
       )
 
       super$initialize(
@@ -79,7 +84,7 @@ LearnerRegrBtgp = R6Class("LearnerRegrBtgp",
         pars$basemax = encoded$basemax
       }
 
-      model = mlr3misc::invoke(tgp::btgp,
+      model = invoke(tgp::btgp,
         X = encoded$data,
         Z = task$truth(),
         .args = pars
@@ -106,7 +111,11 @@ LearnerRegrBtgp = R6Class("LearnerRegrBtgp",
         encoded$data = encoded$data[, self$model$column_names, drop = FALSE]
       }
 
-      pred = mlr3misc::invoke(predict,
+      if (self$predict_type == "se") {
+        pars$krige = TRUE
+      }
+
+      pred = invoke(predict,
         object = self$model$model,
         XX = encoded$data,
         .args = pars