Merge pull request #245 from tidymodels/fix-param-names

Fix param names

Author: topepo

NEWS.md

@@ -12,6 +12,10 @@
 
  * `nnet` was added as an engine to `multinom_reg()` [#209](https://github.com/tidymodels/parsnip/issues/209)
 
+## Breaking Changes
+
+ * There were some mis-mapped parameters (going between `parsnip` and the underlying model function) for `spark` boosted trees and some `keras` models. See [897c927](https://github.com/tidymodels/parsnip/commit/897c92719332caf7344e7c9c8895ac673517d2c8).
+
 
 # parsnip 0.0.4
 

R/boost_tree_data.R

@@ -309,7 +309,7 @@ set_model_arg(
   model = "boost_tree",
   eng = "spark",
   parsnip = "min_info_gain",
-  original = "gamma",
+  original = "loss_reduction",
   func = list(pkg = "dials", fun = "loss_reduction"),
   has_submodel = FALSE
 )

R/linear_reg_data.R

@@ -342,6 +342,15 @@ set_model_engine("linear_reg", "regression", "keras")
 set_dependency("linear_reg", "keras", "keras")
 set_dependency("linear_reg", "keras", "magrittr")
 
+set_model_arg(
+  model = "linear_reg",
+  eng = "keras",
+  parsnip = "penalty",
+  original = "penalty",
+  func = list(pkg = "dials", fun = "penalty"),
+  has_submodel = FALSE
+)
+
 set_fit(
   model = "linear_reg",
   eng = "keras",

R/logistic_reg_data.R

@@ -288,9 +288,9 @@ set_dependency("logistic_reg", "keras", "magrittr")
 set_model_arg(
   model = "logistic_reg",
   eng = "keras",
-  parsnip = "decay",
-  original = "decay",
-  func = list(pkg = "dials", fun = "weight_decay"),
+  parsnip = "penalty",
+  original = "penalty",
+  func = list(pkg = "dials", fun = "penalty"),
   has_submodel = FALSE
 )
 

R/mlp.R

@@ -265,12 +265,12 @@ class2ind <- function (x, drop2nd = FALSE) {
 #' @param x A data frame or matrix of predictors
 #' @param y A vector (factor or numeric) or matrix (numeric) of outcome data.
 #' @param hidden_units An integer for the number of hidden units.
-#' @param decay A non-negative real number for the amount of weight decay. Either
+#' @param penalty A non-negative real number for the amount of weight decay. Either
 #'  this parameter _or_ `dropout` can specified.
 #' @param dropout The proportion of parameters to set to zero. Either
-#'  this parameter _or_ `decay` can specified.
+#'  this parameter _or_ `penalty` can specified.
 #' @param epochs An integer for the number of passes through the data.
-#' @param act A character string for the type of activation function between layers.
+#' @param activation A character string for the type of activation function between layers.
 #' @param seeds A vector of three positive integers to control randomness of the
 #'  calculations.
 #' @param ... Currently ignored.
@@ -279,11 +279,11 @@ class2ind <- function (x, drop2nd = FALSE) {
 #' @export
 keras_mlp <-
   function(x, y,
-           hidden_units = 5, decay = 0, dropout = 0, epochs = 20, act = "softmax",
+           hidden_units = 5, penalty = 0, dropout = 0, epochs = 20, activation = "softmax",
            seeds = sample.int(10^5, size = 3),
            ...) {
 
-    if (decay > 0 & dropout > 0) {
+    if (penalty > 0 & dropout > 0) {
       stop("Please use either dropoput or weight decay.", call. = FALSE)
     }
     if (!is.matrix(x)) {
@@ -307,20 +307,20 @@ keras_mlp <-
 
     model <- keras::keras_model_sequential()
 
-    if (decay > 0) {
+    if (penalty > 0) {
       model %>%
         keras::layer_dense(
           units = hidden_units,
-          activation = act,
+          activation = activation,
           input_shape = ncol(x),
-          kernel_regularizer = keras::regularizer_l2(decay),
+          kernel_regularizer = keras::regularizer_l2(penalty),
           kernel_initializer = keras::initializer_glorot_uniform(seed = seeds[1])
         )
     } else {
       model %>%
         keras::layer_dense(
           units = hidden_units,
-          activation = act,
+          activation = activation,
           input_shape = ncol(x),
           kernel_initializer = keras::initializer_glorot_uniform(seed = seeds[1])
         )
@@ -330,7 +330,7 @@ keras_mlp <-
       model %>%
         keras::layer_dense(
           units = hidden_units,
-          activation = act,
+          activation = activation,
           input_shape = ncol(x),
           kernel_initializer = keras::initializer_glorot_uniform(seed = seeds[1])
         ) %>%

R/mlp_data.R

@@ -24,7 +24,7 @@ set_model_arg(
   eng = "keras",
   parsnip = "penalty",
   original = "penalty",
-  func = list(pkg = "dials", fun = "weight_decay"),
+  func = list(pkg = "dials", fun = "penalty"),
   has_submodel = FALSE
 )
 set_model_arg(
@@ -188,7 +188,7 @@ set_model_arg(
   eng = "nnet",
   parsnip = "penalty",
   original = "decay",
-  func = list(pkg = "dials", fun = "weight_decay"),
+  func = list(pkg = "dials", fun = "penalty"),
   has_submodel = FALSE
 )
 set_model_arg(

R/multinom_reg_data.R

@@ -172,9 +172,9 @@ set_dependency("multinom_reg", "keras", "magrittr")
 set_model_arg(
   model = "multinom_reg",
   eng = "keras",
-  parsnip = "decay",
-  original = "decay",
-  func = list(pkg = "dials", fun = "weight_decay"),
+  parsnip = "penalty",
+  original = "penalty",
+  func = list(pkg = "dials", fun = "penalty"),
   has_submodel = FALSE
 )