more linear regression glmnet updates for #195

Author: topepo

DESCRIPTION

@@ -1,5 +1,5 @@
 Package: parsnip
-Version: 0.0.2.9000
+Version: 0.0.3
 Title: A Common API to Modeling and Analysis Functions
 Description: A common interface is provided to allow users to specify a model without having to remember the different argument names across different functions or computational engines (e.g. 'R', 'Spark', 'Stan', etc). 
 Authors@R: c(

NAMESPACE

@@ -56,6 +56,7 @@ S3method(print,svm_rbf)
 S3method(translate,boost_tree)
 S3method(translate,decision_tree)
 S3method(translate,default)
+S3method(translate,linear_reg)
 S3method(translate,mars)
 S3method(translate,mlp)
 S3method(translate,nearest_neighbor)

NEWS.md

@@ -1,17 +1,26 @@
-# parsnip 0.0.2.9000
+# parsnip 0.0.3
+
+Unplanned release based on CRAN requirements for Solaris.
 
 ## Breaking Changes
 
  * The method that `parsnip` stores the model information has changed. Any custom models from previous versions will need to use the new method for registering models. The methods are detailed in `?get_model_env()` and the [package vignette for adding models](https://tidymodels.github.io/parsnip/articles/articles/Scratch.html).
- * The mode need to be declared for models that can be used for more than one mode prior to fitting and/or translation). 
+
+ * The mode need to be declared for models that can be used for more than one mode prior to fitting and/or translation. 
+
  * For `surv_reg()`, the engine that uses the `survival` package is now called `survival` instead of `survreg`.  
 
+ * For `glmnet` models, the full regularization path is always fit regardless of the value given to `penalty`. Previously, the model was fit with passing `penalty` to `glmnet`'s `lambda` argument and the model could only make predictions at those specific values. [(#195)](https://github.com/tidymodels/parsnip/issues/195)
+
 ## New Features
 
  * `add_rowindex()` can create a column called `.row` to a data frame. 
 
  * If a computational engine is not explicitly set, a default will be used. Each default is documented on the corresponding model page. A warning is issued at fit time unless verbosity is zero.  
+
  * `nearest_neighbor` gained a `multi_predict` method. The `multi_predict()` documentation is a little better organized.  
+ 
+ * A suite of internal functions were added to help with upcoming model tuning features.  
 
 
 # parsnip 0.0.2

R/linear_reg.R

@@ -336,7 +336,12 @@ multi_predict._elnet <-
     object$spec <- eval_args(object$spec)
 
     if (is.null(penalty)) {
-      penalty <- object$fit$lambda
+      # See discussion in https://github.com/tidymodels/parsnip/issues/195
+      if (!is.null(object$spec$args$penalty)) {
+        penalty <- object$spec$args$penalty
+      } else {
+        penalty <- object$fit$lambda
+      }
     }
 
     pred <- predict._elnet(object, new_data = new_data, type = "raw",

tests/testthat/test_linear_reg_glmnet.R

@@ -67,10 +67,10 @@ test_that('glmnet prediction, single lambda', {
     y = iris$Sepal.Length
   )
 
-  uni_pred <- c(5.05124049139868, 4.87103404621362, 4.91028250633598, 4.9399094532023,
-                5.08728178043569)
+  uni_pred <- c(5.05125589060219, 4.86977761622526, 4.90912345599309, 4.93931874108359,
+                5.08755154547758)
 
-  expect_equal(uni_pred, predict(res_xy, iris[1:5, num_pred])$.pred)
+  expect_equal(uni_pred, predict(res_xy, iris[1:5, num_pred])$.pred, tolerance = 0.0001)
 
   res_form <- fit(
     iris_basic,
@@ -79,10 +79,10 @@ test_that('glmnet prediction, single lambda', {
     control = ctrl
   )
 
-  form_pred <- c(5.24228948237804, 5.09448280355765, 5.15636527125752, 5.12592317615935,
-                 5.26930099973607)
+  form_pred <- c(5.23960117346944, 5.08769210344022, 5.15129212608077, 5.12000510716518,
+                 5.26736239856889)
 
-  expect_equal(form_pred, predict(res_form, iris[1:5,])$.pred)
+  expect_equal(form_pred, predict(res_form, iris[1:5,])$.pred, tolerance = 0.0001)
 })
 
 
@@ -132,7 +132,8 @@ test_that('glmnet prediction, multiple lambda', {
     as.data.frame(mult_pred),
     multi_predict(res_xy, new_data = iris[1:5, num_pred], lambda = lams) %>%
       unnest() %>%
-      as.data.frame()
+      as.data.frame(),
+    tolerance = 0.0001
   )
 
   res_form <- fit(
@@ -176,7 +177,8 @@ test_that('glmnet prediction, multiple lambda', {
     as.data.frame(form_pred),
     multi_predict(res_form, new_data = iris[1:5, ], lambda = lams) %>%
       unnest() %>%
-      as.data.frame()
+      as.data.frame(),
+    tolerance = 0.0001
   )
 })
 
@@ -249,7 +251,7 @@ test_that('submodel prediction', {
   )
 
   reg_fit <-
-    linear_reg(penalty = c(0, 0.01, 0.1)) %>%
+    linear_reg() %>%
     set_engine("glmnet") %>%
     fit(mpg ~ ., data = mtcars[-(1:4), ])
 
@@ -274,12 +276,6 @@ test_that('error traps', {
 
   skip_if_not_installed("glmnet")
 
-  expect_error(
-    linear_reg(penalty = .1) %>%
-      set_engine("glmnet") %>%
-      fit(mpg ~ ., data = mtcars[-(1:4), ]) %>%
-      predict(mtcars[-(1:4), ], penalty = .2)
-  )
   expect_error(
     linear_reg() %>%
       set_engine("glmnet") %>%