updates for other glmnet models #195

Author: topepo

NAMESPACE

@@ -57,8 +57,10 @@ S3method(translate,boost_tree)
 S3method(translate,decision_tree)
 S3method(translate,default)
 S3method(translate,linear_reg)
+S3method(translate,logistic_reg)
 S3method(translate,mars)
 S3method(translate,mlp)
+S3method(translate,multinom_reg)
 S3method(translate,nearest_neighbor)
 S3method(translate,rand_forest)
 S3method(translate,surv_reg)

R/linear_reg.R

@@ -68,9 +68,9 @@
 #'
 #' \Sexpr[results=rd]{parsnip:::show_fit(parsnip:::linear_reg(), "keras")}
 #'
-#' When using `glmnet` models, there is the option to pass
-#'  multiple values (or no values) to the `penalty` argument. This
-#'  can have an effect on the model object results. When using the
+#' For `glmnet` models, the full regularization path is always fit regardless
+#' of the value given to `penalty`. Also, there is the option to pass
+#'  multiple values (or no values) to the `penalty` argument. When using the
 #'  `predict()` method in these cases, the return value depends on
 #'  the value of `penalty`. When using `predict()`, only a single
 #'  value of the penalty can be used. When predicting on multiple

R/logistic_reg.R

@@ -66,9 +66,9 @@
 #'
 #' \Sexpr[results=rd]{parsnip:::show_fit(parsnip:::logistic_reg(), "keras")}
 #'
-#' When using `glmnet` models, there is the option to pass
-#'  multiple values (or no values) to the `penalty` argument. This
-#'  can have an effect on the model object results. When using the
+#' For `glmnet` models, the full regularization path is always fit regardless
+#' of the value given to `penalty`. Also, there is the option to pass
+#'  multiple values (or no values) to the `penalty` argument. When using the
 #'  `predict()` method in these cases, the return value depends on
 #'  the value of `penalty`. When using `predict()`, only a single
 #'  value of the penalty can be used. When predicting on multiple
@@ -137,6 +137,9 @@ print.logistic_reg <- function(x, ...) {
   invisible(x)
 }
 
+#' @export
+translate.logistic_reg <- translate.linear_reg
+
 # ------------------------------------------------------------------------------
 
 #' @inheritParams update.boost_tree
@@ -235,7 +238,7 @@ organize_glmnet_prob <- function(x, object) {
 }
 
 # ------------------------------------------------------------------------------
-# glmnet call stack for linear regression using `predict` when object has
+# glmnet call stack for logistic regression using `predict` when object has
 # classes "_lognet" and "model_fit" (for class predictions):
 #
 #  predict()
@@ -247,7 +250,7 @@ organize_glmnet_prob <- function(x, object) {
 #        predict.lognet()
 
 
-# glmnet call stack for linear regression using `multi_predict` when object has
+# glmnet call stack for logistic regression using `multi_predict` when object has
 # classes "_lognet" and "model_fit" (for class predictions):
 #
 # 	multi_predict()
@@ -266,6 +269,11 @@ predict._lognet <- function(object, new_data, type = NULL, opts = list(), penalt
   if (any(names(enquos(...)) == "newdata"))
     stop("Did you mean to use `new_data` instead of `newdata`?", call. = FALSE)
 
+  # See discussion in https://github.com/tidymodels/parsnip/issues/195
+  if (is.null(penalty) & !is.null(object$spec$args$penalty)) {
+    penalty <- object$spec$args$penalty
+  }
+
   object$spec$args$penalty <- check_penalty(penalty, object, multi)
 
   object$spec <- eval_args(object$spec)
@@ -286,8 +294,16 @@ multi_predict._lognet <-
       penalty <- eval_tidy(penalty)
 
     dots <- list(...)
-    if (is.null(penalty))
-      penalty <- eval_tidy(object$fit$lambda)
+
+    if (is.null(penalty)) {
+      # 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
+      }
+    }
+
     dots$s <- penalty
 
     if (is.null(type))

R/multinom_reg.R

@@ -57,9 +57,9 @@
 #'
 #' \Sexpr[results=rd]{parsnip:::show_fit(parsnip:::multinom_reg(), "keras")}
 #'
-#' When using `glmnet` models, there is the option to pass
-#'  multiple values (or no values) to the `penalty` argument. This
-#'  can have an effect on the model object results. When using the
+#' For `glmnet` models, the full regularization path is always fit regardless
+#' of the value given to `penalty`. Also, there is the option to pass
+#'  multiple values (or no values) to the `penalty` argument. When using the
 #'  `predict()` method in these cases, the return value depends on
 #'  the value of `penalty`. When using `predict()`, only a single
 #'  value of the penalty can be used. When predicting on multiple
@@ -112,14 +112,17 @@ print.multinom_reg <- function(x, ...) {
   cat("Multinomial Regression Model Specification (", x$mode, ")\n\n", sep = "")
   model_printer(x, ...)
 
-  if(!is.null(x$method$fit$args)) {
+  if (!is.null(x$method$fit$args)) {
     cat("Model fit template:\n")
     print(show_call(x))
   }
 
   invisible(x)
 }
 
+#' @export
+translate.multinom_reg <- translate.linear_reg
+
 # ------------------------------------------------------------------------------
 
 #' @inheritParams update.boost_tree
@@ -188,7 +191,7 @@ organize_multnet_prob <- function(x, object) {
 }
 
 # ------------------------------------------------------------------------------
-# glmnet call stack for linear regression using `predict` when object has
+# glmnet call stack for multinomial regression using `predict` when object has
 # classes "_multnet" and "model_fit" (for class predictions):
 #
 #  predict()
@@ -199,7 +202,7 @@ organize_multnet_prob <- function(x, object) {
 #       predict.multnet()
 
 
-# glmnet call stack for linear regression using `multi_predict` when object has
+# glmnet call stack for multinomial regression using `multi_predict` when object has
 # classes "_multnet" and "model_fit" (for class predictions):
 #
 # 	multi_predict()
@@ -217,6 +220,11 @@ organize_multnet_prob <- function(x, object) {
 predict._multnet <-
   function(object, new_data, type = NULL, opts = list(), penalty = NULL, multi = FALSE, ...) {
 
+    # See discussion in https://github.com/tidymodels/parsnip/issues/195
+    if (is.null(penalty) & !is.null(object$spec$args$penalty)) {
+      penalty <- object$spec$args$penalty
+    }
+
     object$spec$args$penalty <- check_penalty(penalty, object, multi)
 
     object$spec <- eval_args(object$spec)
@@ -242,8 +250,14 @@ multi_predict._multnet <-
       penalty <- eval_tidy(penalty)
 
     dots <- list(...)
-    if (is.null(penalty))
-      penalty <- eval_tidy(object$fit$lambda)
+    if (is.null(penalty)) {
+      # 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
+      }
+    }
     dots$s <- penalty
 
     if (is.null(type))

R/multinom_reg_data.R

@@ -44,7 +44,7 @@ set_pred(
   mode = "classification",
   type = "class",
   value = list(
-    pre = check_glmnet_lambda,
+    pre = NULL,
     post = organize_multnet_class,
     func = c(fun = "predict"),
     args =

man/linear_reg.Rd

@@ -76,7 +76,6 @@ test_that('primary arguments', {
                  x = expr(missing_arg()),
                  y = expr(missing_arg()),
                  weights = expr(missing_arg()),
-                 lambda = new_empty_quosure(1),
                  family = "gaussian"
                )
   )

man/logistic_reg.Rd

@@ -79,7 +79,6 @@ test_that('primary arguments', {
                  x = expr(missing_arg()),
                  y = expr(missing_arg()),
                  weights = expr(missing_arg()),
-                 lambda = new_empty_quosure(1),
                  family = "binomial"
                )
   )