step_smote and nearest_neighbor have a tuning parameter with same name and causes an error

[rfsaldanha]

Hi!

The functions step_smote from {themis} and nearest_neighbor from {parsnip} have a common tuning parameter called neighbors. When I try to tune both hyperparameters in a workflow, I receive an error about duplicated item. Is there a way to 'rename' a hyperparameter?

Bellow a reprex:

library(tidymodels)
library(themis)
library(palmerpenguins)
#> 
#> Attaching package: 'palmerpenguins'
#> The following object is masked from 'package:modeldata':
#> 
#>     penguins

p_split <- initial_split(data = penguins |> drop_na(), strata = species)

folds <- vfold_cv(data = training(p_split), strata = species)

rec <- recipe(species ~ bill_length_mm + bill_depth_mm, data = training(p_split)) |>
  step_smote(species, over_ratio = tune(), neighbors = tune())

rf_spec <- 
  rand_forest(
    trees = tune(),
    min_n = tune()
  ) |>
  set_engine("ranger") |>
  set_mode("classification")

knn_spec <-
  nearest_neighbor(
    neighbors = tune(),
    dist_power = tune()
  ) |>
  set_engine("kknn") |>
  set_mode("classification")

wf_set <- workflow_set(
  preproc = list(rec),
  models = list(
    rf = rf_spec,
    knn = knn_spec
  ), 
  cross = TRUE
)

tune_res <- 
  wf_set |>
  workflow_map(
    "tune_grid",
    resamples = folds,
    grid = 10,
    verbose = TRUE
  )
#> i 1 of 2 tuning:     recipe_rf
#> ✔ 1 of 2 tuning:     recipe_rf (9.9s)
#> i 2 of 2 tuning:     recipe_knn
#> ✖ 2 of 2 tuning:     recipe_knn failed with: Error in hardhat::extract_parameter_set_dials(wflow) :   Element `id` should have unique values. Duplicates exist for item(s): 'neighbors'

^{Created on 2024-02-22 with reprex v2.1.0}

[EmilHvitfeldt]

Hello @rfsaldanha 👋

In cases were we need to tune the same parameter multiple times, like here with neighbors, can we use the id argument of tune() to distinguish them. I'm adding tune("neighbors_smote") below in step_smote() to denote it to be different than the tune()used innearest_neighbor()`

library(tidymodels)
library(themis)
library(palmerpenguins)
#> 
#> Attaching package: 'palmerpenguins'
#> The following object is masked from 'package:modeldata':
#> 
#>     penguins

p_split <- initial_split(data = penguins |> drop_na(), strata = species)

folds <- vfold_cv(data = training(p_split), strata = species)

rec <- recipe(species ~ bill_length_mm + bill_depth_mm, data = training(p_split)) |>
  step_smote(species, over_ratio = tune(), neighbors = tune("neighbors_smote"))

rf_spec <- 
  rand_forest(
    trees = tune(),
    min_n = tune()
  ) |>
  set_engine("ranger") |>
  set_mode("classification")

knn_spec <-
  nearest_neighbor(
    neighbors = tune(),
    dist_power = tune()
  ) |>
  set_engine("kknn") |>
  set_mode("classification")

wf_set <- workflow_set(
  preproc = list(rec),
  models = list(
    rf = rf_spec,
    knn = knn_spec
  ), 
  cross = TRUE
)

tune_res <- 
  wf_set |>
  workflow_map(
    "tune_grid",
    resamples = folds,
    grid = 10,
    verbose = TRUE
  )
#> i 1 of 2 tuning:     recipe_rf
#> ✔ 1 of 2 tuning:     recipe_rf (9.9s)
#> i 2 of 2 tuning:     recipe_knn
#> ✔ 2 of 2 tuning:     recipe_knn (5.4s)

tune_res
#> # A workflow set/tibble: 2 × 4
#>   wflow_id   info             option    result   
#>   <chr>      <list>           <list>    <list>   
#> 1 recipe_rf  <tibble [1 × 4]> <opts[2]> <tune[+]>
#> 2 recipe_knn <tibble [1 × 4]> <opts[2]> <tune[+]>

[rfsaldanha]

That's great! Thank you.

[github-actions]

This issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with a reprex: https://reprex.tidyverse.org) and link to this issue.