Extract the candidates a final fit actually scored

Returns the candidate parameter settings that nested_final_fit()'s tuning run actually evaluated — the full-data counterpart of the .grid column nested_tune_grid() records for each outer fold.

Usage

extract_scored_candidates(x, ...)

Arguments

x

A nested_final_fit object from nested_final_fit().

...

Not used.

Value

A tibble with one row per candidate scored, carrying one column per tuned parameter plus tune's .config label for the candidate. It is the same shape as one element of nested_tune_grid()'s .grid column, so the two can be compared directly.

This is what was scored, not what was asked for. A grid given as a size is expanded by tune and may reach fewer candidates than the number requested; a candidate that failed everywhere scored nothing. See the .grid discussion in nested_tune_grid() for the full account of how the two records diverge, which holds here too — this record is derived the same way.

One pointer there does not carry over. A candidate that failed on every inner resample is missing from this table, and on a nested_tune_grid() result its failure is recorded in that object's .notes column. A nested_final_fit has no such column. Look instead inside the tuning run itself — tune::collect_notes(extract_tune_results(x)).

See also

extract_tune_results(), nested_final_fit(), nested_tune_grid()

Examples

data(mtcars)

rec <- recipes::step_pca(
  recipes::recipe(mpg ~ ., data = mtcars),
  recipes::all_predictors(),
  num_comp = tune::tune()
)
wf <- workflows::workflow(rec, parsnip::linear_reg())

set.seed(1)
folds <- nested_resamples(
  mtcars,
  outside = rsample::vfold_cv(v = 3),
  inside = rsample::vfold_cv(v = 3)
)

set.seed(3)
final <- nested_final_fit(wf, folds, grid = data.frame(num_comp = 1:3))

extract_scored_candidates(final)
#> # A tibble: 3 × 2
#>   num_comp .config        
#>      <int> <chr>          
#> 1        1 pre1_mod0_post0
#> 2        2 pre2_mod0_post0
#> 3        3 pre3_mod0_post0