Define composition = dgCMatrix in tune::tune_grid()

[paulofelipe]

Hi everyone,

Is it possible to use the composition = dgCMatrix option when I pass a recipe in the tune_grid function?

Currently, you can set the composition = dgCMatrix option in thebake () function.

Thanks!

Here a reprex:

# Packages ----------------------------------------------------------------

library(tidyverse)
library(tidymodels)
#> Registered S3 method overwritten by 'xts':
#>   method     from
#>   as.zoo.xts zoo
#> ── Attaching packages ────────────────────────────────────────────── tidymodels 0.0.3 ──
#> ✓ broom     0.5.2          ✓ recipes   0.1.7.9002
#> ✓ dials     0.0.4          ✓ rsample   0.0.5     
#> ✓ infer     0.5.1          ✓ yardstick 0.0.4     
#> ✓ parsnip   0.0.4.9000
#> ── Conflicts ───────────────────────────────────────────────── tidymodels_conflicts() ──
#> x scales::discard()   masks purrr::discard()
#> x dplyr::filter()     masks stats::filter()
#> x recipes::fixed()    masks stringr::fixed()
#> x dplyr::lag()        masks stats::lag()
#> x dials::margin()     masks ggplot2::margin()
#> x yardstick::spec()   masks readr::spec()
#> x recipes::step()     masks stats::step()
#> x recipes::yj_trans() masks scales::yj_trans()
library(textrecipes)
library(workflows)
library(tune)

# Data  --------------------------------------------------------------------

train <- read_csv('train.csv.zip') %>% 
  sample_frac(0.1)
#> Parsed with column specification:
#> cols(
#>   id = col_character(),
#>   comment_text = col_character(),
#>   toxic = col_double(),
#>   severe_toxic = col_double(),
#>   obscene = col_double(),
#>   threat = col_double(),
#>   insult = col_double(),
#>   identity_hate = col_double()
#> )

train <- train %>% 
  mutate(toxic = ifelse(toxic == 1, "yes", "no"))

# Recipe -----------------------------------------------------------------

rec <- recipe(toxic ~ comment_text, data = train) %>% 
  step_tokenize(comment_text) %>% 
  step_stopwords(comment_text) %>% 
  step_tokenfilter(comment_text, max_tokens = 1000) %>% 
  step_tfidf(comment_text)


# Model ------------------------------------------------------------------

model <- logistic_reg(
  mode = "classification",
  penalty = tune(),
  mixture = tune()
) %>% 
  set_engine("glmnet")

# Workflow ---------------------------------------------------------------

toxic_wflow <- workflow() %>% 
  add_recipe(rec) %>% 
  add_model(model)

# Grid --------------------------------------------------------------------

param_grid <- grid_regular(
  mixture(),
  penalty(),
  levels = 2
)

# CV ----------------------------------------------------------------------

cv_rs <- vfold_cv(
  train %>% select(toxic, comment_text),
  v = 10,
  strata = toxic
)

# Tune --------------------------------------------------------------------

grid_search <- tune_grid(
  toxic_wflow,
  resamples = cv_rs,
  metrics = metric_set(roc_auc),
  grid = param_grid,
)

show_best(grid_search, n = 1)
#> # A tibble: 1 x 7
#>   penalty mixture .metric .estimator  mean     n std_err
#>     <dbl>   <dbl> <chr>   <chr>      <dbl> <int>   <dbl>
#> 1       1       0 roc_auc binary     0.893    10 0.00423

^{Created on 2019-12-12 by the reprex package (v0.3.0)}

[Max]

It won't be able to do that on the first version. Can you put an issue in and we'll put it on the list?

[paulofelipe]

Ok, Max! Thanks for the answer!