tests/testthat/test-relu.R

In recipes: Preprocessing and Feature Engineering Steps for Modeling

library(testthat)
library(recipes)

df <- tibble(val1 = -10:10, val2 = factor(LETTERS[1:21]))

test_that("default relu settings", {
  baked <- recipe(~., data = df) %>%
    step_relu(val1) %>%
    prep(df, verbose = FALSE) %>%
    bake(df)

  expected_baked <- df %>%
    dplyr::mutate(right_relu_val1 = pmax(rep(0, length(val1)), val1))

  expect_equal(baked, expected_baked)
})


test_that("shifted and reversed relu", {
  baked <- recipe(~., data = df) %>%
    step_relu(val1, shift = 5, reverse = TRUE) %>%
    prep(df, verbose = FALSE) %>%
    bake(df)

  expected_baked <- df %>%
    dplyr::mutate(left_relu_val1 = pmax(rep(0, length(val1)), -(val1 - 5)))

  expect_equal(baked, expected_baked)
})


test_that("reversed softplus", {
  baked <- recipe(~., data = df) %>%
    step_relu(val1, smooth = TRUE, reverse = TRUE) %>%
    prep(df, verbose = FALSE) %>%
    bake(df)

  expected_baked <- df %>%
    dplyr::mutate(left_relu_val1 = log1p(exp(-val1)))

  expect_equal(baked, expected_baked)
})


test_that("shifted and prefixed softplus", {
  baked <- recipe(~., data = df) %>%
    step_relu(val1, shift = 5, smooth = TRUE, prefix = "sp_") %>%
    prep(df, verbose = FALSE) %>%
    bake(df)

  expected_baked <- df %>%
    dplyr::mutate(sp_val1 = log1p(exp(val1 - 5)))

  expect_equal(baked, expected_baked)
})

test_that("works with all_predictors() selector", {
  expect_silent({
    rec <- recipe(Species ~ ., data = iris) %>%
      step_relu(all_predictors())
  })

  expect_silent(prepped_rec <- prep(rec, iris))
})


test_that("input checking", {
  expect_snapshot(error = TRUE,
    recipe(~., data = df) %>%
      step_relu(val1, shift = TRUE) %>% # wrong argument type to shift
      prep(df, verbose = FALSE)
  )

  expect_snapshot(error = TRUE,
    recipe(~., data = df) %>%
      step_relu(val1, reverse = 3) %>% # wrong argument type to reverse
      prep(df, verbose = FALSE)
  )

  expect_snapshot(error = TRUE,
    recipe(~., data = df) %>%
      step_relu(val1, smooth = "cat") %>% # wrong argument type to smooth
      prep(df, verbose = FALSE)
  )

  expect_snapshot(error = TRUE,
    recipe(~., data = df) %>%
      step_relu(val2) %>% # apply to non-numeric column
      prep(df, verbose = FALSE)
  )
})

# Infrastructure ---------------------------------------------------------------

test_that("bake method errors when needed non-standard role columns are missing", {
  rec <- recipe(~., data = df) %>%
    step_relu(val1) %>%
    update_role(val1, new_role = "potato") %>%
    update_role_requirements("potato", bake = FALSE) %>%
    prep(df, verbose = FALSE)

  expect_error(bake(rec, df[, 2, drop = FALSE]),
               class = "new_data_missing_column")
})

test_that("empty printing", {
  rec <- recipe(mpg ~ ., mtcars)
  rec <- step_relu(rec)

  expect_snapshot(rec)

  rec <- prep(rec, mtcars)

  expect_snapshot(rec)
})

test_that("empty selection prep/bake is a no-op", {
  rec1 <- recipe(mpg ~ ., mtcars)
  rec2 <- step_relu(rec1)

  rec1 <- prep(rec1, mtcars)
  rec2 <- prep(rec2, mtcars)

  baked1 <- bake(rec1, mtcars)
  baked2 <- bake(rec2, mtcars)

  expect_identical(baked1, baked2)
})

test_that("empty selection tidy method works", {
  rec <- recipe(mpg ~ ., mtcars)
  rec <- step_relu(rec)

  expect <- tibble(
    terms = character(),
    shift = double(),
    reverse = logical(),
    id = character()
  )

  expect_identical(tidy(rec, number = 1), expect)

  rec <- prep(rec, mtcars)

  expect_identical(tidy(rec, number = 1), expect)
})

test_that("printing", {
  rec <- recipe(~., data = mtcars) %>%
    step_relu(disp)

  expect_snapshot(print(rec))
  expect_snapshot(prep(rec))
})