tests/testthat/test-ml-clustering-gaussian-mixture.R

In rstudio/sparklyr: R Interface to Apache Spark

skip_connection("ml-clustering-gaussian-mixture")
skip_on_livy()
skip_on_arrow_devel()

test_that("ml_gaussian_mixture() default params", {
  test_requires_version("3.0.0")
  sc <- testthat_spark_connection()
  test_default_args(sc, ml_gaussian_mixture)
})

test_that("ml_gaussian_mixture() param setting", {
  test_requires_version("3.0.0")
  sc <- testthat_spark_connection()
  test_args <- list(
    k = 5,
    max_iter = 80,
    tol = 0.05,
    seed = 345345,
    features_col = "wefwef",
    prediction_col = "awefawef",
    probability_col = "ewfiuwje"
  )
  test_param_setting(sc, ml_gaussian_mixture, test_args)
})

test_that("ml_gaussian_mixture param setting", {
  sc <- testthat_spark_connection()
  test_requires_version("2.0.0", "gaussian mixture requires 2.0+")
  args <- list(
    x = sc, k = 9, max_iter = 120, tol = 0.02,
    seed = 98, features_col = "fcol",
    prediction_col = "pcol", probability_col = "prcol"
  )
  predictor <- do.call(ml_gaussian_mixture, args)

  expect_equal(ml_params(predictor, names(args)[-1]), args[-1])
})

test_that("ml_gaussian_mixture() default params are correct", {
  sc <- testthat_spark_connection()
  test_requires_version("2.0.0", "gaussian mixture requires 2.0+")
  predictor <- ml_pipeline(sc) %>%
    ml_gaussian_mixture() %>%
    ml_stage(1)

  args <- get_default_args(
    ml_gaussian_mixture,
    c("x", "uid", "...", "seed")
  )

  expect_equal(
    ml_params(predictor, names(args)),
    args
  )
})

test_that("ml_gaussian_mixture() works properly", {
  sc <- testthat_spark_connection()
  test_requires_version("2.0.0", "gaussian mixture requires 2.0+")
  sample_data_path <- get_test_data_path("sample_kmeans_data.txt")
  sample_data <- spark_read_libsvm(sc, "sample_data",
    sample_data_path,
    overwrite = TRUE
  )

  gmm <- ml_gaussian_mixture(sample_data, k = 2, seed = 1)

  expect_equal(gmm$weights, c(0.5, 0.5))

  expect_warning_on_arrow(
    g_mean <- gmm$gaussians_df() %>% pull(mean)
  )

  expect_equal(
    g_mean,
    list(c(0.1, 0.1, 0.1), c(9.1, 9.1, 9.1))
  )

  expect_warning_on_arrow(
    g_cov <- gmm$gaussians_df() %>% pull(cov)
  )

  expect_equal(
    g_cov,
    list(
      matrix(rep(0.00666666667, 9), nrow = 3),
      matrix(rep(0.00666666667, 9), nrow = 3)
    )
  )
})

test_clear_cache()