tests/testthat/test_ResultAssignerSurrogate.R

In mlr3mbo: Flexible Bayesian Optimization

test_that("ResultAssignerSurrogate works", {
  result_assigner = ResultAssignerSurrogate$new()
  expect_null(result_assigner$surrogate)

  instance = MAKE_INST_1D()
  design = generate_design_random(instance$search_space, n = 4L)$data
  instance$eval_batch(design)

  expect_null(instance$result)
  result_assigner$assign_result(instance)
  expect_r6(result_assigner$surrogate, classes = "SurrogateLearner")
  expect_data_table(instance$result, nrows = 1L)
})

test_that("ResultAssignerSurrogate result and best can be different", {
  skip_on_cran()
  skip_if_not_installed("rpart")
  result_assigner = ResultAssignerSurrogate$new(surrogate = SurrogateLearner$new(lrn("regr.rpart")))

  instance = MAKE_INST_1D()
  design = generate_design_grid(instance$search_space, resolution = 4L)$data
  instance$eval_batch(design)

  expect_null(instance$result)
  result_assigner$assign_result(instance)
  expect_data_table(instance$result, nrows = 1L)
  mean = result_assigner$surrogate$predict(design)$mean
  best_index = which.min(mean)  # first one
  expect_equal(instance$result[[instance$archive$cols_x]], design[best_index, ][[instance$archive$cols_x]])
  expect_equal(instance$result[[instance$archive$cols_y]], instance$archive$data[best_index, ][[instance$archive$cols_y]])
  expect_true(abs(instance$result[[instance$archive$cols_y]] - mean[best_index]) > 1e-2)
})

test_that("ResultAssignerSurrogate works with OptimizerMbo and bayesopt_ego", {
  result_assigner = ResultAssignerSurrogate$new()
  expect_null(result_assigner$surrogate)

  instance = MAKE_INST_1D_NOISY()
  surrogate = SurrogateLearner$new(REGR_KM_NOISY)
  acq_function = AcqFunctionAEI$new()
  acq_optimizer = AcqOptimizer$new(opt("random_search", batch_size = 2L), terminator = trm("evals", n_evals = 2L))

  optimizer = opt("mbo", loop_function = bayesopt_ego, surrogate = surrogate, acq_function = acq_function, acq_optimizer = acq_optimizer, result_assigner = result_assigner)
  optimizer$optimize(instance)
  expect_true(nrow(instance$archive$data) == 5L)

  expect_r6(result_assigner$surrogate, classes = "SurrogateLearner")
  expect_r6(result_assigner$surrogate$learner, classes = "Learner")
  expect_data_table(instance$result, nrow = 1L)
})

test_that("ResultAssignerSurrogate works with OptimizerMbo and bayesopt_parego", {
  result_assigner = ResultAssignerSurrogate$new()
  expect_null(result_assigner$surrogate)

  instance = MAKE_INST(OBJ_1D_2, search_space = PS_1D, terminator = trm("evals", n_evals = 5L))
  surrogate = SurrogateLearner$new(REGR_KM_DETERM)
  acq_function = AcqFunctionEI$new()
  acq_optimizer = AcqOptimizer$new(opt("random_search", batch_size = 2L), terminator = trm("evals", n_evals = 2L))

  optimizer = opt("mbo", loop_function = bayesopt_parego, surrogate = surrogate, acq_function = acq_function, acq_optimizer = acq_optimizer, result_assigner = result_assigner)
  optimizer$optimize(instance)
  expect_true(nrow(instance$archive$data) == 5L)

  expect_r6(result_assigner$surrogate, classes = "SurrogateLearnerCollection")
  expect_list(result_assigner$surrogate$learner, types = "Learner")
  expect_data_table(instance$result, min.rows = 1L)
})

test_that("ResultAssignerSurrogate works with OptimizerMbo and bayesopt_smsego", {
  result_assigner = ResultAssignerSurrogate$new()
  expect_null(result_assigner$surrogate)

  instance = MAKE_INST(OBJ_1D_2, search_space = PS_1D, terminator = trm("evals", n_evals = 5L))
  surrogate = SurrogateLearnerCollection$new(list(REGR_KM_DETERM, REGR_KM_DETERM$clone(deep = TRUE)))
  acq_function = AcqFunctionSmsEgo$new()
  acq_optimizer = AcqOptimizer$new(opt("random_search", batch_size = 2L), terminator = trm("evals", n_evals = 2L))

  optimizer = opt("mbo", loop_function = bayesopt_smsego, surrogate = surrogate, acq_function = acq_function, acq_optimizer = acq_optimizer, result_assigner = result_assigner)
  optimizer$optimize(instance)
  expect_true(nrow(instance$archive$data) == 5L)

  expect_r6(result_assigner$surrogate, classes = "SurrogateLearnerCollection")
  expect_list(result_assigner$surrogate$learner, types = "Learner")
  expect_data_table(instance$result, min.rows = 1L)
})

test_that("ResultAssignerSurrogate passes internal tuned values", {
  result_assigner = ResultAssignerSurrogate$new()

  learner = lrn("classif.debug",
    validate = 0.2,
    early_stopping = TRUE,
    x = to_tune(0.2, 0.3),
    iter = to_tune(upper = 1000, internal = TRUE, aggr = function(x) 99))

  instance = ti(
    task = tsk("pima"),
    learner = learner,
    resampling = rsmp("cv", folds = 3),
    measures = msr("classif.ce"),
    terminator = trm("evals", n_evals = 20),
    store_benchmark_result = TRUE
  )
  surrogate = SurrogateLearner$new(REGR_KM_DETERM)
  acq_function = AcqFunctionEI$new()
  acq_optimizer = AcqOptimizer$new(opt("random_search", batch_size = 2L), terminator = trm("evals", n_evals = 2L))

  tuner = tnr("mbo", result_assigner = result_assigner)
  expect_data_table(tuner$optimize(instance), nrows = 1)
  expect_list(instance$archive$data$internal_tuned_values, len = 20, types = "list")
  expect_equal(instance$archive$data$internal_tuned_values[[1]], list(iter = 99))
  expect_false(instance$result_learner_param_vals$early_stopping)
  expect_equal(instance$result_learner_param_vals$iter, 99)
})