def test_reset_classification(make_random_dataset):
"""Make sure the reset of the reclassification status works as expected."""
X, _ = make_random_dataset # pylint: disable=invalid-name
palinstance = PALBase(X, ["model"], 3)
lows = np.zeros((100, 3))
highs = np.zeros((100, 3))
means = np.full((100, 3), 1)
palinstance._means = means
palinstance.std = np.full((100, 3), 0.1)
pareto_optimal = np.array([False] * 98 + [True, True])
sampled = np.array([[False] * 3, [False] * 3, [False] * 3, [False] * 3])
unclassified = np.array([True] * 98 + [False, False])
palinstance.rectangle_lows = lows
palinstance.rectangle_ups = highs
palinstance.sampled = sampled
palinstance.pareto_optimal = pareto_optimal
palinstance.unclassified = unclassified
palinstance._reset_classification()
assert palinstance.number_unclassified_points == len(X)
assert palinstance.number_pareto_optimal_points == 0
assert palinstance.number_discarded_points == 0
def test__update_hyperrectangles(make_random_dataset):
"""Testing if the updating of the hyperrectangles works as expected.
As above, the core functionality is tested in for the function with the logic.
Here, we're more interested in seeing how it works with the class object
"""
X, _ = make_random_dataset # pylint:disable=invalid-name
palinstance = PALBase(X, ["model"], 4, beta_scale=1)
palinstance._means = np.array([[0, 0, 1, 0], [0, 0, 0, 1]])
palinstance.std = np.array([[0, 0, 0, 0], [1, 0, 0, 0]])
with pytest.raises(TypeError):
# Beta is not defined
palinstance._update_hyperrectangles()
palinstance._update_beta()
palinstance._update_hyperrectangles()
assert palinstance.rectangle_lows is not None
assert palinstance.rectangle_ups is not None
assert palinstance.rectangle_lows[0][0] == 0
assert palinstance.rectangle_ups[0][0] == 0
assert palinstance.rectangle_lows[0][2] == 1
assert palinstance.rectangle_ups[0][2] == 1
assert palinstance.rectangle_lows[1][0] < -1
assert palinstance.rectangle_ups[1][0] > 1
assert len(palinstance.hyperrectangle_sizes) == len(palinstance._means)
def test__update_coef_var_mask(make_random_dataset):
"""Test that the coefficient of variation mask works as expected"""
X, _ = make_random_dataset # pylint:disable=invalid-name
palinstance = PALBase(X[:2], ["model"], 3, beta_scale=1)
palinstance._means = np.array([[1, 1, 1, 1], [1, 1, 1, 1]])
palinstance.std = np.array([[0, 0, 0, 0], [3, 1, 1, 1]])
assert (palinstance.coef_var_mask == np.array([True, True])).all()
palinstance._update_coef_var_mask()
assert (palinstance.coef_var_mask == np.array([True, False])).all()
palinstance._means = np.array([[0, 0, 0, 0], [0, 0, 0, 0]])
palinstance.std = np.array([[0, 0, 0, 0], [3, 1, 1, 1]])
assert (palinstance.coef_var_mask == np.array([True, False])).all()
def test__replace_by_measurements(make_random_dataset):
"""Test that replacing the mean/std by the measured ones works"""
X, y = make_random_dataset # pylint:disable=invalid-name
palinstance = PALBase(X, ["model"], 3, beta_scale=1)
assert palinstance.measurement_uncertainty.sum() == 0
sample_idx = np.array([1, 2, 3, 4])
palinstance.update_train_set(sample_idx, y[sample_idx], y[sample_idx])
palinstance._means = palinstance.measurement_uncertainty
palinstance.std = palinstance.measurement_uncertainty
palinstance._replace_by_measurements()
assert (palinstance.y == palinstance.std).all()
def test_augment_design_space(make_random_dataset):
"""Testing the basic functionality of the augmentation method
Does NOT test the re-classification step, which needs a model"""
X, _ = make_random_dataset # pylint: disable=invalid-name
X_augmented = np.vstack([X, X]) # pylint: disable=invalid-name
palinstance = PALBase(X, ["model"], 3)
# Iteration count to low
with pytest.raises(ValueError):
palinstance.augment_design_space(X_augmented)
palinstance.iteration = 2
# Incorrect shape
with pytest.raises(AssertionError):
palinstance.augment_design_space(X_augmented[:, 2])
with pytest.raises(ValueError):
palinstance.augment_design_space(X_augmented[:, :2])
# Mock that we already ran that
lows = np.zeros((100, 3))
highs = np.zeros((100, 3))
means = np.full((100, 3), 1)
palinstance._means = means
palinstance.std = np.full((100, 3), 0.1)
pareto_optimal = np.array([False] * 98 + [True, True])
sampled = np.array([[False] * 3, [False] * 3, [False] * 3, [False] * 3])
unclassified = np.array([True] * 98 + [False, False])
palinstance.rectangle_lows = lows
palinstance.rectangle_ups = highs
palinstance.sampled = sampled
palinstance.pareto_optimal = pareto_optimal
palinstance.unclassified = unclassified
# As we do not have a model, we cannot test the classification
palinstance.augment_design_space(X_augmented, clean_classify=False)
assert palinstance.number_discarded_points == 0
assert palinstance.number_pareto_optimal_points == 2
assert palinstance.number_unclassified_points == 298
assert palinstance.number_sampled_points == 0
assert palinstance.number_design_points == 300
assert len(palinstance.means) == 300
assert len(palinstance.std) == 300
def test_sample(make_random_dataset):
"""Test the sampling"""
X, _ = make_random_dataset # pylint:disable=invalid-name
palinstance = PALBase(X, ["model"], 4)
lows = np.array([
[0.0, 0.0, 0.0, 0.0],
[-1.0, -1.0, -1.0, -1.0],
[-2.0, -2.0, -2.0, -2.0],
[2.0, 2.0, 2.0, 2.0],
])
highs = np.array([
[1.0, 1.0, 1.0, 1.0],
[1.0, 1.0, 1.0, 1.0],
[1.0, 1.0, 1.0, 1.0],
[2.0, 2.0, 2.0, 2.0],
])
means = np.array([[1, 1, 1, 1], [1, 1, 1, 1], [1, 1, 1, 1], [1, 1, 1, 1]])
palinstance._means = means
pareto_optimal = np.array([False, False, True, True])
sampled = np.array([[False] * 4, [False] * 4, [False] * 4, [False] * 4])
unclassified = np.array([True, True, False, False])
palinstance.rectangle_lows = lows
palinstance.rectangle_ups = highs
palinstance.sampled = sampled
palinstance.pareto_optimal = pareto_optimal
palinstance.unclassified = unclassified
sampled_idx = palinstance.sample()
assert sampled_idx == 2
pareto_optimal = np.array([False, False, True, True])
sampled = np.array([[False] * 4, [False] * 4, [False] * 4, [False] * 4])
unclassified = np.array([True, True, False, False])
palinstance.sampled = sampled
palinstance.pareto_optimal = pareto_optimal
palinstance.unclassified = unclassified
sampled_idx = palinstance.sample()
assert sampled_idx == 2
pareto_optimal = np.array([False, False, True, True])
sampled = np.array([[False] * 4, [False] * 4, [True] * 4, [False] * 4])
unclassified = np.array([True, True, False, False])
palinstance.sampled = sampled
palinstance.pareto_optimal = pareto_optimal
palinstance.unclassified = unclassified
sampled_idx = palinstance.sample()
assert sampled_idx == 1
pareto_optimal = np.array([False, False, False, True])
sampled = np.array([[False] * 4, [False] * 4, [True] * 4, [False] * 4])
unclassified = np.array([True, True, False, False])
palinstance.sampled = sampled
palinstance.pareto_optimal = pareto_optimal
palinstance.unclassified = unclassified
sampled_idx = palinstance.sample()
assert sampled_idx == 1