def test_select_random_classifier():
competences = np.random.rand(100)
expected = np.where(competences > 0.25)[0]
probabilistic_test = Probabilistic(create_pool_all_agree(1, 100))
probabilistic_test.n_classes = 4
indices = probabilistic_test.select(competences)
assert np.array_equal(indices, expected)
def test_select_threshold():
competences = np.random.rand(100)
expected = np.where(competences > 0.5)[0]
probabilistic_test = Probabilistic(create_pool_all_agree(1, 100))
probabilistic_test.selection_threshold = 0.5
indices = probabilistic_test.select(competences)
assert np.array_equal(indices, expected)
def test_not_predict_proba():
X = X_dsel_ex1
y = y_dsel_ex1
clf1 = Perceptron()
clf1.fit(X, y)
with pytest.raises(ValueError):
Probabilistic([clf1, clf1])
def test_estimate_competence_ones():
query = np.atleast_2d([1, 1])
probabilistic_test = Probabilistic(create_pool_classifiers())
probabilistic_test.distances = distances_ex1[0, 0:3]
probabilistic_test.neighbors = [0, 2, 1]
probabilistic_test.DFP_mask = np.ones(probabilistic_test.n_classifiers)
probabilistic_test.C_src = np.ones((3, 3))
competence = probabilistic_test.estimate_competence(query)
assert (competence == 1.0).all()
def test_estimate_competence():
query = np.atleast_2d([1, 1])
probabilistic_test = Probabilistic(create_pool_classifiers())
probabilistic_test.distances = [0.5, 1.0, 2.0]
probabilistic_test.neighbors = [0, 1, 2]
probabilistic_test.DFP_mask = np.ones(probabilistic_test.n_classifiers)
probabilistic_test.C_src = np.array([[0.5, 0.2, 0.8], [1.0, 1.0, 1.0],
[1.0, 0.6, 0.3]])
competence = probabilistic_test.estimate_competence(query)
assert np.allclose(competence, [0.665, 0.458, 0.855], atol=0.01)
def test_estimate_competence_batch():
n_samples = 10
query = np.ones((n_samples, 2))
probabilistic_test = Probabilistic(create_pool_classifiers())
probabilistic_test.distances = np.tile([0.5, 1.0, 2.0], (n_samples, 1))
probabilistic_test.neighbors = np.tile([0, 1, 2], (n_samples, 1))
probabilistic_test.DFP_mask = np.ones(
(n_samples, probabilistic_test.n_classifiers))
probabilistic_test.C_src = np.array([[0.5, 0.2, 0.8], [1.0, 1.0, 1.0],
[1.0, 0.6, 0.3]])
expected = np.tile([0.665, 0.458, 0.855], (n_samples, 1))
competence = probabilistic_test.estimate_competence(query)
assert np.allclose(competence, expected, atol=0.01)
def test_select_all_ones():
competences = np.ones(100)
probabilistic_test = Probabilistic(create_pool_all_agree(1, 100))
probabilistic_test.n_classes = 2
indices = probabilistic_test.select(competences)
assert indices == list(range(100))
def test_potential_function():
dists = np.array([1.0, 0.5, 2, 0.33])
value = Probabilistic.potential_func(dists)
assert np.allclose(value, [0.3679, 0.7788, 0.0183, 0.8968], atol=0.001)
def test_potential_function_zeros():
dists = np.zeros(10)
value = Probabilistic.potential_func(dists)
assert np.array_equal(value, np.ones(10))
def test_potential_function_batch():
dists = np.tile([1.0, 0.5, 2, 0.33], (10, 1))
value = Probabilistic.potential_func(dists)
expected = np.tile([0.3679, 0.7788, 0.0183, 0.8968], (10, 1))
assert np.allclose(value, expected, atol=0.001)
def test_select_all_zeros():
competences = np.zeros(100)
probabilistic_test = Probabilistic(create_pool_all_agree(1, 100))
probabilistic_test.n_classes = 2
selected_matrix = probabilistic_test.select(competences)
assert selected_matrix.all()
