def test_predict_proba_instance():
query = np.array([-1, 1])
n_classifiers = 3
pool_classifiers = create_pool_classifiers()
dcs_test = BaseDCS(pool_classifiers)
dcs_test.n_classes_ = 2
competences = np.random.rand(n_classifiers)
dcs_test.estimate_competence = MagicMock(return_value=competences)
expected = pool_classifiers[np.argmax(competences)].predict_proba(query)
predictions = []
probabilities = []
for clf in dcs_test.pool_classifiers:
predictions.append(clf.predict(query)[0])
probabilities.append(clf.predict_proba(query)[0])
query = np.atleast_2d(query)
predictions = np.atleast_2d(predictions)
probabilities = np.array(probabilities)
probabilities = np.expand_dims(probabilities, axis=0)
predicted_proba = dcs_test.predict_proba_with_ds(query, predictions,
probabilities)
assert np.array_equal(predicted_proba, expected)
def test_select_all_batch():
competences = np.array([[1.0, 1.0, 0.5], [0.8, 0.9, 0.9], [0.15, 0.15, 0.15], [0.0, 0.0, 0.0]])
expected = np.array([[True, True, False], [False, True, True], [True, True, True], [True, True, True]])
pool_classifiers = create_pool_classifiers()
dcs_test = BaseDCS(pool_classifiers, selection_method='all')
selected_clf = dcs_test.select(competences)
assert np.array_equal(selected_clf, expected)
def test_selection_method_type(selection_method, create_pool_classifiers):
X = np.random.rand(10, 2)
y = np.ones(10)
with pytest.raises(TypeError):
dcs = BaseDCS(create_pool_classifiers,
selection_method=selection_method)
dcs.fit(X, y)
def test_classify_with_ds_diff_sizes():
query = np.ones((10, 2))
predictions = np.ones((5, 3))
dcs_test = BaseDCS(create_pool_classifiers())
with pytest.raises(ValueError):
dcs_test.classify_with_ds(query, predictions)
def test_select_best_batch():
competences = np.array([[1.0, 0.5, 0.5], [0.8, 0.9, 1.0], [0.0, 0.0, 0.15]])
expected = [0, 2, 2]
pool_classifiers = create_pool_classifiers()
dcs_test = BaseDCS(pool_classifiers, selection_method='best')
selected_clf = dcs_test.select(competences)
assert np.array_equal(selected_clf, expected)
def test_valid_selection_method(selection_method):
X = np.random.rand(10, 2)
y = np.ones(10)
with pytest.raises(ValueError):
dcs = BaseDCS(create_pool_classifiers(),
selection_method=selection_method)
dcs.fit(X, y)
def test_select_diff(competences, expected):
rng = np.random.RandomState(123456)
dcs_test = BaseDCS(selection_method='diff',
diff_thresh=0.15,
random_state=rng)
selected_clf = dcs_test.select(np.array(competences))
assert np.allclose(selected_clf, expected)
def test_proba_with_ds_diff_sizes():
query = np.ones((10, 2))
predictions = np.ones((5, 3))
probabilities = np.ones((5, 3, 2))
dcs_test = BaseDCS()
with pytest.raises(ValueError):
dcs_test.predict_proba_with_ds(query, predictions, probabilities)
def test_select_random(competences, expected):
rng = np.random.RandomState(123456)
pool_classifiers = create_pool_classifiers()
dcs_test = BaseDCS(pool_classifiers, selection_method='random', random_state=rng)
dcs_test.fit(X_dsel_ex1, y_dsel_ex1)
selected_clf = dcs_test.select(np.array(competences))
assert np.allclose(selected_clf, expected)
def test_valid_diff_threshold_value(diff_thresh):
X = np.random.rand(10, 2)
y = np.ones(10)
with pytest.raises(ValueError):
dcs = BaseDCS(create_pool_classifiers(),
selection_method='diff',
diff_thresh=diff_thresh)
dcs.fit(X, y)
def test_select_random_batch():
competences = np.array([[0.5, 0.5, 0.5], [0.8, 0.9, 1.0], [0.0, 0.10,
0.0]])
expected = np.array([1, 2, 1])
rng = np.random.RandomState(123456)
dcs_test = BaseDCS(selection_method='random', random_state=rng)
selected_clf = dcs_test.select(competences)
assert np.array_equal(selected_clf, expected)
def test_select_random_batch():
competences = np.array([[0.5, 0.5, 0.5], [0.8, 0.9, 1.0], [0.0, 0.10, 0.0]])
expected = np.array([1, 2, 1])
rng = np.random.RandomState(123456)
pool_classifiers = create_pool_classifiers()
dcs_test = BaseDCS(pool_classifiers, selection_method='random', random_state=rng)
dcs_test.fit(X_dsel_ex1, y_dsel_ex1)
selected_clf = dcs_test.select(competences)
assert np.array_equal(selected_clf, expected)
def test_classify_instance_all(competences, expected):
query = np.array([-1, 1])
pool_classifiers = create_pool_classifiers()
dcs_test = BaseDCS(pool_classifiers, selection_method='all')
dcs_test.estimate_competence = MagicMock(return_value=np.array(competences))
predictions = []
for clf in dcs_test.pool_classifiers:
predictions.append(clf.predict(query)[0])
predicted_label = dcs_test.classify_with_ds(query, np.array(predictions))
assert predicted_label == expected
def test_select_diff_batch():
competences = np.array([[1.0, 0.5, 0.5], [0.8, 0.9, 1.0], [0.0, 0.0,
0.15]])
expected = np.array([0, 2, 2])
rng = np.random.RandomState(123456)
dcs_test = BaseDCS(selection_method='diff',
diff_thresh=0.15,
random_state=rng)
selected_clf = dcs_test.select(competences)
assert np.array_equal(selected_clf, expected)
def test_classify_instance_all_batch():
competences = np.array([[0.6, 0.2, 0.6], [0.5, 0.8, 0.5]])
expected = [0, 1]
n_samples = 2
query = np.ones((n_samples, 2))
pool_classifiers = create_pool_classifiers()
dcs_test = BaseDCS(pool_classifiers, selection_method='all')
dcs_test.estimate_competence = MagicMock(return_value=np.array(competences))
predictions = []
for clf in dcs_test.pool_classifiers:
predictions.append(clf.predict(query)[0])
predicted_label = dcs_test.classify_with_ds(query, np.tile(predictions, (n_samples, 1)))
assert np.array_equal(predicted_label, expected)
def test_classify_instance():
query = np.array([-1, 1])
n_classifiers = 3
pool_classifiers = create_pool_classifiers()
dcs_test = BaseDCS(pool_classifiers)
competences = np.random.rand(n_classifiers)
dcs_test.estimate_competence = MagicMock(return_value=competences)
expected = pool_classifiers[np.argmax(competences)].predict(query)[0]
predictions = []
for clf in dcs_test.pool_classifiers:
predictions.append(clf.predict(query)[0])
predicted_label = dcs_test.classify_with_ds(query, np.array(predictions))
assert predicted_label == expected
def test_classify_instance_batch():
n_samples = 3
n_classifiers = 3
query = np.ones((n_samples, 2))
pool_classifiers = create_pool_classifiers()
dcs_test = BaseDCS(pool_classifiers)
competences = np.random.rand(n_samples, n_classifiers)
dcs_test.estimate_competence = MagicMock(return_value=competences)
expected = []
for ind in range(n_samples):
expected.append(pool_classifiers[np.argmax(competences[ind, :])].predict(query)[0])
predictions = []
for clf in dcs_test.pool_classifiers:
predictions.append(clf.predict(query)[0])
predicted_label = dcs_test.classify_with_ds(query, np.tile(predictions, (3, 1)))
assert np.array_equal(predicted_label, expected)
def test_predict_proba_instance_all(competences, expected):
query = np.array([-1, 1])
pool_classifiers = create_pool_classifiers()
dcs_test = BaseDCS(pool_classifiers, selection_method='all')
dcs_test.n_classes_ = 2
dcs_test.estimate_competence = MagicMock(return_value=np.array(competences))
predictions = []
probabilities = []
for clf in dcs_test.pool_classifiers:
predictions.append(clf.predict(query)[0])
probabilities.append(clf.predict_proba(query)[0])
query = np.atleast_2d(query)
predictions = np.atleast_2d(predictions)
probabilities = np.array(probabilities)
probabilities = np.expand_dims(probabilities, axis=0)
predicted_proba = dcs_test.predict_proba_with_ds(query, predictions, probabilities)
assert np.isclose(predicted_proba, expected).all()
def test_select_all(competences, expected):
pool_classifiers = create_pool_classifiers()
dcs_test = BaseDCS(pool_classifiers, selection_method='all')
selected_clf = dcs_test.select(np.array(competences))
assert np.allclose(selected_clf, expected)
