def test_classify_with_ds_diff_sizes(create_pool_classifiers):
query = np.ones((10, 2))
predictions = np.ones((5, 3))
des_test = BaseDES(create_pool_classifiers)
with pytest.raises(ValueError):
des_test.classify_with_ds(query, predictions)
def test_classify_instance_selection(create_pool_classifiers):
query = np.array([-1, 1])
pool_classifiers = create_pool_classifiers + create_pool_classifiers
des_test = BaseDES(pool_classifiers, mode='selection')
selected_index = np.array([[True, True, False, False, False, True]])
des_test.select = MagicMock(return_value=selected_index)
predictions = []
for clf in des_test.pool_classifiers:
predictions.append(clf.predict(query)[0])
predicted_label = des_test.classify_with_ds(query, np.array(predictions))
assert predicted_label == 0.0
def test_classify_instance_weighting(create_pool_classifiers):
query = np.array([-1, 1])
pool_classifiers = create_pool_classifiers + create_pool_classifiers
des_test = BaseDES(pool_classifiers, mode='weighting')
des_test.classes_ = np.array([0, 1])
des_test.n_classes_ = 2
competences = np.array([0.55, 1.0, 0.2, 0.60, 0.75, 0.3])
des_test.estimate_competence = MagicMock(return_value=competences)
predictions = []
for clf in des_test.pool_classifiers:
predictions.append(clf.predict(query)[0])
predicted_label = des_test.classify_with_ds(query, np.array(predictions))
assert predicted_label == 1.0
def test_classify_instance_selection_batch(create_pool_classifiers):
n_samples = 3
query = np.ones((n_samples, 2))
pool_classifiers = create_pool_classifiers + create_pool_classifiers
des_test = BaseDES(pool_classifiers, mode='selection')
selected_index = np.array(
[[True, True, False, False, False, True] * n_samples])
des_test.select = MagicMock(return_value=selected_index)
predictions = []
for clf in des_test.pool_classifiers:
predictions.append(clf.predict(query)[0])
predicted_label = des_test.classify_with_ds(query, np.tile(predictions,
(n_samples, 1)))
assert np.allclose(predicted_label, 0) and predicted_label.size == 3
def test_classify_instance_weighting(create_pool_classifiers):
n_samples = 3
query = np.ones((n_samples, 2))
pool_classifiers = create_pool_classifiers + create_pool_classifiers
des_test = BaseDES(pool_classifiers, mode='weighting')
des_test.classes_ = np.array([0, 1])
des_test.n_classes_ = 2
competences = np.tile([0.55, 1.0, 0.2, 0.60, 0.75, 0.3], (3, 1))
des_test.estimate_competence = MagicMock(return_value=competences)
predictions = []
for clf in des_test.pool_classifiers:
predictions.append(clf.predict(query)[0])
predicted_label = des_test.classify_with_ds(np.tile(predictions, (3, 1)))
assert np.allclose(predicted_label, 1) and predicted_label.size == 3
def test_classify_instance_hybrid(create_pool_classifiers):
query = np.array([-1, 1])
expected = 1
pool_classifiers = create_pool_classifiers + create_pool_classifiers
des_test = BaseDES(pool_classifiers, mode='hybrid')
des_test.classes_ = np.array([0, 1])
des_test.n_classes_ = 2
selected_indices = np.array([[True, True, False, False, False, True]])
competences = np.array([[0.55, 1.0, 0.2, 0.60, 0.75, 0.3]])
des_test.estimate_competence = MagicMock(return_value=competences)
des_test.select = MagicMock(return_value=selected_indices)
predictions = []
for clf in des_test.pool_classifiers:
predictions.append(clf.predict(query)[0])
predicted_label = des_test.classify_with_ds(query, np.array(predictions))
assert expected == predicted_label
def test_classify_instance_hybrid_batch():
query = np.ones((3, 2))
expected = 1
pool_classifiers = create_pool_classifiers() + create_pool_classifiers()
des_test = BaseDES(pool_classifiers, mode='hybrid')
des_test.classes_ = np.array([0, 1])
des_test.n_classes_ = 2
selected_indices = np.tile([True, True, False, False, False, True], (3, 1))
competences = np.tile([0.55, 1.0, 0.2, 0.60, 0.75, 0.3], (3, 1))
des_test.estimate_competence = MagicMock(return_value=competences)
des_test.select = MagicMock(return_value=selected_indices)
predictions = []
for clf in des_test.pool_classifiers:
predictions.append(clf.predict(query)[0])
predicted_label = des_test.classify_with_ds(query,
np.tile(predictions, (3, 1)))
assert np.allclose(predicted_label, expected)