def test_estimate_competence_kuncheva_ex_batch():
# considering a batch composed of 10 samples
query = np.ones((10, 2))
a_posteriori_test = APosteriori([create_base_classifier(return_value=1)],
k=k_ex_kuncheva)
a_posteriori_test.fit(dsel_processed_kuncheva,
y_dsel_ex_kuncheva_dependent)
a_posteriori_test.DSEL_processed_ = dsel_processed_kuncheva
a_posteriori_test.dsel_scores_ = dsel_scores_ex_kuncheva
a_posteriori_test.n_classes_ = n_classes_ex_kuncheva
# repeating the same matrix in a new axis to simulate a batch input.
neighbors = np.tile(neighbors_ex_kuncheva, (10, 1))
distances = np.tile(distances_ex_kuncheva, (10, 1))
predictions = []
for clf in a_posteriori_test.pool_classifiers:
predictions.append(clf.predict(query)[0])
competences = a_posteriori_test.estimate_competence(query, neighbors,
distances,
predictions=np.array(
predictions))
assert np.allclose(competences, 0.95, atol=0.01)
def test_estimate_competence_kuncheva_ex():
query = np.array([1, 1])
a_posteriori_test = APosteriori([create_base_classifier(return_value=1)],
k=k_ex_kuncheva)
a_posteriori_test.processed_dsel = dsel_processed_kuncheva
a_posteriori_test.dsel_scores = dsel_scores_ex_kuncheva
a_posteriori_test.DSEL_target = y_dsel_ex_kuncheva_dependent
a_posteriori_test.n_classes = n_classes_ex_kuncheva
a_posteriori_test.neighbors = neighbors_ex_kuncheva
a_posteriori_test.distances = distances_ex_kuncheva
a_posteriori_test.DFP_mask = [1]
competences = a_posteriori_test.estimate_competence(query.reshape(1, -1))
assert np.isclose(competences, 0.95, atol=0.01)
def test_estimate_competence_all_ones(index):
query = np.array([1, 1])
a_posteriori_test = APosteriori(create_pool_classifiers())
a_posteriori_test.processed_dsel = dsel_processed_ex1
a_posteriori_test.dsel_scores = dsel_scores_all_ones
a_posteriori_test.DSEL_target = y_dsel_ex1
a_posteriori_test.n_classes = 2
a_posteriori_test.neighbors = neighbors_ex1[index, :]
a_posteriori_test.distances = distances_all_ones[index, :]
a_posteriori_test.DFP_mask = [1, 1, 1]
expected = [1.0, 1.0, 1.0]
competences = a_posteriori_test.estimate_competence(query.reshape(1, -1))
assert np.isclose(competences, expected).all()
def test_estimate_competence_kuncheva_ex(example_kuncheva):
query = np.atleast_2d([1, 1])
a_posteriori_test = APosteriori(k=example_kuncheva['k'])
a_posteriori_test.n_classifiers_ = 1
a_posteriori_test.DSEL_processed_ = example_kuncheva['dsel_processed']
a_posteriori_test.dsel_scores_ = example_kuncheva['dsel_scores']
a_posteriori_test.DSEL_target_ = example_kuncheva['y_dependent']
a_posteriori_test.n_classes_ = example_kuncheva['n_classes']
neighbors = example_kuncheva['neighbors'].reshape(1, -1)
distances = example_kuncheva['distances'].reshape(1, -1)
predictions = np.array([[1]])
competences = a_posteriori_test.estimate_competence(
query, neighbors, distances, predictions=np.array(predictions))
assert np.isclose(competences, 0.95, atol=0.01)
def test_estimate_competence_diff_target(index, example_all_ones):
_, _, neighbors, distances, dsel_processed, _ = example_all_ones
query = np.atleast_2d([1, 1])
a_posteriori_test = APosteriori()
a_posteriori_test.n_classifiers_ = 3
a_posteriori_test.DSEL_processed_ = dsel_processed
a_posteriori_test.dsel_scores_ = np.ones((15, 3, 3))
a_posteriori_test.DSEL_target_ = np.ones(15, dtype=int) * 2
a_posteriori_test.n_classes_ = 2
neighbors = neighbors[index, :].reshape(1, -1)
distances = distances[index, :].reshape(1, -1)
expected = [0.0, 0.0, 0.0]
predictions = np.array([0, 1, 0])
competences = a_posteriori_test.estimate_competence(
query, neighbors, distances, predictions=np.array(predictions))
assert np.isclose(competences, expected).all()
def test_estimate_competence_kuncheva_ex_batch(example_kuncheva):
# considering a batch composed of 10 samples
query = np.ones((10, 2))
a_posteriori_test = APosteriori(k=example_kuncheva['k'])
a_posteriori_test.fit(example_kuncheva['dsel_processed'],
example_kuncheva['y_dependent'])
a_posteriori_test.DSEL_processed_ = example_kuncheva['dsel_processed']
a_posteriori_test.dsel_scores_ = example_kuncheva['dsel_scores']
a_posteriori_test.n_classes_ = example_kuncheva['n_classes']
# repeating the same matrix in a new axis to simulate a batch input.
neighbors = np.tile(example_kuncheva['neighbors'], (10, 1))
distances = np.tile(example_kuncheva['distances'], (10, 1))
predictions = np.ones((1, 10))
competences = a_posteriori_test.estimate_competence(
query, neighbors, distances, predictions=np.array(predictions))
assert np.allclose(competences, 0.95, atol=0.01)
def test_estimate_competence_all_ones(index):
query = np.atleast_2d([1, 1])
a_posteriori_test = APosteriori(create_pool_classifiers())
a_posteriori_test.fit(X_dsel_ex1, y_dsel_ex1)
a_posteriori_test.DSEL_processed_ = dsel_processed_ex1
a_posteriori_test.dsel_scores_ = dsel_scores_all_ones
neighbors = neighbors_ex1[index, :].reshape(1, -1)
distances = distances_all_ones[index, :].reshape(1, -1)
expected = [1.0, 1.0, 1.0]
predictions = []
for clf in a_posteriori_test.pool_classifiers:
predictions.append(clf.predict(query)[0])
competences = a_posteriori_test.estimate_competence(
query, neighbors, distances, predictions=np.array(predictions))
assert np.isclose(competences, expected).all()
def test_estimate_competence_kuncheva_ex():
query = np.atleast_2d([1, 1])
a_posteriori_test = APosteriori([create_base_classifier(return_value=1)],
k=k_ex_kuncheva)
a_posteriori_test.processed_dsel = dsel_processed_kuncheva
a_posteriori_test.dsel_scores = dsel_scores_ex_kuncheva
a_posteriori_test.DSEL_target = y_dsel_ex_kuncheva_dependent
a_posteriori_test.n_classes = n_classes_ex_kuncheva
a_posteriori_test.neighbors = neighbors_ex_kuncheva
a_posteriori_test.distances = distances_ex_kuncheva
a_posteriori_test.DFP_mask = [1]
predictions = []
for clf in a_posteriori_test.pool_classifiers:
predictions.append(clf.predict(query)[0])
competences = a_posteriori_test.estimate_competence(
query, predictions=np.array(predictions))
assert np.isclose(competences, 0.95, atol=0.01)
def test_estimate_competence_diff_target(index):
query = np.atleast_2d([1, 1])
pool_classifiers = create_pool_classifiers()
a_posteriori_test = APosteriori(pool_classifiers=pool_classifiers)
a_posteriori_test.n_classifiers_ = len(pool_classifiers)
a_posteriori_test.DSEL_processed_ = dsel_processed_ex1
a_posteriori_test.dsel_scores_ = np.ones((15, 3, 3))
a_posteriori_test.DSEL_target_ = np.ones(15, dtype=int) * 2
a_posteriori_test.n_classes_ = 2
neighbors = neighbors_ex1[index, :].reshape(1, -1)
distances = distances_all_ones[index, :].reshape(1, -1)
expected = [0.0, 0.0, 0.0]
predictions = []
for clf in a_posteriori_test.pool_classifiers:
predictions.append(clf.predict(query)[0])
competences = a_posteriori_test.estimate_competence(
query, neighbors, distances, predictions=np.array(predictions))
assert np.isclose(competences, expected).all()
def test_estimate_competence_all_ones(index):
query = np.atleast_2d([1, 1])
a_posteriori_test = APosteriori(create_pool_classifiers())
a_posteriori_test.processed_dsel = dsel_processed_ex1
a_posteriori_test.dsel_scores = dsel_scores_all_ones
a_posteriori_test.DSEL_target = y_dsel_ex1
a_posteriori_test.n_classes = 2
a_posteriori_test.neighbors = neighbors_ex1[index, :]
a_posteriori_test.distances = distances_all_ones[index, :]
a_posteriori_test.DFP_mask = [1, 1, 1]
expected = [1.0, 1.0, 1.0]
predictions = []
for clf in a_posteriori_test.pool_classifiers:
predictions.append(clf.predict(query)[0])
competences = a_posteriori_test.estimate_competence(
query, predictions=np.array(predictions))
assert np.isclose(competences, expected).all()
def test_estimate_competence_kuncheva_ex_batch(example_kuncheva):
# considering a batch composed of 10 samples
query = np.ones((10, 2))
classifier = MagicMock()
classifier.predict.return_value = [1]
classifier.predict_proba.return_value = None
a_posteriori_test = APosteriori(pool_classifiers=classifier,
k=example_kuncheva['k'])
a_posteriori_test.n_classifiers_ = 1
a_posteriori_test.DSEL_processed_ = example_kuncheva['dsel_processed']
a_posteriori_test.DSEL_target_ = example_kuncheva['y_dependent']
a_posteriori_test.dsel_scores_ = example_kuncheva['dsel_scores']
a_posteriori_test.n_classes_ = example_kuncheva['n_classes']
# repeating the same matrix in a new axis to simulate a batch input.
neighbors = example_kuncheva['neighbors']
distances = example_kuncheva['distances']
predictions = [1]
competences = a_posteriori_test.estimate_competence(
query, neighbors, distances, predictions=np.array(predictions))
assert np.allclose(competences, 0.95, atol=0.01)