def test_binary():
Y1_prob = np.array([[0.9, 0.1], [0.2, 0.8], [0.3, 0.7]])
Y2_prob = np.array([[0.7, 0.3], [0.9, 0.1], [0.4, 0.6]])
Y_expected = np.array([0, 0, 1])
est1 = MockEstimator(Y1_prob)
est2 = MockEstimator(Y2_prob)
voting_classifier = WellcomeVotingClassifier(estimators=[est1, est2],
voting="soft")
X = ["mock", "data", "not used"]
Y = voting_classifier.predict(X)
assert np.array_equal(Y, Y_expected)
def test_multilabel():
Y1_prob = np.array([[0.9, 0.5], [0.2, 0.8], [0.3, 0.9]])
Y2_prob = np.array([[0.7, 0.6], [0.6, 0.5], [0.4, 0.3]])
Y_expected = np.array([[1, 1], [0, 1], [0, 1]])
est1 = MockEstimator(Y1_prob, multilabel=True)
est2 = MockEstimator(Y2_prob, multilabel=True)
voting_classifier = WellcomeVotingClassifier(estimators=[est1, est2],
voting="soft",
multilabel=True)
X = ["mock", "data", "not used"]
Y = voting_classifier.predict(X)
assert np.array_equal(Y, Y_expected)
def test_multiclass():
Y1_prob = np.array([[0.6, 0.1, 0.1, 0.2], [0.2, 0.3, 0.1, 0.4],
[0.3, 0.1, 0.4, 0.2]])
Y2_prob = np.array([[0.5, 0.3, 0.2, 0.0], [0.1, 0.6, 0.1, 0.2],
[0.1, 0.3, 0.1, 0.5]])
Y_expected = np.array([0, 1, 3])
est1 = MockEstimator(Y1_prob)
est2 = MockEstimator(Y2_prob)
voting_classifier = WellcomeVotingClassifier(estimators=[est1, est2],
voting="soft")
X = ["mock", "data", "not used"]
Y = voting_classifier.predict(X)
assert np.array_equal(Y, Y_expected)
def test_inc_vectorizer():
Y1_prob = np.array([[0.9, 0.1], [0.2, 0.8], [0.3, 0.7]])
Y2_prob = np.array([[0.7, 0.3], [0.9, 0.1], [0.4, 0.6]])
Y_expected = np.array([0, 0, 1])
pipe1 = Pipeline([('vect', MockVectorizer(size=3)),
('est', MockEstimator(Y1_prob))])
pipe2 = Pipeline([('vect', MockVectorizer(size=5)),
('est', MockEstimator(Y2_prob))])
voting_classifier = WellcomeVotingClassifier(estimators=[pipe1, pipe2],
voting="soft")
X = ["mock", "data", "not used"]
Y = voting_classifier.predict(X)
assert np.array_equal(Y, Y_expected)
def test_multilabel_hard():
# Since no num_agree, majority (>=2/3) agreement needed
Y1_prob = np.array([[0.9, 0.9], [0.1, 0.9], [0.1, 0.9]])
Y2_prob = np.array([[0.9, 0.9], [0.9, 0.9], [0.1, 0.1]])
Y3_prob = np.array([[0.9, 0.9], [0.9, 0.1], [0.1, 0.1]])
Y_expected = np.array([[1, 1], [1, 1], [0, 0]])
est1 = MockEstimator(Y1_prob, multilabel=True)
est2 = MockEstimator(Y2_prob, multilabel=True)
est3 = MockEstimator(Y3_prob, multilabel=True)
voting_classifier = WellcomeVotingClassifier(estimators=[est1, est2, est3],
voting="hard",
multilabel=True)
X = ["mock", "data", "not used"]
Y = voting_classifier.predict(X)
assert np.array_equal(Y, Y_expected)
def test_binary_hard_num_agree():
Y1_prob = np.array([[0.9, 0.1], [0.9, 0.1], [0.9, 0.1]])
Y2_prob = np.array([[0.8, 0.2], [0.8, 0.2], [0.2, 0.8]])
Y3_prob = np.array([[0.7, 0.3], [0.3, 0.7], [0.3, 0.7]])
Y4_prob = np.array([[0.4, 0.6], [0.4, 0.6], [0.4, 0.6]])
Y_expected = np.array([0, 0, 1])
est1 = MockEstimator(Y1_prob)
est2 = MockEstimator(Y2_prob)
est3 = MockEstimator(Y3_prob)
est4 = MockEstimator(Y4_prob)
voting_classifier = WellcomeVotingClassifier(
estimators=[est1, est2, est3, est4], voting="hard", num_agree=2)
X = ["mock", "data", "not used"]
Y = voting_classifier.predict(X)
assert np.array_equal(Y, Y_expected)
def load(self, model_paths):
self.estimators = []
for model_path in model_paths.split(","):
estimator = self._load(model_path)
self.estimators.append(estimator)
self.model = WellcomeVotingClassifier(estimators=self.estimators,
voting="soft",
multilabel=True)
def test_multiclass_hard_num_agree():
Y1_prob = np.array([[0.6, 0.1, 0.1, 0.2], [0.2, 0.3, 0.1, 0.4],
[0.3, 0.1, 0.4, 0.2]])
Y2_prob = np.array([[0.5, 0.3, 0.2, 0.0], [0.1, 0.6, 0.1, 0.2],
[0.1, 0.3, 0.1, 0.5]])
Y3_prob = np.array([[0.5, 0.3, 0.2, 0.0], [0.2, 0.3, 0.1, 0.4],
[0.1, 0.5, 0.1, 0.3]])
# Using highest values:
# 1st data point: 0: 3 votes -> clear winner is 0
# 2nd data point: 1: 1 vote, 3: 2 votes -> no winner, pick first in numerical order, so 1
# 3rd data point: 2: 1 vote, 3: 1 vote, 1: 1 vote -> no winner, pick first in numerical order, so 1 # noqa
Y_expected = np.array([0, 1, 1])
est1 = MockEstimator(Y1_prob)
est2 = MockEstimator(Y2_prob)
est3 = MockEstimator(Y3_prob)
voting_classifier = WellcomeVotingClassifier(estimators=[est1, est2, est3],
voting="hard",
num_agree=3)
X = ["mock", "data", "not used"]
Y = voting_classifier.predict(X)
assert np.array_equal(Y, Y_expected)
X = ["One two", "One", "Three and two", "Three"]
Y = [[1, 1, 0], [1, 0, 0], [0, 1, 1], [0, 0, 1]]
vec = CountVectorizer()
vec.fit(X)
X_vec = vec.transform(X)
sgd = OneVsRestClassifier(SGDClassifier(loss="log"))
nb = OneVsRestClassifier(MultinomialNB())
sgd.fit(X_vec, Y)
nb.fit(X_vec, Y)
voting_classifier = WellcomeVotingClassifier(estimators=[sgd, nb],
voting="soft",
multilabel=True)
Y_pred = voting_classifier.predict(X_vec)
print(Y_pred)
Y = [1, 0, 1, 0]
sgd = SGDClassifier(loss="log")
nb = MultinomialNB()
sgd.fit(X_vec, Y)
nb.fit(X_vec, Y)
voting_classifier = WellcomeVotingClassifier(estimators=[sgd, nb],
voting="soft")