def test_correct_predict(self):
X, y = generate_data()
voter = TreeClassificationVoter()
voter.fit(X, y)
y_hat = np.argmax(voter.predict_proba(X), axis=1)
testing.assert_allclose(y, y_hat)
def task_similarity(datax,
dataz,
transformer_kwargsx={'max_depth': 6},
transformer_kwargsz={'max_depth': 6},
balance=False,
acorn=None):
if acorn is not None:
np.random.seed(acorn)
n, d = datax[0].shape
m, p = dataz[0].shape
if balance:
min_nm = np.min([n, m])
x_inds = np.random.choice(np.arange(n), size=min_nm, replace=False)
z_inds = np.random.choice(np.arange(m), size=min_nm, replace=False)
datax = (datax[0][x_inds], datax[1][x_inds])
dataz = (dataz[0][z_inds], datay[1][z_inds])
# Initialize and fit transformers
transformerx = TreeClassificationTransformer(transformer_kwargsx)
transformerx.fit(*datax)
transformed_datax_x = transformerx.transform(datax[0])
transformerz = TreeClassificationTransformer(transformer_kwargsz)
transformerz.fit(*dataz)
transformed_datax_z = transformerz.transform(datax[0])
# Initialize and fit voters
classesx = np.unique(datax[1])
voterx = TreeClassificationVoter(classes=classesx)
voterx.fit(transformed_datax_x, datax[1])
voterz = TreeClassificationVoter(classes=classesx)
voterz.fit(transformed_datax_z, datax[1])
# Get predictions
yhatx = voterx.predict(transformed_datax_x)
yhatz = voterz.predict(transformed_datax_z)
task_similarity = np.mean(yhatx == yhatz)
return task_similarity
def test_predict_without_fit(self):
X, y = generate_data()
testing.assert_raises(NotFittedError,
TreeClassificationVoter().predict_proba, X)
def test_initialize(self):
TreeClassificationVoter()
assert True
def test_predict_without_fit(self):
X, y = generate_data()
with self.assertRaises(NotFittedError):
voter = TreeClassificationVoter()
voter.vote(X)