def test_decreasing_mse_for_increased_partition_prior(self):
for data_matrix_transform in data_matrix_transforms:
mu = 0
sd_prior = 1
prior_obs = 0.01
kappa = prior_obs
alpha = prior_obs / 2
var_prior = sd_prior**2
tau_prior = 1 / var_prior
beta = alpha / tau_prior
prior = np.array([mu, kappa, alpha, beta])
x = np.linspace(-np.pi / 2, np.pi / 2, 20)
y = np.linspace(-np.pi / 2, np.pi / 2, 20)
X = np.array([x, y]).T
y = np.sin(x) + 3 * np.cos(y)
X = data_matrix_transform(X)
for i_model in range(len(create_regression_trees(prior, 0.5))):
mse_list = []
for partition_prior in [0.1, 0.5, 0.9, 0.99]:
model = create_regression_trees(prior,
partition_prior)[i_model]
print('Testing {}'.format(type(model).__name__))
model.fit(X, y)
print(model)
mse = mean_squared_error(y, model.predict(X))
mse_list.append(mse)
self.assertTrue(mse_list[-1] < mse_list[0])
for i in range(0, len(mse_list) - 1):
self.assertTrue(mse_list[i + 1] <= mse_list[i])
def test_cannot_predict_with_bad_input_dimensions(self):
mu = 0
sd_prior = 1
prior_obs = 0.01
kappa = prior_obs
alpha = prior_obs / 2
var_prior = sd_prior**2
tau_prior = 1 / var_prior
beta = alpha / tau_prior
prior = np.array([mu, kappa, alpha, beta])
for data_matrix_transform in data_matrix_transforms:
for model in create_regression_trees(prior, 0.5):
Xy = np.array([
[0.0, 0.0, 0],
[0.0, 1.0, 1],
[1.0, 1.0, 0],
[1.0, 0.0, 1],
[1.0, 0.0, 0],
])
X = Xy[:, :-1]
y = Xy[:, -1]
X = data_matrix_transform(X)
print('Testing {}'.format(type(model).__name__))
model.fit(X, y)
print(model)
model.predict([0, 0])
try:
model.predict(0)
self.fail()
except ValueError:
pass
try:
model.predict([0])
self.fail()
except ValueError:
pass
try:
model.predict([0, 0, 0])
self.fail()
except ValueError:
pass
def test_cannot_predict_before_training(self):
mu = 0
sd_prior = 1
prior_obs = 0.01
kappa = prior_obs
alpha = prior_obs / 2
var_prior = sd_prior**2
tau_prior = 1 / var_prior
beta = alpha / tau_prior
prior = np.array([mu, kappa, alpha, beta])
for model in create_regression_trees(prior, 0.5):
# can't predict yet
try:
model.predict([])
self.fail()
except ValueError:
pass
def test_no_split(self):
for data_matrix_transform in data_matrix_transforms:
mu = 0
sd_prior = 1
prior_obs = 0.01
kappa = prior_obs
alpha = prior_obs / 2
var_prior = sd_prior**2
tau_prior = 1 / var_prior
beta = alpha / tau_prior
prior = np.array([mu, kappa, alpha, beta])
for model in create_regression_trees(prior, 0.5):
Xy = np.array([
[0.0, 0.0, 0],
[0.1, 0.1, 1.3],
[0.9, 0.9, 0],
[1.0, 1.0, 1.2],
[1.0, 1.0, 0],
])
X = Xy[:, :-1]
y = Xy[:, -1]
X = data_matrix_transform(X)
print('Testing {}'.format(type(model).__name__))
model.fit(X, y)
print(model)
self.assertEqual(model.get_depth(), 0)
self.assertEqual(model.get_n_leaves(), 1)
self.assertEqual(model.n_data, 5)
self.assertIsNone(model.child1)
self.assertIsNone(model.child2)
if isinstance(model, PerpendicularRegressionTree):
self.assertEqual(model.split_dimension, -1)
self.assertEqual(model.split_value, None)
else:
self.assertEqual(model.best_hyperplane_origin, None)
n = len(y)
mean = y.mean()
mu, kappa, alpha, beta = prior
kappa_post = kappa + n
mu_post = (kappa * mu + n * mean) / kappa_post
expected = np.array([mu_post, mu_post, mu_post, mu_post])
self.assertEqual(model.predict([0.0, 0.5]),
np.expand_dims(expected[0], 0))
self.assertEqual(model.predict([0.49, 0.5]),
np.expand_dims(expected[1], 0))
self.assertEqual(model.predict([0.51, 0.5]),
np.expand_dims(expected[2], 0))
self.assertEqual(model.predict([1.0, 0.5]),
np.expand_dims(expected[3], 0))
for data_matrix_transform2 in data_matrix_transforms:
assert_array_equal(
model.predict(
data_matrix_transform2([[0.0, 0.5], [0.49, 0.5],
[0.51, 0.5], [1.0, 0.5]])),
expected)
def test_print_empty_model(self):
for model in create_regression_trees(np.array([1, 1]), 0.5):
print(model)
