def test_iris():
# Check consistency on dataset iris.
classes = np.unique(iris.target)
clf_samme = prob_samme = None
for alg in ['SAMME', 'SAMME.R']:
clf = AdaBoostClassifier(algorithm=alg)
clf.fit(iris.data, iris.target)
assert_array_equal(classes, clf.classes_)
proba = clf.predict_proba(iris.data)
if alg == "SAMME":
clf_samme = clf
prob_samme = proba
assert_equal(proba.shape[1], len(classes))
assert_equal(clf.decision_function(iris.data).shape[1], len(classes))
score = clf.score(iris.data, iris.target)
assert score > 0.9, "Failed with algorithm %s and score = %f" % \
(alg, score)
# Somewhat hacky regression test: prior to
# ae7adc880d624615a34bafdb1d75ef67051b8200,
# predict_proba returned SAMME.R values for SAMME.
clf_samme.algorithm = "SAMME.R"
assert_array_less(0,
np.abs(clf_samme.predict_proba(iris.data) - prob_samme))
def test_graph_lasso(random_state=0):
# Sample data from a sparse multivariate normal
dim = 20
n_samples = 100
random_state = check_random_state(random_state)
prec = make_sparse_spd_matrix(dim, alpha=.95,
random_state=random_state)
cov = linalg.inv(prec)
X = random_state.multivariate_normal(np.zeros(dim), cov, size=n_samples)
emp_cov = empirical_covariance(X)
for alpha in (.1, .01):
covs = dict()
for method in ('cd', 'lars'):
cov_, _, costs = graph_lasso(emp_cov, alpha=.1, return_costs=True)
covs[method] = cov_
costs, dual_gap = np.array(costs).T
# Check that the costs always decrease
assert_array_less(np.diff(costs), 0)
# Check that the 2 approaches give similar results
assert_array_almost_equal(covs['cd'], covs['lars'])
# Smoke test the estimator
model = GraphLasso(alpha=.1).fit(X)
assert_array_almost_equal(model.covariance_, covs['cd'])
def test_normalize_option_multilabel_classification():
# Test in the multilabel case
n_classes = 4
n_samples = 100
# for both random_state 0 and 1, y_true and y_pred has at least one
# unlabelled entry
_, y_true = make_multilabel_classification(n_features=1,
n_classes=n_classes,
random_state=0,
allow_unlabeled=True,
n_samples=n_samples)
_, y_pred = make_multilabel_classification(n_features=1,
n_classes=n_classes,
random_state=1,
allow_unlabeled=True,
n_samples=n_samples)
# To make sure at least one empty label is present
y_true += [0]*n_classes
y_pred += [0]*n_classes
for name in METRICS_WITH_NORMALIZE_OPTION:
metrics = ALL_METRICS[name]
measure = metrics(y_true, y_pred, normalize=True)
assert_array_less(-1.0 * measure, 0,
err_msg="We failed to test correctly the normalize "
"option")
assert_allclose(metrics(y_true, y_pred, normalize=False) / n_samples,
measure, err_msg="Failed with %s" % name)
def test_radius_neighbors():
# Checks whether Returned distances are less than `radius`
# At least one point should be returned when the `radius` is set
# to mean distance from the considering point to other points in
# the database.
# Moreover, this test compares the radius neighbors of LSHForest
# with the `sklearn.neighbors.NearestNeighbors`.
n_samples = 12
n_features = 2
n_iter = 10
rng = np.random.RandomState(42)
X = rng.rand(n_samples, n_features)
lshf = ignore_warnings(LSHForest, category=DeprecationWarning)()
# Test unfitted estimator
assert_raises(ValueError, lshf.radius_neighbors, X[0])
ignore_warnings(lshf.fit)(X)
for i in range(n_iter):
# Select a random point in the dataset as the query
query = X[rng.randint(0, n_samples)].reshape(1, -1)
# At least one neighbor should be returned when the radius is the
# mean distance from the query to the points of the dataset.
mean_dist = np.mean(pairwise_distances(query, X, metric='cosine'))
neighbors = lshf.radius_neighbors(query, radius=mean_dist,
return_distance=False)
assert_equal(neighbors.shape, (1,))
assert_equal(neighbors.dtype, object)
assert_greater(neighbors[0].shape[0], 0)
# All distances to points in the results of the radius query should
# be less than mean_dist
distances, neighbors = lshf.radius_neighbors(query,
radius=mean_dist,
return_distance=True)
assert_array_less(distances[0], mean_dist)
# Multiple points
n_queries = 5
queries = X[rng.randint(0, n_samples, n_queries)]
distances, neighbors = lshf.radius_neighbors(queries,
return_distance=True)
# dists and inds should not be 1D arrays or arrays of variable lengths
# hence the use of the object dtype.
assert_equal(distances.shape, (n_queries,))
assert_equal(distances.dtype, object)
assert_equal(neighbors.shape, (n_queries,))
assert_equal(neighbors.dtype, object)
# Compare with exact neighbor search
query = X[rng.randint(0, n_samples)].reshape(1, -1)
mean_dist = np.mean(pairwise_distances(query, X, metric='cosine'))
nbrs = NearestNeighbors(algorithm='brute', metric='cosine').fit(X)
distances_exact, _ = nbrs.radius_neighbors(query, radius=mean_dist)
distances_approx, _ = lshf.radius_neighbors(query, radius=mean_dist)
def test_predict_rank_normalized(self):
pred_socres = self.clf.decision_function(self.X_test)
pred_ranks = self.clf._predict_rank(self.X_test, normalized=True)
# assert the order is reserved
assert_allclose(rankdata(pred_ranks), rankdata(pred_socres), atol=3)
assert_array_less(pred_ranks, 1.01)
assert_array_less(-0.1, pred_ranks)
def test_predict_rank(self):
pred_socres = self.clf.decision_function(self.X_test)
pred_ranks = self.clf._predict_rank(self.X_test)
# assert the order is reserved
assert_allclose(rankdata(pred_ranks), rankdata(pred_socres), atol=3)
assert_array_less(pred_ranks, self.X_train.shape[0] + 1)
assert_array_less(-0.1, pred_ranks)
def test_predict_rank_normalized(self):
pred_socres = self.clf.decision_function(self.X_test)
pred_ranks = self.clf._predict_rank(self.X_test, normalized=True)
# assert the order is reserved
assert_allclose(rankdata(pred_ranks), rankdata(pred_socres), atol=2)
assert_array_less(pred_ranks, 1.01)
assert_array_less(-0.1, pred_ranks)
def test_predict_rank(self):
pred_socres = self.clf.decision_function(self.X_test)
pred_ranks = self.clf._predict_rank(self.X_test)
# assert the order is reserved
assert_allclose(rankdata(pred_ranks), rankdata(pred_socres), atol=2)
assert_array_less(pred_ranks, self.X_train.shape[0] + 1)
assert_array_less(-0.1, pred_ranks)
def test_radius_neighbors():
"""Checks whether Returned distances are less than `radius`
At least one point should be returned when the `radius` is set
to mean distance from the considering point to other points in
the database.
Moreover, this test compares the radius neighbors of LSHForest
with the `sklearn.neighbors.NearestNeighbors`.
"""
n_samples = 12
n_features = 2
n_iter = 10
rng = np.random.RandomState(42)
X = rng.rand(n_samples, n_features)
lshf = LSHForest()
# Test unfitted estimator
assert_raises(ValueError, lshf.radius_neighbors, X[0])
lshf.fit(X)
for i in range(n_iter):
query = X[rng.randint(0, n_samples)]
mean_dist = np.mean(pairwise_distances(query, X, metric='cosine'))
neighbors = lshf.radius_neighbors(query,
radius=mean_dist,
return_distance=False)
# At least one neighbor should be returned.
assert_greater(neighbors.shape[0], 0)
# All distances should be less than mean_dist
distances, neighbors = lshf.radius_neighbors(query,
radius=mean_dist,
return_distance=True)
assert_array_less(distances[0], mean_dist)
# Multiple points
n_queries = 5
queries = X[rng.randint(0, n_samples, n_queries)]
distances, neighbors = lshf.radius_neighbors(queries, return_distance=True)
assert_equal(neighbors.shape[0], n_queries)
assert_equal(distances.shape[0], n_queries)
# dists and inds should not be 2D arrays
assert_equal(distances.ndim, 1)
assert_equal(neighbors.ndim, 1)
# Compare with exact neighbor search
query = X[rng.randint(0, n_samples)]
mean_dist = np.mean(pairwise_distances(query, X, metric='cosine'))
nbrs = NearestNeighbors(algorithm='brute', metric='cosine')
nbrs.fit(X)
distances_approx, _ = lshf.radius_neighbors(query, radius=mean_dist)
distances_exact, _ = nbrs.radius_neighbors(query, radius=mean_dist)
# Distances of exact neighbors is less than or equal to approximate
assert_true(
np.all(
np.less_equal(np.sort(distances_exact[0]),
np.sort(distances_approx[0]))))
def test_api():
res = dummy_minimize(
branin, [(-5.0, 10.0), (0.0, 15.0)], random_state=0, maxiter=100)
assert_array_equal(res.x.shape, (2,))
assert_array_equal(res.x_iters.shape, (100, 2))
assert_array_equal(res.func_vals.shape, (100,))
assert_array_less(res.x_iters, np.tile([10, 15], (100, 1)))
assert_array_less(np.tile([-5, 0], (100, 1)), res.x_iters)
assert_raises(ValueError, dummy_minimize, lambda x: x, [[-5, 10]])
def test_api():
res = gp_minimize(branin, [[-5, 10], [0, 15]], random_state=0, maxiter=20)
assert_array_equal(res.x.shape, (2, ))
assert_array_equal(res.x_iters.shape, (20, 2))
assert_array_equal(res.func_vals.shape, (20, ))
assert_array_less(res.x_iters, np.tile([10, 15], (20, 1)))
assert_array_less(np.tile([-5, 0], (20, 1)), res.x_iters)
assert_raises(ValueError, gp_minimize, lambda x: x, [[-5, 10]])
def test_radius_neighbors():
"""Checks whether Returned distances are less than `radius`
At least one point should be returned when the `radius` is set
to mean distance from the considering point to other points in
the database.
Moreover, this test compares the radius neighbors of LSHForest
with the `sklearn.neighbors.NearestNeighbors`.
"""
n_samples = 12
n_features = 2
n_iter = 10
rng = np.random.RandomState(42)
X = rng.rand(n_samples, n_features)
lshf = LSHForest()
# Test unfitted estimator
assert_raises(ValueError, lshf.radius_neighbors, X[0])
lshf.fit(X)
for i in range(n_iter):
query = X[rng.randint(0, n_samples)]
mean_dist = np.mean(pairwise_distances(query, X, metric='cosine'))
neighbors = lshf.radius_neighbors(query, radius=mean_dist,
return_distance=False)
# At least one neighbor should be returned.
assert_greater(neighbors.shape[0], 0)
# All distances should be less than mean_dist
distances, neighbors = lshf.radius_neighbors(query,
radius=mean_dist,
return_distance=True)
assert_array_less(distances[0], mean_dist)
# Multiple points
n_queries = 5
queries = X[rng.randint(0, n_samples, n_queries)]
distances, neighbors = lshf.radius_neighbors(queries,
return_distance=True)
assert_equal(neighbors.shape[0], n_queries)
assert_equal(distances.shape[0], n_queries)
# dists and inds should not be 2D arrays
assert_equal(distances.ndim, 1)
assert_equal(neighbors.ndim, 1)
# Compare with exact neighbor search
query = X[rng.randint(0, n_samples)]
mean_dist = np.mean(pairwise_distances(query, X, metric='cosine'))
nbrs = NearestNeighbors(algorithm='brute', metric='cosine')
nbrs.fit(X)
distances_approx, _ = lshf.radius_neighbors(query, radius=mean_dist)
distances_exact, _ = nbrs.radius_neighbors(query, radius=mean_dist)
# Distances of exact neighbors is less than or equal to approximate
assert_true(np.all(np.less_equal(np.sort(distances_exact[0]),
np.sort(distances_approx[0]))))
def test_solution_inside_bounds():
gpr = VGP(kernel=kernel).fit(X, y)
bounds = gpr.kernel.bounds
max_ = np.finfo(gpr.kernel.theta.dtype).max
tiny = 1e-10
bounds[~np.isfinite(bounds[:, 1]), 1] = max_
assert_array_less(bounds[:, 0], gpr.kernel.theta + tiny)
assert_array_less(gpr.kernel.theta, bounds[:, 1] + tiny)
def test_solution_inside_bounds(kernel):
# Test that hyperparameter-optimization remains in bounds#
gpr = GaussianProcessRegressor(kernel=kernel).fit(X, y)
bounds = gpr.kernel_.bounds
max_ = np.finfo(gpr.kernel_.theta.dtype).max
tiny = 1e-10
bounds[~np.isfinite(bounds[:, 1]), 1] = max_
assert_array_less(bounds[:, 0], gpr.kernel_.theta + tiny)
assert_array_less(gpr.kernel_.theta, bounds[:, 1] + tiny)
def test_solution_inside_bounds(kernel):
# Test that hyperparameter-optimization remains in bounds#
gpr = GaussianProcessRegressor(kernel=kernel).fit(X, y)
bounds = gpr.kernel_.bounds
max_ = np.finfo(gpr.kernel_.theta.dtype).max
tiny = 1e-10
bounds[~np.isfinite(bounds[:, 1]), 1] = max_
assert_array_less(bounds[:, 0], gpr.kernel_.theta + tiny)
assert_array_less(gpr.kernel_.theta, bounds[:, 1] + tiny)
def test_normalize_option_multiclass_classification(name):
# Test in the multiclass case
random_state = check_random_state(0)
y_true = random_state.randint(0, 4, size=(20, ))
y_pred = random_state.randint(0, 4, size=(20, ))
n_samples = y_true.shape[0]
metrics = ALL_METRICS[name]
measure = metrics(y_true, y_pred, normalize=True)
assert_array_less(-1.0 * measure, 0,
err_msg="We failed to test correctly the normalize "
"option")
assert_allclose(metrics(y_true, y_pred, normalize=False) / n_samples,
measure)
def test_normalize_option_multiclass_classification(name):
# Test in the multiclass case
random_state = check_random_state(0)
y_true = random_state.randint(0, 4, size=(20, ))
y_pred = random_state.randint(0, 4, size=(20, ))
n_samples = y_true.shape[0]
metrics = ALL_METRICS[name]
measure = metrics(y_true, y_pred, normalize=True)
assert_array_less(-1.0 * measure, 0,
err_msg="We failed to test correctly the normalize "
"option")
assert_allclose(metrics(y_true, y_pred, normalize=False) / n_samples,
measure)
def test_std_bayesian_ridge_ard_with_constant_input():
# Test BayesianRidge and ARDRegression standard dev. for edge case of
# constant target vector
# The standard dev. should be relatively small (< 0.01 is tested here)
n_samples = 4
n_features = 5
random_state = check_random_state(42)
constant_value = random_state.rand()
X = random_state.random_sample((n_samples, n_features))
y = np.full(n_samples, constant_value)
expected_upper_boundary = 0.01
for clf in [BayesianRidge(), ARDRegression()]:
_, y_std = clf.fit(X, y).predict(X, return_std=True)
assert_array_less(y_std, expected_upper_boundary)
def test_class_weights():
# check that the class weights are updated
# simple 3 cluster dataset
X, y = make_blobs(random_state=1)
for Model in [DPGMM, VBGMM]:
dpgmm = Model(n_components=10, random_state=1, alpha=20, n_iter=50)
dpgmm.fit(X)
# get indices of components that are used:
indices = np.unique(dpgmm.predict(X))
active = np.zeros(10, dtype=np.bool)
active[indices] = True
# used components are important
assert_array_less(.1, dpgmm.weights_[active])
# others are not
assert_array_less(dpgmm.weights_[~active], .05)
def test_mem_vec_diff_clusters():
"""
Verify membership vector produces as many clusters as requested
"""
# Ignore user warnings in this function
warnings.filterwarnings("ignore", category=UserWarning)
# Given a flat clustering trained for n_clusters picked by HDBSCAN,
n_clusters_fit = None
clusterer = HDBSCAN_flat(X, n_clusters=n_clusters_fit)
n_clusters_fitted = n_clusters_from_labels(clusterer.labels_)
# When membership_vector_flat is called with new data for some n_clusters,
n_clusters_predict = n_clusters_fitted + 3
memberships = membership_vector_flat(clusterer,
X_test,
n_clusters=n_clusters_predict)
# Then the number of clusters in memberships should be as requested,
assert_equal(memberships.shape[1], n_clusters_predict)
# and the number of points should equal those in the test set
assert_equal(len(memberships), len(X_test))
# and all probabilities are <= 1.
assert_array_less(memberships, np.ones(memberships.shape) + 1.e-14)
# ========================================
# Given a flat clustering for a specified n_clusters,
n_clusters_fit = n_clusters_from_labels(clusterer.labels_) + 2
clusterer = HDBSCAN_flat(X, n_clusters=n_clusters_fit)
# When membership_vector_flat is called with new data for some n_clusters,
n_clusters_predict = n_clusters_fit + 3
memberships = membership_vector_flat(clusterer,
X_test,
n_clusters=n_clusters_predict)
# Then the number of clusters in memberships should be as requested,
assert_equal(memberships.shape[1], n_clusters_predict)
# and the number of points should equal those in the test set
assert_equal(len(memberships), len(X_test))
# and all probabilities are <= 1.
assert_array_less(memberships, np.ones(memberships.shape) + 1.e-14)
return
def test_explained_variance(X_sparse, kind, n_components, solver):
X = X_sparse if kind == 'sparse' else X_sparse.toarray()
svd = TruncatedSVD(n_components, algorithm=solver)
X_tr = svd.fit_transform(X)
# Assert that all the values are greater than 0
assert_array_less(0.0, svd.explained_variance_ratio_)
# Assert that total explained variance is less than 1
assert_array_less(svd.explained_variance_ratio_.sum(), 1.0)
# Test that explained_variance is correct
total_variance = np.var(X_sparse.toarray(), axis=0).sum()
variances = np.var(X_tr, axis=0)
true_explained_variance_ratio = variances / total_variance
assert_allclose(
svd.explained_variance_ratio_,
true_explained_variance_ratio,
)
def test_graphical_lasso(random_state=0):
# Sample data from a sparse multivariate normal
dim = 20
n_samples = 100
random_state = check_random_state(random_state)
prec = make_sparse_spd_matrix(dim, alpha=.95, random_state=random_state)
cov = linalg.inv(prec)
X = random_state.multivariate_normal(np.zeros(dim), cov, size=n_samples)
emp_cov = empirical_covariance(X)
for alpha in (0., .1, .25):
covs = dict()
icovs = dict()
for method in ('cd', 'lars'):
cov_, icov_, costs = graphical_lasso(emp_cov,
return_costs=True,
alpha=alpha,
mode=method)
covs[method] = cov_
icovs[method] = icov_
costs, dual_gap = np.array(costs).T
# Check that the costs always decrease (doesn't hold if alpha == 0)
if not alpha == 0:
assert_array_less(np.diff(costs), 0)
# Check that the 2 approaches give similar results
assert_array_almost_equal(covs['cd'], covs['lars'], decimal=4)
assert_array_almost_equal(icovs['cd'], icovs['lars'], decimal=4)
# Smoke test the estimator
model = GraphicalLasso(alpha=.25).fit(X)
model.score(X)
assert_array_almost_equal(model.covariance_, covs['cd'], decimal=4)
assert_array_almost_equal(model.covariance_, covs['lars'], decimal=4)
# For a centered matrix, assume_centered could be chosen True or False
# Check that this returns indeed the same result for centered data
Z = X - X.mean(0)
precs = list()
for assume_centered in (False, True):
prec_ = GraphicalLasso(
assume_centered=assume_centered).fit(Z).precision_
precs.append(prec_)
assert_array_almost_equal(precs[0], precs[1])
def test_graphical_lasso(random_state=0):
# Sample data from a sparse multivariate normal
dim = 20
n_samples = 100
random_state = check_random_state(random_state)
prec = make_sparse_spd_matrix(dim, alpha=.95,
random_state=random_state)
cov = linalg.inv(prec)
X = random_state.multivariate_normal(np.zeros(dim), cov, size=n_samples)
emp_cov = empirical_covariance(X)
for alpha in (0., .1, .25):
covs = dict()
icovs = dict()
for method in ('cd', 'lars'):
cov_, icov_, costs = graphical_lasso(emp_cov, return_costs=True,
alpha=alpha, mode=method)
covs[method] = cov_
icovs[method] = icov_
costs, dual_gap = np.array(costs).T
# Check that the costs always decrease (doesn't hold if alpha == 0)
if not alpha == 0:
assert_array_less(np.diff(costs), 0)
# Check that the 2 approaches give similar results
assert_array_almost_equal(covs['cd'], covs['lars'], decimal=4)
assert_array_almost_equal(icovs['cd'], icovs['lars'], decimal=4)
# Smoke test the estimator
model = GraphicalLasso(alpha=.25).fit(X)
model.score(X)
assert_array_almost_equal(model.covariance_, covs['cd'], decimal=4)
assert_array_almost_equal(model.covariance_, covs['lars'], decimal=4)
# For a centered matrix, assume_centered could be chosen True or False
# Check that this returns indeed the same result for centered data
Z = X - X.mean(0)
precs = list()
for assume_centered in (False, True):
prec_ = GraphicalLasso(
assume_centered=assume_centered).fit(Z).precision_
precs.append(prec_)
assert_array_almost_equal(precs[0], precs[1])
def test_approx_predict_same_clusters():
"""
Verify that approximate_predict_flat produces as many clusters as clusterer
"""
# Given a flat clustering trained for some n_clusters,
n_clusters = 5
clusterer = HDBSCAN_flat(X,
cluster_selection_method='eom',
n_clusters=n_clusters)
# When using approximate_predict_flat without specifying n_clusters,
labels_flat, proba_flat = approximate_predict_flat(clusterer,
X_test,
n_clusters=None)
# Then, the number of clusters produced must match the original n_clusters
n_clusters_out = n_clusters_from_labels(labels_flat)
assert_equal(n_clusters_out, n_clusters)
# and all probabilities are <= 1.
assert_array_less(proba_flat, np.ones(len(proba_flat)) + 1.e-14)
return
def test_approx_predict_diff_clusters():
"""
Verify that approximate_predict_flat produces as many clusters as asked
"""
# Given a flat clustering trained for some n_clusters,
n_clusters_fit = 5
clusterer = HDBSCAN_flat(X,
cluster_selection_method='eom',
n_clusters=n_clusters_fit,
prediction_data=True)
# When using approximate_predict_flat with specified n_clusters,
n_clusters_predict = 3
labels_flat, proba_flat = approximate_predict_flat(
clusterer, X_test, n_clusters=n_clusters_predict)
# Then, the requested number of clusters must be produced
n_clusters_out = n_clusters_from_labels(labels_flat)
assert_equal(n_clusters_out, n_clusters_predict)
# and all probabilities are <= 1.
assert_array_less(proba_flat, np.ones(len(proba_flat)) + 1.e-14)
# When using approximate_predict_flat with more clusters
# than 'eom' can handle,
n_clusters_predict = 12
with warnings.catch_warnings(record=True) as w:
labels_flat, proba_flat = approximate_predict_flat(
clusterer, X_test, n_clusters=n_clusters_predict)
# Then, a warning is raised saying 'eom' can't get this clustering,
assert len(w) > 0
assert issubclass(w[-1].category, UserWarning)
assert "Cannot predict" in str(w[-1].message)
# But the requested number of clusters must still be produced using 'leaf'
n_clusters_out = n_clusters_from_labels(labels_flat)
assert_equal(n_clusters_out, n_clusters_predict)
# and all probabilities are <= 1.
assert_array_less(proba_flat, np.ones(len(proba_flat)) + 1.e-14)
return
def test_all_points_mem_vec_same_clusters():
"""
Verify membership vector for training set produces same n_clusters
as clusterer
"""
# Given a flat clustering trained for n_clusters picked by HDBSCAN,
n_clusters_fit = None
clusterer = HDBSCAN_flat(X, n_clusters=n_clusters_fit)
# When all_points_membership_vectors_flat is called,
memberships = all_points_membership_vectors_flat(clusterer)
# Then the number of clusters in memberships matches those of clusterer,
assert_equal(memberships.shape[1],
n_clusters_from_labels(clusterer.labels_))
# and the number of points should equal those in the training set
assert_equal(len(memberships), len(X))
# and all probabilities are <= 1.
assert_array_less(memberships, np.ones(memberships.shape) + 1.e-14)
# ========================================
# Given a flat clustering for a specified n_clusters,
n_clusters_fit = n_clusters_from_labels(clusterer.labels_) - 2
clusterer = HDBSCAN_flat(X, n_clusters=n_clusters_fit)
# When all_points_membership_vectors_flat is called,
memberships = all_points_membership_vectors_flat(clusterer)
# Then the number of clusters in memberships matches those of clusterer,
assert_equal(memberships.shape[1],
n_clusters_from_labels(clusterer.labels_))
# and the number of points should equal those in the training set
assert_equal(len(memberships), len(X))
# and all probabilities are <= 1.
assert_array_less(memberships, np.ones(memberships.shape) + 1.e-14)
return
def test_graph_lasso(random_state=0):
# Sample data from a sparse multivariate normal
dim = 20
n_samples = 100
random_state = check_random_state(random_state)
prec = make_sparse_spd_matrix(dim, alpha=.95, random_state=random_state)
cov = linalg.inv(prec)
X = random_state.multivariate_normal(np.zeros(dim), cov, size=n_samples)
emp_cov = empirical_covariance(X)
for alpha in (.1, .01):
covs = dict()
for method in ('cd', 'lars'):
cov_, _, costs = graph_lasso(emp_cov, alpha=.1, return_costs=True)
covs[method] = cov_
costs, dual_gap = np.array(costs).T
# Check that the costs always decrease
assert_array_less(np.diff(costs), 0)
# Check that the 2 approaches give similar results
assert_array_almost_equal(covs['cd'], covs['lars'])
# Smoke test the estimator
model = GraphLasso(alpha=.1).fit(X)
assert_array_almost_equal(model.covariance_, covs['cd'])
def test_forest_interface():
forest, tree1, tree2, i1, i2, x, y, y_ = generate_dummy_forest()
i = _tree_apply(tree1, x)
assert_array_equal(i, i1)
i = _tree_apply(tree2, x)
assert_array_equal(i, i2)
i, ti, ni = _compute_coverage_matrix(forest, x, max_nodecount=np.inf, max_interaction=np.inf)
i_ref, _ = _unique_rows(np.hstack([i1, i2]).T)
i_ref = i_ref[::-1].T
assert_array_equal(i, i_ref)
assert_array_equal(np.unique(ti), [0, 1]) # two trees
assert_array_less(-1, ni)
assert_array_less(ni[ti == 0], tree1.node_count)
assert_array_less(ni[ti == 1], tree2.node_count)
for i, f in enumerate(_tree_featurecount(tree1)):
assert f <= min(2, i)
for i, f in enumerate(_tree_featurecount(tree2)):
assert f <= min(2, i)
def test_radius_neighbors():
# Checks whether Returned distances are less than `radius`
# At least one point should be returned when the `radius` is set
# to mean distance from the considering point to other points in
# the database.
# Moreover, this test compares the radius neighbors of LSHForest
# with the `sklearn.neighbors.NearestNeighbors`.
n_samples = 12
n_features = 2
n_iter = 10
rng = np.random.RandomState(42)
X = rng.rand(n_samples, n_features)
lshf = ignore_warnings(LSHForest, category=DeprecationWarning)()
# Test unfitted estimator
assert_raises(ValueError, lshf.radius_neighbors, X[0])
ignore_warnings(lshf.fit)(X)
for i in range(n_iter):
# Select a random point in the dataset as the query
query = X[rng.randint(0, n_samples)].reshape(1, -1)
# At least one neighbor should be returned when the radius is the
# mean distance from the query to the points of the dataset.
mean_dist = np.mean(pairwise_distances(query, X, metric='cosine'))
neighbors = lshf.radius_neighbors(query, radius=mean_dist,
return_distance=False)
assert_equal(neighbors.shape, (1,))
assert_equal(neighbors.dtype, object)
assert_greater(neighbors[0].shape[0], 0)
# All distances to points in the results of the radius query should
# be less than mean_dist
distances, neighbors = lshf.radius_neighbors(query,
radius=mean_dist,
return_distance=True)
assert_array_less(distances[0], mean_dist)
# Multiple points
n_queries = 5
queries = X[rng.randint(0, n_samples, n_queries)]
distances, neighbors = lshf.radius_neighbors(queries,
return_distance=True)
# dists and inds should not be 1D arrays or arrays of variable lengths
# hence the use of the object dtype.
assert_equal(distances.shape, (n_queries,))
assert_equal(distances.dtype, object)
assert_equal(neighbors.shape, (n_queries,))
assert_equal(neighbors.dtype, object)
# Compare with exact neighbor search
query = X[rng.randint(0, n_samples)].reshape(1, -1)
mean_dist = np.mean(pairwise_distances(query, X, metric='cosine'))
nbrs = NearestNeighbors(algorithm='brute', metric='cosine').fit(X)
distances_exact, _ = nbrs.radius_neighbors(query, radius=mean_dist)
distances_approx, _ = lshf.radius_neighbors(query, radius=mean_dist)
# Radius-based queries do not sort the result points and the order
# depends on the method, the random_state and the dataset order. Therefore
# we need to sort the results ourselves before performing any comparison.
sorted_dists_exact = np.sort(distances_exact[0])
sorted_dists_approx = np.sort(distances_approx[0])
# Distances to exact neighbors are less than or equal to approximate
# counterparts as the approximate radius query might have missed some
# closer neighbors.
assert_true(np.all(np.less_equal(sorted_dists_exact,
sorted_dists_approx)))
def test_explained_variance():
# Test sparse data
svd_a_10_sp = TruncatedSVD(10, algorithm="arpack")
svd_r_10_sp = TruncatedSVD(10, algorithm="randomized", random_state=42)
svd_a_20_sp = TruncatedSVD(20, algorithm="arpack")
svd_r_20_sp = TruncatedSVD(20, algorithm="randomized", random_state=42)
X_trans_a_10_sp = svd_a_10_sp.fit_transform(X)
X_trans_r_10_sp = svd_r_10_sp.fit_transform(X)
X_trans_a_20_sp = svd_a_20_sp.fit_transform(X)
X_trans_r_20_sp = svd_r_20_sp.fit_transform(X)
# Test dense data
svd_a_10_de = TruncatedSVD(10, algorithm="arpack")
svd_r_10_de = TruncatedSVD(10, algorithm="randomized", random_state=42)
svd_a_20_de = TruncatedSVD(20, algorithm="arpack")
svd_r_20_de = TruncatedSVD(20, algorithm="randomized", random_state=42)
X_trans_a_10_de = svd_a_10_de.fit_transform(X.toarray())
X_trans_r_10_de = svd_r_10_de.fit_transform(X.toarray())
X_trans_a_20_de = svd_a_20_de.fit_transform(X.toarray())
X_trans_r_20_de = svd_r_20_de.fit_transform(X.toarray())
# helper arrays for tests below
svds = (svd_a_10_sp, svd_r_10_sp, svd_a_20_sp, svd_r_20_sp, svd_a_10_de,
svd_r_10_de, svd_a_20_de, svd_r_20_de)
svds_trans = (
(svd_a_10_sp, X_trans_a_10_sp),
(svd_r_10_sp, X_trans_r_10_sp),
(svd_a_20_sp, X_trans_a_20_sp),
(svd_r_20_sp, X_trans_r_20_sp),
(svd_a_10_de, X_trans_a_10_de),
(svd_r_10_de, X_trans_r_10_de),
(svd_a_20_de, X_trans_a_20_de),
(svd_r_20_de, X_trans_r_20_de),
)
svds_10_v_20 = (
(svd_a_10_sp, svd_a_20_sp),
(svd_r_10_sp, svd_r_20_sp),
(svd_a_10_de, svd_a_20_de),
(svd_r_10_de, svd_r_20_de),
)
svds_sparse_v_dense = (
(svd_a_10_sp, svd_a_10_de),
(svd_a_20_sp, svd_a_20_de),
(svd_r_10_sp, svd_r_10_de),
(svd_r_20_sp, svd_r_20_de),
)
# Assert the 1st component is equal
for svd_10, svd_20 in svds_10_v_20:
assert_array_almost_equal(
svd_10.explained_variance_ratio_,
svd_20.explained_variance_ratio_[:10],
decimal=5,
)
# Assert that 20 components has higher explained variance than 10
for svd_10, svd_20 in svds_10_v_20:
assert_greater(
svd_20.explained_variance_ratio_.sum(),
svd_10.explained_variance_ratio_.sum(),
)
# Assert that all the values are greater than 0
for svd in svds:
assert_array_less(0.0, svd.explained_variance_ratio_)
# Assert that total explained variance is less than 1
for svd in svds:
assert_array_less(svd.explained_variance_ratio_.sum(), 1.0)
# Compare sparse vs. dense
for svd_sparse, svd_dense in svds_sparse_v_dense:
assert_array_almost_equal(svd_sparse.explained_variance_ratio_,
svd_dense.explained_variance_ratio_)
# Test that explained_variance is correct
for svd, transformed in svds_trans:
total_variance = np.var(X.toarray(), axis=0).sum()
variances = np.var(transformed, axis=0)
true_explained_variance_ratio = variances / total_variance
assert_array_almost_equal(
svd.explained_variance_ratio_,
true_explained_variance_ratio,
)
def test_radius_neighbors():
# Checks whether Returned distances are less than `radius`
# At least one point should be returned when the `radius` is set
# to mean distance from the considering point to other points in
# the database.
# Moreover, this test compares the radius neighbors of LSHForest
# with the `sklearn.neighbors.NearestNeighbors`.
n_samples = 12
n_features = 2
n_iter = 10
rng = np.random.RandomState(42)
X = rng.rand(n_samples, n_features)
lshf = LSHForest()
# Test unfitted estimator
assert_raises(ValueError, lshf.radius_neighbors, X[0])
ignore_warnings(lshf.fit)(X)
for i in range(n_iter):
# Select a random point in the dataset as the query
query = X[rng.randint(0, n_samples)].reshape(1, -1)
# At least one neighbor should be returned when the radius is the
# mean distance from the query to the points of the dataset.
mean_dist = np.mean(pairwise_distances(query, X, metric='cosine'))
neighbors = lshf.radius_neighbors(query, radius=mean_dist,
return_distance=False)
assert_equal(neighbors.shape, (1,))
assert_equal(neighbors.dtype, object)
assert_greater(neighbors[0].shape[0], 0)
# All distances to points in the results of the radius query should
# be less than mean_dist
distances, neighbors = lshf.radius_neighbors(query,
radius=mean_dist,
return_distance=True)
assert_array_less(distances[0], mean_dist)
# Multiple points
n_queries = 5
queries = X[rng.randint(0, n_samples, n_queries)]
distances, neighbors = lshf.radius_neighbors(queries,
return_distance=True)
# dists and inds should not be 1D arrays or arrays of variable lengths
# hence the use of the object dtype.
assert_equal(distances.shape, (n_queries,))
assert_equal(distances.dtype, object)
assert_equal(neighbors.shape, (n_queries,))
assert_equal(neighbors.dtype, object)
# Compare with exact neighbor search
query = X[rng.randint(0, n_samples)].reshape(1, -1)
mean_dist = np.mean(pairwise_distances(query, X, metric='cosine'))
nbrs = NearestNeighbors(algorithm='brute', metric='cosine').fit(X)
distances_exact, _ = nbrs.radius_neighbors(query, radius=mean_dist)
distances_approx, _ = lshf.radius_neighbors(query, radius=mean_dist)
# Radius-based queries do not sort the result points and the order
# depends on the method, the random_state and the dataset order. Therefore
# we need to sort the results ourselves before performing any comparison.
sorted_dists_exact = np.sort(distances_exact[0])
sorted_dists_approx = np.sort(distances_approx[0])
# Distances to exact neighbors are less than or equal to approximate
# counterparts as the approximate radius query might have missed some
# closer neighbors.
assert_true(np.all(np.less_equal(sorted_dists_exact,
sorted_dists_approx)))
def check_minimizer_bounds(result, n_calls):
# no values should be below or above the bounds
eps = 10e-9 # check for assert_array_less OR equal
assert_array_less(result.x_iters, np.tile([10+eps, 15+eps], (n_calls, 1)))
assert_array_less(np.tile([-5-eps, 0-eps], (n_calls, 1)), result.x_iters)
def test_explained_variance():
# Test sparse data
svd_a_10_sp = TruncatedSVD(10, algorithm="arpack")
svd_r_10_sp = TruncatedSVD(10, algorithm="randomized", random_state=42)
svd_a_20_sp = TruncatedSVD(20, algorithm="arpack")
svd_r_20_sp = TruncatedSVD(20, algorithm="randomized", random_state=42)
X_trans_a_10_sp = svd_a_10_sp.fit_transform(X)
X_trans_r_10_sp = svd_r_10_sp.fit_transform(X)
X_trans_a_20_sp = svd_a_20_sp.fit_transform(X)
X_trans_r_20_sp = svd_r_20_sp.fit_transform(X)
# Test dense data
svd_a_10_de = TruncatedSVD(10, algorithm="arpack")
svd_r_10_de = TruncatedSVD(10, algorithm="randomized", random_state=42)
svd_a_20_de = TruncatedSVD(20, algorithm="arpack")
svd_r_20_de = TruncatedSVD(20, algorithm="randomized", random_state=42)
X_trans_a_10_de = svd_a_10_de.fit_transform(X.toarray())
X_trans_r_10_de = svd_r_10_de.fit_transform(X.toarray())
X_trans_a_20_de = svd_a_20_de.fit_transform(X.toarray())
X_trans_r_20_de = svd_r_20_de.fit_transform(X.toarray())
# helper arrays for tests below
svds = (svd_a_10_sp, svd_r_10_sp, svd_a_20_sp, svd_r_20_sp, svd_a_10_de,
svd_r_10_de, svd_a_20_de, svd_r_20_de)
svds_trans = (
(svd_a_10_sp, X_trans_a_10_sp),
(svd_r_10_sp, X_trans_r_10_sp),
(svd_a_20_sp, X_trans_a_20_sp),
(svd_r_20_sp, X_trans_r_20_sp),
(svd_a_10_de, X_trans_a_10_de),
(svd_r_10_de, X_trans_r_10_de),
(svd_a_20_de, X_trans_a_20_de),
(svd_r_20_de, X_trans_r_20_de),
)
svds_10_v_20 = (
(svd_a_10_sp, svd_a_20_sp),
(svd_r_10_sp, svd_r_20_sp),
(svd_a_10_de, svd_a_20_de),
(svd_r_10_de, svd_r_20_de),
)
svds_sparse_v_dense = (
(svd_a_10_sp, svd_a_10_de),
(svd_a_20_sp, svd_a_20_de),
(svd_r_10_sp, svd_r_10_de),
(svd_r_20_sp, svd_r_20_de),
)
# Assert the 1st component is equal
for svd_10, svd_20 in svds_10_v_20:
assert_array_almost_equal(
svd_10.explained_variance_ratio_,
svd_20.explained_variance_ratio_[:10],
decimal=5,
)
# Assert that 20 components has higher explained variance than 10
for svd_10, svd_20 in svds_10_v_20:
assert_greater(
svd_20.explained_variance_ratio_.sum(),
svd_10.explained_variance_ratio_.sum(),
)
# Assert that all the values are greater than 0
for svd in svds:
assert_array_less(0.0, svd.explained_variance_ratio_)
# Assert that total explained variance is less than 1
for svd in svds:
assert_array_less(svd.explained_variance_ratio_.sum(), 1.0)
# Compare sparse vs. dense
for svd_sparse, svd_dense in svds_sparse_v_dense:
assert_array_almost_equal(svd_sparse.explained_variance_ratio_,
svd_dense.explained_variance_ratio_)
# Test that explained_variance is correct
for svd, transformed in svds_trans:
total_variance = np.var(X.toarray(), axis=0).sum()
variances = np.var(transformed, axis=0)
true_explained_variance_ratio = variances / total_variance
assert_array_almost_equal(
svd.explained_variance_ratio_,
true_explained_variance_ratio,
)
def check_minimizer_bounds(result):
# no values should be below or above the bounds
eps = 10e-9 # check for assert_array_less OR equal
assert_array_less(result.x_iters, np.tile([10 + eps, 15 + eps], (7, 1)))
assert_array_less(np.tile([-5 - eps, 0 - eps], (7, 1)), result.x_iters)
