def test_minibatch_k_means_init(data, init):
mb_k_means = MiniBatchKMeans(init=init,
n_clusters=n_clusters,
random_state=42,
n_init=10)
mb_k_means.fit(data)
_check_fitted_model(mb_k_means)
def test_int_input():
X_list = [[0, 0], [10, 10], [12, 9], [-1, 1], [2, 0], [8, 10]]
for dtype in [np.int32, np.int64]:
X_int = np.array(X_list, dtype=dtype)
X_int_csr = sp.csr_matrix(X_int)
init_int = X_int[:2]
fitted_models = [
KMeans(n_clusters=2).fit(X_int),
KMeans(n_clusters=2, init=init_int, n_init=1).fit(X_int),
# mini batch kmeans is very unstable on such a small dataset hence
# we use many inits
MiniBatchKMeans(n_clusters=2, n_init=10, batch_size=2).fit(X_int),
MiniBatchKMeans(n_clusters=2, n_init=10,
batch_size=2).fit(X_int_csr),
MiniBatchKMeans(n_clusters=2,
batch_size=2,
init=init_int,
n_init=1).fit(X_int),
MiniBatchKMeans(n_clusters=2,
batch_size=2,
init=init_int,
n_init=1).fit(X_int_csr),
]
for km in fitted_models:
assert km.cluster_centers_.dtype == np.float64
expected_labels = [0, 1, 1, 0, 0, 1]
scores = np.array([
v_measure_score(expected_labels, km.labels_)
for km in fitted_models
])
assert_array_almost_equal(scores, np.ones(scores.shape[0]))
def test_predict_minibatch_dense_sparse(init):
# check that models trained on sparse input also works for dense input at
# predict time
mb_k_means = MiniBatchKMeans(n_clusters=n_clusters,
init=init,
n_init=10,
random_state=0).fit(X_csr)
assert_array_equal(mb_k_means.predict(X), mb_k_means.labels_)
def test_mini_batch_k_means_random_init_partial_fit():
km = MiniBatchKMeans(n_clusters=n_clusters, init="random", random_state=42)
# use the partial_fit API for online learning
for X_minibatch in np.array_split(X, 10):
km.partial_fit(X_minibatch)
# compute the labeling on the complete dataset
labels = km.predict(X)
assert v_measure_score(true_labels, labels) == 1.0
def test_mb_kmeans_verbose():
mb_k_means = MiniBatchKMeans(init="k-means++",
n_clusters=n_clusters,
random_state=42,
verbose=1)
old_stdout = sys.stdout
sys.stdout = StringIO()
try:
mb_k_means.fit(X)
finally:
sys.stdout = old_stdout
def test_minibatch_sensible_reassign_partial_fit():
zeroed_X, true_labels = make_blobs(n_samples=n_samples,
centers=5,
cluster_std=1.,
random_state=42)
zeroed_X[::2, :] = 0
mb_k_means = MiniBatchKMeans(n_clusters=20, random_state=42, init="random")
for i in range(100):
mb_k_means.partial_fit(zeroed_X)
# there should not be too many exact zero cluster centers
assert mb_k_means.cluster_centers_.any(axis=1).sum() > 10
def test_minibatch_reassign():
# Give a perfect initialization, but a large reassignment_ratio,
# as a result all the centers should be reassigned and the model
# should no longer be good
sample_weight = np.ones(X.shape[0], dtype=X.dtype)
for this_X in (X, X_csr):
mb_k_means = MiniBatchKMeans(n_clusters=n_clusters,
batch_size=100,
random_state=42)
mb_k_means.fit(this_X)
score_before = mb_k_means.score(this_X)
try:
old_stdout = sys.stdout
sys.stdout = StringIO()
# Turn on verbosity to smoke test the display code
_mini_batch_step(this_X,
sample_weight, (X**2).sum(axis=1),
mb_k_means.cluster_centers_,
mb_k_means.counts_,
np.zeros(X.shape[1], np.double),
False,
distances=np.zeros(X.shape[0]),
random_reassign=True,
random_state=42,
reassignment_ratio=1,
verbose=True)
finally:
sys.stdout = old_stdout
assert score_before > mb_k_means.score(this_X)
# Give a perfect initialization, with a small reassignment_ratio,
# no center should be reassigned
for this_X in (X, X_csr):
mb_k_means = MiniBatchKMeans(n_clusters=n_clusters,
batch_size=100,
init=centers.copy(),
random_state=42,
n_init=1)
mb_k_means.fit(this_X)
clusters_before = mb_k_means.cluster_centers_
# Turn on verbosity to smoke test the display code
_mini_batch_step(this_X,
sample_weight, (X**2).sum(axis=1),
mb_k_means.cluster_centers_,
mb_k_means.counts_,
np.zeros(X.shape[1], np.double),
False,
distances=np.zeros(X.shape[0]),
random_reassign=True,
random_state=42,
reassignment_ratio=1e-15)
assert_array_almost_equal(clusters_before, mb_k_means.cluster_centers_)
def test_minibatch_sensible_reassign_fit():
# check if identical initial clusters are reassigned
# also a regression test for when there are more desired reassignments than
# samples.
zeroed_X, true_labels = make_blobs(n_samples=100,
centers=5,
cluster_std=1.,
random_state=42)
zeroed_X[::2, :] = 0
mb_k_means = MiniBatchKMeans(n_clusters=20,
batch_size=10,
random_state=42,
init="random")
mb_k_means.fit(zeroed_X)
# there should not be too many exact zero cluster centers
assert mb_k_means.cluster_centers_.any(axis=1).sum() > 10
# do the same with batch-size > X.shape[0] (regression test)
mb_k_means = MiniBatchKMeans(n_clusters=20,
batch_size=201,
random_state=42,
init="random")
mb_k_means.fit(zeroed_X)
# there should not be too many exact zero cluster centers
assert mb_k_means.cluster_centers_.any(axis=1).sum() > 10
def test_sparse_mb_k_means_callable_init():
def test_init(X, k, random_state):
return centers
# Small test to check that giving the wrong number of centers
# raises a meaningful error
msg = "does not match the number of clusters"
with pytest.raises(ValueError, match=msg):
MiniBatchKMeans(init=test_init, random_state=42).fit(X_csr)
# Now check that the fit actually works
mb_k_means = MiniBatchKMeans(n_clusters=3, init=test_init,
random_state=42).fit(X_csr)
_check_fitted_model(mb_k_means)
def test_minibatch_default_init_size():
mb_k_means = MiniBatchKMeans(init=centers.copy(),
n_clusters=n_clusters,
batch_size=10,
random_state=42,
n_init=1).fit(X)
assert mb_k_means.init_size_ == 3 * mb_k_means.batch_size
_check_fitted_model(mb_k_means)
def test_minibatch_set_init_size():
mb_k_means = MiniBatchKMeans(init=centers.copy(),
n_clusters=n_clusters,
init_size=666,
random_state=42,
n_init=1).fit(X)
assert mb_k_means.init_size == 666
assert mb_k_means.init_size_ == n_samples
_check_fitted_model(mb_k_means)
def test_scaled_weights():
# scaling all sample weights by a common factor
# shouldn't change the result
sample_weight = np.ones(n_samples)
for estimator in [
KMeans(n_clusters=n_clusters, random_state=42),
MiniBatchKMeans(n_clusters=n_clusters, random_state=42)
]:
est_1 = clone(estimator).fit(X)
est_2 = clone(estimator).fit(X, sample_weight=0.5 * sample_weight)
assert_almost_equal(v_measure_score(est_1.labels_, est_2.labels_), 1.0)
assert_almost_equal(_sort_centers(est_1.cluster_centers_),
_sort_centers(est_2.cluster_centers_))
def test_unit_weights_vs_no_weights():
# not passing any sample weights should be equivalent
# to all weights equal to one
sample_weight = np.ones(n_samples)
for estimator in [
KMeans(n_clusters=n_clusters, random_state=42),
MiniBatchKMeans(n_clusters=n_clusters, random_state=42)
]:
est_1 = clone(estimator).fit(X)
est_2 = clone(estimator).fit(X, sample_weight=sample_weight)
assert_almost_equal(v_measure_score(est_1.labels_, est_2.labels_), 1.0)
assert_almost_equal(_sort_centers(est_1.cluster_centers_),
_sort_centers(est_2.cluster_centers_))
def test_minibatch_with_many_reassignments():
# Test for the case that the number of clusters to reassign is bigger
# than the batch_size
n_samples = 550
rnd = np.random.RandomState(42)
X = rnd.uniform(size=(n_samples, 10))
# Check that the fit works if n_clusters is bigger than the batch_size.
# Run the test with 550 clusters and 550 samples, because it turned out
# that this values ensure that the number of clusters to reassign
# is always bigger than the batch_size
n_clusters = 550
MiniBatchKMeans(n_clusters=n_clusters,
batch_size=100,
init_size=n_samples,
random_state=42).fit(X)
def test_weighted_vs_repeated():
# a sample weight of N should yield the same result as an N-fold
# repetition of the sample
rng = np.random.RandomState(0)
sample_weight = rng.randint(1, 5, size=n_samples)
X_repeat = np.repeat(X, sample_weight, axis=0)
estimators = [
KMeans(init="k-means++", n_clusters=n_clusters, random_state=42),
KMeans(init="random", n_clusters=n_clusters, random_state=42),
KMeans(init=centers.copy(), n_clusters=n_clusters, random_state=42),
MiniBatchKMeans(n_clusters=n_clusters, batch_size=10, random_state=42)
]
for estimator in estimators:
est_weighted = clone(estimator).fit(X, sample_weight=sample_weight)
est_repeated = clone(estimator).fit(X_repeat)
repeated_labels = np.repeat(est_weighted.labels_, sample_weight)
assert_almost_equal(
v_measure_score(est_repeated.labels_, repeated_labels), 1.0)
if not isinstance(estimator, MiniBatchKMeans):
assert_almost_equal(_sort_centers(est_weighted.cluster_centers_),
_sort_centers(est_repeated.cluster_centers_))
def test_minibatch_kmeans_partial_fit_int_data():
# Issue GH #14314
X = np.array([[-1], [1]], dtype=np.int)
km = MiniBatchKMeans(n_clusters=2)
km.partial_fit(X)
assert km.cluster_centers_.dtype.kind == "f"
km = factory(n_clusters=n_clusters, init=init, random_state=run_id,
n_init=n_init, **params).fit(X)
inertia[i, run_id] = km.inertia_
p = plt.errorbar(n_init_range, inertia.mean(axis=1), inertia.std(axis=1))
plots.append(p[0])
legends.append("%s with %s init" % (factory.__name__, init))
plt.xlabel('n_init')
plt.ylabel('inertia')
plt.legend(plots, legends)
plt.title("Mean inertia for various k-means init across %d runs" % n_runs)
# Part 2: Qualitative visual inspection of the convergence
X, y = make_data(random_state, n_samples_per_center, grid_size, scale)
km = MiniBatchKMeans(n_clusters=n_clusters, init='random', n_init=1,
random_state=random_state).fit(X)
plt.figure()
for k in range(n_clusters):
my_members = km.labels_ == k
color = cm.nipy_spectral(float(k) / n_clusters, 1)
plt.plot(X[my_members, 0], X[my_members, 1], 'o', marker='.', c=color)
cluster_center = km.cluster_centers_[k]
plt.plot(cluster_center[0], cluster_center[1], 'o',
markerfacecolor=color, markeredgecolor='k', markersize=6)
plt.title("Example cluster allocation with a single random init\n"
"with MiniBatchKMeans")
plt.show()
'alt.atheism', 'comp.graphics', 'comp.sys.ibm.pc.hardware',
'comp.sys.mac.hardware', 'comp.windows.x', 'misc.forsale', 'rec.autos',
'rec.motorcycles', 'rec.sport.baseball', 'rec.sport.hockey', 'sci.crypt',
'sci.electronics', 'sci.med', 'sci.space', 'soc.religion.christian',
'talk.politics.guns', 'talk.politics.mideast', 'talk.politics.misc',
'talk.religion.misc'
]
newsgroups = fetch_20newsgroups(categories=categories)
y_true = newsgroups.target
vectorizer = NumberNormalizingVectorizer(stop_words='english', min_df=5)
cocluster = SpectralCoclustering(n_clusters=len(categories),
svd_method='arpack',
random_state=0)
kmeans = MiniBatchKMeans(n_clusters=len(categories),
batch_size=20000,
random_state=0)
print("Vectorizing...")
X = vectorizer.fit_transform(newsgroups.data)
print("Coclustering...")
start_time = time()
cocluster.fit(X)
y_cocluster = cocluster.row_labels_
print("Done in {:.2f}s. V-measure: {:.4f}".format(
time() - start_time, v_measure_score(y_cocluster, y_true)))
print("MiniBatchKMeans...")
start_time = time()
y_kmeans = kmeans.fit_predict(X)
def test_minibatch_k_means_init_multiple_runs_with_explicit_centers():
mb_k_means = MiniBatchKMeans(init=centers.copy(),
n_clusters=n_clusters,
random_state=42,
n_init=10)
assert_warns(RuntimeWarning, mb_k_means.fit, X)
def test_minibatch_tol():
mb_k_means = MiniBatchKMeans(n_clusters=n_clusters,
batch_size=10,
random_state=42,
tol=.01).fit(X)
_check_fitted_model(mb_k_means)
X, labels_true = make_blobs(n_samples=3000, centers=centers, cluster_std=0.7)
# #############################################################################
# Compute clustering with Means
k_means = KMeans(init='k-means++', n_clusters=3, n_init=10)
t0 = time.time()
k_means.fit(X)
t_batch = time.time() - t0
# #############################################################################
# Compute clustering with MiniBatchKMeans
mbk = MiniBatchKMeans(init='k-means++',
n_clusters=3,
batch_size=batch_size,
n_init=10,
max_no_improvement=10,
verbose=0)
t0 = time.time()
mbk.fit(X)
t_mini_batch = time.time() - t0
# #############################################################################
# Plot result
fig = plt.figure(figsize=(8, 3))
fig.subplots_adjust(left=0.02, right=0.98, bottom=0.05, top=0.9)
colors = ['#4EACC5', '#FF9C34', '#4E9A06']
# We want to have the same colors for the same cluster from the
# MiniBatchKMeans and the KMeans algorithm. Let's pair the cluster centers per
('Sparse comp. - MiniBatchSparsePCA',
decomposition.MiniBatchSparsePCA(n_components=n_components,
alpha=0.8,
n_iter=100,
batch_size=3,
random_state=rng), True),
('MiniBatchDictionaryLearning',
decomposition.MiniBatchDictionaryLearning(n_components=15,
alpha=0.1,
n_iter=50,
batch_size=3,
random_state=rng), True),
('Cluster centers - MiniBatchKMeans',
MiniBatchKMeans(n_clusters=n_components,
tol=1e-3,
batch_size=20,
max_iter=50,
random_state=rng), True),
('Factor Analysis components - FA',
decomposition.FactorAnalysis(n_components=n_components,
max_iter=20), True),
]
# #############################################################################
# Plot a sample of the input data
plot_gallery("First centered Olivetti faces", faces_centered[:n_components])
# #############################################################################
# Do the estimation and plot it
def test_minibatch_init_with_large_k():
mb_k_means = MiniBatchKMeans(init='k-means++', init_size=10, n_clusters=20)
# Check that a warning is raised, as the number clusters is larger
# than the init_size
assert_warns(RuntimeWarning, mb_k_means.fit, X)
import matplotlib.pyplot as plt
import numpy as np
from mrex import datasets
from mrex.cluster import MiniBatchKMeans
from mrex.feature_extraction.image import extract_patches_2d
faces = datasets.fetch_olivetti_faces()
# #############################################################################
# Learn the dictionary of images
print('Learning the dictionary... ')
rng = np.random.RandomState(0)
kmeans = MiniBatchKMeans(n_clusters=81, random_state=rng, verbose=True)
patch_size = (20, 20)
buffer = []
t0 = time.time()
# The online learning part: cycle over the whole dataset 6 times
index = 0
for _ in range(6):
for img in faces.images:
data = extract_patches_2d(img,
patch_size,
max_patches=50,
random_state=rng)
data = np.reshape(data, (len(data), -1))
buffer.append(data)
ax = fig.add_subplot(1, 3, ind + 1)
for this_centroid, k, col in zip(centroids, range(n_clusters), colors_):
mask = labels == k
ax.scatter(X[mask, 0], X[mask, 1],
c='w', edgecolor=col, marker='.', alpha=0.5)
if birch_model.n_clusters is None:
ax.scatter(this_centroid[0], this_centroid[1], marker='+',
c='k', s=25)
ax.set_ylim([-25, 25])
ax.set_xlim([-25, 25])
ax.set_autoscaley_on(False)
ax.set_title('Birch %s' % info)
# Compute clustering with MiniBatchKMeans.
mbk = MiniBatchKMeans(init='k-means++', n_clusters=100, batch_size=100,
n_init=10, max_no_improvement=10, verbose=0,
random_state=0)
t0 = time()
mbk.fit(X)
t_mini_batch = time() - t0
print("Time taken to run MiniBatchKMeans %0.2f seconds" % t_mini_batch)
mbk_means_labels_unique = np.unique(mbk.labels_)
ax = fig.add_subplot(1, 3, 3)
for this_centroid, k, col in zip(mbk.cluster_centers_,
range(n_clusters), colors_):
mask = mbk.labels_ == k
ax.scatter(X[mask, 0], X[mask, 1], marker='.',
c='w', edgecolor=col, alpha=0.5)
ax.scatter(this_centroid[0], this_centroid[1], marker='+',
c='k', s=25)
print("done in %fs" % (time() - t0))
explained_variance = svd.explained_variance_ratio_.sum()
print("Explained variance of the SVD step: {}%".format(
int(explained_variance * 100)))
print()
# #############################################################################
# Do the actual clustering
if opts.minibatch:
km = MiniBatchKMeans(n_clusters=true_k,
init='k-means++',
n_init=1,
init_size=1000,
batch_size=1000,
verbose=opts.verbose)
else:
km = KMeans(n_clusters=true_k,
init='k-means++',
max_iter=100,
n_init=1,
verbose=opts.verbose)
print("Clustering sparse data with %s" % km)
t0 = time()
km.fit(X)
print("done in %0.3fs" % (time() - t0))
print()