def test_not_2_views():
with pytest.raises(ValueError):
view1 = np.random.random((10, ))
view2 = np.random.random((10, ))
view3 = np.random.random((10, ))
kmeans = MultiviewKMeans()
kmeans.fit([view1, view2, view3])
def perform_clustering(seed, m_data, labels, n_clusters):
# Singleview kmeans clustering
# Cluster each view separately
s_kmeans = KMeans(n_clusters=n_clusters, random_state=seed, n_init=100)
s_clusters_v1 = s_kmeans.fit_predict(m_data[0])
s_clusters_v2 = s_kmeans.fit_predict(m_data[1])
# Concatenate the multiple views into a single view
s_data = np.hstack(m_data)
s_clusters = s_kmeans.fit_predict(s_data)
# Compute nmi between true class labels and singleview cluster labels
s_nmi_v1 = nmi_score(labels, s_clusters_v1)
s_nmi_v2 = nmi_score(labels, s_clusters_v2)
s_nmi = nmi_score(labels, s_clusters)
print('Singleview View 1 NMI Score: {0:.3f}\n'.format(s_nmi_v1))
print('Singleview View 2 NMI Score: {0:.3f}\n'.format(s_nmi_v2))
print('Singleview Concatenated NMI Score: {0:.3f}\n'.format(s_nmi))
# Multiview kmeans clustering
# Use the MultiviewKMeans instance to cluster the data
m_kmeans = MultiviewKMeans(n_clusters=n_clusters,
n_init=100,
random_state=seed)
m_clusters = m_kmeans.fit_predict(m_data)
# Compute nmi between true class labels and multiview cluster labels
m_nmi = nmi_score(labels, m_clusters)
print('Multiview NMI Score: {0:.3f}\n'.format(m_nmi))
return m_clusters
def test_predict_no_centroids1():
with pytest.raises(AttributeError):
kmeans = MultiviewKMeans()
kmeans.centroids_ = [None, None]
view1 = np.random.random((10, 11))
view2 = np.random.random((10, 10))
kmeans.predict([view1, view2])
def test_init_not_2_views(data_small):
with pytest.raises(ValueError):
view1 = np.random.random((2, 8))
view2 = np.random.random((2, 9))
view3 = np.random.random((2, 9))
kmeans = MultiviewKMeans(init=[view1, view2])
kmeans.fit(data_small)
def test_fit_predict_n_jobs_all(data_random):
n_clusters = data_random['n_clusters']
kmeans = MultiviewKMeans(n_clusters=n_clusters, n_jobs=-1)
cluster_pred = kmeans.fit_predict(data_random['test_data'])
assert (data_random['n_test'] == cluster_pred.shape[0])
for cl in cluster_pred:
assert (cl >= 0 and cl < data_random['n_clusters'])
def test_final_centroids_no_consensus():
with pytest.raises(ConvergenceWarning):
kmeans = MultiviewKMeans(random_state=RANDOM_SEED)
view1 = np.array([[0, 1], [1, 0]])
view2 = np.array([[1, 0], [0, 1]])
v1_centroids = np.array([[0, 1], [1, 0]])
v2_centroids = np.array([[0, 1], [1, 0]])
centroids = [v1_centroids, v2_centroids]
kmeans._final_centroids([view1, view2], centroids)
def test_final_centroids_less_than_n_clusters():
with pytest.raises(ConvergenceWarning):
kmeans = MultiviewKMeans(n_clusters=3, random_state=RANDOM_SEED)
view1 = np.random.random((2, 5))
view2 = np.random.random((2, 6))
v1_centroids = np.random.random((3, 5))
v2_centroids = np.random.random((3, 6))
centroids = [v1_centroids, v2_centroids]
kmeans._final_centroids([view1, view2], centroids)
def test_fit_predict_max_iter(data_random):
n_clusters = data_random['n_clusters']
max_iter = 5
kmeans = MultiviewKMeans(n_clusters=n_clusters, max_iter=max_iter)
cluster_pred = kmeans.fit_predict(data_random['test_data'])
assert (data_random['n_test'] == cluster_pred.shape[0])
for cl in cluster_pred:
assert (cl >= 0 and cl < data_random['n_clusters'])
def test_fit_predict_patience(data_random):
n_clusters = data_random['n_clusters']
patience = 10
kmeans = MultiviewKMeans(n_clusters=n_clusters, patience=patience)
cluster_pred = kmeans.fit_predict(data_random['test_data'])
assert (data_random['n_test'] == cluster_pred.shape[0])
for cl in cluster_pred:
assert (cl >= 0 and cl < data_random['n_clusters'])
def test_fit_predict_init_random(data_random):
n_clusters = data_random['n_clusters']
init = 'random'
kmeans = MultiviewKMeans(n_clusters=n_clusters, init='random')
cluster_pred = kmeans.fit_predict(data_random['test_data'])
assert (data_random['n_test'] == cluster_pred.shape[0])
for cl in cluster_pred:
assert (cl >= 0 and cl < data_random['n_clusters'])
def test_fit_predict_n_clusters():
n_clusters = 3
v1_data = np.array([[0, 0], [1, 0], [0, 1]])
v2_data = np.array([[0, 0], [1, 0], [0, 1]])
data = [v1_data, v2_data]
kmeans = MultiviewKMeans(n_clusters=n_clusters)
cluster_pred = kmeans.fit_predict(data)
cluster_pred = list(set(cluster_pred))
assert (len(cluster_pred) == n_clusters)
def test_fit_predict_init_predefined():
n_clusters = 2
v1_centroid = np.array([[0, 0], [1, 1]])
v2_centroid = np.array([[0, 0], [1, 1]])
centroids = [v1_centroid, v2_centroid]
v1_data = np.array([[0, 0], [0.3, 0.2], [0.5, 0.5], [0.7, 0.7], [1, 1]])
v2_data = np.array([[0, 0], [0.2, 0.4], [0.5, 0.5], [0.4, 0.7], [1, 1]])
data = [v1_data, v2_data]
kmeans = MultiviewKMeans(n_clusters=n_clusters, init=centroids)
cluster_pred = kmeans.fit_predict(data)
def test_predict_random_small(data_random):
kmeans = MultiviewKMeans()
input_data = [
data_random['fit_data'][0][:2], data_random['fit_data'][1][:2]
]
kmeans.fit(input_data)
cluster_pred = kmeans.predict(data_random['test_data'])
assert (data_random['n_test'] == cluster_pred.shape[0])
for cl in cluster_pred:
assert (cl >= 0 and cl < data_random['n_clusters'])
def test_n_init_not_positive_int():
with pytest.raises(ValueError):
kmeans = MultiviewKMeans(n_init=-1)
kmeans.fit(data_small)
with pytest.raises(ValueError):
kmeans = MultiviewKMeans(n_init=0)
kmeans.fit(data_small)
def test_max_iter_not_positive_int(data_small):
with pytest.raises(ValueError):
kmeans = MultiviewKMeans(max_iter=-1)
kmeans.fit(data_small)
with pytest.raises(ValueError):
kmeans = MultiviewKMeans(max_iter=0)
kmeans.fit(data_small)
def data_random():
num_fit_samples = 200
num_test_samples = 5
n_feats1 = 20
n_feats2 = 18
n_clusters = 2
np.random.seed(RANDOM_SEED)
fit_data = []
fit_data.append(np.random.rand(num_fit_samples, n_feats1))
fit_data.append(np.random.rand(num_fit_samples, n_feats2))
test_data = []
test_data.append(np.random.rand(num_test_samples, n_feats1))
test_data.append(np.random.rand(num_test_samples, n_feats2))
kmeans = MultiviewKMeans(n_clusters=n_clusters, random_state=RANDOM_SEED)
return {
'n_test': num_test_samples,
'n_feats1': n_feats1,
'n_feats2': n_feats2,
'n_clusters': n_clusters,
'kmeans': kmeans,
'fit_data': fit_data,
'test_data': test_data
}
def test_predict_deterministic():
n_clusters = 2
v1_centroid = np.array([[0, 0], [1, 1]])
v2_centroid = np.array([[0, 0], [1, 1]])
centroids = [v1_centroid, v2_centroid]
v1_data = np.array([[0, 0], [0.3, 0.2], [0.5, 0.5], [0.7, 0.7], [1, 1]])
v2_data = np.array([[0, 0], [0.2, 0.4], [0.5, 0.5], [0.4, 0.7], [1, 1]])
data = [v1_data, v2_data]
kmeans = MultiviewKMeans(n_clusters=n_clusters)
kmeans.centroids_ = centroids
cluster_pred = kmeans.predict(data)
true_clusters = [0, 0, 0, 1, 1]
for ind in range(len(true_clusters)):
assert cluster_pred[ind] == true_clusters[ind]
def test_predict_no_centroids2():
kmeans = MultiviewKMeans()
with pytest.raises(ConvergenceWarning):
view1 = np.array([[0, 1], [1, 0]])
view2 = np.array([[1, 0], [0, 1]])
v1_centroids = np.array([[0, 1], [1, 0]])
v2_centroids = np.array([[0, 1], [1, 0]])
centroids = [v1_centroids, v2_centroids]
kmeans._final_centroids([view1, view2], centroids)
with pytest.raises(AttributeError):
kmeans.predict([view1, view2])
def test_patience_not_nonnegative_int(data_small):
with pytest.raises(ValueError):
kmeans = MultiviewKMeans(patience=-1)
kmeans.fit(data_small)
def test_init_clusters_not_same(data_small):
with pytest.raises(ValueError):
view1 = np.random.random((2, 8))
view2 = np.random.random((3, 9))
kmeans = MultiviewKMeans(init=[view1, view2])
kmeans.fit(data_small)
def test_samples_not_2D_2():
with pytest.raises(ValueError):
view1 = np.random.random((10, ))
view2 = np.random.random((10, ))
kmeans = MultiviewKMeans()
kmeans.fit([view1, view2])
def test_init_samples_not_list(data_small):
with pytest.raises(ValueError):
view1 = 1
view2 = 3
kmeans = MultiviewKMeans(init=[view1, view2])
kmeans.fit(data_small)
def test_tol_not_nonnegative_float(data_small):
with pytest.raises(ValueError):
kmeans = MultiviewKMeans(tol=-0.05)
kmeans.fit(data_small)
def test_not_init1(data_small):
with pytest.raises(ValueError):
kmeans = MultiviewKMeans(init='Not_Init')
kmeans.fit(data_small)
def test_predict_not_fit():
with pytest.raises(NotFittedError):
kmeans = MultiviewKMeans()
view1 = np.random.random((10, 11))
view2 = np.random.random((10, 10))
kmeans.predict([view1, view2])
def test_random_state_not_convertible(data_small):
with pytest.raises(ValueError):
kmeans = MultiviewKMeans(random_state='ab')
kmeans.fit(data_small)
plt.tight_layout()
plt.show()
###############################################################################
# Cluster using Multiview KMeans
# ------------------------------
#
# We can compare the estimated clusters from
# :class:`mvlearn.cluster.MultiviewKMeans` to regular
# KMeans on each of the views. Multiview Kmeans clearly finds two clusters
# matching the two different genotype labels observed in the prior plots.
from mvlearn.cluster import MultiviewKMeans # noqa: E402
from sklearn.cluster import KMeans # noqa: E402
Xs_labels = MultiviewKMeans(n_clusters=2, random_state=0).fit_predict(Xs)
X1_labels = KMeans(n_clusters=2, random_state=0).fit_predict(Xs[0])
X2_labels = KMeans(n_clusters=2, random_state=0).fit_predict(Xs[1])
sca_kwargs = {'alpha': 0.7, 's': 20}
colors = np.asarray(['Red', 'Blue'])
f, axes = plt.subplots(1, 3, figsize=(8, 4))
axes[0].scatter(*zip(*X_mvmds), c=colors[Xs_labels], **sca_kwargs)
axes[0].set_title('Multiview Kmeans Clusters')
axes[1].scatter(*zip(*X_mvmds), c=colors[X1_labels], **sca_kwargs)
axes[1].set_title('View 1 Kmeans Clusters')
axes[2].scatter(*zip(*X_mvmds), c=colors[X2_labels], **sca_kwargs)
axes[2].set_title('View 2 Kmeans Clusters')
for ax in axes:
ax.set_xlabel('MVMDS component 1')
def test_samples_not_list():
with pytest.raises(ValueError):
view1 = 1
view2 = 3
kmeans = MultiviewKMeans()
kmeans.fit([view1, view2])
# Concatenate the multiple views into a single view
s_data = np.hstack(Xs)
s_clusters = s_kmeans.fit_predict(s_data)
# Compute nmi between true class labels and singleview cluster labels
s_nmi_v1 = nmi_score(labels, s_clusters_v1)
s_nmi_v2 = nmi_score(labels, s_clusters_v2)
s_nmi = nmi_score(labels, s_clusters)
print('Singleview View 1 NMI Score: {0:.3f}\n'.format(s_nmi_v1))
print('Singleview View 2 NMI Score: {0:.3f}\n'.format(s_nmi_v2))
print('Singleview Concatenated NMI Score: {0:.3f}\n'.format(s_nmi))
# Multiview kmeans clustering
# Use the MultiviewKMeans instance to cluster the data
m_kmeans = MultiviewKMeans(n_clusters=n_class, random_state=RANDOM_SEED)
m_clusters = m_kmeans.fit_predict(Xs)
# Compute nmi between true class labels and multiview cluster labels
m_nmi = nmi_score(labels, m_clusters)
print('Multiview NMI Score: {0:.3f}\n'.format(m_nmi))
###############################################################################
# Comparing predicted cluster labels vs the truth
# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
# We will display the clustering results of the Multiview kmeans clustering
# algorithm below, along with the true class labels.
# Running TSNE to display clustering results via low dimensional embedding
tsne = TSNE()
def test_init_samples_not_2D_2(data_small):
with pytest.raises(ValueError):
view1 = np.random.random((2, ))
view2 = np.random.random((2, ))
kmeans = MultiviewKMeans(init=[view1, view2])
kmeans.fit(data_small)
