def test_ase_three_blocks():
"""
Expect 3 clusters from a 3 block model
"""
np.random.seed(1)
num_sims = 10
# Generate adjacency and labels
n = 50
n_communites = [n, n, n]
p = np.array([[0.8, 0.3, 0.2], [0.3, 0.8, 0.3], [0.2, 0.3, 0.8]])
y = np.repeat([1, 2, 3], repeats=n)
for _ in range(num_sims):
A = sbm(n=n_communites, p=p)
# Embed to get latent positions
ase = AdjacencySpectralEmbed(n_components=5)
X_hat = ase.fit_transform(A)
# Compute clusters
AutoGMM = AutoGMMCluster(max_components=10)
AutoGMM.fit(X_hat, y)
n_components = AutoGMM.n_components_
# Assert that the three cluster model is the best
assert_equal(n_components, 3)
# Asser that we get perfect clustering
assert_allclose(AutoGMM.ari_, 1)
def test_predict_without_fit():
# Generate random data
X = np.random.normal(0, 1, size=(100, 3))
with pytest.raises(NotFittedError):
AutoGMM = AutoGMMCluster(min_components=2)
AutoGMM.predict(X)
def test_two_class():
"""
Easily separable two gaussian problem.
"""
np.random.seed(1)
n = 100
d = 3
num_sims = 10
for _ in range(num_sims):
X1 = np.random.normal(2, 0.5, size=(n, d))
X2 = np.random.normal(-2, 0.5, size=(n, d))
X = np.vstack((X1, X2))
y = np.repeat([0, 1], n)
AutoGMM = AutoGMMCluster(max_components=5)
AutoGMM.fit(X, y)
n_components = AutoGMM.n_components_
# Assert that the two cluster model is the best
assert_equal(n_components, 2)
# Asser that we get perfect clustering
assert_allclose(AutoGMM.ari_, 1)
def test_two_class_aic():
"""
Easily separable two gaussian problem.
"""
np.random.seed(1)
n = 100
d = 3
X1 = np.random.normal(2, 0.5, size=(n, d))
X2 = np.random.normal(-2, 0.5, size=(n, d))
X = np.vstack((X1, X2))
y = np.repeat([0, 1], n)
AutoGMM = AutoGMMCluster(max_components=5, selection_criteria="aic")
AutoGMM.fit(X, y)
n_components = AutoGMM.n_components_
# AIC gets the number of components wrong
assert_equal(n_components >= 1, True)
assert_equal(n_components <= 5, True)
# Assert that the ari value is valid
assert_equal(AutoGMM.ari_ >= -1, True)
assert_equal(AutoGMM.ari_ <= 1, True)
def test_no_y():
np.random.seed(1)
n = 100
d = 3
X1 = np.random.normal(2, 0.5, size=(n, d))
X2 = np.random.normal(-2, 0.5, size=(n, d))
X = np.vstack((X1, X2))
AutoGMM = AutoGMMCluster(max_components=5)
AutoGMM.fit(X)
assert_equal(AutoGMM.n_components_, 2)
def test_five_class():
"""
Easily separable five gaussian problem.
"""
np.random.seed(1)
n = 100
mus = [[i * 5, 0] for i in range(5)]
cov = np.eye(2) # balls
X = np.vstack([np.random.multivariate_normal(mu, cov, n) for mu in mus])
AutoGMM = AutoGMMCluster(min_components=3, max_components=10, covariance_type="all")
AutoGMM.fit(X)
assert_equal(AutoGMM.n_components_, 5)
def test_cosine_with_0():
X = np.array([
[0, 1, 0],
[1, 0, 1],
[0, 0, 0],
[1, 1, 0],
[0, 0, 1],
[0, 1, 1],
[1, 1, 1],
[1, 0, 0],
[0, 1, 1],
[1, 1, 0],
[0, 1, 0],
])
with pytest.warns(UserWarning):
AutoGMM = AutoGMMCluster(min_components=2, affinity="all")
AutoGMM.fit(X)
def test_five_class_aic():
"""
Easily separable five gaussian problem.
"""
np.random.seed(1)
n = 100
mus = [[i * 5, 0] for i in range(5)]
cov = np.eye(2) # balls
X = np.vstack([np.random.multivariate_normal(mu, cov, n) for mu in mus])
AutoGMM = AutoGMMCluster(
min_components=3,
max_components=10,
covariance_type="all",
selection_criteria="aic",
)
AutoGMM.fit(X)
# AIC fails often so there is no assertion here
assert_equal(AutoGMM.n_components_ >= 3, True)
assert_equal(AutoGMM.n_components_ <= 10, True)
results = pd.DataFrame(columns=['ARI', 'Time'])
x_full = x
c_true_full = c_true
n_full = x.shape[0]
for i in range(num_runs):
print('Run number: ' + str(i))
idxs = idxs_full.iloc[:, i].values - 1
x = x_full[idxs, ]
c_true = c_true_full[idxs, ]
start_time = time.time()
pyc = AutoGMMCluster(min_components=ks[0],
max_components=ks[len(ks) - 1],
affinity=affinities,
linkage=linkages,
covariance_type=covariance_types,
random_state=0)
pyc.fit(x, c_true)
best_ari_bic = pyc.ari_
results = results.append(
{
'Time': time.time() - start_time,
'ARI': best_ari_bic
},
ignore_index=True)
results.to_csv(path_or_buf=output_file)
#%%
#Wisconsin Breast Cancer Diagnostic Data
def test_covariances():
"""
Easily separable two gaussian problem.
"""
np.random.seed(1)
n = 100
mu1 = [-10, 0]
mu2 = [10, 0]
# Spherical
cov1 = 2 * np.eye(2)
cov2 = 2 * np.eye(2)
X1 = np.random.multivariate_normal(mu1, cov1, n)
X2 = np.random.multivariate_normal(mu2, cov2, n)
X = np.concatenate((X1, X2))
AutoGMM = AutoGMMCluster(min_components=2, covariance_type="all")
AutoGMM.fit(X)
assert_equal(AutoGMM.covariance_type_, "spherical")
# Diagonal
np.random.seed(10)
cov1 = np.diag([1, 1])
cov2 = np.diag([2, 1])
X1 = np.random.multivariate_normal(mu1, cov1, n)
X2 = np.random.multivariate_normal(mu2, cov2, n)
X = np.concatenate((X1, X2))
AutoGMM = AutoGMMCluster(max_components=2, covariance_type="all")
AutoGMM.fit(X)
assert_equal(AutoGMM.covariance_type_, "diag")
# Tied
cov1 = np.array([[2, 1], [1, 2]])
cov2 = np.array([[2, 1], [1, 2]])
X1 = np.random.multivariate_normal(mu1, cov1, n)
X2 = np.random.multivariate_normal(mu2, cov2, n)
X = np.concatenate((X1, X2))
AutoGMM = AutoGMMCluster(max_components=2, covariance_type="all")
AutoGMM.fit(X)
assert_equal(AutoGMM.covariance_type_, "tied")
# Full
cov1 = np.array([[2, -1], [-1, 2]])
cov2 = np.array([[2, 1], [1, 2]])
X1 = np.random.multivariate_normal(mu1, cov1, n)
X2 = np.random.multivariate_normal(mu2, cov2, n)
X = np.concatenate((X1, X2))
AutoGMM = AutoGMMCluster(max_components=2, covariance_type="all")
AutoGMM.fit(X)
assert_equal(AutoGMM.covariance_type_, "full")
def test_cosine_on_0():
X = np.array([[0, 1, 0], [1, 0, 1], [0, 0, 0], [1, 1, 0], [0, 0, 1]])
with pytest.raises(ValueError):
AutoGMM = AutoGMMCluster(min_components=3, affinity="all")
AutoGMM.fit(X)
def test_inputs():
# Generate random data
X = np.random.normal(0, 1, size=(100, 3))
# min_components < 1
with pytest.raises(ValueError):
AutoGMM = AutoGMMCluster(min_components=0)
# min_components integer
with pytest.raises(TypeError):
AutoGMM = AutoGMMCluster(min_components="1")
# max_components < min_components
with pytest.raises(ValueError):
AutoGMM = AutoGMMCluster(min_components=1, max_components=0)
# max_components integer
with pytest.raises(TypeError):
AutoGMM = AutoGMMCluster(min_components=1, max_components="1")
# affinity is not an array, string or list
with pytest.raises(TypeError):
AutoGMM = AutoGMMCluster(min_components=1, affinity=1)
# affinity is not in ['euclidean', 'manhattan', 'cosine', 'none']
with pytest.raises(ValueError):
AutoGMM = AutoGMMCluster(min_components=1, affinity="graspy")
# linkage is not an array, string or list
with pytest.raises(TypeError):
AutoGMM = AutoGMMCluster(min_components=1, linkage=1)
# linkage is not in ['single', 'average', 'complete', 'ward']
with pytest.raises(ValueError):
AutoGMM = AutoGMMCluster(min_components=1, linkage="graspy")
# euclidean is not an affinity option when ward is a linkage option
with pytest.raises(ValueError):
AutoGMM = AutoGMMCluster(min_components=1, affinity="manhattan", linkage="ward")
# covariance type is not an array, string or list
with pytest.raises(TypeError):
AutoGMM = AutoGMMCluster(min_components=1, covariance_type=1)
# covariance type is not in ['spherical', 'diag', 'tied', 'full']
with pytest.raises(ValueError):
AutoGMM = AutoGMMCluster(min_components=1, covariance_type="graspy")
# min_cluster > n_samples when max_cluster is None
with pytest.raises(ValueError):
AutoGMM = AutoGMMCluster(1000)
AutoGMM.fit(X)
with pytest.raises(ValueError):
AutoGMM = AutoGMMCluster(1000)
AutoGMM.fit_predict(X)
# max_cluster > n_samples when max_cluster is not None
with pytest.raises(ValueError):
AutoGMM = AutoGMMCluster(10, 1001)
AutoGMM.fit(X)
with pytest.raises(ValueError):
AutoGMM = AutoGMMCluster(10, 1001)
AutoGMM.fit_predict(X)
# min_cluster > n_samples when max_cluster is None
with pytest.raises(ValueError):
AutoGMM = AutoGMMCluster(1000)
AutoGMM.fit(X)
with pytest.raises(ValueError):
AutoGMM = AutoGMMCluster(10, 1001)
AutoGMM.fit_predict(X)
# min_cluster > n_samples when max_cluster is not None
with pytest.raises(ValueError):
AutoGMM = AutoGMMCluster(1000, 1001)
AutoGMM.fit(X)
with pytest.raises(ValueError):
AutoGMM = AutoGMMCluster(1000, 1001)
AutoGMM.fit_predict(X)
# label_init is not a 1-D array
with pytest.raises(TypeError):
AutoGMM = AutoGMMCluster(label_init=np.zeros([100, 2]))
# label_init is not 1-D array, a list or None.
with pytest.raises(TypeError):
AutoGMM = AutoGMMCluster(label_init="label")
# label_init length is not equal to n_samples
with pytest.raises(ValueError):
AutoGMM = AutoGMMCluster(label_init=np.zeros([50, 1]))
AutoGMM.fit(X)
with pytest.raises(ValueError):
AutoGMM = AutoGMMCluster(label_init=np.zeros([50, 1]))
AutoGMM.fit_predict(X)
with pytest.raises(TypeError):
AutoGMM = AutoGMMCluster(label_init=np.zeros([100, 2]), max_iter=-2)
def test_labels_init():
X = np.random.normal(0, 1, size=(5, 3))
with pytest.raises(ValueError):
AutoGMM = AutoGMMCluster(min_components=1,
max_components=1,
label_init=np.array([0, 0, 0, 0, 1]))
AutoGMM.fit_predict(X)
with pytest.raises(ValueError):
AutoGMM = AutoGMMCluster(min_components=1,
max_components=2,
label_init=np.array([0, 0, 0, 0, 1]))
AutoGMM.fit_predict(X)
with pytest.raises(ValueError):
AutoGMM = AutoGMMCluster(min_components=2,
max_components=3,
label_init=np.array([0, 0, 0, 0, 1]))
AutoGMM.fit_predict(X)
AutoGMM = AutoGMMCluster(min_components=2,
max_components=2,
label_init=np.array([0, 0, 0, 0, 1]))
AutoGMM.fit_predict(X)
continue
if affinity == 'none' and linkage != 'complete':
continue
for covariance_type in covariance_types:
for i, n in enumerate(ns):
file = ".\data\\" + str(n) + ".csv"
x = np.genfromtxt(file, delimiter=',', skip_header=0)
x = x[:, np.arange(1, x.shape[1])]
c_true = np.genfromtxt(file,
delimiter=',',
usecols=(0),
skip_header=0)
start_time = time.time()
pyc = AutoGMMCluster(min_components=ks[0],
max_components=ks[len(ks) - 1],
affinity=affinity,
linkage=linkage,
covariance_type=covariance_type)
pyc.fit(x, c_true)
entry = {
'N': n,
'Affinity': affinity,
'Linkage': linkage,
'Covariance_Type': covariance_type,
'Time': time.time() - start_time
}
results = results.append(entry, ignore_index=True)
print(entry)
results.to_csv(path_or_buf=output_file)
