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)
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)
plt.xlabel('First Dimension', fontsize=24)
plt.ylabel('Second Dimension', fontsize=24)
plt.title('AutoGMM Clustering', fontsize=24, fontweight='bold')
plt.xticks(fontsize=18)
plt.yticks(fontsize=18)
fname = './autogmm_clustering_dataset' + str(dataset) + '.png'
plt.savefig(fname)
pyc = AutoGMMCluster(min_components=ks[0],
max_components=ks[len(ks) - 1],
affinity=affinities,
linkage=linkages,
covariance_type=covariance_types,
random_state=0)
c_hat_autogmm, ari = pyc.fit_predict(x, c_true)
#np.savetxt('autogmm.csv',labels, delimiter=',')
combo = [pyc.affinity_, pyc.linkage_, pyc.covariance_type_]
k = pyc.n_components_
reg = pyc.reg_covar_
bic = -pyc.criter_
results = pyc.results_
print('Info for table:')
print('Best model: ' + str(combo))
print('Best reg: ' + str(reg))
print('Best k: ' + str(k))
print('Best BIC: ' + str(bic))
print('Best ARI: ' + str(ari))
