def test_affinity_propagation_convergence_warning_dense_sparse(centers):
"""Non-regression, see #13334"""
rng = np.random.RandomState(42)
X = rng.rand(40, 10)
y = (4 * rng.rand(40)).astype(np.int)
ap = AffinityPropagation(random_state=46)
ap.fit(X, y)
ap.cluster_centers_ = centers
with pytest.warns(None) as record:
assert_array_equal(ap.predict(X), np.zeros(X.shape[0], dtype=int))
assert len(record) == 0
def test_affinity_propagation_convergence_warning_dense_sparse(centers):
"""Non-regression, see #13334"""
rng = np.random.RandomState(42)
X = rng.rand(40, 10)
y = (4 * rng.rand(40)).astype(int)
ap = AffinityPropagation(random_state=46)
ap.fit(X, y)
ap.cluster_centers_ = centers
with warnings.catch_warnings():
warnings.simplefilter("error", ConvergenceWarning)
assert_array_equal(ap.predict(X), np.zeros(X.shape[0], dtype=int))
def cluster_bbox(X, clutter=1):
A = -euclidean_distances(X, squared=True)
pref = np.percentile(A, clutter)
# bbox clustering
af = AffinityPropagation(preference=pref)
af.fit(X)
# membership function
if (af.cluster_centers_.ndim == 3):
af.cluster_centers_ = af.cluster_centers_.reshape(
(af.cluster_centers_.shape[0], -1))
M_mem = np.sqrt(pairwise_distances(X, af.cluster_centers_))
M_mem = np.exp(-M_mem)
M_mem /= M_mem.sum(axis=1, keepdims=True)
return M_mem, af.cluster_centers_
