def test_maui_clusters_picks_optimal_k_by_ami():
ami_mock = mock.Mock()
ami_mock.side_effect = [
2,
3,
1,
] # the optimal AMI will be given at the second trial
with mock.patch("sklearn.metrics.adjusted_mutual_info_score", ami_mock):
maui_model = Maui(n_hidden=[10], n_latent=2, epochs=1)
maui_model.z_ = pd.DataFrame(
np.random.randn(10, 2),
index=[f"sample {i}" for i in range(10)],
columns=["LF1", "LF2"],
)
maui_model.x_ = pd.DataFrame(
np.random.randn(20, 10),
index=[f"feature {i}" for i in range(20)],
columns=[f"sample {i}" for i in range(10)],
)
the_y = pd.Series(np.arange(10), index=maui_model.z_.index)
maui_model.cluster(
ami_y=the_y, optimal_k_range=[1, 2, 3]
) # the second trial is k=2
assert maui_model.optimal_k_ == 2
def test_maui_clusters_picks_optimal_k_by_silhouette():
silhouette_mock = mock.Mock()
silhouette_mock.side_effect = [2,3,1] # the optimal silhouette will be given at the second trial
with mock.patch('sklearn.metrics.silhouette_score', silhouette_mock):
maui_model = Maui(n_hidden=[10], n_latent=2, epochs=1)
maui_model.z_ = pd.DataFrame(np.random.randn(10,2),
index=[f'sample {i}' for i in range(10)],
columns=['LF1', 'LF2'])
maui_model.x_ = pd.DataFrame(np.random.randn(20,10),
index=[f'feature {i}' for i in range(20)],
columns=[f'sample {i}' for i in range(10)])
maui_model.cluster(optimal_k_method='silhouette', optimal_k_range=[1,2,3]) # the second trial is k=2
assert maui_model.optimal_k_ == 2
def test_maui_clusters_picks_optimal_k_with_custom_scoring():
scorer = mock.Mock()
scorer.side_effect = [2,3,1] # the optimal AMI will be given at the second trial
scorer.__name__ = 'mock_scorer'
maui_model = Maui(n_hidden=[10], n_latent=2, epochs=1)
maui_model.z_ = pd.DataFrame(np.random.randn(10,2),
index=[f'sample {i}' for i in range(10)],
columns=['LF1', 'LF2'])
maui_model.x_ = pd.DataFrame(np.random.randn(20,10),
index=[f'feature {i}' for i in range(20)],
columns=[f'sample {i}' for i in range(10)])
maui_model.cluster(optimal_k_method=scorer, optimal_k_range=[1,2,3]) # the second trial is k=2
assert maui_model.optimal_k_ == 2
def test_maui_clusters_with_single_k():
maui_model = Maui(n_hidden=[10], n_latent=2, epochs=1)
maui_model.z_ = pd.DataFrame(np.random.randn(10,2),
index=[f'sample {i}' for i in range(10)],
columns=['LF1', 'LF2'])
maui_model.x_ = pd.DataFrame(np.random.randn(20,10),
index=[f'feature {i}' for i in range(20)],
columns=[f'sample {i}' for i in range(10)])
yhat = maui_model.cluster(5)
assert yhat.shape == (10,)
def test_maui_clusters_only_samples_in_y_index_when_optimizing():
maui_model = Maui(n_hidden=[10], n_latent=2, epochs=1)
maui_model.z_ = pd.DataFrame(np.random.randn(10,2),
index=[f'sample {i}' for i in range(10)],
columns=['LF1', 'LF2'])
maui_model.x_ = pd.DataFrame(np.random.randn(20,10),
index=[f'feature {i}' for i in range(20)],
columns=[f'sample {i}' for i in range(10)])
y = pd.Series(['a','a','a','b','b','b'],
index=[f'sample {i}' for i in range(6)])
yhat = maui_model.cluster(ami_y=y, optimal_k_range=[1,2,3])
assert set(yhat.index) == set(y.index)
def test_maui_clusters_only_samples_in_y_index_when_optimizing():
np.random.seed(0)
maui_model = Maui(n_hidden=[10], n_latent=2, epochs=1)
maui_model.z_ = pd.DataFrame(
np.random.randn(10, 2),
index=[f"sample {i}" for i in range(10)],
columns=["LF1", "LF2"],
)
maui_model.x_ = pd.DataFrame(
np.random.randn(20, 10),
index=[f"feature {i}" for i in range(20)],
columns=[f"sample {i}" for i in range(10)],
)
y = pd.Series(["a", "a", "a", "b", "b", "b"],
index=[f"sample {i}" for i in range(6)])
yhat = maui_model.cluster(ami_y=y, optimal_k_range=[1, 2, 3])
assert set(yhat.index) == set(y.index)