def test_1000by6_matrix_auto_tune(self):
matrix = np.array([[1.0, 0.0, 0.0, 0.0, 0.0, 0.0]] * 400 +
[[0.0, 1.0, 0.0, 0.0, 0.0, 0.0]] * 300 +
[[0.0, 0.0, 2.0, 0.0, 0.0, 0.0]] * 200 +
[[0.0, 0.0, 0.0, 1.0, 0.0, 0.0]] * 100)
noisy = np.random.rand(1000, 6) * 2 - 1
matrix = matrix + noisy * 0.1
refinement_sequence = [RefinementName.RowWiseThreshold]
refinement_options = refinement.RefinementOptions(
thresholding_type=ThresholdType.Percentile,
refinement_sequence=refinement_sequence)
auto_tune = autotune.AutoTune(p_percentile_min=0.9,
p_percentile_max=0.95,
init_search_step=0.03,
search_level=1)
clusterer = spectral_clusterer.SpectralClusterer(
max_clusters=4,
refinement_options=refinement_options,
autotune=auto_tune,
laplacian_type=LaplacianType.GraphCut,
row_wise_renorm=True)
labels = clusterer.predict(matrix)
labels = utils.enforce_ordered_labels(labels)
expected = np.array([0] * 400 + [1] * 300 + [2] * 200 + [3] * 100)
self.assertTrue(np.array_equal(expected, labels))
def test_6by2_matrix_auto_tune(self):
matrix = np.array([
[1.0, 0.0],
[1.1, 0.1],
[0.0, 1.0],
[0.1, 1.0],
[0.9, -0.1],
[0.0, 1.2],
])
refinement_sequence = [RefinementName.RowWiseThreshold]
refinement_options = refinement.RefinementOptions(
thresholding_type=ThresholdType.Percentile,
refinement_sequence=refinement_sequence)
auto_tune = autotune.AutoTune(p_percentile_min=0.60,
p_percentile_max=0.95,
init_search_step=0.05,
search_level=1)
clusterer = spectral_clusterer.SpectralClusterer(
max_clusters=2,
refinement_options=refinement_options,
autotune=auto_tune,
laplacian_type=LaplacianType.GraphCut,
row_wise_renorm=True)
labels = clusterer.predict(matrix)
labels = utils.enforce_ordered_labels(labels)
expected = np.array([0, 0, 1, 1, 0, 1])
self.assertTrue(np.array_equal(expected, labels))
def test_1000by6_matrix(self):
matrix = np.array([[1.0, 0.0, 0.0, 0.0, 0.0, 0.0]] * 400 +
[[0.0, 1.0, 0.0, 0.0, 0.0, 0.0]] * 300 +
[[0.0, 0.0, 2.0, 0.0, 0.0, 0.0]] * 200 +
[[0.0, 0.0, 0.0, 1.0, 0.0, 0.0]] * 100)
noisy = np.random.rand(1000, 6) * 2 - 1
matrix = matrix + noisy * 0.1
labels = configs.icassp2018_clusterer.predict(matrix)
labels = utils.enforce_ordered_labels(labels)
expected = np.array([0] * 400 + [1] * 300 + [2] * 200 + [3] * 100)
self.assertTrue(np.array_equal(expected, labels))
def test_1000by6_matrix_euclidean_dist(self):
matrix = np.array([[1.0, 0.0, 0.0, 0.0, 0.0, 0.0]] * 400 +
[[0.0, 1.0, 0.0, 0.0, 0.0, 0.0]] * 300 +
[[0.0, 0.0, 2.0, 0.0, 0.0, 0.0]] * 200 +
[[0.0, 0.0, 0.0, 1.0, 0.0, 0.0]] * 100)
noisy = np.random.rand(1000, 6) * 2 - 1
matrix = matrix + noisy * 0.1
labels = custom_distance_kmeans.run_kmeans(matrix,
n_clusters=4,
max_iter=300,
custom_dist="euclidean")
labels = utils.enforce_ordered_labels(labels)
expected = np.array([0] * 400 + [1] * 300 + [2] * 200 + [3] * 100)
self.assertTrue(np.array_equal(expected, labels))
def test_1000by6_matrix(self):
matrix = np.array([[1.0, 0.0, 0.0, 0.0, 0.0, 0.0]] * 400 +
[[0.0, 1.0, 0.0, 0.0, 0.0, 0.0]] * 300 +
[[0.0, 0.0, 2.0, 0.0, 0.0, 0.0]] * 200 +
[[0.0, 0.0, 0.0, 1.0, 0.0, 0.0]] * 100)
noisy = np.random.rand(1000, 6) * 2 - 1
matrix = matrix + noisy * 0.1
refinement_options = refinement.RefinementOptions(
gaussian_blur_sigma=0,
p_percentile=0.2,
refinement_sequence=ICASSP2018_REFINEMENT_SEQUENCE)
clusterer = spectral_clusterer.SpectralClusterer(
refinement_options=refinement_options, stop_eigenvalue=0.01)
labels = clusterer.predict(matrix)
labels = utils.enforce_ordered_labels(labels)
expected = np.array([0] * 400 + [1] * 300 + [2] * 200 + [3] * 100)
self.assertTrue(np.array_equal(expected, labels))
def test_6by2_matrix(self):
matrix = np.array([
[1.0, 0.0],
[1.1, 0.1],
[0.0, 1.0],
[0.1, 1.0],
[0.9, -0.1],
[0.0, 1.2],
])
speaker_turn_scores = [0, 0, 1.5, 0, 1.5, 1.5]
constraint_matrix = constraint.ConstraintMatrix(
speaker_turn_scores, threshold=1).compute_diagonals()
labels = configs.turntodiarize_clusterer.predict(
matrix, constraint_matrix)
labels = utils.enforce_ordered_labels(labels)
expected = np.array([0, 0, 1, 1, 0, 1])
self.assertTrue(np.array_equal(expected, labels))
def test_6by2_matrix_euclidean_dist(self):
matrix = np.array([
[1.0, 0.0],
[1.1, 0.1],
[0.0, 1.0],
[0.1, 1.0],
[0.9, -0.1],
[0.0, 1.2],
])
labels = custom_distance_kmeans.run_kmeans(matrix,
n_clusters=2,
max_iter=300,
custom_dist="euclidean")
labels = utils.enforce_ordered_labels(labels)
expected = np.array([0, 0, 1, 1, 0, 1])
self.assertTrue(np.array_equal(expected, labels))
def test_6by2_matrix(self):
matrix = np.array([
[1.0, 0.0],
[1.1, 0.1],
[0.0, 1.0],
[0.1, 1.0],
[0.9, -0.1],
[0.0, 1.2],
])
refinement_options = refinement.RefinementOptions(
gaussian_blur_sigma=0,
p_percentile=0.95,
refinement_sequence=ICASSP2018_REFINEMENT_SEQUENCE)
clusterer = spectral_clusterer.SpectralClusterer(
refinement_options=refinement_options)
labels = clusterer.predict(matrix)
labels = utils.enforce_ordered_labels(labels)
expected = np.array([0, 0, 1, 1, 0, 1])
self.assertTrue(np.array_equal(expected, labels))
def test_6by2_matrix_affinity_integration(self):
matrix = np.array([
[1.0, 0.0],
[1.1, 0.1],
[0.0, 1.0],
[0.1, 1.0],
[0.9, -0.1],
[0.0, 1.2],
])
constraint_matrix = np.array([
[1, 0, 0, 0, 0, 0],
[0, 1, 0, 0, 0, 0],
[0, 0, 1, 1, 1, 1],
[0, 0, 1, 1, 1, 1],
[0, 0, 1, 1, 1, 1],
[0, 0, 1, 1, 1, 1],
])
refinement_sequence = [
RefinementName.RowWiseThreshold, RefinementName.Symmetrize
]
refinement_options = refinement.RefinementOptions(
p_percentile=0.95,
thresholding_type=ThresholdType.Percentile,
thresholding_with_binarization=True,
thresholding_preserve_diagonal=True,
symmetrize_type=SymmetrizeType.Average,
refinement_sequence=refinement_sequence)
constraint_options = constraint.ConstraintOptions(
constraint_name=ConstraintName.AffinityIntegration,
apply_before_refinement=False,
integration_type=IntegrationType.Max)
clusterer = spectral_clusterer.SpectralClusterer(
max_clusters=2,
refinement_options=refinement_options,
constraint_options=constraint_options,
laplacian_type=LaplacianType.GraphCut,
row_wise_renorm=True)
labels = clusterer.predict(matrix, constraint_matrix)
labels = utils.enforce_ordered_labels(labels)
expected = np.array([0, 0, 1, 1, 1, 1])
self.assertTrue(np.array_equal(expected, labels))
def test_1000by6_matrix_normalized_laplacian(self):
matrix = np.array([[1.0, 0.0, 0.0, 0.0, 0.0, 0.0]] * 400 +
[[0.0, 1.0, 0.0, 0.0, 0.0, 0.0]] * 300 +
[[0.0, 0.0, 2.0, 0.0, 0.0, 0.0]] * 200 +
[[0.0, 0.0, 0.0, 1.0, 0.0, 0.0]] * 100)
noisy = np.random.rand(1000, 6) * 2 - 1
matrix = matrix + noisy * 0.1
refinement_sequence = []
refinement_options = refinement.RefinementOptions(
p_percentile=0.95, refinement_sequence=refinement_sequence)
clusterer = spectral_clusterer.SpectralClusterer(
max_clusters=4,
refinement_options=refinement_options,
laplacian_type=LaplacianType.GraphCut,
row_wise_renorm=True)
labels = clusterer.predict(matrix)
labels = utils.enforce_ordered_labels(labels)
expected = np.array([0] * 400 + [1] * 300 + [2] * 200 + [3] * 100)
self.assertTrue(np.array_equal(expected, labels))
def test_6by2_matrix_normalized_laplacian(self):
matrix = np.array([
[1.0, 0.0],
[1.1, 0.1],
[0.0, 1.0],
[0.1, 1.0],
[0.9, -0.1],
[0.0, 1.2],
])
refinement_sequence = []
refinement_options = refinement.RefinementOptions(
p_percentile=0.95, refinement_sequence=refinement_sequence)
clusterer = spectral_clusterer.SpectralClusterer(
max_clusters=2,
refinement_options=refinement_options,
laplacian_type=LaplacianType.GraphCut,
row_wise_renorm=True)
labels = clusterer.predict(matrix)
labels = utils.enforce_ordered_labels(labels)
expected = np.array([0, 0, 1, 1, 0, 1])
self.assertTrue(np.array_equal(expected, labels))
def test_small_array(self): labels = np.array([2, 2, 1, 0, 3, 3, 1]) expected = np.array([0, 0, 1, 2, 3, 3, 1]) result = utils.enforce_ordered_labels(labels) self.assertTrue(np.array_equal(expected, result))
