def test_mirac_wrong_args(self):
x = np.zeros((10, 10))
# wrong min_cl_n
with pytest.raises(ValueError) as excinfo:
cluster.MIRAC(x, metric='euclidean', min_cl_n=-0.1)
with pytest.raises(ValueError) as excinfo:
cluster.MIRAC(x, metric='euclidean', min_cl_n=-0.1)
# wrong cl_mdl_scale_factor
with pytest.raises(ValueError) as excinfo:
cluster.MIRAC(x, metric='euclidean', cl_mdl_scale_factor=-0.1)
# wrong encode type
with pytest.raises(ValueError) as excinfo:
cluster.MIRAC(x, metric='euclidean', encode_type='1')
with pytest.raises(ValueError) as excinfo:
cluster.MIRAC(x, metric='euclidean', encode_type=1)
with pytest.raises(ValueError) as excinfo:
cluster.MIRAC(x, metric='euclidean', dim_reduct_method='NONN')
# hac tree n_leaves different from n_samples
z = sch.linkage([[0], [5], [6], [8], [9], [12]],
method='single', optimal_ordering=True)
hct = eda.HClustTree(sch.to_tree(z))
with pytest.raises(ValueError) as excinfo:
cluster.MIRAC(x, metric='euclidean', hac_tree=hct)
def test_mirac_dim_reduction(self):
tx, tlab = skdset.make_blobs(n_samples=100, n_features=2,
centers=10, random_state=8927)
tx = tx - tx.min()
sdm = eda.SampleDistanceMatrix(tx, metric='euclidean')
m = cluster.MIRAC(sdm._x, sdm._d, metric='euclidean', linkage='ward',
min_cl_n=35, cl_mdl_scale_factor=1,
dim_reduct_method='PCA', verbose=True)
assert len(m.labs) == 100
hct = eda.HClustTree.hclust_tree(sdm._d, linkage='ward',
optimal_ordering=True)
m2 = cluster.MIRAC(sdm._x, hac_tree=hct, min_cl_n=35,
cl_mdl_scale_factor=1, encode_type='data',
mdl_method=eda.mdl.ZeroIGKdeMdl,
dim_reduct_method='t-SNE', verbose=False)
assert m.labs == m2.labs
assert m2._sdm._lazy_load_d is None
tx2, tlab2 = skdset.make_blobs(n_samples=500, n_features=50,
centers=5, cluster_std=15,
random_state=8927)
tx2 = tx2 - tx2.min()
cluster.MIRAC(tx2, metric='euclidean', min_cl_n=15,
optimal_ordering=False, cl_mdl_scale_factor=0,
dim_reduct_method='UMAP', verbose=True)
# auto to encode distance
tx3, tlab3 = skdset.make_blobs(n_samples=100, n_features=101,
cluster_std=0.01, centers=5,
random_state=8927)
tx3 = tx3 - tx3.min()
sdm = eda.SampleDistanceMatrix(tx3, metric='euclidean')
m = cluster.MIRAC(sdm._x, sdm._d, metric='euclidean', linkage='ward',
min_cl_n=5, cl_mdl_scale_factor=1,
dim_reduct_method='t-SNE', verbose=True)
# empty
sdm = eda.SampleDistanceMatrix([[], []], metric='euclidean')
with pytest.raises(Exception) as excinfo:
m = cluster.MIRAC(sdm._x, sdm._d, metric='euclidean',
linkage='ward',
min_cl_n=35, cl_mdl_scale_factor=1,
dim_reduct_method='PCA', verbose=True)
def test_mirac_rerun(self):
tx, tlab = skdset.make_blobs(n_samples=100, n_features=2,
centers=10, random_state=8927)
tx = tx - tx.min()
m = cluster.MIRAC(tx, min_cl_n=35,
cl_mdl_scale_factor=1, encode_type='data',
mdl_method=eda.mdl.ZeroIGKdeMdl,
dim_reduct_method='t-SNE', verbose=False)
m_tsne = m._sdm._x
m.tune_parameters(0, 5, 0.1, 1)
assert m_tsne is m._sdm._x
def test_mirac_run_300x1000(self):
mirac_res = cluster.MIRAC(self.x300x1000,
metric="correlation",
encode_type="distance",
min_cl_n=10,
min_split_mdl_red_ratio=0.05,
nprocs=3)
assert len(mirac_res.labs) == 300
ulabs, ulab_cnts = np.unique(mirac_res.labs, return_counts=True)
np.testing.assert_equal(ulab_cnts, [60] * 5)
mirac_res_ulab_ind = [
np.where(mirac_res._labs == ulab)[0] for ulab in ulabs
]
for i in mirac_res_ulab_ind:
assert len(np.unique(self.labs300[i])) == 1
def test_mirac_run_100x500(self):
mirac_res = cluster.MIRAC(self.x100x500,
metric="correlation",
min_cl_n=35,
encode_type="data",
cl_mdl_scale_factor=1,
min_split_mdl_red_ratio=0.1,
nprocs=1,
verbose=True)
assert len(mirac_res.labs) == 100
ulabs, ulab_cnts = np.unique(mirac_res.labs, return_counts=True)
np.testing.assert_equal(ulab_cnts, [50] * 2)
mirac_res_ulab_ind = [
np.where(mirac_res._labs == ulab)[0] for ulab in ulabs
]
for i in mirac_res_ulab_ind:
assert len(np.unique(self.labs100[i])) == 1
def test_mirac_cover_tests(self):
# make all non-neg
x = np.zeros((10, 10))
cluster.MIRAC(x, metric='euclidean', min_cl_n=25)
cluster.MIRAC(x, metric='euclidean', min_cl_n=25)
cluster.MIRAC(x, metric='euclidean', min_cl_n=25, verbose=True)
tx, tlab = skdset.make_blobs(n_samples=100, n_features=2,
centers=10, random_state=8927)
tx = tx - tx.min()
sdm = eda.SampleDistanceMatrix(tx, metric='euclidean')
m = cluster.MIRAC(sdm._x, sdm._d, metric='euclidean', linkage='ward',
min_cl_n=35, cl_mdl_scale_factor=1, verbose=True)
assert len(m.labs) == 100
hct = eda.HClustTree.hclust_tree(sdm._d, linkage='ward',
optimal_ordering=True)
m2 = cluster.MIRAC(sdm._x, hac_tree=hct, min_cl_n=35,
cl_mdl_scale_factor=1, encode_type='data',
mdl_method=eda.mdl.ZeroIGKdeMdl, verbose=False)
assert m.labs == m2.labs
assert m2._sdm._lazy_load_d is None
tx2, tlab2 = skdset.make_blobs(n_samples=500, n_features=5,
centers=5, cluster_std=1,
random_state=8927)
tx2 = tx2 - tx2.min()
cluster.MIRAC(tx2, metric='euclidean', min_cl_n=15,
optimal_ordering=False, cl_mdl_scale_factor=0,
verbose=True)
# auto to encode distance
tx3, tlab3 = skdset.make_blobs(n_samples=100, n_features=101,
cluster_std=0.01, centers=5,
random_state=8927)
tx3 = tx3 - tx3.min()
sdm = eda.SampleDistanceMatrix(tx3, metric='euclidean')
m = cluster.MIRAC(sdm._x, sdm._d, metric='euclidean', linkage='ward',
min_cl_n=5, cl_mdl_scale_factor=1, verbose=True)
assert m._encode_type == 'distance'
# empty
sdm = eda.SampleDistanceMatrix([[], []], metric='euclidean')
m = cluster.MIRAC(sdm._x, sdm._d, metric='euclidean', linkage='ward',
min_cl_n=35, cl_mdl_scale_factor=1, verbose=True)
def test_mirac_dmat_heatmap(self):
# create a dummy MIRAC object
sdm = eda.SampleDistanceMatrix([[], []], metric='euclidean')
m = cluster.MIRAC(sdm._x, sdm._d, metric='euclidean', linkage='ward',
min_cl_n=35, cl_mdl_scale_factor=1, verbose=True)
# empty tree
m._hac_tree = eda.HClustTree(None)
assert m.dmat_heatmap() is None
# normal tree and d
sdm_5x2 = eda.SampleDistanceMatrix([[0, 0],
[100, 100],
[1, 1],
[101, 101],
[80, 80]],
metric='euclidean')
# This tree should be
# _______|_____
# | ____|___
# __|___ | __|___
# | | | | |
# 0 2 4 1 3
# Leaves are in optimal order.
hct = eda.HClustTree.hclust_tree(sdm_5x2.d, linkage='auto')
m._hac_tree = hct
m._sdm = sdm_5x2
# invalid labels
# leaf order: [0, 2, 4, 1, 3]
# leaf labs : [1, 1, 2, 3, 2]
m._labs = [1, 3, 1, 2, 2]
with pytest.raises(ValueError) as excinfo:
m.dmat_heatmap()
m._labs = ['1', '3', '1', '2', '2']
with pytest.raises(ValueError) as excinfo:
m.dmat_heatmap()
# leaf order: [0, 2, 4, 1, 3]
# leaf labs : [5, 1, 2, 1, 2]
m._labs = [5, 1, 1, 2, 2]
with pytest.raises(ValueError) as excinfo:
m.dmat_heatmap()
# leaf order: [0, 2, 4, 1, 3]
# leaf labs : [5, 1, 2, 3, 2]
m._labs = [5, 3, 1, 2, 2]
with pytest.raises(ValueError) as excinfo:
m.dmat_heatmap()
# leaf order: [0, 2, 4, 1, 3]
# leaf labs : [5, 10, 20, 3, 5]
m._labs = [5, 3, 10, 5, 20]
with pytest.raises(ValueError) as excinfo:
m.dmat_heatmap()
# valid labels
# leaf order: [0, 2, 4, 1, 3]
# leaf labs : [2, 2, 1, 1, 0]
m._labs = [2, 1, 2, 0, 1]
m.dmat_heatmap()
m._labs = ['2', '1', '2', '0', '1']
m.dmat_heatmap()
# leaf order: [0, 2, 4, 1, 3]
# leaf labs : [0, 2, 4, 1, 3]
m._labs = [0, 1, 2, 3, 4]
m.dmat_heatmap()
# leaf order: [0, 2, 4, 1, 3]
# leaf labs : [0, 0, 0, 0, 0]
m._labs = [0]*5
m.dmat_heatmap()
# have a figure
# leaf order: [ 0 , 2 , 4 , 1 , 3 ]
# leaf labs : ['b', 'l', 'r', 'e', 'o']
m._labs = ['beginning', 'end', 'l', 'out', 'r']
fig = m.dmat_heatmap(
selected_labels=['beginning', 'end', 'l', 'r'])
assert fig is not None
return fig