def test_sort_features(self):
x = np.array([[0, 2, 30, 10],
[1, 2, 30, 10],
[0, 3, 33, 10],
[2, 5, 30, 7],
[2, 5, 30, 9]])
x = spsp.csr_matrix(x)
sdm = eda.SampleDistanceMatrix(
x, metric='euclidean')
sdm2 = eda.SampleDistanceMatrix(
x, metric='euclidean')
sdm2.sort_features(fdist_metric='euclidean', optimal_ordering=True)
assert sdm2.fids == [2, 3, 1, 0]
def test_empty_init(self):
# sparse matrix has 2 dims
eda.SampleDistanceMatrix(
spsp.csr_matrix(np.empty(0)),
metric='euclidean')
sdm = eda.SampleDistanceMatrix(
spsp.csr_matrix(np.empty((0, 0))), metric='euclidean')
assert len(sdm.sids) == 0
assert len(sdm.fids) == 0
assert sdm._x.shape == (0, 0)
assert sdm._d.shape == (0, 0)
assert sdm._col_sorted_d.shape == (0, 0)
assert sdm._col_argsorted_d.shape == (0, 0)
assert sdm.tsne(n_iter=250).shape == (0, 0)
def test_pca_var_explained(self):
np.random.seed(123)
x5k = np.random.normal(size=5000)
x5k = spsp.csr_matrix(x5k)
sdm = eda.SampleDistanceMatrix(x5k.reshape(20, -1))
assert sdm._skd_pca.explained_variance_.shape == (20,)
assert sdm._skd_pca.explained_variance_ratio_.shape == (20,)
def test_init_wrong_d_type(self):
d_3x3 = np.array([[0, np.sqrt(2), np.sqrt(8)],
['1a1', 0, np.sqrt(2)],
[np.sqrt(8), np.sqrt(2), 0]])
with pytest.raises(Exception) as excinfo:
eda.SampleDistanceMatrix(self.x_3x2, d_3x3)
def test_s_ith_nn_d_dist(self):
nn_sdm = eda.SampleDistanceMatrix(
spsp.csr_matrix([[0, 0, 0], [1, 1, 1], [5, 5, 5],
[6, 6, 6], [10, 10, 10],
[20, 20, 20]]),
metric='euclidean')
nn_sdm.s_ith_nn_d_dist(1)
def test_no_pdist_detect_rare_samples_wrong_args(self):
tsdm = eda.SampleDistanceMatrix(
[[0, 0], [1, 1], [200, 200], [200, 200], [200, 200], [100, 100],
[101, 101], [99, 99], [100, 100], [102, 102]],
metric="euclidean",
use_pdist=False)
rsd = knn.RareSampleDetection(tsdm)
# Invalid parameters
with pytest.raises(ValueError) as excinfo:
rsd.detect_rare_samples(0, 1, 1)
with pytest.raises(ValueError) as excinfo:
rsd.detect_rare_samples(1, 0, 1)
with pytest.raises(ValueError) as excinfo:
rsd.detect_rare_samples(1, 1, 0)
with pytest.raises(ValueError) as excinfo:
rsd.detect_rare_samples(1, 1, 0.5)
with pytest.raises(ValueError) as excinfo:
rsd.detect_rare_samples(0.5, 1, 1)
with pytest.raises(ValueError) as excinfo:
rsd.detect_rare_samples(1, -0.1, 1)
# Parameters of different length
with pytest.raises(ValueError) as excinfo:
rsd.detect_rare_samples([1, 2], 1, 1)
with pytest.raises(ValueError) as excinfo:
rsd.detect_rare_samples(1, [1, 2], 1)
with pytest.raises(ValueError) as excinfo:
rsd.detect_rare_samples(1, 1, [1, 2])
def test_id_x(self):
sids = list("abcdef")
fids = list(range(10, 20))
sdm = eda.SampleDistanceMatrix(
spsp.csr_matrix(np.random.ranf(60).reshape(6, -1)),
sids=sids, fids=fids)
# select sf
ss_sdm = sdm.id_x(['a', 'f'], list(range(10, 15)))
assert ss_sdm._x.shape == (2, 5)
assert ss_sdm.sids == ['a', 'f']
assert ss_sdm.fids == list(range(10, 15))
np.testing.assert_equal(
ss_sdm.d, sdm._d[np.ix_((0, 5), (0, 5))])
# select with Default
ss_sdm = sdm.id_x()
assert ss_sdm._x.shape == (6, 10)
assert ss_sdm.sids == list("abcdef")
assert ss_sdm.fids == list(range(10, 20))
np.testing.assert_equal(ss_sdm.d, sdm._d)
# select with None
ss_sdm = sdm.id_x(None, None)
assert ss_sdm._x.shape == (6, 10)
assert ss_sdm.sids == list("abcdef")
assert ss_sdm.fids == list(range(10, 20))
np.testing.assert_equal(ss_sdm.d, sdm._d)
# select non-existent inds
# id lookup raises ValueError
with pytest.raises(ValueError) as excinfo:
sdm.id_x([6])
with pytest.raises(ValueError) as excinfo:
sdm.id_x(None, ['a'])
def test_impute_features_stat_fun(self):
tsdm = eda.SampleDistanceMatrix(spsp.csr_matrix(self.tx),
metric="euclidean",
use_pdist=False)
tsdm.s_knn_connectivity_matrix(5, use_hnsw=False)
fmp = knn.FeatureImputation(tsdm)
res_sdml = fmp.impute_features([8], [3], [0.5], [10])
res_sdml2 = fmp.impute_features(8, 3, 0.5, 10, 1, np.median)
np.testing.assert_equal(res_sdml2[0]._x.A, res_sdml[0]._x.A)
assert len(fmp._res_lut) == 1
res_sdml3 = fmp.impute_features(
8, 3, 0.5, 10, 1, lambda x, axis=0: np.median(x, axis=axis))
np.testing.assert_equal(res_sdml3[0]._x.A, res_sdml[0]._x.A)
assert len(fmp._res_lut) == 2
res_sdml4 = fmp.impute_features(
8, 3, 0.5, 10, 1, lambda x, axis=0: np.median(x, axis=axis))
np.testing.assert_equal(res_sdml4[0]._x.A, res_sdml[0]._x.A)
assert len(fmp._res_lut) == 3
res_sdml5 = fmp.impute_features(8, 3, 0.5, 10, 1, np.min)
assert not np.array_equal(res_sdml5[0]._x.A, res_sdml[0]._x.A)
assert len(fmp._res_lut) == 4
res_sdml6 = fmp.impute_features(8,
3,
0.5,
10,
1,
lambda x, axis=0: np.min(x, axis=axis))
np.testing.assert_equal(res_sdml5[0]._x.A, res_sdml6[0]._x.A)
assert len(fmp._res_lut) == 5
def test_put_tsne_wrong_args(self):
tmet = 'euclidean'
sdm = eda.SampleDistanceMatrix(self.x_3x2, metric=tmet)
with pytest.raises(ValueError) as excinfo:
sdm.put_tsne(1, [1, 2, 3])
with pytest.raises(ValueError) as excinfo:
sdm.put_tsne({1: 2}, [1, 2, 3])
def test_get_tsne_kv(self):
tmet = 'euclidean'
sdm = eda.SampleDistanceMatrix(self.x_3x2, metric=tmet)
assert sdm.get_tsne_kv(1) is None
assert sdm.get_tsne_kv(1) is None
assert sdm.get_tsne_kv(0) is None
assert sdm.get_tsne_kv(2) is None
def run_test_impute_features(self, nprocs, use_hnsw):
tsdm = eda.SampleDistanceMatrix(spsp.csr_matrix(self.tx),
metric="euclidean",
use_pdist=False)
tsdm.s_knn_connectivity_matrix(5, use_hnsw=use_hnsw)
fmp = knn.FeatureImputation(tsdm)
res_sdml = fmp.impute_features([1, 3, 5], [1, 1, 3], [0.5, 1.5, 0.5],
[1, 1, 5],
nprocs=nprocs)
assert len(fmp._res_lut) == 3
assert res_sdml[0]._x.A.shape == (9, 2)
assert res_sdml[1]._x.A.shape == (9, 2)
assert res_sdml[2]._x.A.shape == (9, 2)
# value
ref_res0 = np.array([[1, 1], [1, 1], [1, 2], [2, 3], [2, 5], [0, 100],
[0, 100], [0, 101], [10, 100]])
np.testing.assert_equal(res_sdml[0]._x.A, ref_res0)
ref_res1 = np.array([[np.median([1, 1, 2]),
np.median([1, 2, 3])],
[np.median([1, 1, 2]),
np.median([1, 2, 3])], [np.median([1, 1, 2]), 2],
[2, 3], [2, 5], [np.median([0, 0, 10]), 100],
[np.median([0, 0, 10]), 100],
[np.median([0, 0, 10]), 101], [10, 100]])
np.testing.assert_equal(res_sdml[1]._x.A, ref_res1)
assert res_sdml[2]._x.A[0, 0] > 0
assert res_sdml[2]._x.A[0, 1] > 0
assert res_sdml[2]._x.A[5, 0] > 0
assert res_sdml[2]._x.A[6, 0] > 0
assert res_sdml[2]._x.A[7, 0] > 0
# lookup
np.testing.assert_equal(fmp._res_lut[(1, 1, 0.5, 1, np.median)][0].A,
res_sdml[0]._x.A)
np.testing.assert_equal(fmp._res_lut[(3, 1, 1.5, 1, np.median)][0].A,
res_sdml[1]._x.A)
np.testing.assert_equal(fmp._res_lut[(5, 3, 0.5, 5, np.median)][0].A,
res_sdml[2]._x.A)
fmp.impute_features([1, 3, 5], [1, 1, 3], [0.5, 1.5, 0.5], [1, 1, 5],
nprocs=nprocs)
assert len(fmp._res_lut) == 3
np.testing.assert_equal(fmp._sdm.x.A.tolist(), self.tx)
# run results should be placed with the correct order
fmp2 = knn.FeatureImputation(tsdm)
res1 = fmp2.impute_features([1, 3, 5], [1, 1, 3], [0.5, 1.5, 0.5],
[1, 1, 5],
nprocs=nprocs)
fmp3 = knn.FeatureImputation(tsdm)
res2 = fmp3.impute_features([2, 3, 4], [1, 1, 3], [0.5, 1.5, 0.5],
[1, 1, 5],
nprocs=nprocs)
res3 = fmp2.impute_features([2, 3, 4], [1, 1, 3], [0.5, 1.5, 0.5],
[1, 1, 5],
nprocs=nprocs)
np.testing.assert_equal(res1[1]._x.A, res3[1]._x.A)
np.testing.assert_equal(res2[0]._x.A, res3[0]._x.A)
np.testing.assert_equal(res2[2]._x.A, res3[2]._x.A)
def test_provide_graph(self):
sdm = eda.SampleDistanceMatrix(self.x_20x5)
knn_conn_mat = sdm.s_knn_connectivity_matrix(5)
knn_aff_graph = eda.SampleDistanceMatrix.knn_conn_mat_to_aff_graph(
knn_conn_mat, 2)
cluster.Community(
self.x_20x5, graph=knn_aff_graph,
partition_method="RBConfigurationVertexPartition").labs
def test_s_ith_nn_d(self):
nn_sdm = eda.SampleDistanceMatrix(
spsp.csr_matrix([[0], [1], [5], [6], [10], [20]]),
metric='euclidean')
np.testing.assert_allclose([0, 0, 0, 0, 0, 0],
nn_sdm.s_ith_nn_d(0))
np.testing.assert_allclose([1, 1, 1, 1, 4, 10],
nn_sdm.s_ith_nn_d(1))
np.testing.assert_allclose([5, 4, 4, 4, 5, 14],
nn_sdm.s_ith_nn_d(2))
def test_cosine_pdist(self):
np.random.seed(222)
x = np.random.ranf(10000).reshape(500, -1)
x = spsp.csr_matrix(x)
skd = sklearn.metrics.pairwise.pairwise_distances(x, metric='cosine')
np.testing.assert_allclose(
eda.SampleDistanceMatrix.cosine_pdist(x), skd)
np.testing.assert_allclose(
eda.SampleDistanceMatrix(x, metric='cosine')._d, skd)
def test_sdm_truncated_pca_plot(self):
sids = list(range(8))
fids = [str(i) for i in range(10)]
np.random.seed(123)
x = np.random.ranf(80).reshape(8, -1)
x_sorted = x[np.argsort(x[:, 5])]
g = x_sorted[:, 5]
x_sorted = spsp.csr_matrix(x_sorted)
sdm = eda.SampleDistanceMatrix(x_sorted, sids=sids, fids=fids)
return sdm.pca_plot(gradient=g, figsize=(10, 10), s=50)
def test_detect_rare_samples_single_run(self):
tsdm = eda.SampleDistanceMatrix(
[[0, 0], [1, 1], [200, 200], [101, 101], [99, 99], [100, 100],
[102, 102]],
metric="euclidean")
rsd = knn.RareSampleDetection(tsdm)
resl = rsd.detect_rare_samples(1, 0.1, 1)
assert resl[0] == []
resl2 = rsd.detect_rare_samples(1, 0.1, 5)
assert resl2[0] == []
def test_knn_filter_samples_single_run(self):
tsdm = eda.SampleDistanceMatrix(
[[0, 0], [1, 1], [200, 200], [101, 101], [99, 99], [100, 100],
[102, 102]],
metric="euclidean")
skf = qc.SampleKNNFilter(tsdm)
resl = skf.knn_filter_samples(1, 0.1, 1)
assert resl[0] == []
resl2 = skf.knn_filter_samples(1, 0.1, 5)
assert resl2[0] == []
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_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_s_ith_nn_ind(self):
nn_sdm = eda.SampleDistanceMatrix(
spsp.csr_matrix([[0, 0, 0], [1, 1, 1], [5, 5, 5],
[6, 6, 6], [10, 10, 10],
[20, 20, 20]]),
metric='euclidean')
np.testing.assert_allclose([0, 1, 2, 3, 4, 5],
nn_sdm.s_ith_nn_ind(0))
np.testing.assert_allclose([1, 0, 3, 2, 3, 4],
nn_sdm.s_ith_nn_ind(1))
np.testing.assert_allclose([2, 2, 1, 4, 2, 3],
nn_sdm.s_ith_nn_ind(2))
def test_hnsw_detect_rare_samples_single_run(self):
tsdm = eda.SampleDistanceMatrix(
[[0.00001, 0.00001], [1, 1], [200, 200], [101, 101], [99, 99],
[100, 100], [102, 102]],
metric="euclidean",
use_pdist=False)
rsd = knn.RareSampleDetection(tsdm)
res = rsd._no_pdist_rare_s_detect(1, 0.1, 1)
resl = rsd.detect_rare_samples(1, 0.1, 1, use_pca=True, use_hnsw=True)
assert resl[0] == []
resl2 = rsd.detect_rare_samples(1, 0.1, 5, use_pca=True, use_hnsw=True)
assert resl2[0] == []
def test_impute_features_single_run(self):
tsdm = eda.SampleDistanceMatrix(spsp.csr_matrix(self.tx),
metric="euclidean",
use_pdist=False)
tsdm.s_knn_connectivity_matrix(5, use_hnsw=False)
fmp = knn.FeatureImputation(tsdm)
res_sdml = fmp.impute_features([8], [3], [0.5], [10])
fmp2 = knn.FeatureImputation(tsdm)
res_sdml2 = fmp2.impute_features(8, 3, 0.5, 10)
np.testing.assert_equal(res_sdml2[0]._x.A, res_sdml[0]._x.A)
tsdm = eda.SampleDistanceMatrix(spsp.csr_matrix(self.tx),
metric="euclidean",
use_pdist=False)
tsdm.s_knn_connectivity_matrix(5, use_hnsw=True)
fmp = knn.FeatureImputation(tsdm)
res_sdml = fmp.impute_features([8], [3], [0.5], [10])
fmp2 = knn.FeatureImputation(tsdm)
res_sdml2 = fmp2.impute_features(8, 3, 0.5, 10, verbose=True)
np.testing.assert_equal(res_sdml2[0]._x.A, res_sdml[0]._x.A)
def test_no_pdist_detect_rare_samples_par(self):
tsdm = eda.SampleDistanceMatrix(
[[0, 0], [1, 1], [200, 200], [200, 200], [200, 200], [100, 100],
[101, 101], [99, 99], [100, 100], [102, 102]],
metric="euclidean",
use_pdist=False)
rsd = knn.RareSampleDetection(tsdm)
resl = rsd.detect_rare_samples([3, 4, 5], [10] * 3, [5, 6, 7], 3)
assert resl == [list(range(5, 10)), list(range(5, 10)), []]
assert len(rsd._res_lut) == 3
assert rsd._res_lut[(3, 10, 5)][1][-1] == resl[0]
assert rsd._res_lut[(4, 10, 6)][1][-1] == resl[1]
assert rsd._res_lut[(5, 10, 7)][1][-1] == resl[2]
def test_s_knn_graph_grad_lab_same_marker(self):
np.random.seed(123)
x = np.concatenate((np.random.normal(0, 1, 10),
np.random.normal(20, 1, 20))).reshape(30, -1)
x = spsp.csr_matrix(x)
sdm = eda.SampleDistanceMatrix(x, metric='euclidean')
sdm.s_knn_graph(5, figsize=(5, 5))
assert (5, 1) in sdm._knn_ng_lut
assert len(sdm._knn_ng_lut) == 1
gradient = np.array([1] * 10 + [10] * 20)
labs = gradient = np.array([1] * 10 + [2] * 20)
return sdm.s_knn_graph(5, gradient=gradient, labels=labs,
different_label_markers=False,
figsize=(5, 5),
alpha=0.8, random_state=123)
def test_tsne_default_init(self):
tmet = 'euclidean'
tsne_kwargs = {'metric': tmet, 'n_iter': 250,
'random_state': 123}
ref_tsne = eda.tsne(self.x_3x2.toarray(), **tsne_kwargs)
sdm = eda.SampleDistanceMatrix(self.x_3x2, metric=tmet)
init_tsne = sdm._last_tsne
assert init_tsne.shape == (3, 2)
assert len(sdm.tsne_lut) == 1
tsne2 = sdm.tsne(store_res=True, **tsne_kwargs)
np.testing.assert_allclose(ref_tsne, tsne2)
assert len(sdm.tsne_lut) == 2
def test_no_pdist_detect_rare_samples_empty_subset(self):
tsdm = eda.SampleDistanceMatrix(
[[0, 0], [1, 1], [200, 200], [200, 200], [200, 200], [100, 100],
[101, 101], [99, 99], [100, 100], [102, 102]],
metric="euclidean",
use_pdist=False)
rsd = knn.RareSampleDetection(tsdm)
resl = rsd.detect_rare_samples(1, 10, 5)
resl2 = rsd.detect_rare_samples([1], [10], 5)
# scalar and list params should have the same results
assert resl == resl2
# result lut should be the same length
assert len(rsd._res_lut) == 1
assert rsd._res_lut[(1, 10, 5)][1][-1] == resl[0]
def test_init_wrong_d_size(self):
d_2x2 = np.array([[0, np.sqrt(2)],
[np.sqrt(2), 0]])
d_2x2 = np.array([[0, np.sqrt(2)],
[np.sqrt(2), 0]])
d_1x6 = np.arange(6)
d_3x2 = np.array([[0, np.sqrt(2)],
[np.sqrt(2), 0],
[1, 2]])
with pytest.raises(Exception) as excinfo:
eda.SampleDistanceMatrix(self.x_3x2, d_2x2)
with pytest.raises(Exception) as excinfo:
eda.SampleDistanceMatrix(self.x_3x2, d_3x2)
with pytest.raises(Exception) as excinfo:
eda.SampleDistanceMatrix(self.x_3x2, d_3x2)
with pytest.raises(Exception) as excinfo:
eda.SampleDistanceMatrix(self.x_3x2, d_1x6)
def test_sdm_umap_feature_gradient_plus10_plot_sparse(self):
sids = list(range(8))
fids = [str(i) for i in range(10)]
np.random.seed(123)
x = np.random.ranf(80).reshape(8, -1)
x_sorted = x[np.argsort(x[:, 5])]
x_sorted = spsp.csr_matrix(x_sorted)
sdm = eda.SampleDistanceMatrix(
x_sorted, sids=sids, fids=fids)
fig = sdm.umap_feature_gradient_plot(
'5', transform=lambda x: x + 10, figsize=(10, 10), s=50)
assert (sdm._x != x_sorted).nnz == 0
np.testing.assert_equal(sdm._sids, sids)
np.testing.assert_equal(sdm._fids, fids)
return fig
def test_knn_ind_lut(self):
nn_sdm = eda.SampleDistanceMatrix(
spsp.csr_matrix([[0, 0, 0], [1, 1, 1], [5, 5, 5],
[6, 6, 6], [10, 10, 10],
[20, 20, 20]]),
metric='euclidean')
assert nn_sdm.s_knn_ind_lut(0) == dict(zip(range(6), [[]]*6))
assert (nn_sdm.s_knn_ind_lut(1) ==
dict(zip(range(6), [[1], [0], [3], [2], [3], [4]])))
assert (nn_sdm.s_knn_ind_lut(2) ==
dict(zip(range(6), [[1, 2], [0, 2], [3, 1],
[2, 4], [3, 2], [4, 3]])))
assert (nn_sdm.s_knn_ind_lut(3) ==
dict(zip(range(6), [[1, 2, 3], [0, 2, 3], [3, 1, 0],
[2, 4, 1], [3, 2, 1], [4, 3, 2]])))
nn_sdm.s_knn_ind_lut(5)
def test_sdm_truncated_pca_feature_gradient_plot_sslabs_empty(self):
sids = list(range(8))
fids = [str(i) for i in range(10)]
np.random.seed(123)
x = np.random.ranf(80).reshape(8, -1)
x_sorted = x[np.argsort(x[:, 5])]
x_sorted = spsp.csr_matrix(x_sorted)
sdm = eda.SampleDistanceMatrix(
x_sorted, sids=sids, fids=fids)
fig = sdm.pca_feature_gradient_plot(
'5', labels=list('abcdefgh'), selected_labels=[],
figsize=(10, 10), s=50)
assert (sdm._x != x_sorted).nnz == 0
np.testing.assert_equal(sdm._sids, sids)
np.testing.assert_equal(sdm._fids, fids)
return fig
