def test_return_dask_array(self, client):
Z = da.random.permutation(da.arange(4)).reshape(2, 2)
ncl_row = 2
ncl_col = 2
da.random.seed(1234)
_, _, row_cl, col_cl, _ = coclustering(Z, ncl_row, ncl_col, 1.e-5, 1,
1.e-8)
assert isinstance(row_cl, da.core.Array)
assert isinstance(col_cl, da.core.Array)
def test_as_many_clusters_as_elements(self, client):
# it should immediately converge (2 iterations)
ncl_row = 8
ncl_col = 7
da.random.seed(1234)
Z = da.random.randint(100, size=(ncl_row, ncl_col)).astype('float64')
conv, niterations, _, _, _ = coclustering(Z, ncl_row, ncl_col, 1.e-5,
100, 1.e-8)
assert conv
assert niterations == 2
def test_bigger_matrix(self, client):
Z = da.random.randint(100, size=(20, 15)).astype('float64')
ncl_row = 5
ncl_col = 6
da.random.seed(1234)
_, _, row_cl, col_cl, _ = coclustering(Z, ncl_row, ncl_col, 1.e-5, 100,
1.e-8)
np.testing.assert_array_equal(np.sort(np.unique(row_cl.compute())),
np.arange(ncl_row))
np.testing.assert_array_equal(np.sort(np.unique(col_cl.compute())),
np.arange(ncl_col))
def test_zero_matrix(self, client):
# special case for the error - and no nan/inf
Z = np.zeros((8, 7))
Z = da.from_array(Z)
ncl_row = 3
ncl_col = 4
epsilon = 1.e-6
da.random.seed(1234)
_, _, _, _, error = coclustering(Z, ncl_row, ncl_col, 1.e-5, 100,
epsilon)
assert np.isclose(error, Z.size * epsilon)
def test_constant_row_matrix(self, client):
# should give one cluster in columns
Z = np.repeat(np.arange(8), 7).reshape(8, 7)
Z = da.from_array(Z)
ncl_row = 3
ncl_col = 4
da.random.seed(1234)
_, _, row_cl, col_cl, _ = coclustering(Z, ncl_row, ncl_col, 1.e-5, 100,
1.e-8)
assert np.unique(row_cl.compute()).size == ncl_row
assert np.unique(col_cl.compute()).size == 1
def test_small_matrix(self, client):
da.random.seed(1234)
Z = da.random.permutation(da.arange(12)).reshape(3, 4)
ncl_row = 2
ncl_col = 3
conv, niterations, row_cl, col_cl, error = coclustering(
Z, ncl_row, ncl_col, 1.e-5, 100, 1.e-8)
assert conv
assert niterations == 3
np.testing.assert_array_equal(row_cl.compute(), np.array([1, 0, 0]))
np.testing.assert_array_equal(col_cl.compute(), np.array([1, 1, 2, 0]))
np.testing.assert_almost_equal(error, -56.457907947376775)
def test_constant_col_matrix(self, client):
# should give one cluster in rows
Z = np.tile(np.arange(7), (8, 1))
Z = da.from_array(Z)
ncl_row = 3
ncl_col = 4
da.random.seed(1234)
_, _, row_cl, col_cl, _ = coclustering(Z, ncl_row, ncl_col, 1.e-5, 100,
1.e-8)
print(col_cl.compute())
assert np.unique(row_cl.compute()).size == 1
assert np.unique(col_cl.compute()).size == ncl_col