def test_sparse_input():
"""Test that sparse matrices are accepted as input"""
from scipy.sparse import csr_matrix
A = np.abs(random_state.randn(10, 10))
A[:, 2 * np.arange(5)] = 0
T1 = nmf.ProjectedGradientNMF(n_components=5,
init='random',
random_state=999).fit_transform(A)
A_sparse = csr_matrix(A)
pg_nmf = nmf.ProjectedGradientNMF(n_components=5,
init='random',
random_state=999)
T2 = pg_nmf.fit_transform(A_sparse)
assert_array_almost_equal(
pg_nmf.reconstruction_err_,
linalg.norm(A - np.dot(T2, pg_nmf.components_), 'fro'))
assert_array_almost_equal(T1, T2)
# same with sparseness
T2 = nmf.ProjectedGradientNMF(n_components=5,
init='random',
sparseness='data',
random_state=999).fit_transform(A_sparse)
T1 = nmf.ProjectedGradientNMF(n_components=5,
init='random',
sparseness='data',
random_state=999).fit_transform(A)
def test_projgrad_nmf_sparseness():
# Test sparseness
# Test that sparsity constraints actually increase sparseness in the
# part where they are applied.
A = np.abs(random_state.randn(10, 10))
m = nmf.ProjectedGradientNMF(n_components=5, random_state=0).fit(A)
data_sp = nmf.ProjectedGradientNMF(n_components=5, sparseness='data',
random_state=0).fit(A).data_sparseness_
comp_sp = nmf.ProjectedGradientNMF(n_components=5, sparseness='components',
random_state=0).fit(A).comp_sparseness_
assert_greater(data_sp, m.data_sparseness_)
assert_greater(comp_sp, m.comp_sparseness_)
def test_sparse_input():
"""Test that sparse matrices are accepted as input"""
from scipy.sparse import csr_matrix
A = np.abs(random_state.randn(10, 10))
A[:, 2 * np.arange(5)] = 0
T1 = nmf.ProjectedGradientNMF(n_components=5,
init='random',
random_state=999).fit_transform(A)
A = csr_matrix(A)
T2 = nmf.ProjectedGradientNMF(n_components=5,
init='random',
random_state=999).fit_transform(A)
assert_array_almost_equal(T1, T2)
def test_projgrad_nmf_fit_nn_output():
"""Test that the decomposition does not contain negative values"""
A = np.c_[5 * np.ones(5) - xrange(1, 6), 5 * np.ones(5) + xrange(1, 6)]
for init in (None, 'nndsvd', 'nndsvda', 'nndsvdar'):
model = nmf.ProjectedGradientNMF(n_components=2, init=init)
transf = model.fit_transform(A)
assert_false((model.components_ < 0).any() or (transf < 0).any())
def test_projgrad_nmf_transform():
# Test that NMF.transform returns close values
# (transform uses scipy.optimize.nnls for now)
A = np.abs(random_state.randn(6, 5))
m = nmf.ProjectedGradientNMF(n_components=5, init='nndsvd', random_state=0)
transf = m.fit_transform(A)
assert_true(np.allclose(transf, m.transform(A), atol=1e-2, rtol=0))
def test_projgrad_nmf_fit_close():
"""Test that the fit is not too far away"""
assert_true(
nmf.ProjectedGradientNMF(5, init='nndsvda').fit(
np.abs(random_state.randn(6, 5))).reconstruction_err_ < 0.05)
def test_projgrad_nmf_fit_nn_input():
"""Test model fit behaviour on negative input"""
A = -np.ones((2, 2))
m = nmf.ProjectedGradientNMF(n_components=2, init=None)
m.fit(A)
def test_n_components_greater_n_features():
"""Smoke test for the case of more components than features."""
A = np.abs(random_state.randn(30, 10))
nmf.ProjectedGradientNMF(n_components=15, sparseness='data').fit(A)
def test_projgrad_nmf_fit_close():
"""Test that the fit is not too far away"""
pnmf = nmf.ProjectedGradientNMF(5, init='nndsvda', random_state=0)
X = np.abs(random_state.randn(6, 5))
assert_less(pnmf.fit(X).reconstruction_err_, 0.05)
def test_projgrad_nmf_fit_nn_input():
# Test model fit behaviour on negative input
A = -np.ones((2, 2))
m = nmf.ProjectedGradientNMF(n_components=2, init=None, random_state=0)
m.fit(A)
