def test_itkrmm(initial_data):
signals, dico = initial_data()
err1 = LA.norm(signals - np.dot(dico, sparse.omp_batch(signals, dico, 2)))
masks = np.ones_like(signals)
new_dico = optimize.itkrmm(signals.copy(),
masks,
dico,
n_nonzero=2,
iters=20)
new_signals = sparse.omp_batch(signals, new_dico, 2)
err2 = LA.norm(signals - np.dot(new_dico, new_signals))
assert dico.shape == new_dico.shape
assert err2 < err1
lr = optimize.reconstruct_low_rank(signals.copy(),
masks.copy(),
n_low_rank=2)
newer_dico = optimize.itkrmm(signals.copy(),
masks,
dico,
2,
20,
low_rank=lr)
newer_signals = sparse.omp_batch(signals.copy(), newer_dico, 2)
err3 = LA.norm(signals - np.dot(newer_dico, newer_signals))
assert newer_dico.shape == (dico.shape[0], dico.shape[1] + 2 * lr.shape[1])
assert err3 < err1
with pytest.warns(RuntimeWarning):
masks[:, 1] = 0
optimize.itkrmm(signals.copy(), masks, dico, 2, 20)
def test_odl(initial_data):
signals, dico = initial_data()
n_nonzero = 4
init_codes = sparse.omp_batch(signals, dico, n_nonzero)
new_dict = optimize.odl(signals, dico, n_nonzero=n_nonzero)
new_codes = sparse.omp_batch(signals, new_dict, n_nonzero)
new1 = np.dot(dico, init_codes)
new2 = np.dot(new_dict, new_codes)
err1 = LA.norm(signals - new1)
err2 = LA.norm(signals - new2)
assert err2 < err1
def test_omp_batch_sparse_signal_tol(sparse_signal):
signals = sparse_signal[:, :20]
n_atoms = 256
TOL = 100
n_threads = 1
dictionary = utils.dct_dict(n_atoms, 8)
sparse_c = sparse.omp_batch(signals, dictionary,
tol=TOL, n_threads=n_threads)
c = np.dot(dictionary, sparse_c)
assert np.all(np.linalg.norm(c - signals, axis=0) < TOL)
def test_omp_batch_sparse_signal_n_nonzero(sparse_signal):
signals = sparse_signal[:, :200]
n_atoms = 256
n_nonzero = 10
n_threads = 1
dictionary = utils.dct_dict(n_atoms, 8)
sparse_c = sparse.omp_batch(signals, dictionary,
n_nonzero=n_nonzero, n_threads=n_threads)
# Compare with sklearn
Xy = np.dot(dictionary.T, signals)
Gram = np.dot(dictionary.T, dictionary)
sparse_sk = orthogonal_mp_gram(Gram, Xy, n_nonzero_coefs=n_nonzero)
assert sparse_c.shape == sparse_sk.shape
assert np.count_nonzero(sparse_c) <= n_nonzero * signals.shape[1]
c = np.dot(dictionary, sparse_c)
sk = np.dot(dictionary, sparse_sk)
assert np.all(np.linalg.norm(c - sk, axis=0) < 1e-6)