def test_whitening_on_whitened():
data = np.random.normal(size=(1000, 50))
from sklearn.decomposition import PCA
data = PCA(whiten=True).fit_transform(data)
cov = Covariance().fit(data).fetch_model()
whitened = cov.whiten(data)
np.testing.assert_array_almost_equal(whitened, data)
def test_XX_weighted_meanconst(self):
est = Covariance(lagtime=self.lag, compute_c0t=False, bessels_correction=False)
cc = est.fit(self.data - self.mean_const, weights=self.data_weights).fetch_model()
np.testing.assert_allclose(cc.mean_0, self.mx_c_wobj_lag0)
np.testing.assert_allclose(cc.cov_00, self.Mxx_c_wobj_lag0)
cc = est.fit(self.data - self.mean_const, weights=self.data_weights, column_selection=self.cols_2).fetch_model()
np.testing.assert_allclose(cc.cov_00, self.Mxx_c_wobj_lag0[:, self.cols_2])
def test_XX_weightobj_meanfree(self):
# many passes
est = Covariance(lagtime=self.lag, compute_c0t=False, remove_data_mean=True, bessels_correction=False)
cc = est.fit(self.data, weights=self.data_weights, n_splits=10).fetch_model()
np.testing.assert_allclose(cc.mean_0, self.mx_wobj_lag0)
np.testing.assert_allclose(cc.cov_00, self.Mxx0_wobj_lag0)
cc = est.fit(self.data, column_selection=self.cols_2, weights=self.data_weights).fetch_model()
np.testing.assert_allclose(cc.cov_00, self.Mxx0_wobj_lag0[:, self.cols_2])
def test_XX_with_mean(self):
# many passes
est = Covariance(lagtime=self.lag, compute_c0t=False, remove_data_mean=False, bessels_correction=False)
cc = est.fit(self.data).fetch_model()
np.testing.assert_allclose(cc.mean_0, self.mx_lag0)
np.testing.assert_allclose(cc.cov_00, self.Mxx_lag0)
cc = est.fit(self.data, column_selection=self.cols_2).fetch_model()
np.testing.assert_allclose(cc.cov_00, self.Mxx_lag0[:, self.cols_2])
def test_XY_sym_meanconst(self):
est = Covariance(lagtime=self.lag, compute_c0t=True, reversible=True, bessels_correction=False)
cc = est.fit(self.Xc_lag0).fetch_model()
np.testing.assert_allclose(cc.mean_0, self.m_c_sym)
np.testing.assert_allclose(cc.cov_00, self.Mxx_c_sym)
np.testing.assert_allclose(cc.cov_0t, self.Mxy_c_sym)
cc = est.fit(self.Xc_lag0, column_selection=self.cols_2).fetch_model()
np.testing.assert_allclose(cc.cov_00, self.Mxx_c_sym[:, self.cols_2])
np.testing.assert_allclose(cc.cov_0t, self.Mxy_c_sym[:, self.cols_2])
def test_multiple_fetch():
# checks that the model instance does not change when the estimator was not updated
data = np.random.normal(size=(5000, 3))
est = Covariance(5, compute_c00=True, compute_c0t=False, compute_ctt=False)
m1 = est.fit(data).model
m2 = est.model
m3 = est.partial_fit(np.random.normal(size=(50, 3))).model
np.testing.assert_(m1 is m2)
np.testing.assert_(m1 is not m3)
np.testing.assert_(m2 is not m3)
def test_fit_from_cov():
data = np.random.normal(size=(500, 3))
# fitting with C0t and symmetric covariances, should pass
TICA().fit(Covariance(1, compute_c0t=True, reversible=True).fit(data))
with np.testing.assert_raises(ValueError):
TICA().fit(Covariance(1, compute_c0t=True, reversible=False).fit(data))
with np.testing.assert_raises(ValueError):
TICA().fit(Covariance(1, compute_c0t=False, reversible=True).fit(data))
def test_XXXY_meanfree(self):
# many passes
est = Covariance(lagtime=self.lag, remove_data_mean=True, compute_c0t=True, bessels_correction=False)
cc = est.fit(self.data).fetch_model()
np.testing.assert_allclose(cc.mean_0, self.mx)
np.testing.assert_allclose(cc.mean_t, self.my)
np.testing.assert_allclose(cc.cov_00, self.Mxx0)
np.testing.assert_allclose(cc.cov_0t, self.Mxy0)
cc = est.fit(self.data, column_selection=self.cols_2).fetch_model()
np.testing.assert_allclose(cc.cov_00, self.Mxx0[:, self.cols_2])
np.testing.assert_allclose(cc.cov_0t, self.Mxy0[:, self.cols_2])
def test_XXXY_sym_withmean(self):
# many passes
est = Covariance(lagtime=self.lag, remove_data_mean=False, compute_c0t=True, reversible=True,
bessels_correction=False)
cc = est.fit(self.data).fetch_model()
np.testing.assert_allclose(cc.mean_0, self.m_sym)
np.testing.assert_allclose(cc.cov_00, self.Mxx_sym)
np.testing.assert_allclose(cc.cov_0t, self.Mxy_sym)
cc = est.fit(self.data, column_selection=self.cols_2).fetch_model()
np.testing.assert_allclose(cc.cov_00, self.Mxx_sym[:, self.cols_2])
np.testing.assert_allclose(cc.cov_0t, self.Mxy_sym[:, self.cols_2])
def test_weights():
weights = np.concatenate([np.ones((1001,)) * 1e-16, np.ones((3999,))])
np.testing.assert_equal(len(weights), 5000)
data = np.random.normal(size=(5000, 2))
cov = Covariance(lagtime=5, compute_c00=True, compute_c0t=True, compute_ctt=False)
model = cov.fit(data, weights=weights, n_splits=64).fetch_model()
model2 = cov.fit(data[1002:], weights=weights[1002:], n_splits=55).fetch_model()
np.testing.assert_array_almost_equal(model.cov_00, model2.cov_00, decimal=2)
np.testing.assert_array_almost_equal(model.cov_0t, model2.cov_0t, decimal=2)
np.testing.assert_array_almost_equal(model.mean_0, model2.mean_0, decimal=2)
np.testing.assert_array_almost_equal(model.mean_t, model2.mean_t, decimal=2)
def test_XXXY_weightobj_sym_meanfree(self):
# many passes
est = Covariance(lagtime=self.lag, remove_data_mean=True, compute_c0t=True, reversible=True,
bessels_correction=False)
cc = est.fit(self.data, weights=self.data_weights).fetch_model()
np.testing.assert_allclose(cc.mean_0, self.m_sym_wobj)
np.testing.assert_allclose(cc.cov_00, self.Mxx0_sym_wobj)
np.testing.assert_allclose(cc.cov_0t, self.Mxy0_sym_wobj)
cc = est.fit(self.data, weights=self.data_weights, column_selection=self.cols_2).fetch_model()
np.testing.assert_allclose(cc.cov_00, self.Mxx0_sym_wobj[:, self.cols_2])
np.testing.assert_allclose(cc.cov_0t, self.Mxy0_sym_wobj[:, self.cols_2])
def test_non_matching_length(self):
n = 100
data = [np.random.random(size=(n, 2)) for n in range(3)]
data = (data[:-1], data[1:])
weights = [np.random.random(n) for _ in range(3)]
weights[0] = weights[0][:-3]
with self.assertRaises(ValueError):
Covariance(1, compute_c0t=True).fit(data, weights=weights)
with self.assertRaises(ValueError):
Covariance(1, compute_c0t=True).fit(data, weights=weights[:10])
def test_XXXY_withmean(self):
# many passes
est = Covariance(lagtime=self.lag, remove_data_mean=False, compute_c0t=True, bessels_correction=False)
cc = est.fit(self.data, n_splits=1).fetch_model()
assert not cc.bessels_correction
np.testing.assert_allclose(cc.mean_0, self.mx)
np.testing.assert_allclose(cc.mean_t, self.my)
np.testing.assert_allclose(cc.cov_00, self.Mxx)
np.testing.assert_allclose(cc.cov_0t, self.Mxy)
cc = est.fit(self.data, n_splits=1, column_selection=self.cols_2).fetch_model()
np.testing.assert_allclose(cc.cov_00, self.Mxx[:, self.cols_2])
np.testing.assert_allclose(cc.cov_0t, self.Mxy[:, self.cols_2])
def test_XXXY_weightobj_meanfree(self):
for n_splits in [1, 2, 3, 4, 5, 6, 7]:
est = Covariance(lagtime=self.lag, remove_data_mean=True, compute_c0t=True, bessels_correction=False)
cc = est.fit(self.data, weights=self.data_weights, n_splits=n_splits).fetch_model()
np.testing.assert_allclose(cc.mean_0, self.mx_wobj)
np.testing.assert_allclose(cc.mean_t, self.my_wobj)
np.testing.assert_allclose(cc.cov_00, self.Mxx0_wobj)
np.testing.assert_allclose(cc.cov_0t, self.Mxy0_wobj)
cc = est.fit(self.data, weights=self.data_weights, column_selection=self.cols_2,
n_splits=n_splits).fetch_model()
np.testing.assert_allclose(cc.cov_00, self.Mxx0_wobj[:, self.cols_2])
np.testing.assert_allclose(cc.cov_0t, self.Mxy0_wobj[:, self.cols_2])
def test_re_estimate_weight_types(self):
# check different types are allowed and re-estimation works
x = np.random.random((100, 2))
c = Covariance(lagtime=1, compute_c0t=True)
c.fit(x, weights=np.ones((len(x),))).fetch_model()
c.fit(x, weights=np.ones((len(x),))).fetch_model()
c.fit(x, weights=None).fetch_model()
c.fit(x, weights=x[:, 0]).fetch_model()
def test_eig_corr(epsilon, method, canonical_signs, return_rank,
hermitian_ctt):
data = np.random.normal(size=(5000, 3))
from deeptime.covariance import Covariance
covariances = Covariance(lagtime=10, compute_c00=True,
compute_ctt=True).fit(data).fetch_model()
if not hermitian_ctt:
covariances.cov_tt[0, 1] += 1e-6
assert_(not np.allclose(covariances.cov_tt, covariances.cov_tt.T))
out = eig_corr(covariances.cov_00,
covariances.cov_tt,
epsilon=epsilon,
method=method,
canonical_signs=canonical_signs,
return_rank=return_rank)
eigenvalues = out[0]
eigenvectors = out[1]
if return_rank:
rank = out[2]
assert_equal(rank, len(eigenvalues))
for r in range(len(out[0])):
assert_array_almost_equal(covariances.cov_00 @ eigenvectors[r] *
eigenvalues[r],
covariances.cov_tt @ eigenvectors[r],
decimal=2)
def test_weights_close_to_zero(self):
n = 1000
data = [np.random.random(size=(n, 2)) for _ in range(5)]
# create some artificial correlations
data[0][:, 0] *= np.random.randint(n)
data = np.asarray(data)
weights = [np.ones(n, dtype=np.float32) for _ in range(5)]
# omit the first trajectory by setting a weight close to zero.
weights[0][:] = 1E-44
weights = np.asarray(weights)
est = Covariance(lagtime=1, compute_c0t=True)
for data_traj, weights_traj in zip(data, weights):
est.partial_fit((data_traj[:-3], data_traj[3:]), weights=weights_traj[:-3])
cov = est.fetch_model()
# cov = covariance_lagged(data, lag=3, weights=weights, chunksize=10)
assert np.all(cov.cov_00 < 1)
def test_sanity(fixed_seed, two_state_hmm, model):
traj, traj_rot, loader = two_state_hmm
tae = setup_tae() if model == 'tae' else setup_tvae()
tae.fit(loader, n_epochs=40)
out = tae.transform(traj_rot).reshape((-1, 1))
out = Covariance().fit(out).fetch_model().whiten(out)
dtraj = dt.clustering.Kmeans(2).fit(out).transform(out)
msm = dt.markov.msm.MaximumLikelihoodMSM().fit_from_discrete_timeseries(
dtraj, 1).fetch_model()
assert_array_almost_equal(msm.transition_matrix, [[.9, .1], [.1, .9]],
decimal=1)
def test_weights_incompatible():
data = np.random.normal(size=(5000, 3))
est = Covariance(5)
with np.testing.assert_raises(ValueError):
est.fit(data, weights=np.arange(10)) # incompatible shape
with np.testing.assert_raises(ValueError):
est.fit(data, weights=np.ones((len(data), 2))) # incompatible shape
def test_whitening():
data = np.random.normal(size=(5000, 50))
data = Covariance().fit(data).fetch_model().whiten(data)
cov = Covariance().fit(data).fetch_model()
np.testing.assert_array_almost_equal(cov.cov_00, np.eye(50), decimal=2)
np.testing.assert_array_almost_equal(cov.mean_0, np.zeros_like(cov.mean_0))