def test_pca_bad_solver():
X = np.random.RandomState(0).rand(5, 4)
pca = dd.PCA(n_components=3, svd_solver="bad_argument")
assert_raises(ValueError, pca.fit, X)
def test_whitening():
# Check that PCA output has unit-variance
rng = np.random.RandomState(0)
n_samples = 100
n_features = 80
n_components = 30
rank = 50
# some low rank data with correlated features
X = np.dot(
rng.randn(n_samples, rank),
np.dot(np.diag(np.linspace(10.0, 1.0, rank)),
rng.randn(rank, n_features)),
)
# the component-wise variance of the first 50 features is 3 times the
# mean component-wise variance of the remaining 30 features
X[:, :50] *= 3
assert X.shape == (n_samples, n_features)
# the component-wise variance is thus highly varying:
assert X.std(axis=0).std() > 43.8
dX = da.from_array(X, chunks=(50, n_features))
for solver, copy in product(solver_list, (True, False)):
# whiten the data while projecting to the lower dim subspace
X_ = dX.copy() # make sure we keep an original across iterations.
pca = dd.PCA(
n_components=n_components,
whiten=True,
copy=copy,
svd_solver=solver,
random_state=0,
iterated_power=4,
)
# test fit_transform
X_whitened = pca.fit_transform(X_.copy())
assert X_whitened.shape == (n_samples, n_components)
# X_whitened2 = pca.transform(X_)
# XXX: These differ for randomized.
# assert_eq(X_whitened.compute(), X_whitened2.compute(),
# atol=tol, rtol=tol)
assert_almost_equal(X_whitened.std(ddof=1, axis=0),
np.ones(n_components),
decimal=6)
assert_almost_equal(X_whitened.mean(axis=0), np.zeros(n_components))
X_ = dX.copy()
pca = dd.PCA(
n_components=n_components,
whiten=False,
copy=copy,
svd_solver=solver,
random_state=0,
).fit(X_)
X_unwhitened = pca.transform(X_)
assert X_unwhitened.shape == (n_samples, n_components)
# in that case the output components still have varying variances
assert_almost_equal(X_unwhitened.std(axis=0).std(), 74.1, 1)
def test_basic():
a = dd.PCA()
b = sd.PCA()
a.fit(dX)
b.fit(X)
assert_estimator_equal(a, b)