def test_pca():
l = np.arange(0, 1, 0.001).reshape((1000, 1))
expand = np.array([2.5, 1.2]).reshape((1, 2))
X = np.dot(l, expand)
X += np.random.normal(0, 1E-6, X.shape)
pca = Pca(1)
pca.fit(X)
desired = np.array([[-0.9015], [-0.4327]])
assert np.allclose(pca.weights, desired, 1e-3)
def test_pca_white():
l = np.arange(0, 1, 0.001).reshape((1000, 1))
expand = np.array([2.5, 1.2]).reshape((1, 2))
X = np.dot(l, expand)
X += np.random.normal(0, 1E-6, X.shape)
pca = Pca(2, whiten=True)
pca.fit(X)
w, s = pca.weights, pca.singular_values
X = pca.transform(X)
assert np.allclose(np.cov(X, rowvar=0), np.eye(2), 1e-2), 'covariance not white'