def test_extract_pixels():
"""Test extraction to downsampled pixels.
"""
c = CoAdd1D(100., 200., 0.5, 50.)
size = 8
sigma_f = 1.5
edges, mu, cov = c.extract_pixels(size, sigma_f, return_cov=True)
edges2, mu2 = c.extract_pixels(size, sigma_f, return_cov=False)
assert np.array_equal(edges, edges2)
assert np.array_equal(mu, mu2)
n = len(edges) - 1
assert mu.shape == (n, )
assert cov.shape == (n, n)
cov = cov.toarray()
assert np.array_equal(cov.T, cov)
assert edges[0] == c.grid[0]
assert edges[-1] == c.grid[n * size]
assert np.array_equal(mu, np.zeros(n))
assert np.allclose(cov, sigma_f**2 * size * np.identity(n))
data = [1, 3, 2], [150, 160, 170, 180], [0.1, 0.2, 0.1]
c.add(*data, 5)
edges, mu, cov = c.extract_pixels(size, sigma_f, return_cov=True)
cov = cov.toarray()
assert np.array_equal(cov.T, cov)
assert np.allclose(np.sum(mu), 5.7583230)
assert np.allclose(np.sum(np.diagonal(cov)), 414.52807)
def test_ctor_rounding():
"""Test wlen_max rounding up.
"""
c = CoAdd1D(100., 199.5, 1., 50.)
assert c.grid[0] == 100.
assert c.grid[-1] == 200.
assert c.grid_scale == 1.
assert c.n_grid == 101
def test_get_f():
"""Test calculation of deconvolved true flux f.
"""
c = CoAdd1D(100., 200., 0.5, 50.)
assert np.all(c.get_f(sigma_f=1) == 0)
data = [1, 3, 2], [150, 160, 170, 180], [0.1, 0.2, 0.1]
c.add(*data, 5)
assert np.allclose(np.sum(c.get_f(sigma_f=1)), 4.950472)
def test_get_phi():
"""Test calculation of phi vector summary statistic.
"""
c = CoAdd1D(100., 200., 0.5, 50.)
assert np.all(c.get_phi() == 0)
data = [1, 3, 2], [150, 160, 170, 180], [0.1, 0.2, 0.1]
c.add(*data, 5)
assert np.allclose(np.mean(c.get_phi()), 0.0895522)
assert np.allclose(np.std(c.get_phi()), 0.1422825)
def test_ctor():
"""Normal construction.
"""
c = CoAdd1D(100., 200., 1., 50.)
assert c.grid[0] == 100.
assert c.grid[-1] == 200.
assert c.grid_scale == 1.
assert np.all(c.phi_sum == 0)
assert np.all(c.A_sum.csr.toarray() == 0)
assert c.nbytes == 93836
def test_extract_downsampled():
"""Test extraction of downsampled baad.
"""
c = CoAdd1D(100., 200., 0.5, 50.)
n = 10
sigma_f = 1.5
coefs = scipy.sparse.identity(c.n_grid, format='csr')[:n]
mu, cov = c.extract_downsampled(coefs, sigma_f, return_cov=True)
assert np.array_equal(mu, np.zeros(n))
assert np.array_equal(cov.toarray(), sigma_f**2 * np.identity(n))
def test_extract_whitened():
c = CoAdd1D(100., 200., 0.5, 50.)
sigma_f = 1.5
psfs, mu = c.extract_whitened(sigma_f)
n = len(mu)
assert psfs.shape == (n, n)
assert np.array_equal(mu, np.zeros(n))
data = [1, 3, 2], [150, 160, 170, 180], [0.1, 0.2, 0.1]
c.add(*data, 5)
psfs, mu = c.extract_whitened(sigma_f)
assert np.allclose(np.sum(mu), 10.262560)
def test_reset():
"""Reset after adding.
"""
c = CoAdd1D(100., 200., 0.5, 50.)
psf = np.zeros(21)
psf[10] = 1
data = [1, 3, 2], [150, 160, 170, 180], [0.1, 0.2, 0.1]
c.add(*data, 5)
assert not np.all(c.phi_sum == 0)
assert not np.all(c.A_sum.csr.toarray() == 0)
c.reset()
assert np.all(c.phi_sum == 0)
assert np.all(c.A_sum.csr.toarray() == 0)
def test_get_log_evidence():
"""Test calculation of log evidence.
"""
c = CoAdd1D(100., 200., 0.5, 50.)
assert c.get_log_evidence(sigma_f=1) == 0
data = [1, 3, 2], [150, 160, 170, 180], [0.1, 0.2, 0.1]
c.add(*data, 5)
assert np.allclose(c.get_log_evidence(sigma_f=1), -0.32704586)
assert np.allclose(c.get_log_evidence(sigma_f=[1]), [-0.32704586])
assert np.allclose(c.get_log_evidence(sigma_f=[0.1, 1, 10]),
[0.00843854, -0.32704586, -5.79924322])
with pytest.raises(ValueError):
c.get_log_evidence(sigma_f=0)
with pytest.raises(ValueError):
c.get_log_evidence(sigma_f=[1, -1])
def test_add_psf():
"""Addition with different types of PSF inputs.
"""
c = CoAdd1D(100., 240., 0.5, 50.)
psf = np.zeros(21)
psf[10] = 1
psfs = np.tile(psf, [3, 1])
data = [1, 3, 2], [150, 160, 170, 180], [0.1, 0.2, 0.1]
def psf_model(w, dw):
return np.exp(-dw**2) + 0 * w
for convolve in True, False:
c.add(*data, 5, convolve)
c.add(*data, [4, 5, 6], convolve)
c.add(*data, psf, convolve)
c.add(*data, psfs, convolve)
c.add(*data, psf_model, convolve)
def test_add_analytic_vs_tabulated():
"""Compare analytic vs tabulated Gaussian PSFs.
"""
c = CoAdd1D(100., 200., 0.5, 50.)
data = [1, 3, 2], [150, 160, 170, 180], [0.1, 0.2, 0.1]
psf_grid = c.grid_scale * np.arange(-10, +11)
for convolve in True, False:
# Common PSF for all pixels.
rms = 1.5
gp0, _, _ = c.add(*data, rms, convolve, retval=True)
psf = np.exp(-0.5 * (psf_grid / rms)**2)
gp1, _, _ = c.add(*data, psf, convolve, retval=True)
assert np.allclose(gp0.toarray(), gp1.toarray(), atol=0.05, rtol=0.05)
# Individual PSFs for each pixel.
rms = np.array([1.4, 1.5, 1.6])
gp0, _, _ = c.add(*data, rms, convolve, retval=True)
psf = np.exp(-0.5 * (psf_grid / rms.reshape(-1, 1))**2)
gp1, _, _ = c.add(*data, psf, convolve, retval=True)
assert np.allclose(gp0.toarray(), gp1.toarray(), atol=0.05, rtol=0.05)
def test_extract_gaussian():
c = CoAdd1D(100., 200., 0.5, 50.)
size = 8
spacing = size * c.grid_scale
rms = spacing
sigma_f = 1.5
centers, mu, cov = c.extract_gaussian(spacing, rms, sigma_f)
n = len(centers)
assert mu.shape == (n, )
assert cov.shape == (n, n)
assert np.array_equal(cov.T, cov)
assert centers[0] == c.grid[0] + 0.5 * spacing
assert centers[-1] == centers[0] + (n - 1) * spacing
assert np.array_equal(mu, np.zeros(n))
data = [1, 3, 2], [150, 160, 170, 180], [0.1, 0.2, 0.1]
c.add(*data, 5)
centers, mu, cov = c.extract_gaussian(spacing, rms, sigma_f)
assert np.array_equal(cov.T, cov)
assert np.allclose(np.sum(mu), 5.7581933)
assert np.allclose(np.sum(np.diagonal(cov)), 96.95508)
def test_get_A():
"""Test calculation of A matrix summary statistic.
"""
c = CoAdd1D(100., 200., 0.5, 50.)
assert np.all(c.get_A(sparse=False) == 0)
assert np.array_equal(
c.get_A(sparse=True).toarray(), c.get_A(sparse=False))
assert np.array_equal(c.get_A(sigma_f=2, sparse=False),
0.25 * np.identity(c.n_grid))
data = [1, 3, 2], [150, 160, 170, 180], [0.1, 0.2, 0.1]
c.add(*data, 5)
A = c.get_A(sparse=True).toarray()
assert np.all(A.T == A)
assert np.allclose(np.trace(A), 3.89099485)
assert np.linalg.slogdet(A)[1] == -np.inf
A = c.get_A(sigma_f=1.1, sparse=False)
assert np.all(A.T == A)
assert np.allclose(np.trace(A), 170.006697)
assert np.allclose(np.linalg.slogdet(A)[1], -35.734250)
A = c.get_A(sigma_f=0.5, sparse=True)
assert A.getformat() == 'csr'
assert A.T.getformat() == 'csc'