def test_estimate_parameters_binned():
s = Spectrum(np.empty((100,)))
axis = s.axes_manager.signal_axes[0]
axis.scale = 2.
axis.offset = -30
g1 = GaussianHF(50015.156, 23, 10)
s.data = g1.function(axis.axis)
s.metadata.Signal.binned = True
g2 = GaussianHF()
g2.estimate_parameters(s, axis.low_value, axis.high_value, True)
nt.assert_almost_equal(
g1.height.value / axis.scale,
g2.height.value)
nt.assert_almost_equal(g2.centre.value, g1.centre.value, delta=1e-3)
nt.assert_almost_equal(g2.fwhm.value, g1.fwhm.value, delta=0.1)
def test_integral_as_signal():
s = Spectrum(np.zeros((2, 3, 100)))
g1 = GaussianHF(fwhm=3.33, centre=20.)
h_ref = np.linspace(0.1, 3.0, s.axes_manager.navigation_size)
for d, h in zip(s._iterate_signal(), h_ref):
g1.height.value = h
d[:] = g1.function(s.axes_manager.signal_axes[0].axis)
m = s.create_model()
g2 = GaussianHF()
m.append(g2)
g2.estimate_parameters(s, 0, 100, True)
m.multifit()
s_out = g2.integral_as_signal()
ref = (h_ref * 3.33 * sqrt2pi / sigma2fwhm).reshape(s_out.data.shape)
np.testing.assert_almost_equal(
s_out.data, ref)
class Test2D():
def setUp(self):
self.s = Spectrum(np.random.random((2, 3)))
def test_to_image(self):
im = self.s.to_image()
assert_true(isinstance(im, Image))
assert_equal(im.data.shape, self.s.data.T.shape)
assert_true(im.data.flags["C_CONTIGUOUS"])
class Test2D:
def setUp(self):
self.s = Spectrum(np.random.random((2, 3)))
def test_to_image(self):
im = self.s.to_image()
nose.tools.assert_true(isinstance(im, Image))
nose.tools.assert_equal(im.data.shape, self.s.data.T.shape)
nose.tools.assert_true(im.data.flags["C_CONTIGUOUS"])
class Test4D():
def setUp(self):
self.s = Spectrum(np.random.random((2, 3, 4, 5)))
def test_to_image(self):
im = self.s.to_image()
assert_true(isinstance(im, Image))
assert_equal(im.data.shape, (5, 2, 3, 4))
assert_true(im.data.flags["C_CONTIGUOUS"])
def setUp(self):
np.random.seed(0) # Same random every time, Line2DROi test requires it
self.s_s = Spectrum(np.random.rand(50, 60, 4))
self.s_s.axes_manager[0].scale = 5
self.s_s.axes_manager[0].units = 'nm'
self.s_s.axes_manager[1].scale = 5
self.s_s.axes_manager[1].units = 'nm'
# 4D dataset
self.s_i = Image(np.random.rand(100, 100, 4, 4))
class Test4D():
def setUp(self):
self.s = Spectrum(np.random.random((2, 3, 4, 5)))
def test_to_image(self):
im = self.s.to_image()
nose.tools.assert_true(isinstance(im, Image))
nose.tools.assert_equal(im.data.shape, (5, 2, 3, 4))
nose.tools.assert_true(im.data.flags["C_CONTIGUOUS"])
def setUp(self):
self.s = Spectrum(np.random.random((2, 3)))
def setUp(self):
self.s = Spectrum(np.random.random((2, 3)))
def setUp(self):
ics = np.random.laplace(size=(3, 1000))
np.random.seed(1)
mixing_matrix = np.random.random((100, 3))
self.s = Spectrum(np.dot(mixing_matrix, ics))
self.s.decomposition()
class TestBSS1D:
def setUp(self):
ics = np.random.laplace(size=(3, 1000))
np.random.seed(1)
mixing_matrix = np.random.random((100, 3))
self.s = Spectrum(np.dot(mixing_matrix, ics))
self.s.decomposition()
def test_on_loadings(self):
if not sklearn_installed:
raise SkipTest
self.s.blind_source_separation(
3, diff_order=0, fun="exp", on_loadings=False)
s2 = self.s.as_spectrum(0)
s2.decomposition()
s2.blind_source_separation(
3, diff_order=0, fun="exp", on_loadings=True)
nose.tools.assert_true(are_bss_components_equivalent(
self.s.get_bss_factors(), s2.get_bss_loadings()))
def test_mask_diff_order_0(self):
if not sklearn_installed:
raise SkipTest
# This test, unlike most other tests, either passes or raises an error.
# It is designed to test if the mask is correctly dilated inside the
# `blind_source_separation_method`. If the mask is not correctely
# dilated the nan in the loadings should raise an error.
mask = self.s._get_signal_signal(dtype="bool")
mask[5] = True
self.s.learning_results.factors[5, :] = np.nan
self.s.blind_source_separation(3, diff_order=0, mask=mask)
def test_mask_diff_order_1(self):
if not sklearn_installed:
raise SkipTest
# This test, unlike most other tests, either passes or raises an error.
# It is designed to test if the mask is correctly dilated inside the
# `blind_source_separation_method`. If the mask is not correctely
# dilated the nan in the loadings should raise an error.
mask = self.s._get_signal_signal(dtype="bool")
mask[5] = True
self.s.learning_results.factors[5, :] = np.nan
self.s.blind_source_separation(3, diff_order=1, mask=mask)
def test_mask_diff_order_0_on_loadings(self):
if not sklearn_installed:
raise SkipTest
# This test, unlike most other tests, either passes or raises an error.
# It is designed to test if the mask is correctly dilated inside the
# `blind_source_separation_method`. If the mask is not correctely
# dilated the nan in the loadings should raise an error.
mask = self.s._get_navigation_signal(dtype="bool")
mask[5] = True
self.s.learning_results.loadings[5, :] = np.nan
self.s.blind_source_separation(3, diff_order=0, mask=mask,
on_loadings=True)
def test_mask_diff_order_1_on_loadings(self):
if not sklearn_installed:
raise SkipTest
# This test, unlike most other tests, either passes or raises an error.
# It is designed to test if the mask is correctly dilated inside the
# `blind_source_separation_method`. If the mask is not correctely
# dilated the nan in the loadings should raise an error.
mask = self.s._get_navigation_signal(dtype="bool")
mask[5] = True
self.s.learning_results.loadings[5, :] = np.nan
self.s.blind_source_separation(3, diff_order=1, mask=mask,
on_loadings=True)