def normalize_l2(spectrum: Spectrum, in_place: bool = False):
sum_norm = np.linalg.norm(spectrum.y, ord=2)
if not in_place:
spectrum = spectrum.copy()
spectrum.y /= sum_norm
return spectrum
def normalize_max(spectrum: Spectrum, in_place: bool = False):
maximum = np.max(spectrum.y)
if not in_place:
spectrum = spectrum.copy()
spectrum.y /= maximum
return spectrum
def normalize_sum(spectrum: Spectrum, value=1.0, in_place: bool = False):
total = np.sum(spectrum.y)
if not in_place:
spectrum = spectrum.copy()
spectrum.y /= total
spectrum.y *= value
return spectrum
def normalize_z_max(spectrum: Spectrum, in_place: bool = False):
mu = np.mean(spectrum.y)
sig = np.std(spectrum.y)
if not in_place:
spectrum = spectrum.copy()
spectrum.y = np.array([(y - mu) / sig for y in spectrum.y])
spectrum.y /= np.max(spectrum.y)
return spectrum
def normalize_minmax(spectrum: Spectrum, in_place: bool = False):
spec_max, spec_min = max(spectrum.y), min(spectrum.y)
mins_array = spec_min * np.ones(shape=spectrum.y.shape, dtype='float')
if not in_place:
spectrum = spectrum.copy()
spectrum.y -= mins_array
spectrum.y /= (spec_max - spec_min)
return spectrum
def normalize_z(spectrum: Spectrum, in_place: bool = False):
"""
Normalize by treating the spectrum as a normal distribution and finding the
'z-score' associated with each point
:param spectrum:
:param in_place:
:return:
"""
mu = np.mean(spectrum.y)
sig = np.std(spectrum.y)
if not in_place:
spectrum = spectrum.copy()
spectrum.y = np.array([(y - mu) / sig for y in spectrum.y])
return spectrum
def normalize_0left_1right(spectrum: Spectrum, in_place: bool = False):
"""
Normalize by setting the input value to 0 and the final value to about 1.
:param spectrum:
:param in_place:
:return:
"""
to_sub = spectrum.y[0]
if not in_place:
spectrum = spectrum.copy()
spectrum.y -= to_sub
spectrum_final_avg = np.mean(spectrum.y[-5:])
spectrum.y /= spectrum_final_avg
return spectrum
class SpectrumTest(PymatgenTest):
def setUp(self):
self.spec1 = Spectrum(np.arange(0, 10, 0.1), np.random.randn(100))
self.spec2 = Spectrum(np.arange(0, 10, 0.1), np.random.randn(100))
self.multi_spec1 = Spectrum(np.arange(0, 10, 0.1),
np.random.randn(100, 2))
self.multi_spec2 = Spectrum(np.arange(0, 10, 0.1),
np.random.randn(100, 2))
def test_normalize(self):
self.spec1.normalize(mode="max")
self.assertAlmostEqual(np.max(self.spec1.y), 1)
self.spec1.normalize(mode="sum")
self.assertAlmostEqual(np.sum(self.spec1.y), 1)
self.multi_spec1.normalize(mode="sum")
self.assertAlmostEqual(np.sum(self.multi_spec1.y[:, 0]), 1)
self.assertAlmostEqual(np.sum(self.multi_spec1.y[:, 1]), 1)
# XRD style mode.
self.spec1.normalize(mode="max", value=100)
self.assertAlmostEqual(np.max(self.spec1.y), 100)
def test_operators(self):
scaled_spect = 3 * self.spec1 + self.spec2
self.assertTrue(
np.allclose(scaled_spect.y, 3 * self.spec1.y + self.spec2.y))
self.assertAlmostEqual(
self.spec1.get_interpolated_value(0.05),
(self.spec1.y[0] + self.spec1.y[1]) / 2,
)
scaled_spect = self.spec1 - self.spec2
self.assertTrue(
np.allclose(scaled_spect.y, self.spec1.y - self.spec2.y))
scaled_spect = self.spec1 / 3
self.assertTrue(np.allclose(scaled_spect.y, self.spec1.y / 3))
scaled_spect = 3 * self.multi_spec1 + self.multi_spec2
self.assertTrue(
np.allclose(scaled_spect.y,
3 * self.multi_spec1.y + self.multi_spec2.y))
self.assertArrayAlmostEqual(
self.multi_spec1.get_interpolated_value(0.05),
(self.multi_spec1.y[0, :] + self.multi_spec1.y[1, :]) / 2,
)
def test_smear(self):
y = np.array(self.spec1.y)
self.spec1.smear(0.2)
self.assertFalse(np.allclose(y, self.spec1.y))
self.assertAlmostEqual(sum(y), sum(self.spec1.y))
y = np.array(self.multi_spec1.y)
self.multi_spec1.smear(0.2)
self.assertFalse(np.allclose(y, self.multi_spec1.y))
self.assertArrayAlmostEqual(np.sum(y, axis=0),
np.sum(self.multi_spec1.y, axis=0))
def test_str(self):
# Just make sure that these methods work.
self.assertIsNotNone(str(self.spec1))
self.assertIsNotNone(str(self.multi_spec1))
def test_copy(self):
spec1copy = self.spec1.copy()
spec1copy.y[0] = spec1copy.y[0] + 1
self.assertNotEqual(spec1copy.y[0], self.spec1.y[0])
self.assertEqual(spec1copy.y[1], self.spec1.y[1])
class SpectrumTest(PymatgenTest):
def setUp(self):
self.spec1 = Spectrum(np.arange(0, 10, 0.1), np.random.randn(100))
self.spec2 = Spectrum(np.arange(0, 10, 0.1), np.random.randn(100))
self.multi_spec1 = Spectrum(np.arange(0, 10, 0.1), np.random.randn(100, 2))
self.multi_spec2 = Spectrum(np.arange(0, 10, 0.1), np.random.randn(100, 2))
def test_normalize(self):
self.spec1.normalize(mode="max")
self.assertAlmostEqual(np.max(self.spec1.y), 1)
self.spec1.normalize(mode="sum")
self.assertAlmostEqual(np.sum(self.spec1.y), 1)
self.multi_spec1.normalize(mode="sum")
self.assertAlmostEqual(np.sum(self.multi_spec1.y[:, 0]), 1)
self.assertAlmostEqual(np.sum(self.multi_spec1.y[:, 1]), 1)
# XRD style mode.
self.spec1.normalize(mode="max", value=100)
self.assertAlmostEqual(np.max(self.spec1.y), 100)
def test_operators(self):
scaled_spect = 3 * self.spec1 + self.spec2
self.assertTrue(np.allclose(scaled_spect.y, 3 * self.spec1.y + self.spec2.y))
self.assertAlmostEqual(
self.spec1.get_interpolated_value(0.05),
(self.spec1.y[0] + self.spec1.y[1]) / 2,
)
scaled_spect = self.spec1 - self.spec2
self.assertTrue(np.allclose(scaled_spect.y, self.spec1.y - self.spec2.y))
scaled_spect = self.spec1 / 3
self.assertTrue(np.allclose(scaled_spect.y, self.spec1.y / 3))
scaled_spect = 3 * self.multi_spec1 + self.multi_spec2
self.assertTrue(np.allclose(scaled_spect.y, 3 * self.multi_spec1.y + self.multi_spec2.y))
self.assertArrayAlmostEqual(
self.multi_spec1.get_interpolated_value(0.05),
(self.multi_spec1.y[0, :] + self.multi_spec1.y[1, :]) / 2,
)
def test_smear(self):
y = np.zeros(100)
y[25] = 1
y[50] = 1
y[75] = 1
spec = Spectrum(np.linspace(-10, 10, 100), y)
spec.smear(0.3)
self.assertFalse(np.allclose(y, spec.y))
self.assertAlmostEqual(sum(y), sum(spec.y))
# Test direct callable use of smearing.
spec2 = Spectrum(np.linspace(-10, 10, 100), y)
spec2.smear(0, func=lambda x: stats.norm.pdf(x, scale=0.3))
self.assertTrue(np.allclose(spec.y, spec2.y))
spec = Spectrum(np.linspace(-10, 10, 100), y)
spec.smear(0.3, func="lorentzian")
self.assertFalse(np.allclose(y, spec.y))
self.assertAlmostEqual(sum(y), sum(spec.y))
y = np.array(self.multi_spec1.y)
self.multi_spec1.smear(0.2)
self.assertFalse(np.allclose(y, self.multi_spec1.y))
self.assertArrayAlmostEqual(np.sum(y, axis=0), np.sum(self.multi_spec1.y, axis=0))
def test_str(self):
# Just make sure that these methods work.
self.assertIsNotNone(str(self.spec1))
self.assertIsNotNone(str(self.multi_spec1))
def test_copy(self):
spec1copy = self.spec1.copy()
spec1copy.y[0] = spec1copy.y[0] + 1
self.assertNotEqual(spec1copy.y[0], self.spec1.y[0])
self.assertEqual(spec1copy.y[1], self.spec1.y[1])
class SpectrumTest(PymatgenTest):
def setUp(self):
self.spec1 = Spectrum(np.arange(0, 10, 0.1), np.random.randn(100))
self.spec2 = Spectrum(np.arange(0, 10, 0.1), np.random.randn(100))
self.multi_spec1 = Spectrum(np.arange(0, 10, 0.1),
np.random.randn(100, 2))
self.multi_spec2 = Spectrum(np.arange(0, 10, 0.1),
np.random.randn(100, 2))
def test_normalize(self):
self.spec1.normalize(mode="max")
self.assertAlmostEqual(np.max(self.spec1.y), 1)
self.spec1.normalize(mode="sum")
self.assertAlmostEqual(np.sum(self.spec1.y), 1)
self.multi_spec1.normalize(mode="sum")
self.assertAlmostEqual(np.sum(self.multi_spec1.y[:, 0]), 1)
self.assertAlmostEqual(np.sum(self.multi_spec1.y[:, 1]), 1)
# XRD style mode.
self.spec1.normalize(mode="max", value=100)
self.assertAlmostEqual(np.max(self.spec1.y), 100)
def test_operators(self):
scaled_spect = 3 * self.spec1 + self.spec2
self.assertTrue(np.allclose(scaled_spect.y,
3 * self.spec1.y + self.spec2.y))
self.assertAlmostEqual(self.spec1.get_interpolated_value(0.05),
(self.spec1.y[0] + self.spec1.y[1]) / 2)
scaled_spect = self.spec1 - self.spec2
self.assertTrue(np.allclose(scaled_spect.y,
self.spec1.y - self.spec2.y))
scaled_spect = self.spec1 / 3
self.assertTrue(np.allclose(scaled_spect.y,
self.spec1.y / 3))
scaled_spect = 3 * self.multi_spec1 + self.multi_spec2
self.assertTrue(np.allclose(scaled_spect.y,
3 * self.multi_spec1.y + self.multi_spec2.y))
self.assertArrayAlmostEqual(
self.multi_spec1.get_interpolated_value(0.05),
(self.multi_spec1.y[0, :] + self.multi_spec1.y[1, :]) / 2)
def test_smear(self):
y = np.array(self.spec1.y)
self.spec1.smear(0.2)
self.assertFalse(np.allclose(y, self.spec1.y))
self.assertAlmostEqual(sum(y), sum(self.spec1.y))
y = np.array(self.multi_spec1.y)
self.multi_spec1.smear(0.2)
self.assertFalse(np.allclose(y, self.multi_spec1.y))
self.assertArrayAlmostEqual(np.sum(y, axis=0),
np.sum(self.multi_spec1.y, axis=0))
def test_str(self):
# Just make sure that these methods work.
self.assertIsNotNone(str(self.spec1))
self.assertIsNotNone(str(self.multi_spec1))
def test_copy(self):
spec1copy = self.spec1.copy()
spec1copy.y[0] = spec1copy.y[0] + 1
self.assertNotEqual(spec1copy.y[0], self.spec1.y[0])
self.assertEqual(spec1copy.y[1], self.spec1.y[1])