def test_gaussian(self):
g = Gaussian(center=4, width=4)
self.assertArrayAlmostEqual([1, np.exp(-1)], g([4, 0]), decimal=4)
self.assertAlmostEqual(g.volume(20), 2118, places=0)
# Make sure the name makes sense
name = g.name()
self.assertIn('Gaussian', name)
self.assertIn('width=4', name)
self.assertIn('center=4', name)
def test_grdf(self):
f1 = Gaussian(1, 0)
f2 = Gaussian(1, 1)
f3 = Gaussian(1, 5)
s_tuples = [(self.sc, 0), (self.cscl, 0)]
# test fit, transform, and featurize dataframe for both run modes GRDF mode
grdf = GeneralizedRadialDistributionFunction(bins=[f1, f2, f3],
mode='GRDF')
grdf.fit(s_tuples)
features = grdf.transform(s_tuples)
self.assertArrayAlmostEqual(
features,
[[4.4807e-06, 0.00031, 0.02670], [3.3303e-06, 0.00026, 0.01753]],
3)
features = grdf.featurize_dataframe(pd.DataFrame(s_tuples), [0, 1])
self.assertArrayEqual(list(features.columns.values), [
0, 1, 'Gaussian center=0 width=1', 'Gaussian center=1 width=1',
'Gaussian center=5 width=1'
])
# pairwise GRDF mode
grdf = GeneralizedRadialDistributionFunction(bins=[f1, f2, f3],
mode='pairwise_GRDF')
grdf.fit(s_tuples)
features = grdf.transform(s_tuples)
self.assertArrayAlmostEqual(features[0],
[4.4807e-06, 3.1661e-04, 0.0267], 3)
self.assertArrayAlmostEqual(
features[1],
[2.1807e-08, 6.1119e-06, 0.0142, 3.3085e-06, 2.5898e-04, 0.0032],
3)
features = grdf.featurize_dataframe(pd.DataFrame(s_tuples), [0, 1])
self.assertArrayEqual(list(features.columns.values), [
0, 1, 'site2 0 Gaussian center=0 width=1',
'site2 1 Gaussian center=0 width=1',
'site2 0 Gaussian center=1 width=1',
'site2 1 Gaussian center=1 width=1',
'site2 0 Gaussian center=5 width=1',
'site2 1 Gaussian center=5 width=1'
])
# test preset
grdf = GeneralizedRadialDistributionFunction.from_preset('gaussian')
grdf.featurize(self.sc, 0)
self.assertArrayEqual([bin.name() for bin in grdf.bins], [
'Gaussian center={} width=1.0'.format(i) for i in np.arange(10.0)
])
def from_preset(preset, width=0.5, spacing=0.5, cutoff=10):
"""
Preset bin functions for this featurizer. Example use:
>>> AFS = AngularFourierSeries.from_preset('gaussian')
>>> AFS.featurize(struct, idx)
Args:
preset (str): shape of bin (either 'gaussian' or 'histogram')
width (float): bin width. std dev for gaussian, width for histogram
spacing (float): the spacing between bin centers
cutoff (float): maximum distance to look for neighbors
"""
# Generate bin functions
if preset == "gaussian":
bins = []
for center in np.arange(0., cutoff, spacing):
bins.append(Gaussian(width, center))
elif preset == "histogram":
bins = []
for start in np.arange(0, cutoff, spacing):
bins.append(Histogram(start, width))
else:
raise ValueError('Not a valid preset condition.')
return AngularFourierSeries(bins, cutoff=cutoff)
def from_preset(preset, width=1.0, spacing=1.0, cutoff=10, mode='GRDF'):
"""
Preset bin functions for this featurizer. Example use:
>>> GRDF = GeneralizedRadialDistributionFunction.from_preset('gaussian')
>>> GRDF.featurize(struct, idx)
Args:
preset (str): shape of bin (either 'gaussian' or 'histogram')
width (float): bin width. std dev for gaussian, width for histogram
spacing (float): the spacing between bin centers
cutoff (float): maximum distance to look for neighbors
mode (str): featurizing mode. either 'GRDF' or 'pairwise_GRDF'
"""
# Generate bin functions
if preset == "gaussian":
bins = []
for center in np.arange(0., cutoff, spacing):
bins.append(Gaussian(width, center))
elif preset == "histogram":
bins = []
for start in np.arange(0, cutoff, spacing):
bins.append(Histogram(start, width))
else:
raise ValueError('Not a valid preset condition.')
return GeneralizedRadialDistributionFunction(bins,
cutoff=cutoff,
mode=mode)
def test_afs(self):
f1 = Gaussian(1, 0)
f2 = Gaussian(1, 1)
f3 = Gaussian(1, 5)
s_tuples = [(self.sc, 0), (self.cscl, 0)]
# test transform,and featurize dataframe
afs = AngularFourierSeries(bins=[f1, f2, f3])
features = afs.transform(s_tuples)
self.assertArrayAlmostEqual(
features, [[
-1.0374e-10, -4.3563e-08, -2.7914e-06, -4.3563e-08,
-1.8292e-05, -0.0011, -2.7914e-06, -0.0011, -12.7863
],
[
-1.7403e-11, -1.0886e-08, -3.5985e-06, -1.0886e-08,
-6.0597e-06, -0.0016, -3.5985e-06, -0.0016, -3.9052
]], 3)
features = afs.featurize_dataframe(pd.DataFrame(s_tuples), [0, 1])
self.assertArrayEqual(list(features.columns.values), [
0, 1, 'AFS (Gaussian center=0 width=1, Gaussian center=0 width=1)',
'AFS (Gaussian center=0 width=1, Gaussian center=1 width=1)',
'AFS (Gaussian center=0 width=1, Gaussian center=5 width=1)',
'AFS (Gaussian center=1 width=1, Gaussian center=0 width=1)',
'AFS (Gaussian center=1 width=1, Gaussian center=1 width=1)',
'AFS (Gaussian center=1 width=1, Gaussian center=5 width=1)',
'AFS (Gaussian center=5 width=1, Gaussian center=0 width=1)',
'AFS (Gaussian center=5 width=1, Gaussian center=1 width=1)',
'AFS (Gaussian center=5 width=1, Gaussian center=5 width=1)'
])
# test preset
afs = AngularFourierSeries.from_preset('gaussian')
afs.featurize(self.sc, 0)
self.assertArrayEqual([bin.name() for bin in afs.bins], [
'Gaussian center={} width=0.5'.format(i)
for i in np.arange(0, 10, 0.5)
])
afs = AngularFourierSeries.from_preset('histogram')
afs.featurize(self.sc, 0)
self.assertArrayEqual([bin.name() for bin in afs.bins], [
'Histogram start={} width=0.5'.format(i)
for i in np.arange(0, 10, 0.5)
])