def fractional_anisotropy(d, dperp0, dperp1):
"""Calculate the fractional anisotropy (FA).
Returns:
ndarray: the fractional anisotropy for each voxel.
"""
def compute(d, dperp0, dperp1):
d, dperp0, dperp1 = map(
lambda el: np.squeeze(el).astype(np.float64),
[d, dperp0, dperp1])
return np.sqrt(1 / 2.) * np.sqrt(
((d - dperp0)**2 + (dperp0 - dperp1)**2 +
(dperp1 - d)**2) / (d**2 + dperp0**2 + dperp1**2))
if len(d.shape) > 1 and d.shape[1] > 1:
fa = np.zeros(d.shape[:2])
for batch_start, batch_end in split_in_batches(d.shape[1],
max_batch_size=100):
fa[:, batch_start:batch_end] = compute(
d[:, batch_start:batch_end], dperp0[:,
batch_start:batch_end],
dperp1[:, batch_start:batch_end])
return fa
else:
return compute(d, dperp0, dperp1)
def sample(self, nmr_samples, burnin=0, thinning=1):
"""Take additional samples from the given likelihood and prior, using this sampler.
This method can be called multiple times in which the sample state is stored in between.
Args:
nmr_samples (int): the number of samples to return
burnin (int): the number of samples to discard before returning samples
thinning (int): how many sample we wait before storing a new one. This will draw extra samples such that
the total number of samples generated is ``nmr_samples * (thinning)`` and the number of samples
stored is ``nmr_samples``. If set to one or lower we store every sample after the burn in.
Returns:
SamplingOutput: the sample output object
"""
if not thinning or thinning < 1:
thinning = 1
if not burnin or burnin < 0:
burnin = 0
max_samples_per_batch = max(1000 // thinning, 100)
with self._logging(nmr_samples, burnin, thinning):
if burnin > 0:
for batch_start, batch_end in split_in_batches(
burnin, max_samples_per_batch):
self._sample(batch_end - batch_start, return_output=False)
if nmr_samples > 0:
outputs = []
for batch_start, batch_end in split_in_batches(
nmr_samples, max_samples_per_batch):
outputs.append(
self._sample(batch_end - batch_start,
thinning=thinning))
return SimpleSampleOutput(*[
np.concatenate([o[ind] for o in outputs], axis=-1)
for ind in range(3)
])
def get_division(self, cl_environments, nmr_instances):
return list(
split_in_batches(nmr_instances, nmr_batches=len(cl_environments)))