def _distance_2(misorientation, verbose, split_size=100):
num_orientations = misorientation.shape[0]
S_1, S_2 = misorientation._symmetry
distance = np.full(misorientation.shape + misorientation.shape, np.infty)
split_size = split_size // S_1.shape[0]
outer_range = range(0, num_orientations, split_size)
if verbose:
from tqdm import tqdm
outer_range = tqdm(outer_range,
total=np.ceil(num_orientations / split_size))
S_1_outer_S_1 = S_1.outer(S_1)
# Calculate the upper half of the distance matrix block by block
for start_index_b in outer_range:
# we use slice object for compactness
index_slice_b = slice(
start_index_b, min(num_orientations, start_index_b + split_size))
o_sub_b = misorientation[index_slice_b]
for start_index_a in range(0, start_index_b + split_size, split_size):
index_slice_a = slice(
start_index_a, min(num_orientations,
start_index_a + split_size))
o_sub_a = misorientation[index_slice_a]
axis = (len(o_sub_a.shape), len(o_sub_a.shape) + 1)
mis2orientation = (~o_sub_a).outer(S_1_outer_S_1).outer(o_sub_b)
# This works through all the identity rotations
for s_2_1, s_2_2 in icombinations(S_2, 2):
m = s_2_1 * mis2orientation * s_2_2
angle = m.angle.data.min(axis=axis)
distance[index_slice_a, index_slice_b] = np.minimum(
distance[index_slice_a, index_slice_b], angle)
# Symmetrize the matrix for convenience
i_lower = np.tril_indices(distance.shape[0], -1)
distance[i_lower] = distance.T[i_lower]
return distance
def _distance(misorientation, verbose, split_size=100):
"""Private function to find the symmetry reduced distance between
all pairs of (mis)orientations
Parameters
----------
misorientation : orix.quaternion.Misorientation
The misorientation to be considered.
verbose : bool
Output progress bar while computing.
split_size : int
Size of block to compute at a time.
Returns
-------
distance : numpy.ndarray
2D matrix containing the angular distance between every
orientation, considering symmetries.
"""
num_orientations = misorientation.shape[0]
S_1, S_2 = misorientation._symmetry
distance = np.full(misorientation.shape + misorientation.shape, np.infty)
split_size = split_size // S_1.shape[0]
outer_range = range(0, num_orientations, split_size)
if verbose:
outer_range = tqdm(outer_range, total=np.ceil(num_orientations / split_size))
S_1_outer_S_1 = S_1.outer(S_1)
# Calculate the upper half of the distance matrix block by block
for start_index_b in outer_range:
# we use slice object for compactness
index_slice_b = slice(
start_index_b, min(num_orientations, start_index_b + split_size)
)
o_sub_b = misorientation[index_slice_b]
for start_index_a in range(0, start_index_b + split_size, split_size):
index_slice_a = slice(
start_index_a, min(num_orientations, start_index_a + split_size)
)
o_sub_a = misorientation[index_slice_a]
axis = (len(o_sub_a.shape), len(o_sub_a.shape) + 1)
mis2orientation = (~o_sub_a).outer(S_1_outer_S_1).outer(o_sub_b)
# This works through all the identity rotations
for s_2_1, s_2_2 in icombinations(S_2, 2):
m = s_2_1 * mis2orientation * s_2_2
angle = m.angle.data.min(axis=axis)
distance[index_slice_a, index_slice_b] = np.minimum(
distance[index_slice_a, index_slice_b], angle
)
# Symmetrize the matrix for convenience
i_lower = np.tril_indices(distance.shape[0], -1)
distance[i_lower] = distance.T[i_lower]
return distance
def getRelationsChampions(dataframe):
columns = dataframe.columns.tolist()
betterRelations = {}
for column in columns:
notThisColumn = list(set(columns) - set([column]))
combinations = set()
for i in range(2, 6):
comb = list(icombinations(notThisColumn, i))
for value in comb:
combinations.add(value)
for otherColumn in notThisColumn:
combinations.add((otherColumn, None))
for combinationTuple in combinations:
combination = list(combinationTuple)
if None in combination:
combination = [combination[0]]
result = convert.filterAndSortedApriori(dataframe, combination,
column)
if len(result) == 0:
continue
score = result[0].ordered_statistics[0].confidence
relation = {'combination': combination, 'confidence': score}
print(f'Relation found for {column}: {relation}')
if column not in betterRelations.keys():
betterRelations[column] = [relation]
else:
betterRelations[column].append(relation)
for key in betterRelations:
relationList = betterRelations[key]
betterRelations[key] = sorted(relationList,
key=lambda x: x['confidence'],
reverse=True)
return betterRelations
def _distance_1(misorientation, verbose):
from itertools import combinations_with_replacement as icombinations
s_1, s_2 = misorientation._symmetry
distance = np.empty((misorientation.size, misorientation.size))
index_pairs = icombinations(range(misorientation.size), 2)
if verbose:
from tqdm import tqdm
index_pairs = tqdm(index_pairs, total=misorientation.size**2)
for i, j in index_pairs:
idxi = np.unravel_index(i, misorientation.shape)
idxj = np.unravel_index(j, misorientation.shape)
m_1, m_2 = misorientation[idxi], misorientation[idxj]
mis2orientation = (
s_2.outer(~m_1).outer(s_1).outer(s_1).outer(m_2).outer(s_2))
axis = (0, len(misorientation.shape) + 1,
len(misorientation.shape) + 2, -1)
d = mis2orientation.angle.data.min(axis=axis)
distance[i, j] = d
distance[j, i] = d
return distance
def _distance_1(misorientation, verbose):
warnings.warn("Use _distance_2 instead", DeprecationWarning)
s_1, s_2 = misorientation._symmetry
distance = np.empty((misorientation.size, misorientation.size))
index_pairs = icombinations(range(misorientation.size), 2)
if verbose:
from tqdm import tqdm
index_pairs = tqdm(index_pairs, total=misorientation.size**2)
for i, j in index_pairs:
idxi = np.unravel_index(i, misorientation.shape)
idxj = np.unravel_index(j, misorientation.shape)
m_1, m_2 = misorientation[idxi], misorientation[idxj]
mis2orientation = s_2.outer(~m_1).outer(s_1).outer(s_1).outer(
m_2).outer(s_2)
axis = (0, len(misorientation.shape) + 1,
len(misorientation.shape) + 2, -1)
d = mis2orientation.angle.data.min(axis=axis)
distance[i, j] = d
distance[j, i] = d
return distance
def combinations(lst):
return chain(*(icombinations(lst, i) for i in xrange(1, len(lst)+1)))
def combinations(lst):
return chain(*(icombinations(lst, i) for i in xrange(1, len(lst) + 1)))