def data(self, data):
data = self._preprocess_data(data)
if isinstance(data, dict):
labeled_by = data.pop('labeled_by', 'rows')
data_contiguous = data.pop('contiguous', None)
if data_contiguous is not None:
self.data = data_contiguous
if len(data) == 0:
return
if labeled_by not in ('rows', 'columns'):
raise ValueError(
'when data provided as a dictionary of '
'matrices, the optional dictionary entry '
'`labeled_by` must be one of `columns` or `rows` '
'(default)')
if not all(sparse_or_dense(m) for m in data.values()):
raise TypeError('when data provided as a dictionary of '
'matrices, each value must be one of the '
'following matrix types: {}'
''.format(CONRAD_MATRIX_TYPES))
if labeled_by == 'columns':
for mat in data.values():
rows = mat.shape[0]
break
if not all([m.shape[0] == rows for m in data.values()]):
raise ValueError('all submatrices must have consistent '
'number of rows when a dictionary of '
'horizontally concatenable matrices is '
'provided')
columns = sum([m.shape[1] for m in data.values()])
else:
rows = sum([m.shape[0] for m in data.values()])
for mat in data.values():
columns = mat.shape[1]
break
if not all([m.shape[1] == columns for m in data.values()]):
raise ValueError('all submatrices must have consistent '
'number of columns when a dictionary of '
'vertically concatenable matrices is '
'provided')
self.__shape_check((rows, columns))
self.__dim1, self.__dim2 = rows, columns
if labeled_by == 'columns':
self.__column_slices.update(data)
else:
self.__row_slices.update(data)
else:
if not sparse_or_dense(data):
raise TypeError('when data provided as a singleton matrix, '
'it must be formatted as one of the following '
'matrix types {}'.format(CONRAD_MATRIX_TYPES))
self.__shape_check(data.shape)
self.__dim1, self.__dim2 = data.shape
self.__data = data
def frame0_to_1(self, in_): """ Allocating version of :meth`AbstractMapping.frame1_to_0`. If input array is a matrix, mapping operates on rows. If array is a vector, mapping operates on entries. Arguments: in_: Input array with :attr:`AbstractMapping.n_frame0` rows and ``k`` >= ``1`` columns. Returns: :class:`numpy.ndarray`: Array with :attr:`AbstractMapping.n_frame0` rows and same number of columns as input. Input entries are mapped one-to-one or many-to-one *into* output. """ if is_vector(in_): out_ = zeros(self.__n_frame0) elif sparse_or_dense(in_): dim1 = self.__n_frame0 dim2 = in_.shape[1] out_ = zeros((dim1, dim2)) return self.frame1_to_0_inplace(in_, out_)
def frame0_to_1(self, in_): """ Map elements of array ``in_`` to a new array. If input array is a matrix, mapping operates on rows. If array is a vector, mapping operates on entries. A new empty vector or matrix Arguments: in_: Input array with :attr:`AbstractMapping.n_frame0` rows and ``k`` >= ``1`` columns. Returns: :class:`numpy.ndarray`: Array with :attr:`AbstractMapping.n_frame1` rows and ``k`` columns. Input entries are mapped one-to-one or one-to-many *into* output. """ if is_vector(in_): out_ = zeros(self.__n_frame1) elif sparse_or_dense(in_): dim1 = self.__n_frame1 dim2 = in_.shape[1] out_ = zeros((dim1, dim2)) return self.frame0_to_1_inplace(in_, out_)
def to_unsafe_data(self, unsafe_file_entry):
if isinstance(unsafe_file_entry, dict):
unsafe_file_entry = UnsafeFileEntry(**unsafe_file_entry)
if not isinstance(unsafe_file_entry, UnsafeFileEntry):
raise TypeError('input should be of type (or parsable as) {}'
''.format(UnsafeFileEntry))
if not unsafe_file_entry.complete:
raise ValueError(
'no numpy file assigned to {}'.format(UnsafeFileEntry))
typed_data = self.read_all(unsafe_file_entry.file)
if isinstance(typed_data, np.ndarray):
return typed_data
if sparse_or_dense(typed_data):
return typed_data
if isinstance(typed_data, dict):
sme = SparseMatrixEntry(**typed_data)
if sme.complete:
return self.to_sparse_matrix(sme)
dme = DenseMatrixEntry(**typed_data)
if dme.complete:
return self.to_dense_matrix(dme)
ve = VectorEntry(**typed_data)
if ve.complete:
return self.to_vector(ve)
return typed_data
def A_mean(self, A_mean=None):
if A_mean is not None:
if not isinstance(A_mean, np.ndarray):
raise TypeError('if argument "A_mean" is provided, it must be '
'of type {}'.format(np.ndarray))
elif not A_mean.size in A_mean.shape:
raise ValueError(
'if argument "A_mean" is provided, it must be '
'a row or column vector. shape of argument: {}'
''.format(A_mean.shape))
else:
if self.__A_full is not None:
if len(A_mean) != self.__A_full.shape[1]:
raise ValueError(
'field "A_full" already set; proposed '
'value for "A_mean" must have same '
'number of entries ({}) as columns in '
'A_full ({})'.format(len(A_mean),
self.__A_full.shape[1]))
self.__A_mean = vec(A_mean)
elif self.__A_full is not None:
if not sparse_or_dense(self.A_full):
raise TypeError('cannot calculate structure.A_mean from'
'structure.A_full: A_full must be one of '
'({},{},{})'.format(np.ndarray, sp.csc_matrix,
sp.csr_matrix))
else:
if isinstance(self.A_full, np.ndarray):
self.__A_mean = np.dot(self.voxel_weights, self.A_full)
else:
self.__A_mean = vec(self.voxel_weights * self.A_full)
self.__A_mean /= float(self.weighted_size)
def A_mean(self, A_mean=None):
if A_mean is not None:
if not isinstance(A_mean, ndarray):
raise TypeError('if argument "A_mean" is provided, it '
'must be of type {}'.format(ndarray))
elif not A_mean.size in A_mean.shape:
raise ValueError('if argument "A_mean" is provided, it must be'
' a row or column vector. shape of argument: '
'{}'.format(A_mean.shape))
else:
if self.__A_full is not None:
if len(A_mean) != self.__A_full.shape[1]:
raise ValueError('field "A_full" already set; '
'proposed value for "A_mean" '
'must have same number of entries '
'({}) as columns in A_full ({})'
''.format(len(A_mean),
self.__A_full.shape[1]))
self.__A_mean = vec(A_mean)
elif self.__A_full is not None:
if not sparse_or_dense(self.A_full):
raise TypeError('cannot calculate structure.A_mean from'
'structure.A_full: A_full must be one of'
' ({},{},{})'.format(ndarray, csc_matrix,
csr_matrix))
else:
self.__A_mean = self.A_full.sum(0) / self.A_full.shape[0]
if not isinstance(self.A_full, ndarray):
self.__A_mean = vec(self.__A_mean)
def frame1_to_0(self, in_): """ Allocating version of :meth`DiscreteMapping.frame1_to_0`. If input array is a matrix, mapping operates on rows. If array is a vector, mapping operates on entries. Arguments: in_: Input array with :attr:`DiscreteMapping.n_frame0` rows and ``k`` >= ``1`` columns. Returns: :class:`numpy.ndarray`: Array with :attr:`DiscreteMapping.n_frame0` rows and same number of columns as input. Input entries are mapped one-to-one or many-to-one *into* output. """ if is_vector(in_): out_ = np.zeros(self.__n_frame0) elif sparse_or_dense(in_): dim1 = self.__n_frame0 dim2 = in_.shape[1] out_ = np.zeros((dim1, dim2)) return self.frame1_to_0_inplace(in_, out_)
def frame0_to_1(self, in_): """ Map elements of array ``in_`` to a new array. If input array is a matrix, mapping operates on rows. If array is a vector, mapping operates on entries. A new empty vector or matrix Arguments: in_: Input array with :attr:`DiscreteMapping.n_frame0` rows and ``k`` >= ``1`` columns. Returns: :class:`numpy.ndarray`: Array with :attr:`DiscreteMapping.n_frame1` rows and ``k`` columns. Input entries are mapped one-to-one or one-to-many *into* output. """ if is_vector(in_): out_ = np.zeros(self.__n_frame1) elif sparse_or_dense(in_): dim1 = self.__n_frame1 dim2 = in_.shape[1] out_ = np.zeros((dim1, dim2)) return self.frame0_to_1_inplace(in_, out_)
def write_matrix(self, directory, name, matrix, overwrite=False):
if not sparse_or_dense(matrix):
raise TypeError('matrix to save must be one of {}'
''.format(CONRAD_MATRIX_TYPES))
if isinstance(matrix, np.ndarray):
return self.write_dense_matrix(directory, name, matrix, overwrite)
else:
return self.write_sparse_matrix(directory, name, matrix, overwrite)
def plannable(self):
"""
True if structure's attached data is sufficient for optimization.
Minimum requirements:
- Structure size determined, and
- Dose matrix assigned, *or*
- Structure collapsable and mean dose matrix assigned.
"""
size_determined = positive_real_valued(self.size)
full_mat_usable = sparse_or_dense(self.A_full)
if full_mat_usable:
full_mat_usable &= self.size == self.A_full.shape[0]
collapsed_mat_usable = bool(
isinstance(self.A_mean, np.ndarray) and self.collapsable)
usable_matrix_loaded = full_mat_usable or collapsed_mat_usable
return size_determined and usable_matrix_loaded
def plannable(self): """ True if structure's attached data is sufficient for optimization. Minimum requirements: - Structure size determined, and - Dose matrix assigned, *or* - Structure collapsable and mean dose matrix assigned. """ size_set = positive_real_valued(self.size) full_mat_usable = sparse_or_dense(self.A_full) if full_mat_usable: full_mat_usable &= self.size == self.A_full.shape[0] collapsed_mat_usable = bool( isinstance(self.A_mean, ndarray) and self.collapsable) usable_matrix_loaded = full_mat_usable or collapsed_mat_usable return size_set and usable_matrix_loaded
def A_full(self, A_full):
if A_full is None:
return
# verify type of A_full
if not sparse_or_dense(A_full):
raise TypeError('input A must by a numpy or scipy csr/csc '
'sparse matrix')
if self.size is not None:
if A_full.shape[0] != self.size:
raise ValueError('# rows of "A_full" must correspond to '
'value of property size ({}) of {} '
'object'.format(self.size, Structure))
else:
self.size = A_full.shape[0]
self.__A_full = A_full
# Pass "None" to self.A_mean setter to trigger calculation of
# mean dose matrix from full dose matrix.
self.A_mean = None
def A_full(self, A_full):
if A_full is None:
return
# verify type of A_full
if not sparse_or_dense(A_full):
raise TypeError('input A must by a numpy or scipy csr/csc '
'sparse matrix')
if self.size is not None:
if A_full.shape[0] != self.size:
raise ValueError('# rows of "A_full" must correspond to '
'value of property size ({}) of {} '
'object'.format(self.size, Structure))
else:
self.size = A_full.shape[0]
self.__A_full = A_full
# Pass "None" to self.A_mean setter to trigger calculation of
# mean dose matrix from full dose matrix.
self.A_mean = None
def frame0_to_1_inplace(self, in_, out_, clear_output=False):
"""
Map elements of array ``in_`` to elements of array ``out_``.
Procedure::
# let forward_map be the map: SET_0 --> SET_1.
# for each INDEX_0, INDEX_1 in forward_map, do
# out_[INDEX_1] += in_[INDEX_0]
# end for
If arrays are matrices, mapping operates on rows. If arrays are
vectors, mapping operates on entries.
Arguments:
in_: Input array with :attr:`AbstractMapping.n_frame0` rows
and ``k`` >= ``1`` columns.
out_: Output array with :attr:`AbstractMapping.n_frame1`
rows and ``k`` >= 1 columns. Modified in-place.
clear_output (:obj:`bool`, optional): If ``True``, set
output array to ``0`` before adding input values.
Returns:
Vector `out_`, after in-place modification.
Raises:
TypeError: If input and output arrays are not (jointly)
vectors or matrices.
ValueError: If input and output array dimensions are not
compatible with each other or consistent with the
dimensions of the mapping.
"""
vector_processing = is_vector(in_) and is_vector(out_)
matrix_processing = sparse_or_dense(in_) and sparse_or_dense(out_)
if not vector_processing or matrix_processing:
raise TypeError('arguments "in_" and "out_" be numpy or '
'scipy vectors or matrices')
dim_in1 = in_.shape[0]
dim_out1 = out_.shape[0]
if vector_processing:
dim_in2 = 1
dim_out2 = 1
else:
dim_in2 = in_.shape[1]
dim_out2 = out_.shape[1]
if bool(
dim_in_1 != self.n_frame0 or
dim_out1 != self.n_frame1 or
dim_in2 != dim_out2):
raise ValueError('arguments "in_" and "out_" be vectors or '
'matrices of dimensions M x N, and K x N, '
'respectively, with:\nM = {}\nK={}\n'
'Provided:\n input: {}x{}\noutput: {}x{}'
''.format(self.n_frame0, self.n_frame1,
dim_in1, dim_in2, dim_out1, dim_out2))
if clear_output:
out_ *= 0
if vector_processing:
for idx_0, idx_1 in enumerate(self.vec):
out_[idx_1] += in_[idx_0]
if matrix_processing:
for idx_0, idx_1 in enumerate(self.vec):
out_[idx_1, :] += in_[idx_0, :]
return out_
def frame0_to_1_inplace(self, in_, out_, clear_output=False):
"""
Map elements of array ``in_`` to elements of array ``out_``.
Procedure::
# let forward_map be the map: SET_0 --> SET_1.
# for each INDEX_0, INDEX_1 in forward_map, do
# out_[INDEX_1] += in_[INDEX_0]
# end for
If arrays are matrices, mapping operates on rows. If arrays are
vectors, mapping operates on entries.
Arguments:
in_: Input array with :attr:`DiscreteMapping.n_frame0` rows
and ``k`` >= ``1`` columns.
out_: Output array with :attr:`DiscreteMapping.n_frame1`
rows and ``k`` >= 1 columns. Modified in-place.
clear_output (:obj:`bool`, optional): If ``True``, set
output array to ``0`` before adding input values.
Returns:
Vector `out_`, after in-place modification.
Raises:
TypeError: If input and output arrays are not (jointly)
vectors or matrices.
ValueError: If input and output array dimensions are not
compatible with each other or consistent with the
dimensions of the mapping.
"""
vector_processing = is_vector(in_) and is_vector(out_)
matrix_processing = sparse_or_dense(in_) and sparse_or_dense(out_)
if not (vector_processing or matrix_processing):
raise TypeError('arguments "in_" and "out_" be numpy or '
'scipy vectors or matrices')
dim_in1 = in_.shape[0]
dim_out1 = out_.shape[0]
if vector_processing:
dim_in2 = 1
dim_out2 = 1
else:
dim_in2 = in_.shape[1]
dim_out2 = out_.shape[1]
if bool(
dim_in1 != self.n_frame0 or
dim_out1 != self.n_frame1 or
dim_in2 != dim_out2):
raise ValueError('arguments "in_" and "out_" be vectors or '
'matrices of dimensions M x N, and K x N, '
'respectively, with:\nM = {}\nK={}\n'
'Provided:\n input: {}x{}\noutput: {}x{}'
''.format(self.n_frame0, self.n_frame1,
dim_in1, dim_in2, dim_out1, dim_out2))
if clear_output:
out_ *= 0
if vector_processing:
for idx_0, idx_1 in enumerate(self.vec):
out_[idx_1] += in_[idx_0]
else:
for idx_0, idx_1 in enumerate(self.vec):
out_[idx_1, :] += in_[idx_0, :]
# TODO: use efficient slicing when input is sparse. warn when output sparse??
return out_