mcmodels.models.voxel.VoxelConnectivityArray¶

class mcmodels.models.voxel.VoxelConnectivityArray(weights, nodes)[source]¶

Class for implicit construction of the voxel model.

VoxelConnectivityArray is used to perfom analysis on the large (~200,000 by 400,000 element) voxel-scale connectivity matrix on normal* machines (loading the entire matrix would take hundreds of GB of working memory). You can access elements of the VoxelConnectivityArray object just like a list or numpy array, where each call to __getitem__ will implicitly construct the given slice of the array. Additionally, several numpy.ndarray methods have been implemented.

See VoxelModel for weights/nodes descriptions.

Parameters:	weights : array-like, shape (n_voxels, n_exps) Weights matrix from fitted VoxelModel. nodes : array-like, shape (n_exps, n_voxels) Nodes matrix from fitted VoxelModel.

Examples

>>> from mcmodels.core import VoxelModelCache
>>> cache = VoxelModelCache()
>>> voxel_array, source_mask, target_mask = cache.get_voxel_connectivity_array()
>>> # VoxelConnectivityArray has several numpy.ndarray like methods
>>> # get some arbitrary model weights
>>> voxel_array[20:22, 10123]
array([0.000145, 0.000098])
>>> voxel_array.shape
(226346, 448962)
>>> voxel_array.T
VoxelConnectivityArray(dtype=float32, shape=(448962, 226346))

Attributes:	`T` Short for transpose() `dtype` numpy.dtype of full array `shape` numpy.shape of full array `size` numpy.size of full array

Methods

`astype`(self, dtype, \\kwargs)	Consistent with numpy.ndarray.astype.
`from_csv`(weights_file, nodes_file, \\kwargs)	Alternative constructor loading weights, nodes from .csv files.
`from_fitted_voxel_model`(voxel_model)	Alternative constructor using fitted `VoxelModel` object.
`from_npy`(weights_file, nodes_file, \\kwargs)	Alternative constructor loading weights, nodes from npy, npz files.
`itercolumns`(self)	Generator for yielding columns of the voxel matrix.
`itercolumns_blocked`(self[, n_blocks])	Generator for yielding blocked columns of the voxel matrix.
`iterrows`(self)	Generator for yielding rows of the voxel matrix.
`iterrows_blocked`(self, n_blocks)	Generator for yielding blocked rows of the voxel matrix.
`mean`(self[, axis])	Consistent with numpy.ndarray.mean.
`sum`(self[, axis])	Consistent with numpy.ndarray.sum.
`transpose`(self)	Returns transpose of full array.

__init__(self, weights, nodes)[source]¶

Methods

`__init__`(self, weights, nodes)
`astype`(self, dtype, \\kwargs)	Consistent with numpy.ndarray.astype.
`from_csv`(weights_file, nodes_file, \\kwargs)	Alternative constructor loading weights, nodes from .csv files.
`from_fitted_voxel_model`(voxel_model)	Alternative constructor using fitted `VoxelModel` object.
`from_npy`(weights_file, nodes_file, \\kwargs)	Alternative constructor loading weights, nodes from npy, npz files.
`itercolumns`(self)	Generator for yielding columns of the voxel matrix.
`itercolumns_blocked`(self[, n_blocks])	Generator for yielding blocked columns of the voxel matrix.
`iterrows`(self)	Generator for yielding rows of the voxel matrix.
`iterrows_blocked`(self, n_blocks)	Generator for yielding blocked rows of the voxel matrix.
`mean`(self[, axis])	Consistent with numpy.ndarray.mean.
`sum`(self[, axis])	Consistent with numpy.ndarray.sum.
`transpose`(self)	Returns transpose of full array.

T¶: Short for transpose()

astype(self, dtype, **kwargs)[source]¶

Consistent with numpy.ndarray.astype.

see numpy.ndarray.astype for more info.

Parameters:	dtype : string Data type to convert array. **kwargs : Keyword arguments to numpy.ndarray.astype
Returns:	self : VoxelArray with new dtype.

dtype¶: numpy.dtype of full array

classmethod from_csv(weights_file, nodes_file, **kwargs)[source]¶

Alternative constructor loading weights, nodes from .csv files.

Parameters:	weights_file : string or path Path to the .csv file containing the model weights. This file can have .gz or .bz2 compression nodes_file : string or path Path to the .csv file containing the model nodes. This file can have .gz or .bz2 compression **kwargs Optional keyword arguments supplied to numpy.loadtxt
Returns:	An instantiated VoxelConnectivityArray object.

classmethod from_fitted_voxel_model(voxel_model)[source]¶

Alternative constructor using fitted VoxelModel object.

Parameters:	voxel_model : A fitted `VoxelModel` object.
Returns:	An instantiated VoxelConnectivityArray object.

classmethod from_npy(weights_file, nodes_file, **kwargs)[source]¶

Alternative constructor loading weights, nodes from npy, npz files.

Parameters:	weights_file : string or path Path to the .npy or .npz file containing the model weights. nodes_file : string or path Path to the .npy or .npz file containing the model nodes. **kwargs Optional keyword arguments supplied to numpy.load
Returns:	An instantiated VoxelConnectivityArray object.

itercolumns(self)[source]¶

Generator for yielding columns of the voxel matrix.

Yields:	array : shape = (n_rows,) Single column of the voxel-scale connectivity matrix.

itercolumns_blocked(self, n_blocks=0)[source]¶

Generator for yielding blocked columns of the voxel matrix.

Parameters:	n_blocks : int The number of blocks of columns that is wished to be returned. Must be on the interval [1, n_columns].
Yields:	array : A block of columns of the full voxel-scale connectivity matrix.

iterrows(self)[source]¶

Generator for yielding rows of the voxel matrix.

Yields:	array : shape = (n_columns,) Single row of the voxel-scale connectivity matrix.

iterrows_blocked(self, n_blocks)[source]¶

Generator for yielding blocked rows of the voxel matrix.

Parameters:	n_blocks : int The number of blocks of rows that is wished to be returned. Must be on the interval [1, n_rows]
Yields:	array : A block of rows of the full voxel-scale connectivity matrix.

mean(self, axis=None)[source]¶

Consistent with numpy.ndarray.mean.

see numpy.ndarray.mean for more info.

Parameters:	axis - None, int, optional (default=None) Axis over which to take mean.
Returns:	array Mean along axis.

shape¶: numpy.shape of full array

size¶: numpy.size of full array

sum(self, axis=None)[source]¶

Consistent with numpy.ndarray.sum.

see numpy.ndarray.sum for more info.

Parameters:	axis - None, int, optional (default=None) Axis over which to sum.
Returns:	array Sum along axis.

transpose(self)[source]¶: Returns transpose of full array.