neurotools.util.array module

Helper functions related to Numpy arrays and other indexing tasks.

neurotools.util.array.lmap(function, *args)

Map, collecting results immediately in a list. Also aliased as lap

neurotools.util.array.slist(x)

Convert iterable to sorted list. Also aliased as sls.

neurotools.util.array.amap(function, *args)

Map, collecting results immediately in a Numpy array. This will try to create a numeric or boolean array first, and fall-back to an object array if a ValueError is encountered. Also aliased as aap.

neurotools.util.array.lap(function, *args)

Map, collecting results immediately in a list. Also aliased as lap

neurotools.util.array.aap(function, *args)

Map, collecting results immediately in a Numpy array. This will try to create a numeric or boolean array first, and fall-back to an object array if a ValueError is encountered. Also aliased as aap.

neurotools.util.array.sls(x)

Convert iterable to sorted list. Also aliased as sls.

neurotools.util.array.arraymap(f, *iterables, **kwargs)

Map functionality over numpy arrays replaces depricated arraymap from pylab.

neurotools.util.array.find(x)

Replacement to Pylab’s lost find() function. Synonym for np.where(np.array(x).ravel())[0]

Parameters:

x (np.array)

neurotools.util.array.ezip(*args)

Enumerate and zip, i.e. enumerate(zip(*args))

neurotools.util.array.asiterable(x)

Attempt to convert an iterable object to a list. This mat eventually be replaced with something fancier, but for now just calls list(iter(x)).

Parameters:

x (iterable)

Return type:

list

neurotools.util.array.invert_permutation(p)

Invert a a permutation

Parameters:

x (list of ints) – Permutation to invert

neurotools.util.array.ndargmax(x)

Get coordinates of largest value in a multidimensional array

Parameters:

x (np.array)

neurotools.util.array.complex_to_nan(x, value=nan)

Replce complex entries with NaN or other value

Parameters:

• x (np.array)

• value (float; default np.NaN) – Value to replace complex entries with

neurotools.util.array.verify_axes(x, axis)

neurotools.util.array.axes_complement(x, axes)

Set of all axes indeces for x not contained in axes.

neurotools.util.array.reslice(naxes, expand_into)

Generate a slice object to expand an array along expand_into to broadcast with an array with naxes dimenstions.

neurotools.util.array.make_rebroadcast_slice(x, axis=0, verbose=False)

Generate correct slice object for broadcasting stastistics averaged over the given axis(es) back to the original shape.

Parameters:

• x (np.array)

• axis (int or tuple; default 0)

• verbose (boolean; default False)

neurotools.util.array.deep_tuple(x)

Convert x to tuple, deeply. Defaults to the identity function if x is not iterable

Parameters:

x (nested iterable)

Returns:

x

Return type:

nested iterable

neurotools.util.array.deep_map(f, tree)

Maps over a tree like structure

Parameters:

• f (function)

• tree (nested iterable)

Return type:

nested iterable

neurotools.util.array.to_indices(x)

There are two ways to extract a subset from numpy arrays: 1. providing a boolean array of the same shape 2. providing a list of indecies

This function is designed to accept either, and return a list of indecies.

Parameters:

x (np.array)

neurotools.util.array.onehot(ids, n=None, dense=False)

Generate “one-hot” representation from integer class labels.

Parameters:

• ids (np.array)

• n (int or None; default None) – Total number of labels. If None, will default to the largest value in ids.

• dense (boolean; default False) – Whether to create missing labels. Ignored if the parameter n is specified.

Returns:

• labels – labels corresponding to each index

• r – One-hot label format

neurotools.util.array.zeroslike(x)

Create numpy array of zeros the same shape and type as x

Parameters:

x (np.array)

neurotools.util.array.oneslike(x)

Create numpy array of ones the same shape and type as x

Parameters:

x (np.array)

neurotools.util.array.split_into_groups(x, group_sizes)

Split np.array x into len(group_sizes) groups, with the size of the groups specified by group_sizes.

This operates along the last axis of x

Parameters:

• x (np.array) – Numpy array to split; Last axis should have the same length as sum(group_sizes)

• group_sizes (iterable of positive ints) – Group sizes

Returns:

List of sub-arrays for each group

Return type:

list

neurotools.util.array.maybe_integer(x)

Cast a float numpy array to int32 or int64, if it would not result in loss of precision.

Parameters:

x (np.array)

Returns:

x – np.int32 or np.int64 if possible, otherwise the original value of x.

Return type:

np.array

neurotools.util.array.widths_to_edges(widths, startat=0)

Convert a list of widths into a list of edges delimiting consecutive bands of the given width

Parameters:

• widths (list of numbers) – Width of each band

• startat (number, default 0) – Starting position of bands

Returns:

edges

Return type:

1D np.array

neurotools.util.array.widths_to_limits(widths, startat=0)

Convert a list of integer widths into a list of [a,b) indecies delimiting the concatenated width

Parameters:

• widths (list of integers) – Width of each band

• startat (int; default 0) – Starting index

Returns:

limits – List of [start,stop) indecies

Return type:

N×2 np.int32

neurotools.util.array.centers(edges)

Get center of histogram bins given as a list of edges.

Parameters:

edges (list of numbers) – Edges of histogram bins

Returns:

centers – Center of histogram bins

Return type:

1D np.array

neurotools.util.array.widths_to_centers(widths, startat=0)

Get centers of a consecutive collection of histogram widths.

Parameters:

• widths (list of numbers) – Width of each band

• startat (number, default 0) – Starting position of bands

neurotools.util.array.extract(zerod)

Extract the scalar value from zero-dimensional np.ndarray.

neurotools.util.array.binspace(start, stop, nbins, eps=1e-09)

neurotools.util.array.binto(x, y, start, stop, nbins=50)

neurotools.util.array.remove_nonincreasing(x, y)

neurotools.util.array.remove_nans(*args)

neurotools.util.array.allclose_recursive(a, b)

Version of np.allclose that recurses through a tuple/ list structure until it reaches an np.ndarray to compare via np.allclose as a base-case.