kdcount package

Subpackages

• kdcount.tests package
  • Submodules
    • kdcount.tests.test_cluster module
    • kdcount.tests.test_correlate module
    • kdcount.tests.test_sphere module
    • kdcount.tests.test_tree module
    • kdcount.tests.test_utils module
  • Module contents

Submodules

• kdcount.cluster module
• kdcount.correlate module
• kdcount.models module
• kdcount.pykdcount module
• kdcount.sphere module
• kdcount.utils module

Module contents

class kdcount.KDAttr

Bases: kdcount.pykdcount.KDAttr

class kdcount.KDNode

Bases: kdcount.pykdcount.KDNode

count(other, r, attrs=None, info={})

Gray & Moore based fast dual tree counting.

r is the edge of bins:

-inf or r[i-1] < count[i] <= r[i]

attrs: None or tuple

if tuple, attrs = (attr_self, attr_other)

Returns: count,

count, weight of attrs is not None

enum(other, rmax, process=None, bunch=100000, **kwargs)

cross correlate with other, for all pairs closer than rmax, iterate.

>>> def process(r, i, j, **kwargs):
>>>    ...

>>> A.enum(... process, **kwargs):
>>>   ...

where r is the distance, i and j are the original input array index of the data. arbitrary args can be passed to process via kwargs.

enumiter(other, rmax, bunch=100000)

cross correlate with other, for all pairs closer than rmax, iterate.

for r, i, j in A.enumiter(...):

...

where r is the distance, i and j are the original input array index of the data.

This uses a thread to convert from KDNode.enum.

fof(linkinglength, out=None)

Friend-of-Friend clustering with linking length.

Returns: the label

integrate(min, max, attr=None)

Calculate the total number of points between [min, max).

If attr is given, also calculate the sum of the weight.

This is a M log(N) operation, where M is the number of min/max queries and N is number of points.

make_forest(chunksize)

Divide a tree branch to a forest, each subtree of size at most chunksize

class kdcount.KDTree(input, boxsize=None, thresh=10)

Bases: kdcount.pykdcount.KDTree

KDTree.root is the root node. The algorithms are implemented as methods of the node.

Parameters:

input : array_like

single or double array of shape (N, ndims).

boxsize : array_like or scalar

If given, the input data is on a torus with periodic boundry. the size of the torus is given by boxsize.

thresh : int

minimal size of a leaf.

kdcount.makeiter(feeder)