multidim.covertree.CoverLevel¶

class multidim.covertree.CoverLevel(covertree, exponent)[source]¶

A thin class to represent one level of the filtration in a CoverTree. A CoverLevel is essentially a collection of dictionaries of adults, friends, children, and other attributes of a particular level.

The various attributes have different orderings, optimized for typical usage and minimal algorithmic complexity.

Notes

The user should never create a CoverLevel directly. Instead, create a CoverTree and access its \(i^{\text{th}}\) level with covertree[i].

Attributes:

Attributes:	covertree : `CoverTree` The CoverTree to which this CoverLevel belongs. pointcloud : `multidim.PointCloud` The PointCloud used to make the CoverTree exponent : int The exponent (that is, index or depth or level) of this CoverLevel in the CoverTree. radius : `numpy.float64` The ball radius T1 : `numpy.float64` The type-1 friends radius T2 : `numpy.float64` The type-2 friends radius T3 : `numpy.float64` The type-3 friends radius adults : list List of adult indices, in order they were born friends1 : `collections.OrderedDict` An ordered dictionary to keep track of type-1 friends. Keyed by the adults, in birth order. The values are lists, in index order. friends2 : `collections.OrderedDict` An ordered dictionary to keep track of type-2 friends. Keyed by the adults, in birth order. The values are lists, in index order. friends3 : `collections.OrderedDict` An ordered dictionary to keep track of type-3 friends. Keyed by the adults, in birth order. The values are lists, in index order. guardians : `numpy.ndarray` An array of `numpy.int64`, which keeps track of the guardians of each point in the underlying `PointCloud`. Adults are their own guardians. predecessor : `collections.OrderedDict` An ordered dictionary to keep track of predecessors of the adults. Keyed by the adults, in birth order. The values are the indices of adults at the previous `CoverLevel`. successors : `collections.OrderedDict` An ordered dictionary to keep track of predecessors of the adults. Keyed by the adults, in birth order. The values are NumPy arrays of indices of adults in the next `CoverLevel`. This is computed only at the next level! children : `collections.OrderedDict` An ordered dictionary to keep track of children1 friends, keyed by the adults, in birth order. The values are NumPy boolean arrays, which allows for easy extraction of subsets of children. weights : `collections.OrderedDict` An ordered dictionary to keep track of total weight of children, keyed by the adults, in birth order. The values are NumPy arrays, with one entry per label. This is computed as part of `cleanup()` entropy : `collections.OrderedDict` An ordered dictionary to keep track of overall entropy of children, keyed by the adults, in birth order. The values are `numpy.float64` numbers, of overall entropy of weights across labels. This is computed and stored via `multidim.models.CDER`, but it otherwise unused.

covertree : CoverTree: The CoverTree to which this CoverLevel belongs.
pointcloud : multidim.PointCloud: The PointCloud used to make the CoverTree
exponent : int: The exponent (that is, index or depth or level) of this CoverLevel in the CoverTree.
radius : numpy.float64: The ball radius
T1 : numpy.float64: The type-1 friends radius
T2 : numpy.float64: The type-2 friends radius
T3 : numpy.float64: The type-3 friends radius
adults : list: List of adult indices, in order they were born
friends1 : collections.OrderedDict: An ordered dictionary to keep track of type-1 friends. Keyed by the adults, in birth order. The values are lists, in index order.
friends2 : collections.OrderedDict: An ordered dictionary to keep track of type-2 friends. Keyed by the adults, in birth order. The values are lists, in index order.
friends3 : collections.OrderedDict: An ordered dictionary to keep track of type-3 friends. Keyed by the adults, in birth order. The values are lists, in index order.
guardians : numpy.ndarray: An array of numpy.int64, which keeps track of the guardians of each point in the underlying PointCloud. Adults are their own guardians.
predecessor : collections.OrderedDict: An ordered dictionary to keep track of predecessors of the adults. Keyed by the adults, in birth order. The values are the indices of adults at the previous CoverLevel.
successors : collections.OrderedDict: An ordered dictionary to keep track of predecessors of the adults. Keyed by the adults, in birth order. The values are NumPy arrays of indices of adults in the next CoverLevel. This is computed only at the next level!
children : collections.OrderedDict: An ordered dictionary to keep track of children1 friends, keyed by the adults, in birth order. The values are NumPy boolean arrays, which allows for easy extraction of subsets of children.
weights : collections.OrderedDict: An ordered dictionary to keep track of total weight of children, keyed by the adults, in birth order. The values are NumPy arrays, with one entry per label. This is computed as part of cleanup()
entropy : collections.OrderedDict: An ordered dictionary to keep track of overall entropy of children, keyed by the adults, in birth order. The values are numpy.float64 numbers, of overall entropy of weights across labels. This is computed and stored via multidim.models.CDER, but it otherwise unused.

`CoverLevel.check`()	Perform basic sanity checks on children, friends, etc.
`CoverLevel.cleanup`()	Internal method – remove duplicate friends, and compute weights and entropy.
`CoverLevel.find_entropy`(adult)	Compute the entropy of the labelled children on an adult.
`CoverLevel.find_label_weights`(adult)	Compute the weights of labelled children of an adult.
`CoverLevel.plot`(canvas[, show_balls, …])	Plot a single level of a `CoverTree`
`CoverLevel.plot_data_title`([show_balls, …])	Internal method – Make source data for plot.

Previous topic

multidim.covertree

Next topic

multidim.covertree.CoverLevel.check