cnextend : Functionality for the hiearchical tree¶

class idpflex.cnextend.ClusterNodeX(*args, **kwargs)[source]¶

Bases: scipy.cluster.hierarchy.ClusterNode

Extension of ClusterNode to accommodate a parent reference and a dictionary-like of properties.

distance_submatrix(dist_mat)[source]¶

Extract matrix of distances between leafs under the node.

Parameters: dist_mat (numpy.ndarray) – Distance matrix (square or in condensed form) among all N leaves of the tree to which the node belongs to. The row indexes of dist_mat must correspond to the node IDs of the leaves.
Returns: square distance matrix MxM between the M leafs under the node
Return type: ndarray

property leaf_ids¶

ID’s of the leafs under the tree, ordered by increasing ID.

Returns
Return type: list

property leafs¶

Find the leaf nodes under this cluster node.

Returns: node leafs ordered by increasing ID
Return type: list

representative(dist_mat, similarity=<function mean>)[source]¶

Find leaf under node that is most similar to all other leaves under the node

Find the leaf that minimizes the similarity between itself and all the other leaves under the node. For instance, the average of all distances between one leaf and all the other leaves results in a similarity scalar for the leaf.

Parameters

dist_mat (ndarray) – condensed or square distance matrix MxM or NxN among all N leaves in the tree or among all M leaves under the node. If dealing with the distance matrix among all leaves in the tree, self.distance_submatrix is first applied.
similarity (function object) – reduction operation on a the list of distances between one leaf and the other (M-1) leaves.

Returns

representative leaf node

Return type

ClusterNodeX

property tree¶

Tree object owning the node

Returns
Return type: Tree

class idpflex.cnextend.Tree(z=None, dm=None)[source]¶

Bases: object

Hierarchical binary tree.

Parameters

z (ndarray) – linkage matrix from which to create the tree. See linkage()
dm (ndarray) – distance matrix from which to create the linkage matrix and tree

from_linkage_matrix(z, node_class=<class 'idpflex.cnextend.ClusterNodeX'>)[source]¶

Refactored to_tree() converts a hierarchical clustering encoded in matrix z (by linkage) into a convenient tree object.

Each node_class instance has a left, right, dist, id, and count attribute. The left and right attributes point to node_class instances that were combined to generate the cluster. If both are None then node_class is a leaf node, its count must be 1, and its distance is meaningless but set to 0.

Parameters

z (ndarray) – linkage matrix. See linkage()
node_class (ClusterNodeX) – the type of nodes composing the tree. Now supports ClusterNodeX and parent class ClusterNode

property leafs¶

Returns: leaf nodes ordered by increasing ID
Return type: list

nodes_above_depth(depth=0)[source]¶

Nodes at or above depth from the root node

Parameters: depth (int) – Depth level starting from the root level (depth=0)
Returns: List of nodes ordered by increasing ID. Last one is the root node
Return type: list

nodes_at_depth(depth=0)[source]¶

Nodes at a given depth from the root node

Parameters: depth (int) – Depth level starting from the root level (depth=0)
Returns: List of nodes corresponding to that particular level
Return type: list

save(filename)[source]¶

Serialize the tree and save to file

Parameters: filename (str) – File name

idpflex.cnextend.load_tree(filename)[source]¶

Load a previously saved tree

Parameters: filename (str) – File name containing the serialized tree
Returns: Tree instance stored in file
Return type: Tree

idpflex.cnextend.random_distance_tree(n_leafs)[source]¶

Instantiate a tree where leafs and nodes have random distances to each other.

Distances randomly retrieved from a flat distribution of numbers between 0 and 1

Parameters

n_leafs (int) – Number of tree leaves

Returns

Elements of the named tuple: - tree: Tree

Tree instance

distance_matrix: ndarray
square distance matrix in between pair of tree leafs

Return type

namedtuple