dendrogram.lattice module

An N-dimensional lattice class with an identify_cluster method.

class dendrogram.lattice.latticeND(data, level)

Bases: object

An N-dimensional lattice class.

property shape

Shape of the lattice.

Type

numpy.ndarray of int

property dim

Dimension of the lattice.

Type

int

property len

Total number of elements in the lattice.

Type

int

property lattice

Area ocupied by the lattice.

Type

numpy.ndarray of int

property label

Label of clusters.

Type

numpy.ndarray of int

identify_cluster()

Identify clusters in the lattice.

A cluster is a group of connected (neighboring) pixels.

Returns

Label of clusters.

Return type

numpy.ndarray of int

dendrogram.structureTree module

A tree class and a makeTree function.

class dendrogram.structureTree.clusterTree(label, mask, isleaf=True)

Bases: object

A tree class.

Note

clusterTree can be either the tree itself or the branch of a tree.

property label

Label of this tree/branch.

Type

int

property mask

Area ocupied by this tree/branch.

Type

numpy.ndarray of bool

property isleaf

Whether this is a leaf or not.

Type

bool

property branches

All branches of the tree. Empty if this is a branch.

Type

dict of int

property children

Children of this branch/tree. Empty if this is a leaf.

Type

dict of int

property parent

Parent of this branch. Empty if this is a tree.

Type

dict of int

create_leaf(label, mask)

Create a new leaf to the tree.

Note

This method does not connect the new leaf to any of the presenting branch!

Parameters

• label (int) – Label of the new leaf.

• mask (numpy.ndarray of bool) – The area ocupied by the new leaf.

merge_branch(label, mask, branch)

Merge many old branches to a new branch.

Note

Here, “merge” means setting the new branch to be the parent of old branches, not removing them!

Parameters

• label (int) – Label of the new branch.

• mask (numpy.ndarray of bool) – The area ocupied by the new branch.

• branch (set or list of int) – Labels of branches to be merged.

merge_final(branch)

Merge final branches to the tree.

Note

Even if there is only one final branch, we still need to merge (or, more properly, “link”) it to the tree. Mind the difference between “tree” and “branch”.

Parameters

branch (set or list of int) – Labels of branches to be merged.

topology(stdout=True)

Print topology of the tree in pure text.

Note

Applying this method to large (with more than 100 branches) or deep (with more than 100 levels) trees is not recommend.

Parameters

stdout (bool) – Whether to print to screen or not.

Returns

Visualized topology of this tree.

Return type

str

dendrogram.structureTree.makeTree(data, min_value, min_delta=0, min_npix=1, num_level=100, print_progress=False)

Make dendrogram tree from N-dimensional data.

Note

Applying this method to large (with more than 100 branches) or deep (with more than 100 levels) trees is not recommend.

Parameters

• data (numpy.ndarray of scalar) – Data to make dendrogram tree.

• min_value (scalar) – Minimum value to consider.

• min_delta (scalar, default to 0) – Lag to be ignored.

• min_npix (int, default to 1) – Minimum number of pixels to form a cluster.

• num_level (int, default to 100) – Number of levels.

• print_progress (bool, default to False) – Whether to print progress or not.

Returns

Tree for the dendrogram.

Return type

clusterTree

Examples

Consider a simple two-dimensional bimodal data, let’s generate the tree with min_value=0:

>>> data = np.array([[2,1], [1,2]])
>>> tree = makeTree(data, min_value=0)

To check the result, we can print the topology of tree:

>>> tp = tree.topology()
└──(-1)
    └──(2)
        ├──(0)
        └──(1)

As expected, two branches 0 and 1 illustrates the bimodality. Similarly, the script below generates dendrogram for a three-peak distribution:

>>> data = np.array([[3,1,1],[1,1,1],[2,1,3]])
>>> tree = makeTree(data, min_value=0)
>>> tp = tree.topology()
└──(-1)
    └──(3)
        ├──(0)
        ├──(1)
        └──(2)