Components of dendrogram

The make_components() function allows obtaining dendrogram components without attaching them to a figure.

import scipy

from pdendro import make_components

X = [[0.0], [1.0], [3.0], [6.0]]
Z = scipy.cluster.hierarchy.linkage(X)

components = make_components(Z, subtrees={2, 4})

The function returns a tuple object containing four items: the first item represents nodes, the second links, the third an axis, and the fourth the other axis.

The following figure shows the items separately.

import plotly.graph_objects as go
import plotly.io as pio

fig = go.Figure()
fig.set_subplots(rows=2, cols=2, subplot_titles=[f"out[{i}]" for i in range(len(components))])
fig.add_trace(components[0], row=1, col=1)
fig.add_traces(list(components[1].values()), rows=1, cols=2)
fig.update_xaxes(components[2], row=2, col=1)
fig.update_yaxes(components[3], row=2, col=2)

# for better clarity
for row, col in [(1, 1), (1, 2), (2, 2)]:
    fig.update_xaxes(matches="x3", visible=False, row=row, col=col)
for row, col in [(1, 1), (1, 2), (2, 1)]:
    fig.update_yaxes(matches="y4", visible=False, row=row, col=col)
for col in [1, 2]:
    fig.add_scatter(x=[], y=[], row=2, col=col)

pio.show(fig)

Note

Tooltips appear for all nodes, though markers are visible only for root nodes of subtrees.

Note

The third and fourth items do not always represent the x- and y-axes. They depend on the direction parameter of the make_components() function.

Additionally, the second item contains separate subtrees.

fig = go.Figure()
fig.set_subplots(
    rows=1,
    cols=len(components[1]),
    subplot_titles=[f"components[1][{node_id}]" for node_id in components[1]],
)
for col, subtree in enumerate(components[1].values(), start=1):
    fig.add_trace(subtree, row=1, col=col)

# for better clarity
fig.update_xaxes(components[2], visible=False, row=1, col=1)
fig.update_yaxes(components[3], visible=False, row=1, col=1)
fig.update_xaxes(matches="x", visible=False, row=1, col=2)
fig.update_yaxes(matches="y", visible=False, row=1, col=2)

pio.show(fig)