KDEGlyph¶

KDEGlyph evaluates an isotropic Gaussian kernel-density estimate of an (x, y) point cloud on a grid (NumPy only, no scipy) and draws it as filled (shade=True, default) or line density contours.

In [1]:

import numpy as np

from cleopatra.glyphs.stats.kde_glyph import KDEGlyph

rng = np.random.default_rng(0)
x = np.concatenate([rng.normal(-1, 0.6, 400), rng.normal(1.5, 0.5, 300)])
y = np.concatenate([rng.normal(0, 0.6, 400), rng.normal(1.0, 0.5, 300)])
from cleopatra.styling.params import Contour, DataStyle

import numpy as np from cleopatra.glyphs.stats.kde_glyph import KDEGlyph rng = np.random.default_rng(0) x = np.concatenate([rng.normal(-1, 0.6, 400), rng.normal(1.5, 0.5, 300)]) y = np.concatenate([rng.normal(0, 0.6, 400), rng.normal(1.0, 0.5, 300)]) from cleopatra.styling.params import Contour, DataStyle

1. Filled density contours¶

In [2]:

kde = KDEGlyph(x, y, cmap='viridis')
fig, ax, cs = kde.plot(title='Filled KDE')

kde = KDEGlyph(x, y, cmap='viridis') fig, ax, cs = kde.plot(title='Filled KDE')

2. Line contours with a wider kernel¶

In [3]:

kde = KDEGlyph(x, y, shade=False, bw_method=1.4, cmap='inferno')
fig, ax, cs = kde.plot(contour=Contour(levels=12), title='Line KDE')

kde = KDEGlyph(x, y, shade=False, bw_method=1.4, cmap='inferno') fig, ax, cs = kde.plot(contour=Contour(levels=12), title='Line KDE')

3. Relief-shaded density surface (hillshade)¶

A kernel-density estimate is itself a smooth surface -- z = density -- so hillshade can shade it like a DEM, making the density peaks and ridges stand out. It renders the density as a relief-shaded image (with the colorbar on a proxy mappable) instead of contour bands. It is the same hillshade option as the other surface glyphs, reusing the raster (shade_grid) geometry on the density grid.

In [4]:

kde = KDEGlyph(x, y, cmap='magma', gridsize=120)
fig, ax, im = kde.plot(data_style=DataStyle(hillshade={'vert_exag': 25}), title='Relief-shaded density')

kde = KDEGlyph(x, y, cmap='magma', gridsize=120) fig, ax, im = kde.plot(data_style=DataStyle(hillshade={'vert_exag': 25}), title='Relief-shaded density')

The two clusters read as shaded 'hills' -- the taller, denser peak clearly rising above the smaller one -- which a flat colour ramp alone would not convey as vividly.

Data-style presets: style= and apply_style()¶

Colour the density with a named preset. style= works at construction or plot() time; apply_style(name) restyles the glyph in place and glyph.style reads it back. The preset is sticky (a later plain plot() keeps it) and clearable (plot(style=None)). Only continuous presets apply -- a categorical preset is rejected, since a density is continuous.

In [5]:

kde = KDEGlyph(x, y, gridsize=80)
kde.plot(data_style=DataStyle(style='temperature'))
print('style:', kde.style)

kde.apply_style('precipitation')  # restyle the density in place
print('style after apply_style:', kde.style)

kde = KDEGlyph(x, y, gridsize=80) kde.plot(data_style=DataStyle(style='temperature')) print('style:', kde.style) kde.apply_style('precipitation') # restyle the density in place print('style after apply_style:', kde.style)

style: temperature
style after apply_style: precipitation