Plotting¶

VBPCApy provides three convenience plotting functions. They require matplotlib, installed via the plot extra.

Install¶

pip install vbpca_py[plot]

Scree plot¶

Shows the explained variance ratio for each component with an optional cumulative line:

from vbpca_py import VBPCA
from vbpca_py.plotting import scree_plot

model = VBPCA(n_components=10, maxiters=200)
model.fit(X)

fig = scree_plot(model, cumulative=True)

Pass an existing axes to embed in a subplot:

import matplotlib.pyplot as plt

fig, ax = plt.subplots()
scree_plot(model, ax=ax, cumulative=True)
ax.set_title("My scree plot")
plt.show()

Loadings bar plot¶

Visualise which features contribute most to a given component:

from vbpca_py.plotting import loadings_barplot

fig = loadings_barplot(model, component=0, top_n=15)

Parameter	Description
`component`	Zero-indexed component to plot (default: 0)
`top_n`	Show only the N highest-magnitude features
`feature_names`	List of feature name strings for the x-axis

Variance explained plot¶

Absolute variance (not ratio) per component, useful for comparing across models:

from vbpca_py.plotting import variance_explained_plot

fig = variance_explained_plot(model)

Customisation¶

All three functions return a matplotlib.figure.Figure and accept an optional ax parameter. Use standard matplotlib API to customise colours, labels, saving, etc.:

fig = scree_plot(model)
fig.savefig("scree.png", dpi=150, bbox_inches="tight")