mlquantify.visualization

Visualization utilities for quantification.

A small collection of matplotlib *Display classes, following the scikit-learn Display API (from_predictions / from_estimator / from_protocol constructors, a plot method, and stored ax_ / figure_ attributes). The plots are quantification-specific and do not duplicate scikit-learn’s classification/regression displays.

Multiple-sample diagnostics (summarise an evaluation protocol run):

• DiagonalDisplay — true vs. predicted prevalence scatter.

• BiasDisplay — signed-error boxplots, global or binned.

• ErrorByShiftDisplay — error vs. prior-probability shift.

Single-sample displays (inspect one prediction):

• PrevalenceDisplay — per-class predicted prevalence bars.

• ConfidenceRegionDisplay — confidence interval / ternary ellipse.

This subpackage is intentionally not imported by import mlquantify so that matplotlib stays off the top-level import path; import it explicitly:

from mlquantify.visualization import DiagonalDisplay

Multiple-sample displays

DiagonalDisplay	True vs.
BiasDisplay	Boxplots of signed prevalence-estimation error.
ErrorByShiftDisplay	Estimation error as a function of prior-probability shift.

Single-sample displays

PrevalenceDisplay	Bar chart of a single sample's predicted class prevalence.
ConfidenceRegionDisplay	Confidence region around a single prevalence prediction.