Skip to content

Visualization

Matplotlib is optional and imported only when one of these functions is called. Install the viz extra or synchronize all development extras.

plot_partition is a geometric view and therefore accepts only effective rank one or two. For higher rank, plot_summary uses a projection-free retained- eigenvalue spectrum, and plot_optimization summarizes center displacement norms across every informative coordinate.

plot_optimization 

plot_optimization(trace: OptimizationTrace) -> Figure

Plot objective, hard retention, occupancies, and center motion.

Parameters:

Returns:

  • Figure –

    Four-panel optimization summary.

plot_partition 

plot_partition(result: QuantizerResult, scores: ArrayLike, weights: ArrayLike | None = None) -> Figure

Plot observations in the fitted informative coordinate system.

Parameters:

  • result (QuantizerResult) –

    Fitted score-space partition.

  • scores (ArrayLike) –

    Raw score matrix compatible with result.

  • weights (ArrayLike | None, default: None ) –

    Optional weights used only to scale marker sizes.

Returns:

  • Figure –

    One- or two-dimensional partition view.

Raises:

  • ValueError –

    If the fitted informative space has rank above two. Use :func:plot_summary for a projection-free diagnostic instead.

plot_information 

plot_information(report: InformationReport) -> Figure

Plot retained information, its spectrum, and bin occupancy.

Parameters:

  • report (InformationReport) –

    Information report for one fixed partition and sample.

Returns:

  • Figure –

    Matrix, eigenvalue, and weighted-occupancy panels. The matrix uses a signed scale so negative off-diagonal values remain visible.

plot_summary 

plot_summary(result: QuantizerResult, scores: ArrayLike, weights: ArrayLike | None = None) -> Figure

Create a compact final-partition and optimization summary.

Parameters:

  • result (QuantizerResult) –

    Fitted score-space partition.

  • scores (ArrayLike) –

    Raw score matrix compatible with result.

  • weights (ArrayLike | None, default: None ) –

    Optional evaluation weights.

Returns:

  • Figure –

    Partition, retained matrix, trace, and occupancy panels.