ml-vizkit

ml-vizkit provides small, reusable visualization utilities for inspecting, comparing, and explaining trained machine-learning models and completed experiments.

It is designed for analysts who want clear, consistent visual evidence without rewriting common plotting mechanics for every project.

Purpose

Applied machine-learning work repeatedly uses the same kinds of visualizations:

• decision boundaries
• confusion matrices
• prediction errors
• class distributions
• actual-versus-predicted plots
• residual plots
• feature importance
• train/test splits
• model comparisons
• split comparisons

ml-vizkit provides concise functions for these common views so analysts can spend more time interpreting models and experimental results.

Design

ml-vizkit follows a small set of design rules:

• work with already-trained models, predictions, and completed experiment results
• do not train models or choose models, features, metrics, or experimental settings
• use established Python visualization and machine-learning libraries underneath
• return Matplotlib Axes objects rather than rendering automatically
• keep visualizations inspectable, adaptable, and easy to replace
• automate plotting mechanics without reducing analytical agency

The caller retains control over display, composition, annotation, export, and interactive use.

Example

```python from ml_vizkit import show_decision_boundary

ax = show_decision_boundary( model, X_test, y_test, )

ax.set_title("Penguin Species Decision Boundary") ```

See Also

• API