Principle:Interpretml Interpret Interactive Visualization Rendering
| Metadata | |
|---|---|
| Sources | InterpretML, InterpretML Docs |
| Domains | Visualization, User_Interface |
| Last Updated | 2026-02-07 12:00 GMT |
Overview
A rendering mechanism that converts explanation data into interactive visual representations for exploration and analysis.
Description
Interactive Visualization Rendering takes Explanation objects (from explain_global or explain_local) and renders them as interactive charts. For EBMs, global explanations produce continuous bar charts (shape functions) and horizontal bar charts (feature importances). Local explanations produce horizontal bar charts showing per-sample feature contributions. The rendering system supports Plotly for interactive charts, Matplotlib as fallback, and a custom Dash-based dashboard for comparing multiple explanations. The key=-1 convention shows the overall summary view.
Usage
Use this principle when you need to present model explanations to users in an interactive, explorable format. It applies to any Explanation object produced by the interpret library.
Theoretical Basis
The visualization rendering follows a provider pattern:
- Receive Explanation object with data_dicts, selector, and visualization type
- Delegate to the detected provider (InlineProvider or DashProvider)
- Provider renders the appropriate chart type based on explanation metadata
- Charts are interactive (hover, zoom, pan) in Plotly-backed environments