Module explain

Interpretability layer: drift attribution, trajectory projection, dimension heatmaps.

Provides human-readable explanations of temporal vector evolution:

• DriftAttribution: per-dimension drift analysis with Pareto ranking
• TrajectoryProjector: PCA-based 2D/3D projection
• DimensionHeatmap: time × dimension change intensity matrix

Structs

DimensionContribution

A single dimension’s contribution to overall drift.

DimensionHeatmap

Time × dimension change intensity matrix.

DriftAttribution

Per-dimension drift attribution.

ProjectedPoint

A 2D projected point.

Functions

dimension_heatmap

Build a dimension heatmap from a trajectory.

drift_attribution

Compute drift attribution between two vectors.

power_iteration 🔒

Power iteration for finding a principal component.

project_trajectory_2d

Project a trajectory to 2D using PCA.