Crate cvx_analytics

cvx-analytics — Advanced temporal analytics for ChronosVector.

Provides analytical capabilities for understanding vector evolution over time:

• calculus: Vector differential calculus (velocity, acceleration, drift, volatility)
• ode: Neural ODE solver (future)
• pelt: PELT offline change point detection (future)
• bocpd: BOCPD online streaming change point detection (future)

Modules

anchor

Anchor-relative trajectory analysis.

anchor_index

Anchor-space invariant index (RFC-011).

backend

Implementation of the AnalyticsBackend trait.

bocpd

Online change point detection via exponentially weighted statistics.

calculus

Vector differential calculus over temporal trajectories.

cohort

Cohort-level temporal drift analytics.

counterfactual

Counterfactual trajectory analysis.

explain

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

fisher_rao

Fisher-Rao distance on the statistical manifold.

granger

Granger causality testing for embedding trajectories.

motifs

Temporal motif discovery via Matrix Profile.

multiscale

Multi-scale temporal analysis: resampling and per-scale drift.

ode

ODE solver and prediction engine.

pelt

PELT (Pruned Exact Linear Time) offline change point detection.

point_process

Temporal point process features extracted from event timestamps.

procrustes

Procrustes alignment for cross-model embedding comparison (RFC-012 P10).

signatures

Path signatures for trajectory characterization.

temporal_join

Temporal join — find time windows where entities are semantically close.

temporal_ml

Differentiable temporal feature extraction.

topology

Topological features via persistent homology.

trajectory

Trajectory distance measures for comparing entity evolutions.

wasserstein

Wasserstein (optimal transport) distance between distributions.