ARGUS: Adaptive Rotation-Invariant Geometric Unsupervised System

TL;DR

Argus reframes drift detection in high-dimensional data streams as a spatial tracking problem on a fixed tessellated latent manifold. It fixes a canonical frame, builds persistent Voronoi cells from representative data, and tracks per-cell statistics (mass, mean, covariance) to derive local drift metrics that aggregate into global and regional indicators. The approach is proven orthogonally invariant under coordinate rotations and scalable to very high dimensions via product quantization, with a graph-based smoothing step that separates coherent drift from shocks. Across synthetic and real-world scenarios, Argus achieves correct drift detection under rotations, precise localization of drift regions, and linear-time per-snapshot performance, offering a principled, geometry-preserving alternative to traditional global tests and costly transport-based methods.

Abstract

Detecting distributional drift in high-dimensional data streams presents fundamental challenges: global comparison methods scale poorly, projection-based approaches lose geometric structure, and re-clustering methods suffer from identity instability. This paper introduces Argus, A framework that reconceptualizes drift detection as tracking local statistics over a fixed spatial partition of the data manifold. The key contributions are fourfold. First, it is proved that Voronoi tessellations over canonical orthonormal frames yield drift metrics that are invariant to orthogonal transformations. The rotations and reflections that preserve Euclidean geometry. Second, it is established that this framework achieves O(N) complexity per snapshot while providing cell-level spatial localization of distributional change. Third, a graph-theoretic characterization of drift propagation is developed that distinguishes coherent distributional shifts from isolated perturbations. Fourth, product quantization tessellation is introduced for scaling to very high dimensions (d>500) by decomposing the space into independent subspaces and aggregating drift signals across subspaces. This paper formalizes the theoretical foundations, proves invariance properties, and presents experimental validation demonstrating that the framework correctly identifies drift under coordinate rotation while existing methods produce false positives. The tessellated approach offers a principled geometric foundation for distribution monitoring that preserves high-dimensional structure without the computational burden of pairwise comparisons.