Drift-Aware Variational Autoencoder-based Anomaly Detection with Two-level Ensembling
Jin Li, Kleanthis Malialis, Christos G. Panayiotou, Marios M. Polycarpou · 2026
Read Drift-Aware Variational Autoencoder-based Anomaly Detection with Two-level Ensembling by Jin Li, Kleanthis Malialis, Christos G. Panayiotou, Marios M. Polycarpou (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-12976.txt.
Description
Streaming data arrive unlabelled and ever-changing, so anomaly detectors must cope with concept drift that erodes scores over time. Jin Li and three co-authors present VAE++ESDD, a method combining incremental learning with two-level ensembling: an ensemble of varying autoencoders predicts anomalies, while an ensemble of number-based drift detectors watches for change. Each detector uses a number-based workings to flag drift.
Tests on real and mock datasets with severely or extremely low anomaly rates and varied drift patterns show the method beating strong plain rivals. It also beats state-of-the-art rivals by a significant margin. For operators monitoring streams where anomalies are rare and conduct shifts, the two-level design offers a practical answer: detect drift explicitly, adapt incrementally, and let ensembles stabilise both tasks at once.
Catalog record
| CREATOR | Jin Li, Kleanthis Malialis, Christos G. Panayiotou, Marios M. Polycarpou |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 83 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-12976