Pi-Transformer: A Physics-informed Attention Mechanism for Time Series Anomaly Detection
Sepehr Maleki, Negar Pourmoazemi · 2025
Read Pi-Transformer: A Physics-informed Attention Mechanism for Time Series Anomaly Detection by Sepehr Maleki, Negar Pourmoazemi (2025) 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-2509-19985.txt.
Description
Anomalies in sensor networks often hide in timing and coordination, not in single wild values. Sepehr Maleki and Negar Pourmoazemi build Pi-Transformer, which watches multivariate series with two attentions at once. One pathway learns from data; the other follows a smooth prior encoding habits such as self-similarity across scales and phase synchrony. The prior ignores amplitude, making it a steady temporal reference.
Training pairs reconstruction with a divergence term that keeps the two attentions close yet distinct. At test time, an alignment-weighted energy score merges with a mismatch score that flags phase and timing breaks. Industrial monitoring and health telemetry both produce such series. A prior that knows normal rhythm can catch disorders that value-only detectors miss.
Catalog record
| CREATOR | Sepehr Maleki, Negar Pourmoazemi |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 57 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-19985