Hyperbolic adaptive spatial-aware multivariate time series anomaly detection
Jiaxin Han, Xuanrong Huo, Yuzhi Xiao, Zhonglin Ye, Yuhui Zheng, Haixing Zhao, Maosong Sun, Zhen Liu · 2026
Read Hyperbolic adaptive spatial-aware multivariate time series anomaly detection by Jiaxin Han, Xuanrong Huo, Yuzhi Xiao, Zhonglin Ye, Yuhui Zheng, Haixing Zhao, Maosong Sun, Zhen Liu (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/PMC13518828.txt.
Description
Jiaxin Han leads an eight-author team proposing a 2026 anomaly detector for multivariate industrial sensor streams. Factories need early warnings from many linked sensors, yet real data shift constantly, true system wiring is unknown, and labeled failures are scarce. Older detectors lean on rigid assumptions and fixed network maps.
Their method pre-trains in hyperbolic space, a curved geometry that fits layered data, to capture cross-variable links. A graph module then learns the connection structure itself and prunes it over time, pinpointing where and when anomalies arise. Attention layers plus a small-sample threshold rule steady the output. Benchmarks showed gains of 10 to 30 percent, with detection, diagnosis, and explanation unified in one framework.
Catalog record
| CREATOR | Jiaxin Han, Xuanrong Huo, Yuzhi Xiao, Zhonglin Ye, Yuhui Zheng, Haixing Zhao, Maosong Sun, Zhen Liu |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 240 KB |
| LICENSE | https://creativecommons.org/licenses/by/4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |