VISTA: Unsupervised 2D Temporal Dependency Representations for Time Series Anomaly Detection Journal: Pattern Recognition
Sinchee Chin, Fan Zhang, Xiaochen Yang, Jing-Hao Xue, Wenming Yang, Peng Jia, Guijin Wang, Luo Yingqun · 2025
Read VISTA: Unsupervised 2D Temporal Dependency Representations for Time Series Anomaly Detection Journal: Pattern Recognition by Sinchee Chin, Fan Zhang, Xiaochen Yang, Jing-Hao Xue, Wenming Yang, Peng Jia, Guijin Wang, Luo Yingqun (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-2504-02498.txt.
Description
Finding anomalies in streams of measurements, a broken sensor or a failing machine, matters in industry, yet labeled examples are rare and real data is noisy. Sinchee Chin and co-authors present VISTA, a detection method that needs no training. It first splits each series into trend, seasonal, and leftover parts to strip out noise. Then it turns the one-dimensional series into a two-dimensional map of how time points relate to each other, which exposes patterns that flat views miss.
A third module folds information across variables into one compact representation using a pretrained feature extractor. The authors report top results on five standard anomaly benchmarks. Because nothing is trained, the method deploys quickly and its settings are easy to tune, which they argue suits factory and monitoring use.
Catalog record
| CREATOR | Sinchee Chin, Fan Zhang, Xiaochen Yang, Jing-Hao Xue, Wenming Yang, Peng Jia, Guijin Wang, Luo Yingqun |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 50 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2504-02498