π™Άπ™³πšπš˜πš›πš–πšŽπš› \mathtt{GDformer} : Going Beyond Subsequence Isolation for Multivariate Time Series Anomaly Detection

Qingxiang Liu, Xiaoliang Luo, Chenghao Liu, Sheng Sun, Di Yao, Lvchun Wang, Wei Yu, Yuxuan Liang Β· 2025

Read π™Άπ™³πšπš˜πš›πš–πšŽπš› \mathtt{GDformer} : Going Beyond Subsequence Isolation for Multivariate Time Series Anomaly Detection by Qingxiang Liu, Xiaoliang Luo, Chenghao Liu, Sheng Sun, Di Yao, Lvchun Wang, Wei Yu, Yuxuan Liang (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-2501-18196.txt.

Description

Qingxiang Liu and seven co-authors present an anomaly detector for multivariate time series in a paper from two thousand twenty-five. With no labeled anomalies to learn from, detectors must draw a tight boundary around normal behavior. Older methods judge short slices in isolation and miss the whole-series picture.

Their model, GDformer, builds a global dictionary of patterns shared by all normal points. Cross-attention maps then show how each point relates to that global picture, giving a natural detection score. Added prototypes sharpen the boundary around normal correlation patterns. On five real-world benchmarks, GDformer reached state-of-the-art results. Its global dictionary also transferred well between data sets, suggesting the shared patterns are broadly reusable.

Catalog record

CREATORQingxiang Liu, Xiaoliang Luo, Chenghao Liu, Sheng Sun, Di Yao, Lvchun Wang, Wei Yu, Yuxuan Liang
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE44 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-18196

Browse the full library Β· About the Tower of Babel