STGAD: Self-temporal generative adversarial framework with transformer attention for unsupervised multivariate time-series anomaly detection and localization

Xiao Liao, Wei Deng, Hongyue Ma, Yihan Mu · 2026

Read STGAD: Self-temporal generative adversarial framework with transformer attention for unsupervised multivariate time-series anomaly detection and localization by Xiao Liao, Wei Deng, Hongyue Ma, Yihan Mu (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/plos-pone-0349223.txt.

Description

Xiao Liao and colleagues present STGAD, a system for spotting anomalies in streams of sensor data where many measurements move together, such as in servers or spacecraft. Nobody labels the failures in advance, so the system must learn what normal looks like and flag departures. Their design pits a generator against a critic in a training contest, with a transformer reading the time patterns.

At test time, each data window is compared with freshly generated normal versions, and the closest match leaves a residual that signals trouble. Two scores, from generation and from the critic, are fused with a threshold that adapts to each dataset. Tests spanned five benchmark sets, from server monitoring to aerospace telemetry, where the system also points to which variables went wrong.

Catalog record

CREATORXiao Liao, Wei Deng, Hongyue Ma, Yihan Mu
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE56098
LICENSECC_BY
RIGHTSCC BY
SOURCEPLOS

Source: https://doi.org/10.1371/journal.pone.0349223

Browse the full library · About the Tower of Babel