Gfm4Mtsad adapts graph foundation models for MTSAD
Original title: Graph Foundation Models Can Become Effective Multivariate Time Series Anomaly Detectors
Gfm4Mtsad bridges pretrained GFMs with time series data using SICA, MAPS and TGEC modules for superior anomaly detection.
Multivariate time series anomaly detection requires modeling temporal dynamics and sensor dependencies. Existing graph-based methods train task-specific networks from scratch. Gfm4Mtsad uses the Structure-and-Information Contract Alignment module to harmonize feature-topology mismatch, Mechanism-Aligned Pseudo-Anomaly Screening to steer unsupervised learning, and Tail-Gated Evidence Calibration to eliminate aggregation distortion. It achieves superior detection performance over existing baselines with two GFM backbones.
- Uses SICA module for structure alignment
- Applies MAPS for pseudo-anomaly screening
- Develops TGEC for evidence calibration
- Achieves superior detection with two GFM backbones
From the original article
1 min read at arxiv.orgMultivariate time series anomaly detection (MTSAD) relies on modeling both temporal dynamics and inter-sensor dependencies to identify complex system faults. Although graph-based detectors explicitly model sensor topologies, they train task-specific networks from scratch on limited target-domain data, remaining vulnerable to spurious correlations and unable to reuse transferable priors.
By Zhuangyan Hu · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.