Book
Open access
Aug 2026
DAMR: Dual Adaptive Multi-Head Representation Learning for Multivariate Time Series Anomaly Detection
DAMR, a novel dual adaptive multi-head representation learning framework for the MTSAD task that achieves significant performance improvements and strong noise robustness and requires far less FLOPs and GPU memory cost than Transformer-based temporal modeling works.
Yi-Ning Wang, Fujun Han, Ke Li et al.
· Proceedings of the 32nd ACM... · 0 citations