Prototype-oriented unsupervised anomaly detection for multivariate time series

Unsupervised anomaly detection (UAD) of mul- tivariate time series (MTS) aims to learn robust representations of normal multivariate temporal patterns. Existing UAD methods try to learn a fixed set of mappings for each MTS, entailing expensive computation and limited model adap- tation. To address this pivotal issue, we propose a prototype-oriented UAD (PUAD) method un- der a probabilistic framework. Specifically, in- stead of learning the mappings for each MTS, the proposed PUAD views multiple MTSs as the dis- tribution over a group of prototypes, which are extracted to represent a diverse set of normal pat- terns. To learn and regulate the prototypes, PUAD introduces a reconstruction-based unsupervised anomaly detection approach, which incorporates a prototype-oriented optimal transport method into a Transformer-powered probabilistic dynamical generative framework. Leveraging meta-learned transferable prototypes, PUAD can achieve high model adaptation capacity for new MTSs. Experi- ments on five public MTS datasets all verify the effectiveness of the proposed UAD method.

Recommended citation: Li, Y., Chen, W., Chen, B., Wang, D., Tian, L., & Zhou, M. (2023, July). Prototype-oriented unsupervised anomaly detection for multivariate time series. In International Conference on Machine Learning (pp. 19407-19424). PMLR. https://openreview.net/pdf?id=3vO4lS6PuF