<ul><li>Utilizing complex inter-variable causal relationships in multivariate time-series can enhance anomaly detection.</li><li>A new approach called Causality-Aware contrastive learning for Robust multivariate Time-Series (CAROTS) is proposed for MTSAD.</li><li>CAROTS incorporates causality into contrastive learning by using data augmentors to obtain causality-preserving and -disturbing samples for training.</li><li>Experiments on real-world and synthetic datasets show that CAROTS with causal relationships integration improves MTSAD capabilities.</li></ul>

Causality-Aware Contrastive Learning for Robust Multivariate Time-Series Anomaly Detection

Discover more