channel-aware

Proposes CALAD, a channel-aware contrastive learning framework for multivariate time series anomaly detection that uses estimated channel relevance to construct contrastive samples, achieving state-of-the-art performance.