<ul><li>A new unsupervised Federated Learning framework called FedKO has been introduced for Multivariate Time-Series (MVTS) anomaly detection.</li><li>FedKO leverages the linear predictive capabilities of Koopman operator theory and the dynamic adaptability of Reservoir Computing.</li><li>It enables effective spatiotemporal processing and privacy preservation for MVTS data in large-scale, distributed environments.</li><li>Experimental results show that FedKO outperforms state-of-the-art methods in MVTS anomaly detection, while also reducing communication size and memory usage.</li></ul>

Federated Koopman-Reservoir Learning for Large-Scale Multivariate Time-Series Anomaly Detection

Discover more