DESCRIPTION: This repository contains code and data for uFedHy-DisMTSADD, an unsupervised multivariate time series anomaly detection and diagnosis algorithm based on federated hyernetworks.
The article is under revision and the reviewers' recognition of the work is much appreciated!
Our experiments are divided into seven sections, namely, detection performance, communication overhead analysis, scalability analysis, heterogeneity analysis, anomaly diagnosis, ablation experiments, and parameter sensitivity analysis. All the experiments are performed on 9 datasets. The detection performance is illustrated in the figure.
Run it under the file Hyper_Fed.py:
Hyper_Fed.py --alg Hyper --tsadalg SC Nor-transformer --dataset psm