Bayesian Tensor Autoencoder with Physics-informed Predictive Prior for Multi-dimensional Time Series Anomaly Detection
To incorporate the predictive prior into the reconstruction-based AE, a Bayesian fusion approach is proposed and this approach can enhance the modeling capability of the model for normal data and mitigate the over-generalization problem of AE.
Jia-Nan Liu, Chun-Guang Li
· 0 citations