MST-VAE is an unsupervised learning approach for anomaly detection in multivariate time series. Inspired by InterFusion paper, we propose a simple yet effective multi-scale convolution kernels applied ...