Xian-Kun Shi

Anomaly Detection in Long-term Time Series with Transformer and Variational Autoencoder

Transformer with Variational AutoEncoder is presented, a framework that combines a dual-embedding representation pipeline, a Transformer encoder for long-range dependency modeling, and VAE-based latent regularization of encoded features to provide an effective and robust solution for long-term time-series anomaly detec...