MA-VAE: Multi-head attention-based variational autoencoder approach for anomaly detection in multivariate time-series applied to automotive endurance powertrain testing
Paper corresponding to source code accepted as a regular paper in the 15th International Conference on Neural Computation Theory and Applications (NCTA 2023).
Scripts can be found under the src folder. The data script outlines the pre-processing of the data. Each model has its own training script. The evaluation script outlines the steps taken to evaluate the models discussed. requirements.txt contains all libraries used.