Generative Data by β-Variational Autoencoders Help Build Stronger Classifiers: ECG Use Case