Abstract: Variational autoencoders (VAEs) are generative models which combine deep learning and Bayesian machine learning. The VAEs are trained via minimizing the loss function, and the most popular ...