variational lossy autoencoder (vlae)

An extension of the Variational Autoencoder (VAE) that introduces an additional hierarchy of latent variables, improving the model's ability to capture complex data distributions.