Combining the mean and log-variance in this way is called the reparameterization trick. The discovery of this idea in the original 2013 research paper ("Auto-Encoding Variational Bayes" by D.P. Kingma ...

A variational autoencoder (VAE) is a deep neural system that can be used to generate synthetic data. VAEs share some architectural similarities with regular neural autoencoders (AEs) but an AE is not ...