Variational Autoencoder (VAE) has been proposed for two years. During the past two years, some good papers related variational autoencoder come up time to time. And I think it is a good tool worth investigating. Recently, I decide to do something about collaborative recommendation with cross-modality multimedia content using Bayesian deep learning. I think VAE could be a good help. In this post, I’ll investigate and explain VAE in my way.