Adversarial Variational Bayes: Unifying Variational Autoencoders and Generative Adversarial Networks
This provides a theoretically justified method for improving VAE training, addressing a bottleneck in generative modeling for machine learning researchers.
The paper tackles the problem of training Variational Autoencoders (VAEs) with limited inference model expressiveness by introducing Adversarial Variational Bayes (AVB), which unifies VAEs and Generative Adversarial Networks (GANs) through a two-player game, achieving exact maximum-likelihood parameter assignments and exact posterior distributions in the nonparametric limit.
Variational Autoencoders (VAEs) are expressive latent variable models that can be used to learn complex probability distributions from training data. However, the quality of the resulting model crucially relies on the expressiveness of the inference model. We introduce Adversarial Variational Bayes (AVB), a technique for training Variational Autoencoders with arbitrarily expressive inference models. We achieve this by introducing an auxiliary discriminative network that allows to rephrase the maximum-likelihood-problem as a two-player game, hence establishing a principled connection between VAEs and Generative Adversarial Networks (GANs). We show that in the nonparametric limit our method yields an exact maximum-likelihood assignment for the parameters of the generative model, as well as the exact posterior distribution over the latent variables given an observation. Contrary to competing approaches which combine VAEs with GANs, our approach has a clear theoretical justification, retains most advantages of standard Variational Autoencoders and is easy to implement.