Statistical foundation of Variational Bayes neural networks

Adversarial Variational Bayes: Unifying Variational Autoencoders and Generative Adversarial Networks

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 mo...

Variational Bayes Solution of Linear Neural Networks and Its Generalization Performance

It is well known that in unidentifiable models, the Bayes estimation provides much better generalization performance than the maximum likelihood (ML) estimation. However, its accurate approximation by Markov chain Monte Carlo methods requires huge computational costs. As an alternative, a tractable approximation method, called the variational Bayes (VB) approach, has recently been proposed and ...

Auto-Encoding Variational Bayes

How can we perform efficient inference and learning in directed probabilistic models, in the presence of continuous latent variables with intractable posterior distributions, and large datasets? We introduce a stochastic variational inference and learning algorithm that scales to large datasets and, under some mild differentiability conditions, even works in the intractable case. Our contributi...

Streaming Variational Bayes

Overview • Large, streaming data sets are increasingly the norm • Inference for Big Data has generally been non-Bayesian • Advantages of Bayes: complex models, coherent treatment of uncertainty, etc. We deliver: • SDA-Bayes, a framework for Streaming, Distributed, Asynchronous Bayesian inference • Experiments demonstrating streaming topic discovery with comparable predictive performance to non-...

Chapter 24 : Variational Bayes